3.2 Estimation and Statistical Significance
The sampling error of an estimate is the error caused by the selection of a sample instead of interviewing every person in the population. The sampling error may be reduced by selecting a large sample and/or by using efficient sample design and estimation strategies (such as stratification, optimal allocation, and ratio estimation). The use of probability sampling methods in NSDUH allows estimation of sampling error from the survey data.
Estimates based on NSDUH data are presented in national reports and tables available at https://www.samhsa.gov/data/data-we-collect/nsduh-national-survey-drug-use-and-health. The national estimates, along with the associated standard errors (SEs, which are the square roots of the variances), were computed for all tables using a multiprocedure package, SUDAAN® Software for Statistical Analysis of Correlated Data. This software uses a Taylor series linearization approach to account for the effects of NSDUH’s complex sample design features in estimating the SEs (RTI International, 2013). The SEs can be used to identify unreliable estimates and also to test for the statistical significance of differences between estimates. The final, nonresponse-adjusted, and poststratified analysis weights were used in SUDAAN to compute unbiased, design-based estimates. See Section 2.3.4 for more information on the analysis weights.
3.2.1 Variance Estimation for Estimated Numbers of People
The variances and SEs of estimates of both means and proportions can be calculated reasonably well in SUDAAN using a Taylor series linearization approach. Estimates of proportions, 
(such as drug use prevalence estimates), for a domain (i.e., subgroups for which the estimates were calculated) d, can be expressed as a ratio estimate,
where 
is a linear statistic estimating the number of people with the characteristic of interest (e.g., substance users) in the domain d, and 
is a linear statistic estimating the total number of people in domain d (including people with or without the characteristic of interest, such as substance users and nonusers). The SUDAAN software package is used to calculate direct estimates of and (and, therefore, ) and can also be used to estimate their respective SEs. A Taylor series approximation method implemented in SUDAAN provides the estimate for the SE of .
When the domain size, , is assumed to be free of sampling error because the domain size has been fixed (see below), the following formula is an alternative to using SUDAAN to estimate the SE for the total number of people with a characteristic of interest (e.g., substance users):
This alternative SE estimation method is theoretically correct when the domain size estimates, , are fixed (i.e., among those domains forced to match their respective U.S. Census Bureau or ACS population estimates through the weight calibration process). In these situations, is not subject to a sampling error induced by the NSDUH design. That is, the U.S. Census Bureau and ACS population estimates are assumed to be free of sampling error induced by the NSDUH design. Section 2.3.4 in this report contains further information about the weight calibration process. In addition, more detailed information about the weighting procedures for 2022 will appear in the 2022 National Survey on Drug Use and Health Methodological Resource Book (Center for Behavioral Health Statistics and Quality [CBHSQ], forthcoming a). Until that volume becomes available, refer to the 2021 National Survey on Drug Use and Health Methodological Resource Book (CBHSQ, 2023a) for general weighting information; however, specific information about the 2022 weighting procedures will not be available in the 2021 volume.
For an estimated number where the domain is nonfixed (i.e., where domain size estimates are not forced to match the U.S. Census Bureau or ACS population estimates), this alternative SE estimation method may still provide a good approximation if it can be assumed the sampling variation in is negligible relative to the sampling variation in . This is a reasonable assumption for many NSDUH analyses.
For some subsets of domain estimates, using this alternative SE estimation method where domain sizes are nonfixed yielded an underestimate of the SE of the total when was subject to considerable variation. Because of this underestimation, the alternative SE estimation method was not implemented for NSDUH products when was nonfixed.
For the 2022 NSDUH, a “mixed-method” approach was used for all detailed tables to improve the accuracy of SEs for the estimated numbers of people and to better reflect the effects of poststratification on the variance of the total estimated numbers of people. This approach assigns the methods of SE calculation to domains within tables so that all estimates among a select set of domains with fixed were calculated using the alternative SE estimation method, and all other estimates were calculated directly in SUDAAN, regardless of what the other estimates are within the same table. The set of domains with a fixed was restricted to main effects and two-way interactions to maintain continuity between years.35 The use of the mixed method did not affect the SE estimates for the corresponding proportions presented in the same sets of tables because all SEs for means and proportions are calculated directly in SUDAAN.
Table 3.1 includes the domains that employed the alternative SE estimation method, including the main domains and the two-way interactions.36 However, Table 3.1 does not include an exhaustive list of domains and interactions that are included in NSDUH reports and detailed tables. For domains not included in Table 3.1, SEs for the estimates of totals are calculated directly in SUDAAN. For example, Tables 6.2 and 6.5 in the Results from the 2022 National Survey on Drug Use and Health: Detailed Tables (CBHSQ, 2023k) present estimates of any mental illness (AMI) and serious mental illness (SMI), respectively, among adults aged 18 or older within the domains of gender, Hispanic origin and race, and current employment. The estimated numbers of adults with AMI or SMI among the total population and age group (age group is the main effect), males and females (age group by gender interaction), and people who were Hispanic or not Hispanic (age group by Hispanic origin interaction) used the alternative SE estimation method to calculate the SEs.
The SEs for all other estimated numbers of people in Tables 6.2 and 6.5 in the 2022 Detailed Tables, including current employment, were calculated directly in SUDAAN because estimates presented in the 2022 Detailed Tables for racial groups are among people who were not Hispanic, unless noted otherwise. For example, the SEs by age group for White people were calculated directly in SUDAAN. The domain for White people is actually for White people who were not Hispanic and is a two-way interaction. Therefore, age group for White people is considered a three-way interaction, and the SEs by age group for White people were calculated directly in SUDAAN. Current employment is also not a fixed domain, and the SE of the estimated number of people would be calculated directly in SUDAAN.
3.2.2 Suppression Criteria for Unreliable Estimates
As has been done in past survey years, direct estimates from NSDUH designated as unreliable are not shown in reports or tables and are noted by asterisks (*). The criteria used to define unreliability of direct estimates from NSDUH are based on the following:
- prevalence (for proportion estimates),
- relative standard error (RSE, defined as the ratio of the SE over the estimate),
- nominal (actual) sample size, and
- effective sample size for each estimate.
These suppression criteria for various NSDUH estimates are summarized in Table 3.2.
3.2.2.1 Suppression Rules for Proportions
The suppression rule applied to the 2022 direct estimates involved suppressing proportions under the following specific situations:
- Extreme prevalence estimates were suppressed if they were close to 0 or 100 percent (i.e., if
< .00005 or > .99995).
- Proportion estimates (), or rates, within the ranges (0.00005, 0.05) and (0.95, 0.99995) were suppressed if37
or
- Proportion estimates between 0.05 and 0.95 were suppressed when Effective n < 68, where
, or
- Any proportion estimates in the range (0,1) were suppressed if the domain size n < 100.
Estimates are suppressed if any of these rules apply. The extreme proportion rule (Rule 1) takes precedence over the other rules for < 0.00005 or > 0.99995. In addition, the RSE and effective n rules (Rules 2 and 3) are interrelated. The two rules are identical for = 0.05, 0.5, or 0.95. The effective n rule (Rule 3) is more stringent for 0.05 < < 0.95, and the RSE rule (Rule 2) is more stringent for < 0.05 or > 0.95. The more stringent of these two rules is used for suppressing estimates, depending on the proportion. Finally, the minimum sample size rule (Rule 4) also can take precedence over the RSE rule.
Under Rule 1, prevalence estimates were suppressed if they were close to 0 or 100 percent (i.e., if < .00005 or > .99995). Extremely small or large proportions are often unreliable and may disclose information about a small number of respondents.
Rule 2 suppresses prevalence estimates with relatively large SEs that indicate low precision. The suppression rule for proportions based on 
is used because it ensures a more uniformly stringent application across the whole range of (i.e., from 0 to 1). The separate formulas for 
and 
produce a symmetric suppression rule; that is, if is suppressed, 
will be suppressed as well (Figure 3.1).38 In contrast, the commonly used rule that suppresses estimates when 
imposes a stringent application for suppressing estimates when is small but imposes a lax application when is large. In addition, a rule based only on is asymmetric in the sense that suppression occurs only in terms of ; that is, there is no complementary rule for ().
Rule 3 suppresses prevalence estimates when there is not enough effective sample size, which is another indicator of low precision. The threshold of .175 in Rule 3 was chosen because it equates with a suppression threshold in Rule 2 based on an effective sample size of 68 when = .05, .50, or .95. If the threshold in Rule 3 were increased, then that would equate with a lower suppression threshold based on an effective sample size, and vice versa.
Figure 3.1 also illustrates how this suppression rule can equivalently be expressed as a suppression rule based on the effective sample size as a function of . The figure shows that when .05 < < .95, the symmetric properties of the rule produce a local minimum effective sample size of 50 at = .2 and = .8; however, as moves away from these two points, then the suppression threshold increases to a maximum of an effective sample size of 68 reached at = .05 or .95, or at the local maximum, = .50. Therefore, to simplify requirements and maintain a conservative suppression rule, estimates of between .05 and .95 were suppressed if they had an effective sample size below 68 (indicated by a horizontal line at 68 in Figure 3.1); the suppression rule was left unchanged for estimates of outside of this range, which will require increasingly larger effective sample sizes in order to avoid suppression. For example, an effective sample size of 153, 232, and 684 is needed when = .01, .005, and .001, respectively.
In addition to the criteria based on effective sample size, estimates were also suppressed under Rule 4 if the nominal sample size was less than 100. This is done to protect against unreliable estimates caused by small design effects and small nominal sample sizes; Table 3.2 shows a formula for calculating design effects.
3.2.2.2 Suppression Rules for Estimated Totals, Means, and Sample Sizes
Estimates of totals (i.e., estimated numbers of people) were suppressed if the corresponding prevalence rates were suppressed. Because of this rule, data users may encounter some unexpected results after applying the suppression rules. For instance, equivalent estimates of totals corresponding to different estimated percentages, , are suppressed differently.
Consider a situation where estimates of the misuse of prescription drugs in the past year are presented among the population aged 12 or older and among people who used prescription drugs for any reason in the past year. Because the associated percentages have different denominators, may not be suppressed for the population aged 12 or older but could be suppressed for the percentage among past year users. In this situation, the estimated total would be displayed for the population aged 12 or older. However, the same estimated total that is associated with the suppressed percentage among past year users would be suppressed. For example, Table 1.22 in the 2022 Detailed Tables (CBHSQ, 2023k) shows among the total population that an estimated 203,000 adolescents aged 12 to 17 in 2021 misused benzodiazepines in the past year. That estimated number was shown as being suppressed for misuse among people who used benzodiazepines for any reason in the past year because the corresponding percentage was suppressed for benzodiazepine misuse among people who used benzodiazepines for any reason.
Another unexpected result may occur when is not suppressed, but the estimated total is displayed as a zero (0). Because the estimated totals are shown as numbers in thousands, a zero actually represents an estimated number greater than zero but less than 500, which is appropriately displayed because was not suppressed.
Estimates of means that are not bounded between 0 and 1 (e.g., mean of age at first use, mean number of days of use in the past 30 days or past 12 months) were suppressed if the RSEs of the estimates were larger than .5 or if the nominal sample size was smaller than 10 respondents. This rule was based on an empirical examination of the estimates of mean age of first use and their SEs for various empirical sample sizes. Although arbitrary, a sample size of 10 appeared to provide sufficient precision and still allow reporting by age at first use for many substances.
In sample size and population tables, such as those in Section 9 of the 2022 Detailed Tables, final respondent sample sizes greater than 100 were rounded to the nearest 10, and sample sizes less than 100 were not reported (i.e., are shown as “<100” in tables). This suppression was done to provide additional confidentiality protection.
3.2.3 Statistical Significance of Differences
This section describes the methods used to compare prevalence estimates. Customarily, one way that observed differences between estimates are evaluated is in terms of their statistical significance. Statistical significance is based on the probability that a difference as large as that observed would occur due to random variability in the estimates if there were no differences in the prevalence estimates being compared; this probability is known as the p value. The significance of observed differences is generally reported at the 0.05 and 0.01 levels when the p value is defined as less than or equal to the designated significance level. Although significance tests are often used to distinguish whether a difference is “real” or simply occurring due to sampling, it is important to note that (1) a “real” difference does not necessarily mean a policy-relevant difference, and (2) tests are based on probability and may give a false impression of certainty.
Section 3.2.3.1 presents the general formula for testing for statistically significant differences. This formula applied to all significance tests between estimates for 2021 and 2022, between subgroups within a single year (e.g., between adolescents aged 12 to 17 and young adults aged 18 to 25 in 2022), and tests of annual averages between subgroups using pooled data (e.g., between adolescents aged 12 to 17 and young adults aged 18 to 25 using pooled data from 2021 and 2022). Significance testing was also conducted to compare estimates for individual subgroups with the corresponding estimate among the overall population (e.g., northeast region vs. all regions).
3.2.3.1 General Significance Testing Formula
When comparing prevalence estimates, the null hypothesis (no difference between prevalence estimates) was tested against the alternative hypothesis (there is a difference in prevalence estimates) using the standard t test (with the appropriate degrees of freedom) for the difference in proportions test, expressed as
where df = the appropriate degrees of freedom, 
= the first prevalence estimate, 
= the second prevalence estimate, 
= the estimated variance of the first prevalence estimate, 
= the estimated variance of the second prevalence estimate, and 
= the estimated covariance between and .
Under the null hypothesis, the test statistic t is a random variable approximately following a t distribution for moderate to large sample sizes. Therefore, calculated values of t, along with the appropriate degrees of freedom, can be used to determine the corresponding probability level (i.e., p value). Whether testing for differences in estimates between years, from different populations within the same year, or from different populations in pooled data, the covariance term in the formula for t will, in general, not be equal to zero. SUDAAN was used to compute estimates of t along with the associated p values using the analysis weights and accounting for the sample design, as described in Chapter 2 of this report.
A similar procedure and a similar formula for t were used for estimated numbers of people with a characteristic of interest, except when the alternative SEs calculation was used (see Section 3.2.1). Whenever the SE for an estimated number of people was calculated outside of SUDAAN using the alternative SE estimation method, the corresponding test statistics also were computed outside of SUDAAN. The 2021 National Survey on Drug Use and Health (NSDUH) Methodological Resource Book, Section 13: Statistical Inference Report (CBHSQ, 2023d) includes examples of code for calculating tests of differences for these forced domains.
Under the null hypothesis, a test statistic with known variances asymptotically follows a standard normal (Z) distribution. However, because the variances of the test statistic are estimated, its distribution is more accurately described by the t distribution for finite sample sizes. As the degrees of freedom approach infinity, the t distribution approaches the Z distribution. Because most tests performed for the 2022 NSDUH had 750 degrees of freedom,39 the t tests performed produce approximately the same numerical results as if a Z test had been performed.
If data users perform independent t tests for the difference of proportions using published estimates and SEs that are shown to a smaller number of decimal places, the test results will typically be similar to tests performed in SUDAAN. However, testing results may differ for two reasons: (1) the covariance term is included in SUDAAN tests, whereas it is not included in independent t tests; and (2) the reduced number of significant digits shown in the published estimates may cause rounding errors in the independent t tests.
3.2.3.2 Significance Testing of Estimates in Different Years
For the 2022 NSDUH, statistical testing was conducted between a given estimate from 2022 and corresponding comparable estimate from 2021 (e.g., past month alcohol use in 2022 vs. the estimate in 2021). These statistical tests indicated whether an estimate in 2022 was lower than, greater than, or similar to the corresponding estimate in 2021. However, estimates from 2021 and 2022 are not comparable with estimates from previous years due to methodological changes to the survey, as described in Section 6 of the 2021 National Survey on Drug Use and Health (NSDUH): Methodological Summary and Definitions (CBHSQ, 2022a).
When significance testing involves comparison of estimates in different years, the formula for the t test described in Section 3.2.3.1 is applied as follows:
- The prevalence estimate and variance from the earlier year (e.g., 2021) become the first prevalence estimate and variance.
- The prevalence estimate and variance from the later year (e.g., 2022) become the second prevalence estimate and variance.
The covariance term represents the covariance between the estimates from each year (e.g., the covariance between estimates in 2021 and 2022). Generally, the covariance term for adjacent years is very small and positive. NSDUH estimates, along with the associated covariance term, are computed using SUDAAN to take into account the NSDUH sample design with overlapping area segments between successive survey years. Ignoring the correlation may result in a slightly more conservative test outcome, with a greater likelihood of assuming no difference between the two estimates. However, a negative correlation is possible and would result in a more liberal test, with a greater likelihood of assuming that the two estimates are different.
Small differences in estimates between years can be statistically significant because of NSDUH’s large sample sizes. These large sample sizes in each year reduce the size of the variances and increase the likelihood that the t test will yield a statistically significant difference. As stated previously, however, small differences between estimates that are not explained by sampling variability are not necessarily relevant from a policy perspective.
Caution is needed when interpreting changes across years in the estimated numbers of people with a characteristic of interest. Respondents with large analysis weights can greatly influence the estimated number in a given year when the number of people in the population with that characteristic is relatively small (e.g., past month heroin users). Large analysis weights for some respondents in a single year can result in the estimated numbers of people with a given characteristic showing an increase between Year 1 and Year 2 (i.e., the year that had the respondents with large analysis weights). The potential for these kinds of year-to-year variations in estimated numbers of people also underscores the importance of reviewing changes across a larger range of years when possible, especially for outcome measures corresponding to a relatively small proportion of the total population.
Caution also is needed when interpreting changes in estimated numbers of people. A change in the estimated number of people with a characteristic of interest could reflect a change in the size of the overall population. Therefore, changes in estimated numbers of people should be considered in conjunction with the corresponding estimated percentages because percentages will control for changes in both the number of people with the characteristic of interest and the total number of people in the population. If corresponding percentages are not available (e.g., for estimates of the number of past year initiates), caution should be taken in interpreting increases over time, which may be explained by population increases rather than by true increases in the characteristic of interest.
3.2.3.3 Significance Testing among Categorical Subdomains in a Single Year
In addition to statistical tests of estimates across years, statistical tests were conducted among population subgroups within a single year. For subpopulations defined by three or more levels of a categorical variable (e.g., age group, race/ethnicity), starting with a test of whether there is any distinction at all between levels is recommended to first control the error level for multiple comparisons. For this purpose, log-linear chi-square tests of independence of the subgroups and the prevalence variables were conducted using SUDAAN.
If Shah’s Wald F test (transformed from the standard Wald chi-square) indicated overall significant differences, then the significance of each particular pairwise comparison of interest was tested using SUDAAN analytic procedures to properly account for the sample design, as described in Section 3.2.3.1 (RTI International, 2013). This two-step procedure protected against inappropriate inferences being drawn due to the number of pairwise differences tested.40
For example, if the overall log-linear chi-square test of independence among age groups is statistically significant, then the following pairwise tests would be conducted between age group estimates: (1) 12 to 17 vs. 18 to 25, (2) 12 to 17 vs. 26 or older, and (3) 18 to 25 vs. 26 or older. For the pairwise comparison between estimates for adolescents aged 12 to 17 and young adults aged 18 to 25, the prevalence estimate and variance for adolescents become the first prevalence estimate and variance, and , respectively. The prevalence estimate and variance for young adults become the second prevalence estimate and variance, and , respectively. The covariance term is 0 because there is no overlap across age groups.
The differences in age groups (and differences among most population subgroups) were deemed to be statistically significant if assessed to be so at the 0.05 level for the 2022 Detailed Tables (CBHSQ, 2023k) and the Key Substance Use and Mental Health Indicators in the United States: Results from the 2022 National Survey on Drug Use and Health report (CBHSQ, 2023j). The sole exception was for testing among racial or ethnic groups for the 2022 Key Substance Use and Mental Health Indicators report. A more conservative level of 0.01 was used for these subgroups because a more conservative level is less likely to assert a significant difference when there is none. The relatively large number of race/ethnicity subgroups being compared (seven) and their varying sample sizes might otherwise result in spurious determinations of significant differences. Although these test results are generally not published in the detailed tables or the Key Substance Use and Mental Health Indicators report, they are used for NSDUH reports to determine whether to highlight differences across subgroups (e.g., by age group).
3.2.3.4 Significance Testing among Categorical Subdomains in Pooled Data from Multiple Years
Some NSDUH reports present estimates using pooled data from 2 or more survey years to improve the precision of estimates. These estimates represent annual averages across the number of years of data being pooled. For example, estimates based on pooled 2021 and 2022 NSDUH data reflect an annual average across 2 years.
The testing procedures discussed for single-year data in Section 3.2.3.3 are applied to testing of estimates among subdomains using pooled data. Specifically, for subpopulations defined by three or more levels of a categorical variable (e.g., age group, race/ethnicity), overall log-linear chi-square tests of independence of the subgroups are first performed. If the test result indicates statistical significance, then the significance of each particular pairwise comparison of interest is tested, as described in Section 3.2.3.1.
3.2.3.5 Significance Testing between a Subdomain and the Overall Population
Significance testing also was conducted using SUDAAN to compare estimates for individual subgroups with the corresponding estimate among the overall population (e.g., northeast region vs. all regions). In the 2022 Detailed Tables (CBHSQ, 2023k), these significance tests were conducted for selected demographic measures (i.e., race/Hispanic origin and region). However, comparing estimates between a subgroup and the overall population increases the covariance in the denominator of the t test formula described in Section 3.2.3.1. Subtracting this covariance term from the sum of the variance terms for the individual estimates will decrease the size of the denominator and increase the size of the t statistic. For this reason, small differences between a subgroup and the overall population can be statistically significant. Again, these tests could be used to aid authors in writing NSDUH reports, but they are not published in the detailed tables.
3.3 Other Information on Data Accuracy
The accuracy of survey estimates can be affected by nonresponse, coding errors, computer processing errors, errors in the sampling frame, reporting errors, and other errors not due to sampling. They are sometimes referred to as “nonsampling errors.” These types of errors and their impact can be reduced through data editing, statistical adjustments for nonresponse, close monitoring and periodic retraining of interviewers, and improvement in various quality control procedures.
Although these types of errors often can be much larger than sampling errors, measurement of most of these errors is difficult. However, the effects of some types of these errors can be examined through proxy measures, such as response rates, and from other research studies. For effects of different modes of data collection within a survey (Section 3.3.3), analyses can compare data across modes.
3.3.1 Screening and Interview Response Rate Patterns
All dwelling units (DUs41) in the sample are screened to confirm eligibility and to select zero, one, or two members to participate in the survey. The weighted screening response rate (SRR) is defined as the weighted number of successfully screened DUs42 divided by the weighted number of eligible DUs (as defined in Table 3.3), or
where 
is the inverse of the unconditional probability of selection for the DU and excludes all adjustments for nonresponse and poststratification defined in Section 2.3.4 of this report.
For the 2022 NSDUH, DU eligibility was imputed for sampled DUs whose members did not initiate the web screening interview and that were not visited by a field interviewer (i.e., DUs with unknown eligibility). Thus, the weighted SRR is equivalent to the response rate 4 (RR4) in the American Association for Public Opinion Research (AAPOR) standard definitions (AAPOR, 2016), which estimates the proportion of sampled DUs with unknown eligibility that are actually eligible. That is,
where I is the weighted sum of the successfully screened DUs, P is the weighted sum of the partially screened DUs, R is the weighted sum of the refusals and break-offs, NC is the weighted sum of the noncontacts, O is the weighted sum of the other eligible nonresponding DUs, e is the estimated proportion of sampled DUs with unknown eligibility that are eligible (estimated using historical NSDUH data), UH is the weighted sum of the sampled addresses in which it is unknown whether an eligible DU exists, and UO is the weighted sum of the DUs in which it is unknown whether an eligible person is present in the DU. According to the definition of a successfully screened DU, no partially screened DUs are in NSDUH’s SRRs (i.e., the letter P in AAPOR’s RR4). Thus, RR4 becomes 
, or
In successfully screened DUs, eligible DU members who were selected were then asked to complete the interview. The weighted interview response rate (IRR) for NSDUH is defined as the weighted number of respondents divided by the weighted number of selected people (Table 3.4), or
where 
is the inverse of the probability of selection for the person and includes DU-level nonresponse and poststratification adjustments (adjustments 1, 2, and 3 in Section 2.3.4.1). In an effort to maximize the IRR, all respondents were offered at least a $30 incentive to encourage them to complete the 2022 NSDUH interview. Some Quarter 4 respondents were given a $5 prepaid screening incentive and/or a $50 interview incentive as part of an incentives experiment to test whether these changes increased the screening and interview response rates (see Section 2.1.2). To be considered a completed interview, a respondent for the 2021 NSDUH needed to provide enough data to pass the usability criteria described in Section 2.3.1.
Similar to the weighted SRR, the weighted IRR is equivalent to the AAPOR standard definition RR2, except that all of the respondents have known eligibility. Thus, the weighted IRR can be written as 
, which is based on the AAPOR definition, or
where I is the weighted sum of the completed interviews, P is the weighted sum of the partial interviews (with enough data to pass the usable interview rule), R is the weighted sum of the refusals and break-offs failing the usable interview rule, NC is the weighted sum of the noncontacts, and O is the weighted sum of the other eligible nonrespondents.
The overall weighted response rate (ORR) is defined as the product of the weighted SRR and weighted IRR, or
3.3.1.1 2022 NSDUH Response Rates
Of the 864,285 eligible sampled DUs in 2022, 217,457 were screened successfully, for a weighted SRR of 25.5 percent (Table 3.3). In these screened DUs, a total of 150,789 people were selected, and completed interviews were obtained from 71,369 of these sampled people, for a weighted IRR of 47.4 percent and a weighted ORR of 12.1 percent (Table 3.4). A total of 41,998 sampled people (29.3 percent) were classified as refusals or parental refusals; 32,040 (18.9 percent) were not available, never at home, or did not respond to the web survey; and 5,382 (4.3 percent) did not participate for various other reasons, such as partially completed but unusable interviews (Section 2.3.1), physical or mental incapacity, or a language barrier (Table 3.4). Among demographic subgroups, the weighted IRR was higher among people aged 26 or older (48.5 percent), females (50.4 percent), Black people (50.5 percent), people in the South (48.5 percent), and residents of nonmetropolitan areas (51.0 percent) than among their corresponding counterparts (Table 3.5).
3.3.1.2 Bias Due to Unit Nonresponse
Maximizing NSDUH response rates is intended to minimize bias in estimates due to different characteristics of respondents and nonrespondents. Drug use surveys may be particularly vulnerable to nonresponse bias if recent or frequent drug users are less likely to participate in the survey, especially for less commonly used substances such as crack cocaine or heroin. However, potential sources of nonresponse bias in one direction (e.g., bias that would decrease estimates) could be offset by corresponding sources of bias in the opposite direction (e.g., bias that would increase estimates), such that overall effects on prevalence because of nonresponse bias could be minimal. Further research on this important topic with recent NSDUH data would be useful.
3.3.2 Item Nonresponse and Inconsistent Responses
3.3.2.1 Item Nonresponse
Among survey participants, item response rates were generally very high for most mental health and drug use items. With the use of multimode data collection in 2022, however, item nonresponse rates were higher than what they were in survey years prior to 2020 when data were collected through in-person interviews only. The increase is due to web respondents discontinuing the survey prior to completion (i.e., breaking off). See Section 2.3.4 for details.
Prior to the introduction of the web mode of data collection in NSDUH, item nonresponse was predominantly caused by responses of “don’t know” or “refused” either in the specific question or in earlier questions that governed skip logic.43 In 2022, item nonresponse for usable interviews among adults was driven by survey break-offs. Item nonresponse tended to be lower for drug use items because of the usability criteria described in Section 2.3.1 and higher for mental health items.44
The way a question is asked or the response options that are offered can also contribute to item nonresponse. In 2022, for example, revisions were made to question QD05, which asked respondents to report the race category that best described them. A notable change was that respondents were no longer offered a response option for “other” race.45 The unweighted number of respondents who did not know or refused to answer QD05 increased from about 200 respondents in 2021 to about 800 in 2022. Because of the large sample sizes, however, the unweighted percentages of respondents who did not know or refused to answer QD05 were relatively low in both 2021 (0.3 percent) and 2022 (1.1 percent).
Prior to 2022, if respondents chose the “other” race response, they were asked to specify what this other race was. Beginning in 2022, respondents were asked to specify another race only if they answered question QD05 as “don’t know.” Consequently, the number of respondents who were given the option to specify another race decreased from more than 3,000 respondents in 2021 to about 500 in 2022. Only a small percentage of “other” race responses in 2021 matched one of the existing race categories.46 Moreover, there was a higher overall rate of missing data in the edited version of race prior to imputation in 2021 (3.3 percent, or about 2,300 respondents) than there was in 2022 (1.0 percent, or about 700 respondents) because of the “other” race data for 2021. Stated another way, the race variable required imputation for about one third as many respondents in 2022 than in 2021. The increased number of respondents in 2022 who did not know or refused to report their race was offset by the decreased number of respondents who were given the opportunity to choose “other” for their race.
Occurrences of missing data in mental health items (e.g., treatment for mental health issues, major depressive episode [MDE]) are important to examine because these questions occurred later in the NSDUH interview when respondents could break off but their interviews could remain usable. Furthermore, not all mental health variables were statistically imputed for 2022 (see Sections 2.3.3 and 3.4.6).
For the receipt of mental health treatment in the past 12 months, between 0.5 and 0.6 percent of the adult respondents who were interviewed in person in 2022 had missing data (e.g., no response, responses of “don’t know” or “refused”) for whether they received treatment at specific locations in the past 12 months or used prescription medication in that period to help with their mental health.47 In comparison, between 3.6 and 3.8 percent of adults who completed the survey via the web had missing data for the same measures, leading to overall rates of 2.0 to 2.1 percent for these items for all adult respondents in 2022. Among adults in 2022 with missing data for whether they received mental health treatment in the past 12 months for these measures, 15.9 to 22.0 percent of adult in-person respondents had missing data because of break-offs compared with 94.7 to 98.6 percent of adult web respondents with missing data. Although final variables for adults’ receipt of mental health treatment were statistically imputed for 2022, these percentages indicate the amount of imputation that was required.
Break-offs contributed less to item nonresponse for respondents aged 12 to 17 than for adults in the 2022 NSDUH. Unlike adults, item nonresponse was generally lower for youths who completed the survey via the web instrument than it was for those who were interviewed in person.
As an example, among youths in 2022, between 0.8 and 1.7 percent in the total sample (regardless of data collection mode) had missing data for whether they received mental health treatment in the past 12 months at specific locations or whether they received specific types of treatment during that period. Among youths who were in-person respondents, 0.9 to 1.9 percent had missing data for the specific locations of mental health treatment in the past 12 months or the use of prescription medication in that period to help with their mental health. In comparison, 0.6 to 1.0 percent of youths who completed the survey via the web had missing data for these measures.
Between 3.0 and 5.2 percent of youths in the total sample had missing data for questions about specific lifetime symptoms of depression (3.3 to 5.6 percent for in-person respondents and 1.8 to 3.4 percent for web respondents). The depression item with the highest proportion of missing data (regardless of data collection mode) was the question about the specific number of pounds youths lost without trying to lose weight.48
Occurrences of missing data in items for the receipt of substance use treatment also are important to examine because these questions occurred later in the NSDUH interview (but earlier in the interview than the mental health items); respondents could break off before these questions and their interviews could remain usable. Furthermore, not all of the substance use treatment variables were statistically imputed for 2022 (see Section 3.4.5). Therefore, potential bias due to missing data are relevant to NSDUH estimates for substances for which treatment was received. Among respondents aged 12 or older in 2022, between 3.0 and 6.6 percent had missing data for the receipt of treatment at specific locations or specific types of treatment in the past 12 months for their use of drugs or alcohol.49 Percentages were between 1.7 and 2.1 for in-person respondents and between 2.1 and 2.9 percent for web respondents.
Patterns of missing data for the receipt of substance use treatment in the past year differed somewhat for in-person and web respondents. Among in-person respondents, 0.05 percent used alcohol or drugs in their lifetime but had missing data because they broke off the interview before they reached the alcohol and drug treatment section, and 1.3 percent had missing data because their status as lifetime alcohol or drug users was unknown. Among web respondents, 2.3 percent had missing data because they broke off the interview before the section, and 0.3 percent had missing data because their status for lifetime alcohol or drug use was unknown.
For questions that ask for an opinion or a perception, such as the perceived risk of harm from the use of different substances or the perceived availability of substances, responses of “don’t know” may be a valid indication of respondents not having an opinion. For example, respondents may not know how much people might risk harming themselves by using specific substances if respondents have not formed an opinion on the topic or if they have no knowledge of the substance.
In 2021 and 2022, fewer than 1 percent of respondents aged 12 or older and fewer than 2 percent of adolescent respondents aged 12 to 17 answered “don’t know” to questions about the perceived risk of harm from using cigarettes or alcohol.50 For the perceived risk of harm from using specific illicit drugs, the percentages of respondents who answered “don’t know” were highest for the perceived risk of harm from trying or using lysergic acid diethylamide (LSD) (about 3 percent of respondents aged 12 or older for trying LSD or using LSD once or twice a week, about 7 percent of respondents aged 12 to 17 for trying LSD).
Responses of “don’t know” to questions about the perceived risk of harm from substance use was the predominant source of missing data in most of these questions. Responses of “don’t know” to questions about the perceived risk of harm from using these substances could reflect a general lack of knowledge about these substances. Similarly, responses of “don’t know” to questions about how easy or difficult it would be to get illicit drugs or marijuana were the predominant source of missing data for those questions. Aside from issues of potential bias discussed in Section 3.3.2.2, excluding respondents with missing data for perceived risk and availability measures (especially those who answered “don’t know”) could have other implications for published estimates. For example, excluding respondents who answered “don’t know” to these questions might create the impression that all people in the population have an opinion about the perceived risk of harm from substance use or the perceived availability of different substances. For these measures, the percentage of people who did not know how to answer these questions could be useful information. However, the estimates on perceived risk and availability of substances published in the 2022 Detailed Tables (CBHSQ, 2023k) exclude respondents from analyses who answered “don’t know” to these questions; thus, this issue is a limitation to consider when interpreting these results.
3.3.2.2 Effects of Missing Data on Estimates
When statistical imputation was not used to replace missing values with nonmissing values (see Section 2.3.3), NSDUH estimates were based on variables that have some missing data. Generally, observations with missing values are excluded from standard NSDUH analyses, including a portion (but not all) of the analyses used to create the annual detailed tables.
Since 2021, additional variables have been statistically imputed to mitigate the effect of increased rates of item nonresponse due to break-offs. See Table 2.5 for a list of measures that were statistically imputed for the 2022 NSDUH. However, not all mental health variables or other variables in later sections of the 2022 interview were statistically imputed. There may be bias in estimates from these variables due to the presence of missing data. Specifically, there may be bias when respondents with missing data are excluded from the analysis.
For estimated numbers of people with a given characteristic, there will always be a negative bias if there are missing values in the domain variable(s), the outcome variable, or both. For example, consider an analysis that estimates the percentage and number of youths between the ages of 12 and 17 who were exposed to school-based substance use prevention messages. This analysis includes a domain variable for youths who attended school in the past 12 months (including those who were homeschooled) and multiple outcome variables, which consist of whether youths received substance use prevention messages in various school settings. Both the domain and the outcome variables may have missing data, and respondents with missing data for either school attendance or exposure to school-based prevention activities would be excluded from the analyses for both the percentage and estimated number of exposed youths.
When a population mean or a population proportion is estimated, there may or may not be bias, and the bias can be either negative or positive. The direction and magnitude of the bias for means and proportions depend on how different the item respondents are from the item nonrespondents with respect to the outcome of interest. For example, if “true” perceptions of the risk of harm from the use of different substances (i.e., no risk, slight risk, moderate risk, great risk) among respondents with missing data matched the distribution of respondents who did not have missing data, then excluding missing data (and decreasing the number of respondents in the denominator) would be expected to increase the estimated percentage of people in the population who perceived great risk of harm from using a substance. However, if the actual perceived risk of harm among respondents with missing data was skewed in favor of perceived great risk of harm, for example, then excluding these missing data might introduce bias in published estimates.
These missing data effects are discussed in more detail in the 2021 Statistical Inference Report (CBHSQ, 2023d).
3.3.2.3 Inconsistent Responses
In order to minimize respondent confusion, inconsistent responses, and item nonresponse, the NSDUH instrument for in-person and web-based interviews was programmed to skip respondents out of inapplicable questions based on their previous answers. This skip logic reduced the potential for inconsistent data by limiting respondents’ opportunity to provide answers that were inconsistent with previous answers. For example, if respondents did not report that they last used alcohol in the past 30 days, they were not asked questions about their use of alcohol in the past 30 days. Similarly, if respondents did not report the use of alcohol or drugs in their lifetime, they were not asked questions about the receipt of substance use treatment.
However, programming of skip patterns within the computer-assisted interviewing (CAI) and web instruments could not eliminate all occurrences of missing or inconsistent data. For example, when asked about cocaine use, respondents who reported not knowing whether they had ever used cocaine are not asked further questions about this substance, resulting in missing data for their most recent use and when they initiated use. Respondents also could report lifetime use of cocaine but give inconclusive information (i.e., responses of “don’t know” or “refused”) for when they last used it. Consequently, information is unknown for whether these lifetime users used cocaine in the past year or past month. Similarly, respondents could give inconsistent responses, such as reporting they last used any form of cocaine more than 12 months ago but they last used crack cocaine in the past 30 days or last used it more than 30 days ago but within the past 12 months; both answers logically cannot be true.
These missing or inconsistent responses in the substance use data are first resolved (where possible) through a logical editing process (e.g., logically inferring more recent reported use of crack cocaine applies to any cocaine). Additionally, missing or inconsistent responses for variables used to establish the prevalence of substance use are usually imputed using statistical methodology. These imputation procedures in NSDUH are based on responses to multiple questions, so that other relevant information is utilized through statistical modeling when determining whether a respondent is classified as a user or nonuser—and, if the respondent is classified as a user, whether the respondent is classified as having used a substance in the past year or the past month. For example, nonspecific data on the most recent use of cocaine are statistically imputed based on a respondent’s data for use (or most recent use) of tobacco products, alcohol, and marijuana. Nevertheless, editing and imputation of missing responses are potential sources of measurement error.
The CAI skip logic also did not eliminate all opportunities for inconsistent reports in the mental health questions. Consequently, the logical editing procedures for the mental health data could slightly increase the amount of missing data when inconsistent answers were given. For example, if adult or adolescent respondents who met the criteria for a lifetime MDE (see Section 3.4.9) reported an age at onset for depression symptoms51 greater than their current age, the inconsistent age-at-onset variable was set to a missing value. However, the number of respondents in 2022 with this inconsistency was small (i.e., fewer than 10 respondents aged 12 or older).
Not all inconsistencies in the data are resolved through editing or imputation. Inconsistencies could remain when questions are asked in different questionnaire sections. For example, respondents could indicate in the coronavirus disease 2019 (COVID-19) section that they experienced specific situations in accessing substance use treatment because of the COVID-19 pandemic, but they previously did not report lifetime use of alcohol or illicit drugs. In situations such as these, data users will need to decide how to handle these inconsistencies in their analyses.
For more information on editing and statistical imputation, see Sections 2.3.2 and 2.3.3 in this report. Details of the editing and imputation procedures for 2022 also will appear in the 2022 National Survey on Drug Use and Health (NSDUH) Methodological Resource Book, Section 10: Editing and Imputation Report (CBHSQ, forthcoming b). Until that report becomes available, refer to the 2021 Editing and Imputation Report (CBHSQ, 2023b) for the most recent documentation of general principles and procedures for editing and imputation.
3.3.3 Revised Estimates for 2021 to Account for Data Collection Mode
Since October 2020, NSDUH has used multimode data collection, in which respondents could complete the interview via the web or in person. Analyses conducted for the 2021 NSDUH indicated that key substance use and mental health estimates differed between data collection modes (i.e., web or in person); these differences are known as “mode effects.” Specifically, web respondents tended to be less likely than in-person respondents to report most substance use or mental health issues. Weighting reduced the mode effects somewhat but not enough to eliminate statistically significant differences. For more information, see Chapter 6 in the 2021 Methodological Summary and Definitions report (CBHSQ, 2022a).
As long as the proportion of interviews completed via these two different modes is stable across years, then mode effects would have little effect on differences in estimates over time. As discussed in Section 2.3.4, however, the proportions of interviews completed via the web or in person had not stabilized in 2021 and 2022. Specifically, more than half of interviews in 2021 (54.6 percent) were completed via the web, but less than half in 2022 (42.4 percent) were completed via the web. Moreover, less than 40 percent of interviews in Quarter 2 (April to June) and Quarter 4 (October to December) of 2022 were completed via the web. Consequently, mode effects could distort differences in estimates between 2021 to 2022, unless the analysis weights are adjusted to take these different mode proportions into account.
The analysis weights used for reports and tables for the 2021 NSDUH included two-way interactions of quarter with other demographic characteristics in the adjustment models as a means of partially accounting for quarterly variations in interview mode proportions. However, including these interaction terms did not adjust the proportions of interviews that were completed via the web or in person to a set of fixed proportions. Without further adjustment to the weights, apparent increases in estimates between 2021 and 2022 could be due to a greater proportion of in-person interviews in 2022 rather than real changes in the population. Stated another way, apparent increases in estimates for 2022 could be partially due to the greater proportion of in-person respondents in 2022, and not just to true changes in prevalence in the population. Similarly, decreases in prevalence may be partially obscured by the changes in proportions.
For this reason, the 2022 Detailed Tables (CBHSQ, 2023k) include revised estimates for 2021 that were created using the updated person-level weights described in Section 2.3.4.3. These updated weights for 2021 were adjusted so that 2021 and 2022 estimates were based on consistent annual mode proportions for comparison. Consequently, NSDUH estimates for 2021 that are presented in the 2022 Detailed Tables may differ from the previously published estimates in national reports and tables for the 2021 NSDUH. The following sections discuss the investigations to update the 2021 weights and the effects of the recalibrated 2021 weights on the 2021 estimates.
3.3.3.1 Investigations to Update the 2021 Estimates
As shown in Figure 2.3, the unweighted proportions of respondents who completed the NSDUH interview via the web or in person varied by quarter in 2021 and 2022, but the proportions showed less variability in Quarters 2 through 4 of 2022. The proportion of interviews completed via the web also trended downward through 2022, with less than 40 percent of interviews being completed via the web in Quarters 2 and 4. When the 2022 weights were poststratified to demographic control totals (see Section 2.3.4 and Table 3.1) but not adjusted for differences in proportions by mode, the resulting weighted percentage of interviews completed via the web was 31.8 percent. Based on this overall weighted percentage for 2022, it could be reasonable to assume that proportions in future years might stabilize in the vicinity of 30 percent of interviews being completed via the web and 70 percent being completed in person.
Because of the differences across quarters in the proportions of interviews completed via the web or in person, analyses were conducted to evaluate the sensitivity of substance use and mental health estimates to differing proportions of interviews that were completed via each mode. If estimates were relatively insensitive to changes in these proportions, then it could be reasonable to compare estimates across years without the need to adjust the analysis weights for the differing proportions of interviews that were completed via each mode. If the estimates were sensitive to these differences, it could be assumed that relatively simple weight adjustments would be feasible to facilitate comparison of estimates between years.
To test sensitivity of estimates to differing proportions of interviews completed via each mode, 2021 weighted estimates of 22 outcomes52 were compared with alternative estimates for which the weights were calibrated or scaled to the following assumed proportions:
- 30 percent of interviews completed via the web and 70 percent completed in person (i.e., closer to anticipated future proportions),
- 40 percent of interviews completed via the web and 60 percent completed in person (i.e., closer to the unweighted proportion in Quarter 4 of 2021), and
- 70 percent of interviews completed via the web and 30 percent completed in person (i.e., closer to the unweighted proportion in Quarter 1 of 2021), the reverse of the first scenario.
Estimates were produced for the 22 outcomes for domains defined by gender, age, race or ethnicity, and levels of urbanization.
These tests determined that the estimates were sensitive to scaling the weights to different proportions completed via each mode. Scaling the weights according to the assumed proportions at the extremes (i.e., 70 percent completed via the web and 30 percent completed in person; 30 percent completed via the web and 70 percent completed in person) yielded the most notable differences relative to published estimates for the 2021 NSDUH that were produced using the standard 2021 person-level analysis weights, with no adjustment for varying proportions via each mode. Of the three scenarios, the weights that were scaled to assume completion of 40 percent of interviews via the web and 60 percent in person were closer to the unweighted 2022 proportions and produced estimates closer to those published for the 2021 NSDUH with unadjusted weights. However, 40 percent of interviews completed via the web and 60 percent completed in person did not appear to be stable proportions. Consequently, a 40/60 adjustment could require further modification in future years.
Because estimates were found to be sensitive to the proportions of interviews that were completed via each mode, two methods were tested for calibrating the 2021 weights. Method A forced the relative proportions of web respondents to be the same in every geographic and demographic domain, and not just for the overall population. Although this method would ensure consistent proportions across years for each domain, changing the weights by Method A was expected to result in larger changes to the 2021 subdomain estimates and weight variations. Method A also assumed that people in each subdomain had the same likelihood to respond via the web or in person. However, this assumption would be violated if people in different subdomains were more likely or less likely to respond via the web.
Method B adjusted the weights to a standard proportion for web and in-person respondents for the overall population and by census division, but it allowed the proportions to vary across other domains. This method also had the flexibility of allowing proportions to change somewhat from year to year in these other domains.
Based on these investigations, the final method that was chosen involved updating the person-level analysis weights for 2021 so that the weighted proportions overall and by census division were the expected future proportions of 30 percent of interviews completed via the web and 70 percent completed in person. Method B was selected for adjusting the 2021 weights. The 2021 estimates with the updated weights were used to compare estimates between 2021 and 2022 in the 2022 Detailed Tables (CBHSQ, 2023k). See Section 2.3.4.3 for more information on the adjustment of the 2021 person-level weights.
3.3.3.2 Effect of Updated Weights on 2021 Estimates
Because analyses indicated that adjusting the 2021 weighted proportions of interviews completed via the web or in person had an effect on published estimates, 134 of the 2021 Detailed Tables were repopulated using the updated weights to test the extent of the differences (CBHSQ, 2022d). The tables covered a range of drug use and mental health outcomes and included estimates among demographic and geographic domains. After exclusion of suppressed estimates (see Section 3.2.2), more than 10,800 distinct prevalence estimates were produced with the updated person-level weights and compared with the original estimates.
Most of the adjusted estimates (70 percent) were larger than the original estimates. This result was expected because prior analyses for the 2021 NSDUH indicated that web respondents were less likely to report most substance use and mental health issues; see Chapter 6 in the 2021 Methodological Summary and Definitions report (CBHSQ, 2022a). The adjustment reduced the contribution of the web interviews, on average, and increased the contribution of the in-person interviews to the final estimates. Consequently, these adjustments created a tendency for prevalence estimates in 2021 to increase relative to previously published estimates. However, changes in the revised 2021 estimates were relatively modest. The majority of the estimates based on the updated weights (92.6 percent) were within one SE of the original estimates. The mean of the absolute difference between paired estimates was only 0.25 percentage points, and the absolute differences ranged from 0 to 3.42 percentage points. The mean relative change was 1.48 percent. The range of relative differences was larger than that for the absolute differences because of the inclusion of some very small prevalence estimates.
3.3.4 Validity of Self-Reported Substance Use
Most estimates of substance use, including those produced for NSDUH, are based on self-reports of use. This section focuses on the validity of NSDUH respondents’ self-reports of substance use and is not intended to provide a comprehensive discussion of issues associated with the validity of any self-report in NSDUH. Factors such as the length of time between an event and the interview date or respondents’ interpretation of a question also can affect respondents’ recall or reporting, independent of the potential sensitivity of the topic covered by a question. An additional factor discussed in this section is the use of multimode data collection (i.e., in person or web) in 2022.
Survey questions about topics such as substance use are considered to be sensitive because respondents may think the questions are intrusive (“none of your business”), pose risks for negative social or legal consequences if their answers were to become known, or require them to provide socially undesirable answers (Tourangeau & Yan, 2007). Although studies generally have supported the validity of self-report data for sensitive topics, the potential for these data to be biased (underreported or overreported) is well documented. The bias varies by several factors, including the mode of administration, the setting, perceptions of privacy, the population under investigation, and for substance use, the type of drug (Aquilino, 1994; Brener et al., 2006; CBHSQ, 2012; Harrison & Hughes, 1997; Lindberg & Scott, 2018; Tourangeau & Smith, 1996; Tourangeau & Yan, 2007; Turner et al., 1992). NSDUH utilizes widely accepted methodological practices for increasing the accuracy of self-reports, such as encouraging privacy through self-administration of questions about sensitive topics—including audio computer-assisted self-interviewing (ACASI) for in-person data collection—and providing assurances that individual responses will remain confidential. Comparisons using these methods within NSDUH data (collected in person) have shown they reduce reporting bias (Gfroerer et al., 2002).
A special study cosponsored by the Substance Abuse and Mental Health Services Administration (SAMHSA) and the National Institute on Drug Abuse (NIDA) examined the validity of NSDUH self-report data on drug use among people aged 12 to 25. The study found urine and hair specimens can be collected with a relatively high response rate in a general population survey, and most youths and young adults reported their recent drug use accurately in self-reports (Harrison et al., 2007).53
3.3.4.1 Reporting of Sensitive Behaviors in Web Surveys
Prior studies spanning multiple decades (e.g., Aquilino, 1994; Lindberg & Scott, 2018; Tourangeau & Smith, 1996) have established that respondents are more likely to report sensitive behaviors when questions are self-administered than when they need to report their answers to interviewers. With the use of multimode data collection in the 2022 NSDUH, including some interviews being completed via the web, it is important to examine respondent willingness to report sensitive behaviors via the web. Given the use of the web for interactions that involve the sharing of sensitive data (e.g., Social Security numbers, credit card numbers), people may be expected to accurately report information about their substance use in a web-based survey. However, the reliability of self-reports in surveys administered via the web remains an open and evolving question.
Kreuter and colleagues (2008) assessed social desirability bias in the reporting of potentially sensitive academic information using interviewer administration via computer-assisted telephone interviewing (CATI) and self-administration using interactive voice recognition (IVR) and web administration. Under a randomized experimental design, web administration increased the reporting of socially undesirable academic information such as a cumulative grade point average below 2.5 compared with CATI, with IVR yielding results between those of web administration and CATI. Web respondents also were less likely than CATI respondents to falsely deny socially undesirable outcomes relative to information from external data sources. Although this study focused on academic outcomes rather than substance use, it is consistent with other literature showing that self-administered data collection modes—including web-based data collection—yield increased reports of sensitive behaviors compared with interviewer-administered modes.
Cernat and colleagues (2016) used data from the 2010 to 2012 waves of the Health and Retirement Study to compare estimates of depression symptoms from the Center for Epidemiologic Studies Depression Scale (CES-D) in interviewer-administered and self-administered web data collection modes. Depression estimates based on the CES-D were lower in interviewer-administered modes than in the web mode. However, the study did not compare the web data collection mode with other self-administration modes.
The Monitoring the Future (MTF) study for 2022 used a web-based questionnaire to collect data from 8th, 10th, and 12th graders. Similar to the 2021 survey, students for the 2022 survey used their own electronic devices (rather than tablets provided by the study) to connect to the web and complete the survey during class time. Some students might have completed the survey via the web in their homes if classes were conducted remotely. Analyses found that substance use trends were replicated when the sample for 2021 was restricted to students who reported that all their classes were in school (46 percent of the sample). Investigators concluded that web administration of the survey at home produced results similar to those for students who completed the web survey in school (Johnston et al., 2022). Because of this, the MTF compared substance use estimates for 2022 with those from prior years. However, the 2022 MTF report for secondary school students (Miech et al., 2023) did not compare substance use estimates for students who completed the survey in school or remotely.
One issue for web-based data collection in NSDUH concerns privacy in settings such as within households. As noted previously, respondents’ perceptions of privacy can influence the reporting of sensitive behaviors, as described by Brener and colleagues (2006) and in a report comparing youth substance use estimates across national surveys (CBHSQ, 2012). For the 2022 NSDUH, field interviewers asked in-person respondents to find a private location to complete the survey. Web respondents were asked to be in a private location within the home and to affirm before starting an interview that they were in a private location. Unlike in-person interviews, however, no privacy ratings were available for web-based interviews to indicate whether interviews remained private throughout the entire interview, or if not, the extent of time for which the interview was less than private and who else might have been present. It also is not known whether NSDUH web respondents perceived the web mode to be a more private method for answering sensitive questions compared with in-person data collection in a private setting using ACASI, but with a field interviewer present. As web-based interviewing is increasingly used for collecting survey data—including data on sensitive topics such as substance use—methodological research comparing data collection via the web with other modes that collect self-administered data would be useful for establishing the factors that encourage or discourage the reporting of sensitive behaviors via the web.
3.3.4.2 Issues for Self-Reporting of Prescription Drug Misuse
The emphasis on past year rather than lifetime misuse of specific prescription drugs in the NSDUH questionnaire appears to affect the validity of estimates for lifetime misuse of prescription psychotherapeutic drugs (see Section C in the 2015 Methodological Summary and Definitions report; CBHSQ, 2016). Respondents in 2022 who did not misuse prescription psychotherapeutic drugs in the past 12 months were asked only about lifetime misuse of general categories of prescription psychotherapeutic drugs (e.g., prescription pain relievers). Respondents also did not have cues for recalling misuse more than 12 months ago of drugs no longer available by prescription in the United States (e.g., sedatives containing methaqualone, such as those with the brand names Quaalude® or Sopor®). Field testing of these prescription drug questions suggested that the emphasis on the past year misuse of specific prescription drugs can result in underreporting of lifetime misuse of prescription psychotherapeutic drugs. For more information, see the National Survey on Drug Use and Health: 2012 Questionnaire Field Test Final Report and the National Survey on Drug Use and Health: 2013 Dress Rehearsal Final Report (CBHSQ, 2014b, 2014c). For this reason, estimates of lifetime misuse of prescription psychotherapeutic drugs are not included in the 2022 national reports and tables.
The prescription drug questions in 2022 allowed respondents to report any use or misuse in the past 12 months for specific medications within a given psychotherapeutic category (e.g., the benzodiazepine tranquilizers Xanax®, Xanax® XR, generic alprazolam, and generic extended-release alprazolam). These details were presented to respondents to aid them with recall and recognition. However, respondents could have difficulty knowing or remembering whether they took a generic or brand name drug or what type of formulation they took (i.e., immediate release or extended release). Therefore, recall of the use or misuse of prescription drugs containing a given active ingredient has been assumed to be more accurate than recall of the exact drugs respondents took. For example, respondents who took the generic benzodiazepine alprazolam (brand name drug Xanax®) could recognize the drug by its brand name and report use or misuse of “Xanax.” This issue may be especially relevant for respondents who misused prescription drugs by taking them without a prescription of their own. However, self-reports of the use or misuse of Xanax® or alprazolam are equivalent analytically for estimating the use or misuse of tranquilizers containing alprazolam, even if respondents may have misreported the exact drug they used or misused in the past year. Therefore, 2022 NSDUH estimates for the use or misuse of prescription psychotherapeutic drugs in the past year are reported for overall psychotherapeutic drug categories (e.g., tranquilizers) or for subtypes of related drugs (e.g., benzodiazepine tranquilizers, tranquilizers containing alprazolam), but they are generally not reported for specific individual prescription drugs from the NSDUH questionnaire.54
3.4 Measurement Issues
Several measurement issues for the 2022 NSDUH are discussed in this section. Issues addressed include the methods for measuring the use and misuse of prescription drugs, the use and misuse of fentanyl (including illegally made fentanyl [IMF]), the initiation of substance use or misuse of prescription drugs, SUDs, substance use treatment (including medication-assisted substance use treatment), mental health treatment, and the definition of county type. Additionally, this section discusses the mental health measures AMI, SMI, and MDE. Starting with Section 3.4.10, measures are discussed for perceived recovery, nicotine vaping, central nervous system (CNS) stimulant use, suicidality, specific COVID-19-related measures, and modes of marijuana use.
This section also discusses how missing data were handled analytically to produce the estimates found in the national reports and tables for the 2022 NSDUH. Refer to Section 3.3.2 for a discussion of potential bias in estimates because of missing data.
Except where noted, estimates for measures described in this section used the main analysis weight described in Section 2.3.4. The section specifically mentions when estimates used the break-off analysis weight because the measures were not statistically imputed.
3.4.1 Use and Misuse of Prescription Drugs
Prescription drug questions in 2022 first included a set of “screener” questions that asked respondents to report any use of specific prescription drugs in the past 12 months, regardless of the reason. Respondents were then asked about misuse in the past 12 months for the specific prescription drugs they reported using in that period (see the next paragraph). This structure simplifies the cognitive task for respondents by separating whether they used a specific prescription drug for any reason and, if so, whether they used it in a way constituting misuse.
Misuse of prescription psychotherapeutic drugs was defined as use “in any way a doctor did not direct you to use it/them.” Respondents were presented with examples of use in any way not directed by a doctor, including (1) use without a prescription of one’s own; (2) use in greater amounts, more often, or longer than told to take a drug; and (3) use in any other way not directed by a doctor. This definition of misuse focuses solely on behaviors constituting misuse, independent of respondents’ motivations for those behaviors. This definition also includes overuse of prescribed medication.
The 2022 NSDUH focused on the misuse of specific prescription drugs in the past 12 months. Focusing on the past year better addresses the information needs of policymakers in federal and state agencies who are concerned with recent misuse of prescription drugs that were available by prescription in the United States. A 12-month time frame also is closer to the interview date and has better recall. Thus, a 12-month time frame is expected to yield more accurate estimates for the recent misuse of prescription drugs. As noted in Section 3.3.4.2, however, the emphasis on the past year misuse of specific prescription drugs is assumed to result in underreporting of lifetime misuse of prescription drugs. Therefore, the 2022 Detailed Tables (CBHSQ, 2023k) do not show estimates for the lifetime misuse of prescription drugs.
Although a form of methamphetamine (Desoxyn®) can be prescribed in the United States, most methamphetamine used in the United States is manufactured illegally. Therefore, questions about methamphetamine use are in a questionnaire section that is separate from questions about prescription stimulants.
In 2022, variables for any use of prescription drugs in overall psychotherapeutic categories (e.g., pain relievers) in the lifetime and past year periods underwent statistical imputation to remove missing values (see Section 2.3.3). Variables for the lifetime, past year, and past month misuse of prescription drugs in overall psychotherapeutic categories also have been imputed. Similarly, variables for the past year use and misuse of benzodiazepines and fentanyl products were imputed for 2022. However, certain variables for any use and misuse of prescription drugs have not been imputed and therefore retained missing values. Section 4.2 discusses these prescription drug measures and how missing data were handled in those situations. Also, see Section 3.3.2 for a discussion of the potential bias in estimates depending on how missing data were handled.
Respondents were not counted as having misused “any other” prescription drug in the past year if the only other drugs they specified were over-the-counter (OTC) drugs.55 Respondents also were not counted as having misused any other prescription drug if (1) the only other drugs they specified corresponded to prescription drug subtypes in the NSDUH questionnaire for that psychotherapeutic category (e.g., pain relievers containing hydrocodone), or (2) they specified only drugs in these subtypes and OTC drugs. For example, respondents who specified Vicodin® and OTC drugs (e.g., acetaminophen) as the only other prescription pain relievers they misused in the past year were counted in estimates for the past year misuse of hydrocodone products but were not counted in estimates for the past year misuse of any other pain reliever.
This procedure could not be applied to estimates for any use in the past year (i.e., not necessarily misuse) of other prescription drugs in a psychotherapeutic category because respondents were not asked to specify the names of other drugs they used. For example, if respondents reported the use of only other prescription drugs in a category, it cannot be determined whether the other drugs that respondents used were all OTC drugs. It also cannot be determined whether the “other” prescription drugs that respondents used belonged only to existing subtypes in that section of the interview. Additional details about these procedures are described in the 2017 Methodological Summary and Definitions report (CBHSQ, 2018).
3.4.2 Use and Misuse of Fentanyl, Including Illegally Made Fentanyl
Data from the Centers for Disease Control and Prevention’s (CDC’s) National Vital Statistics System indicate that overdose deaths involving synthetic opioids other than methadone have been the leading drugs for opioid-involved overdose deaths since the end of 2016 (Ahmad et al., 2023). This involvement of synthetic opioids in fatal overdoses has been attributed to fentanyl and fentanyl analogues that are made illegally in clandestine laboratories rather than by the pharmaceutical industry. These forms of IMF and analogues have been mixed with heroin as an adulterant, substituted for heroin entirely, or sold as counterfeit prescription drugs (Ciccarone, 2019). From 2016 to 2021, age-adjusted rates of drug overdose deaths more than tripled for deaths involving fentanyl (Spencer et al., 2023).
From 2015 to 2021, NSDUH asked about the use and misuse of pharmaceutical forms of fentanyl but did not ask specifically about the use of IMF. However, respondents could specify in the special drugs section of the NSDUH interview that fentanyl was another drug that they injected with a needle. Because many pharmaceutical fentanyl products that are available by prescription in the United States (e.g., Duragesic®, Fentora®, generic fentanyl) are in forms such as skin patches or “lollipops” that would require extraction of the fentanyl before it can be injected, most reports in NSDUH of the injection of fentanyl are assumed to involve the injection of IMF. However, reports of fentanyl injection that require respondents to specify it as some other drug that they injected are likely to yield underestimates of IMF use (Kroutil et al., 2010). Moreover, heroin that contains IMF as an adulterant could be used in ways other than injection, such as smoking or sniffing (Mars et al., 2018).
Therefore, the 2022 NSDUH included questions in the emerging issues section for the first time that specifically asked about the use of IMF. Respondents aged 12 or older were asked whether they ever used IMF and, if so, how long it had been since they last used it. The questionnaire explained that IMF is fentanyl that people cannot get from a doctor or pharmacy and that IMF can come in forms such as powder, pills, blotter paper, or mixed with heroin or other drugs. If respondents reported that they used IMF in any way within the past 12 months or they did not know or refused to report when they last used IMF, they also were asked whether they ever used a needle to inject it and, if so, how long it had been since they last injected it. Respondents who reported using IMF with a needle in their lifetime and reported last using it in any way more than 12 months ago were not asked when they last used a needle to inject IMF; logically, these respondents last injected IMF more than 12 months ago.
Variables for the lifetime and most recent use of IMF overall were imputed and should be analyzed using the main analysis weight. However, variables for the lifetime and most recent use of IMF with a needle were not imputed. Therefore, analysis of variables for the injection of IMF should use the break-off analysis weight.
Beginning in 2022, national reports and tables included new estimates for aggregate measures that included past year use of pharmaceutical fentanyl products or IMF. The measure for any use of fentanyl in the past year included reports of any use of pharmaceutical fentanyl or use of IMF in the past year. The measure for fentanyl misuse included misuse of pharmaceutical fentanyl or use of IMF in the past year.
3.4.3 Initiation of Substance Use or Misuse
In NSDUH, initiation refers to the first use or misuse of a particular substance. For prescription psychotherapeutic drugs (pain relievers, tranquilizers, stimulants, and sedatives), initiation refers to the first time misuse ever occurred.56 All of the initiation variables used to create published estimates for the 2022 NSDUH underwent statistical imputation to remove missing values (see Section 2.3.3). Therefore, these variables were not subject to the kinds of potential bias because of missing data described in Section 3.3.2.
In 2022, the survey questionnaire collected year and month of first use for recent initiates—that is, people who used a particular substance for the first time at their current age or the year before their current age. Month, day, and year of birth also were obtained directly or were imputed for item nonrespondents as part of the data postprocessing. Additionally, the date of the interview was recorded for both in-person and web respondents.
Past year initiation for the 2022 NSDUH referred to respondents whose date of first use of a substance (or misuse for psychotherapeutic drugs) was within the 12 months prior to their interview date.57 Past year initiation was determined by self-reported past year use, the age at first use, the year and month of recent new use, and the interview date.
Calculations of estimates of past year initiation did not take into account whether respondents initiated substance use while a resident of the United States. This method of calculation allowed for direct comparability with other standard measures of substance use because the populations of interest for the measures will be the same (i.e., both measures examined all possible respondents and were not restricted to those initiating substance use only in the United States).
One important note for initiation estimates is the relationship between the main categories and subcategories of substances (e.g., hallucinogens would be a main category, and LSD, phencyclidine [PCP], and Ecstasy would be subcategories in relation to hallucinogens). For most measures of substance use from the 2022 NSDUH, any member of a subcategory was by necessity a member of the main category (e.g., if respondents were past month users of Ecstasy, then they were also past month users of any hallucinogen). However, this situation was not true for estimates for the initiation of substance use. For example, an individual can initiate use of any hallucinogen, LSD, PCP, or Ecstasy only once. Respondents who initiated use of any hallucinogen more than 12 months ago by definition were not past year initiates of hallucinogen use, even if they initiated use of LSD, PCP, or Ecstasy in the past year.
A similar issue applied to initiation estimates for the aggregate substance use categories for the use of any illicit drug and the misuse of prescription psychotherapeutic drugs, tranquilizers or sedatives as a combined category, benzodiazepines, opioids (i.e., heroin or prescription pain relievers), and CNS stimulants (i.e., cocaine, methamphetamine, or prescription stimulants). People who first misused prescription stimulants in the past 12 months but who first misused prescription pain relievers more than 12 months prior to the interview date would be past year initiates for the misuse of stimulants. These people would not be past year initiates for the misuse of prescription psychotherapeutic drugs or illicit drugs because they had already misused pain relievers more than 12 months ago. Because of the potential for respondents to underreport lifetime (but not past year) misuse of prescription psychotherapeutic drugs (see the section below for the initiation of misuse of prescription psychotherapeutic drugs), lifetime (but not past year) misusers of prescription drugs could be misclassified as past year initiates for illicit drugs or other aggregate substance use categories (e.g., opioids) if they reported past year initiation of another illicit drug (e.g., heroin) but failed to report their lifetime misuse of a prescription psychotherapeutic drug (e.g., pain relievers). Section 4.5.3 discusses additional issues for the measurement of initiation of benzodiazepine misuse. For these reasons, the 2022 national reports and tables do not show initiation estimates for the aggregate categories of illicit drugs, prescription psychotherapeutic drugs, opioids, benzodiazepines, the aggregate category for tranquilizers or sedatives, or CNS stimulants.
Also, NSDUH does not collect initiation data for pipe tobacco. Therefore, the aggregate risk for initiation of use of any tobacco product cannot be determined (i.e., including pipe tobacco). For the same reason, the aggregate measure of any tobacco product or nicotine vaping cannot be determined, even though nicotine vaping initiation data are available starting with the 2022 NSDUH. The 2022 NSDUH also did not collect initiation data on marijuana vaping or vaping of flavoring. Thus, the 2022 national reports and tables did not show initiation estimates for the aggregate category of any tobacco product use, any tobacco product use or nicotine vaping, marijuana vaping, or vaping of flavoring.
In addition to estimates of the number of people initiating use of a substance in the past year, 2022 NSDUH estimates were computed for the mean age at first use or misuse among past year initiates of these substances. Unless specified otherwise, estimates of the mean age at initiation in the past 12 months were restricted to people aged 12 to 49 so that these mean age estimates were not influenced by those few respondents who were past year initiates and were aged 50 or older. As a measure of central tendency, means are influenced by the presence of extreme values in the data. Therefore, constraining the mean age estimates to past year initiates aged 12 to 49 was expected to increase the utility of these results to health researchers and analysts by providing a less biased picture of the substance use initiation behaviors among the civilian, noninstitutionalized population in the United States. This constraint was applied only to estimates of mean ages at first use and did not affect estimates for the 2022 NSDUH of the numbers of new users or associated percentages (e.g., the percentage of past year users who initiated use in the past year).
3.4.3.1 Initiation of Use of Substances from the Nicotine Section through the Methamphetamine Section
For cigarettes, nicotine vaping, smokeless tobacco, cigars, alcohol, marijuana, cocaine, crack cocaine, heroin, hallucinogens, inhalants, and methamphetamine, respondents were classified as past year initiates if there were fewer than 365 days between the interview date in 2022 and the imputed month, day, and year of first use of the relevant substance. (If respondents reported first use in calendar year 2021, their first use could be within 365 days of the interview date.) The total number of past year initiates can be used in the estimation of different percentages. For these substances, denominators for the percentages vary according to whether rates are being estimated for (1) all people in the population (or all people in a subgroup of the population, such as people in a given age group), (2) people who are at risk for initiation because they have not used the substance of interest prior to the past 12 months, or (3) past year users of the substance. The 2022 Detailed Tables (CBHSQ, 2023k) show all three of these percentages.
3.4.3.2 Initiation of Misuse of Prescription Psychotherapeutic Drugs
Respondents were asked about the initiation of misuse of prescription psychotherapeutic drugs only for the individual prescription drugs they misused in the past 12 months. Asking respondents to recall their first misuse of any prescription drug in an overall category (e.g., pain relievers) would require them to think about all prescription drugs that could have been available to them when they initiated misuse. However, some of these drugs may no longer have been available when respondents were interviewed.
If respondents reported they first misused one or more prescription drugs at an age or in a year and month more than 12 months prior to the interview date, they logically were not past year initiates for misuse of any drug in that psychotherapeutic category (e.g., pain relievers). For each psychotherapeutic drug category, as the number of prescription drugs that respondents misused in the past 12 months increases, the likelihood also increases that respondents will have first misused some of them more than 12 months ago.
If respondents reported only past year initiation of the drugs they misused in the past 12 months, they were asked in a follow-up question whether they ever misused any drug in that category more than 12 months prior to the interview.58 If respondents answered the follow-up question as “yes,” meaning that they had misused a drug in that category more than 12 months ago, then they were classified as not having initiated misuse in the past year. Respondents who answered the follow-up question as “no” were classified as past year initiates. If respondents answered the follow-up question on initiation as “don’t know” or “refused,” then their status as a past year initiate (or not) was resolved through imputation (see Section 2.3.3).
Because of this question structure for identifying respondents who initiated misuse of any psychotherapeutic drug in a given category in the past year, measures of the age and date of first misuse of any psychotherapeutic drug in that category were created only for respondents who were past year initiates. If past year initiates had no missing data for the age, year, and month when they first misused any psychotherapeutic drug in that category, then the age, year, and month of first misuse logically were assigned from the earliest reports.59 If past year initiates did not know or refused to report the age when they first misused some drugs in that category but they reported first misuse of at least one psychotherapeutic drug in the category at an age 1 year younger than their current age, then it nevertheless could be logically inferred that this was the age when these past year initiates first misused any drug in that category. Similarly, if past year initiates did not know or refused to report the year when they first misused some drugs in that category but they reported first misuse of at least one psychotherapeutic drug in the previous calendar year (e.g., 2021 for respondents in the 2022 NSDUH), then it could be logically inferred respondents initiated misuse of any drug in that category in the previous calendar year. If it was not possible to assign a definite age, year, and month of first misuse for a past year initiate based on the respondent’s questionnaire data, then these values were assigned through imputation.
Field testing in 2013 of redesigned questions for the initiation of misuse of prescription drugs indicated that estimated numbers and percentages of people who initiated the misuse of prescription drugs in the past year were similar based on the redesigned questions and questions from the main survey that measured past year initiation. For more information, see the 2013 Dress Rehearsal Final Report (CBHSQ, 2014c). Therefore, the 2022 national reports and tables present estimated numbers of people who initiated the misuse of prescription pain relievers, tranquilizers, stimulants, or sedatives in the past year.
In addition, the total number of past year initiates of misuse of any psychotherapeutic drug in a category can be used in the estimation of percentages among (1) all people in the population (or all people in a subgroup of the population, such as those in a given age group) and (2) people who were past year users of the substance. The 2022 Detailed Tables (CBHSQ, 2023k) show estimates for these two percentages.
As noted in Section 3.3.4, however, respondents who last misused prescription psychotherapeutic drugs in a category more than 12 months ago may underreport misuse, especially if they were not presented with examples of drugs formerly available by prescription in the United States but were no longer available at the time respondents were interviewed. These respondents who did not report misuse occurring more than 12 months ago would be misclassified as still being “at risk” for initiation of misuse of prescription drugs in that psychotherapeutic category.60 For this reason, the 2022 Detailed Tables do not show percentages for initiation of misuse of psychotherapeutic drugs among people who were at risk for initiation.
3.4.4 Substance Use Disorders
The NSDUH questionnaire included questions designed to measure dependence on nicotine in the form of cigarettes. The questionnaire also included questions to measure SUDs for alcohol and drugs. Estimates for nicotine dependence, SUD, and SUD severity for alcohol and drugs are included in national reports and tables for the 2022 NSDUH.
3.4.4.1 Nicotine Dependence
For nicotine (cigarettes), questions pertaining to dependence were based on the Nicotine Dependence Syndrome Scale (NDSS; Shiffman et al., 1995, 2004) and the Fagerstrom Test of Nicotine Dependence (FTND; Fagerstrom, 1978; Heatherton et al., 1991). Nicotine dependence was not measured for respondents who did not smoke cigarettes in the past month but used other products containing nicotine.
To identify patterns of nicotine (cigarette) dependence within the 2022 NSDUH data, questions measured dependence on nicotine through the use of cigarettes. Respondents were classified as being dependent if they met either the NDSS or the FTND classifications for dependence. The 2022 NSDUH contained 19 NDSS questions addressing five aspects of dependence: (1) smoking drive (compulsion to smoke driven by nicotine craving and withdrawal), (2) nicotine tolerance, (3) continuous smoking, (4) behavioral priority (i.e., preferring smoking over other reinforcing activities), and (5) stereotypy (i.e., fixed patterns of smoking). The 2022 NSDUH contained two mutually exclusive questions (DRCGE19a and DRCGE19b) addressing the FTND measure of dependence. These questions ask respondents who reported smoking cigarettes in the past month if the first cigarette they smoked was within 30 minutes of waking up on the days they smoked.
Missing data for nicotine dependence were statistically imputed for past month cigarette smokers. See Section 2.3.3 for more information on the imputation procedures.
3.4.4.2 Substance Use Disorder for Drugs and Alcohol
SUD estimates for drugs and alcohol in the 2022 NSDUH were based on the criteria in the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5; American Psychiatric Association [APA], 2013). Respondents were asked SUD questions separately for any drugs or alcohol they used in the 12 months prior to the survey.61 SUD questions for drugs applied to marijuana, cocaine (including crack), heroin, hallucinogens, inhalants, methamphetamine, and any use of psychotherapeutic drugs, including prescription pain relievers, tranquilizers, stimulants, or sedatives. Beginning in 2021, NSDUH respondents who reported any use of prescription psychotherapeutic drugs in the past year (i.e., not just misuse of prescription drugs) were asked the respective SUD questions for that category of prescription drugs. More information on how the current SUD questions were developed based on DSM-5 criteria can be found in the 2020 Methodological Summary and Definitions report (CBHSQ, 2021).
Users of NSDUH reports and tables also need to be aware that questions about the use of IMF appeared after SUD questions in the 2022 NSDUH questionnaire. See Section 3.4.2 for additional information about the IMF questions. For this reason, overall SUD, drug use disorder, and opioid use disorder measures do not capture disorders arising solely from the use of IMF.
DSM-5 includes the following SUD criteria (as measured in the 2022 NSDUH):
- The substance is often taken in larger amounts or over a longer period than intended.
- There is a persistent desire or unsuccessful efforts to cut down or control substance use.
- A great deal of time is spent in activities necessary to obtain the substance, use the substance, or recover from its effects.
- There is craving, or a strong desire or urge, to use the substance.
- There is recurrent substance use that results in a failure to fulfill major role obligations at work, school, or home.
- There is continued substance use despite having persistent or recurrent social or interpersonal problems caused by or exacerbated by the effects of the substance.
- Important social, occupational, or recreational activities are given up or reduced because of substance use.
- There is recurrent substance use in situations in which it is physically hazardous.
- Substance use is continued despite knowledge of having a persistent or recurrent physical or psychological problem that is likely to have been caused or exacerbated by the substance.
- There is a need for markedly increased amounts of the substance to achieve intoxication or the desired effect, or there is a markedly diminished effect with continued use of the same amount of the substance (i.e., tolerance).
- There are two components of withdrawal symptoms, either of which meet the overall criterion for withdrawal symptoms:
- There is a required number of withdrawal symptoms that occur when substance use is cut back or stopped following a period of prolonged use.62
- The substance or a related substance is used to get over or avoid withdrawal symptoms.63
Table 3.6 shows how these 11 DSM-5 SUD criteria apply to substances in NSDUH. For prescription psychotherapeutic drugs (i.e., prescription pain relievers, tranquilizers, stimulants, or sedatives), Table 3.6 also shows how these criteria apply if respondents misused prescription psychotherapeutic drugs, or if they simply used but did not misuse prescription drugs in the past year. For consistency with the DSM-5 criteria, NSDUH respondents were classified as having an SUD if they met two or more of the applicable criteria in a 12-month period.
For alcohol, marijuana, cocaine, heroin, and methamphetamine, shown in Table 3.6, respondents were classified as having an SUD if they had at least 2 of the 11 criteria in a 12-month period. However, for hallucinogens and inhalants, respondents were classified with a disorder if they had at least 2 of the first 10 criteria in the past 12 months; the withdrawal criterion does not apply to hallucinogens and inhalants.
For the prescription drugs shown in Table 3.6, the applicable DSM-5 criteria for classifying respondents as having a prescription drug use disorder depends on whether respondents misused prescription drugs or used but did not misuse prescription drugs in the past year. If respondents misused prescription drugs in the past year, they were classified as having a prescription drug use disorder if they had at least 2 of the 11 criteria shown in Table 3.6. However, if respondents used but did not misuse prescription drugs in the past year, they were classified as having a prescription drug use disorder if they had at least two of the first nine criteria shown in Table 3.6. Criteria 10 (tolerance) and 11 (withdrawal) do not apply to respondents who used but did not misuse these prescription drugs in the past year; tolerance and withdrawal can occur as normal physiological adaptations when people use these prescription drugs appropriately under medical supervision (Hasin et al., 2013).
Substances64 that are included in selected SUD measures in the 2022 NSDUH are as follows:
- Any SUD in the past year includes data from past year users of alcohol, marijuana, cocaine (including crack), heroin, hallucinogens, inhalants, and methamphetamine, and any past year users of prescription psychotherapeutic drugs.
- Alcohol use disorder includes only data from past year users of alcohol.
- Drug use disorder includes data from past year users of marijuana, cocaine, heroin, hallucinogens, inhalants, and methamphetamine, and any past year users of prescription psychotherapeutic drugs.
- Opioid use disorder includes data from past year users of heroin or prescription pain relievers. Respondents were not counted as having an opioid use disorder if they did not meet the full SUD criteria for heroin or prescription pain relievers individually. The opioid use disorder estimates do not capture symptoms that arose solely from the use of IMF.
- Tranquilizer or sedative use disorder includes data from past year users of prescription tranquilizers or sedatives. Respondents were not counted as having a tranquilizer or sedative use disorder if they did not meet the full SUD criteria for prescription tranquilizers or sedatives individually.
- CNS stimulant use disorder includes data from past year users of cocaine, methamphetamine, or prescription stimulants. Respondents were not counted as having a CNS stimulant use disorder if they did not meet the full SUD criteria individually for cocaine, methamphetamine, or prescription stimulants.
3.4.4.2.1 Substance Use Disorder Severity
The severity of an SUD according to DSM-5 criteria was determined by the number of individual criteria that a respondent met for a particular substance (or category of substances). The number of criteria required for each severity level was the same for every substance, regardless of how many criteria were applicable for the substance.
- People were classified in the mild SUD category if they met two or three applicable SUD criteria for that substance.
- People were classified in the moderate SUD category if they met four or five applicable SUD criteria for that substance.
- People were classified in the severe SUD category if they met six or more applicable SUD criteria for that substance.
In addition to the three levels of SUD severity for individual substances, measures of severity were created for aggregate SUD categories starting with the 2022 NSDUH. These aggregate SUD categories include the following: any SUD, drug use disorder, prescription drug use disorder, tranquilizer or sedative use disorder, opioid use disorder, and CNS stimulant use disorder. SUD severity measures for these aggregate SUD categories were defined from the maximum severity level (i.e., mild, moderate, or severe) across the multiple SUDs that were included in the category. For example, if people had a moderate alcohol use disorder and a mild marijuana (cannabis) use disorder as their only SUDs in the past year, then they were classified as having moderate SUD.
3.4.4.2.2 Imputation Issues for Substance Use Disorder Measures
Missing values in the DSM-5 SUD data for drugs and alcohol were replaced using the modified predictive mean neighborhood (modPMN) imputation method described in Section 2.3.3. Consequently, the variables for specific SUDs (e.g., marijuana use disorder) as well as the composite SUD variables (e.g., drug use disorder) were not subject to the kinds of potential biases due to missing data described in Section 3.3.2.
Imputation of the SUD data reflected imputation of the variables from the corresponding substance use sections for the most recent use of alcohol, marijuana, cocaine, heroin, hallucinogens, inhalants, methamphetamine, or prescription psychotherapeutic drugs. Only the final SUD and severity outcomes were imputed; individual SUD criterion variables for a given substance were not. For example, if the edited variable for the most recent use of any hallucinogen was missing and then imputed to indicate use more than 12 months ago, then no imputation was necessary for hallucinogen use disorder. On the other hand, if that variable was missing and imputed to indicate past year use, then the overall SUD measures for hallucinogens were also imputed, but not the individual criterion variables for hallucinogen use disorder. ModPMN was also used to determine the respondent’s final hallucinogen use disorder outcome if the respondent reported fewer than two SUD symptoms for hallucinogens, and there were enough individual hallucinogen use disorder questions with missing data to potentially classify the respondent as having a past year hallucinogen use disorder.
3.4.4.2.3 Miscellaneous Data Issues for Substance Use Disorder Measures
For alcohol and marijuana, respondents were asked the SUD questions if they reported substance use on more than 5 days in the past year or if they reported any substance use in the past year but had missing data for the frequency of use in the past year. Therefore, inconsistencies could occur where respondents were classified as having an alcohol or marijuana use disorder but their imputed frequency of use was fewer than 6 days in the past year. However, this situation was uncommon. For example, about 42,000 respondents (unweighted) reported past year alcohol use in 2022.65 Of these, fewer than 100 respondents had missing frequency data and were asked the alcohol use disorder questions, but their final imputed frequency of use indicated they used alcohol on 5 or fewer days in the past year.
For methamphetamine, cocaine, and heroin, respondents were asked the respective SUD questions if they reported past year use in the corresponding substance use sections or if they reported use in the past year in the special drugs section (i.e., use of methamphetamine, cocaine, or heroin with a needle in the past year or smoking or sniffing of heroin in the past year). Thus, the questionnaire logic allowed some respondents to be asked the SUD questions for these drugs even if they did not report past year use when they were asked previously about their most recent use of methamphetamine, cocaine, crack cocaine, or heroin. For example, about 1,350 respondents in 2022 were asked the questions about cocaine use disorder because they reported past year use when asked directly about their most recent use of cocaine or crack. Fewer than 10 additional respondents were asked these questions because they reported past year use of cocaine with a needle in the special drugs section despite not having previously reported past year use of cocaine or crack.
Missing or incomplete responses were not imputed in the special drugs section for the use of cocaine, heroin, or methamphetamine with a needle or for smoking or sniffing heroin. Therefore, if respondents were imputed to have last used cocaine, heroin, or methamphetamine more than 12 months ago and there was nothing from the special drugs section to indicate use of these substances in the past year, then the final imputed recency was used to infer that SUD questions did not apply.
3.4.5 Substance Use Treatment
The substance use treatment questions underwent considerable revisions for the 2022 NSDUH (see Section 2.2.2). These revisions were intended to reflect contemporary changes in the delivery of substance use treatment services. The following is a summary of key changes to these questions:
- Respondents who reported any lifetime use of prescription drugs (pain relievers, tranquilizers, stimulants, or sedatives) were eligible to be asked questions about the receipt of substance use treatment. Before 2022, respondents who reported lifetime use but not misuse of prescription drugs were not asked about the receipt of substance use treatment unless they reported lifetime use of alcohol or other drugs (marijuana, cocaine, heroin, hallucinogens, inhalants, or methamphetamine).
- All lifetime users of alcohol or drugs were asked whether they received specific types of treatment services in the past 12 months. Before 2022, respondents who reported the use of alcohol or illicit drugs were not asked about the specific types of treatment they received in the past 12 months unless they first reported that they received any substance use treatment in their lifetime and, if so, whether they received substance use treatment in the past 12 months. These changes meant that more respondents in 2022 were eligible to be asked the questions about the types of treatment they received in the past 12 months.
- Additional inpatient and outpatient locations were provided for respondents to report where they received treatment in the past 12 months.
- The question about the receipt of treatment over the phone or through video (i.e., telehealth treatment) was revised to emphasize that the treatment was from a therapist or other healthcare professional.
- Questions about medication-assisted treatment (MAT) for alcohol or opioid use were added to the section.
Because of these changes, the definition for the receipt of substance use treatment changed for 2022. Estimates based on these outcomes should not be compared with estimates from 2021 because of these changes.66
Questions for whether respondents perceived an unmet need for substance use treatment and the reasons for respondents not receiving treatment were also revised for the 2022 NSDUH. Specific changes are discussed in Section 3.4.5.5 in the context of measurement of these outcomes.
3.4.5.1 Locations and Types of Services for Substance Use Treatment
Receipt of substance use treatment includes the receipt of treatment in the past year for the use of alcohol or drugs in an inpatient location, in an outpatient location, via telehealth, or in a prison, jail, or juvenile detention center. The definition also includes the receipt of MAT for alcohol use or opioid use.
Inpatient treatment was defined as treatment in locations where people stayed overnight or longer and included the following locations:
- a hospital,
- a residential drug or alcohol rehab or treatment center,
- a residential mental health treatment center, or
- some other place the respondent stayed overnight or longer.
Outpatient treatment was defined as treatment people received in locations that did not require them to stay overnight or longer. These measures included the following locations:
- a drug or alcohol rehab or treatment center as an outpatient;
- a mental health treatment center as an outpatient;
- the office of a therapist, psychologist, psychiatrist, or other substance use treatment professional;
- a general medical clinic or doctor’s office;
- a hospital as an outpatient;
- a school health or counseling center; or
- some other place as an outpatient.
Questions on MAT were asked in the emerging issues section of the questionnaire in 2021. MAT questions were asked in the alcohol or drug treatment section for 2022 in the context of questions about other substance use treatment in the past 12 months. The skip logic also changed for these questions in the alcohol and drug treatment section. Therefore, estimates for MAT in 2022 are not comparable with MAT estimates from the 2021 NSDUH.67
Variables for the receipt of substance use treatment in the past year in inpatient locations, in outpatient locations, via telehealth, or in a prison, jail, or juvenile detention center were statistically imputed. Variables also were statistically imputed for the receipt of MAT in the past year for alcohol use or opioid use. Consequently, variables in 2022 for the receipt of substance use treatment in the past year do not have missing data.
3.4.5.2 Receipt of Other Services for Substance Use
The 2022 NSDUH also collected information on the receipt of other services in the past year for the use of alcohol or drugs, including the following:
- services from a support group,
- services from a peer support specialist or recovery coach,
- services in an emergency room or emergency department, or
- detoxification or withdrawal support services.
NSDUH did not classify these other services as “substance use treatment.” However, they were included in a separate aggregate measure created to cover the receipt of substance use treatment or other services. Variables for the receipt of other services also were statistically imputed, so these variables do not have missing data.
3.4.5.3 Substances for Which People Received Treatment or Other Services
If respondents in the 2022 NSDUH reported the use of alcohol or drugs in their lifetime and reported that they received substance use treatment in the past year, then they were asked to report the substances for which they received treatment. Questions varied depending on the locations in which people received treatment or whether people received treatment via telehealth. Respondents who received other services in the past year also were asked to report the substances for which they received these services.
For MAT, the questions applied specifically to the use of alcohol or opioids. Respondents who reported lifetime use of alcohol were asked whether they received MAT specifically for their use of alcohol. Although respondents were asked about MAT for their use of “drugs,” respondents were asked this MAT question only if they reported lifetime use of opioids (i.e., heroin or prescription pain relievers).
3.4.5.3.1 Substances for Which People Received Treatment as Inpatients or Outpatients
Respondents who reported receiving treatment in inpatient or outpatient locations in the past year were asked to report whether they received treatment for the specific substances that they used in their lifetime; respondents were asked about treatment only for the specific substances they reported using in their lifetime. Respondents could be asked about the receipt of inpatient or outpatient treatment for the following substances:
- alcohol,
- marijuana,
- cocaine,
- heroin,
- hallucinogens,
- inhalants,
- methamphetamine,
- prescription pain relievers,
- prescription tranquilizers,
- prescription stimulants, and
- prescription sedatives.
For example, if respondents reported the use of alcohol, marijuana, cocaine, heroin, and prescription pain relievers in their lifetime, and they reported the receipt of treatment in one or more outpatient locations in the past year, then they were asked whether they received treatment as an outpatient for their use of these five specific substances. These respondents were not asked whether they received treatment as an outpatient for their use of hallucinogens, inhalants, methamphetamine, prescription tranquilizers, prescription stimulants, or prescription sedatives. The same logic applied to questions for the receipt of treatment as an inpatient in the past year for the use of specific substances respondents used in their lifetime.
In addition to these substances, respondents who reported substance use treatment as an inpatient or an outpatient were asked whether they received treatment in the respective setting for their use of “some other drug.” Respondents were asked to specify the other drug or drugs for which they received treatment. However, respondents might not report that they received treatment as an inpatient or an outpatient for their use of a particular substance (e.g., heroin), but they could specify it as “some other drug” for which they received treatment. In this situation, respondents were logically inferred to have received treatment in the respective setting for their use of that substance.
In 2022, relatively large proportions of people who reported that they received inpatient or outpatient treatment in the past 12 months did not indicate the specific substance(s) for which they received treatment in these locations, including treatment for the use of some other drug. Stated another way, these reports of inpatient or outpatient treatment were not substantiated by reports of treatment for the use of specific substances. Specifically, more than one fourth (26.2 percent) of respondents who reported inpatient treatment in the past year did not report the specific substances for which they received treatment as inpatients. Among respondents who reported outpatient treatment in the past year, more than one third (35.0 percent) did not report the specific substances for which they received treatment as outpatients. A “substance unspecified” category was created for these respondents. Because of this issue with unsubstantiated data for inpatient or outpatient treatment for the use of specific substances, estimates of treatment for the use of alcohol, drugs, or both alcohol and drugs in these locations are likely to be underestimates.
3.4.5.3.2 Substances for Which People Received Treatment via Telehealth or in a Prison, Jail, or Juvenile Detention Center
Respondents who reported treatment via telehealth or treatment in a prison, jail, or juvenile detention center and who reported lifetime use of alcohol and drugs were asked for these types of treatment whether they received treatment for their use of alcohol only, drugs only, or both alcohol and drugs. Respondents were not asked these follow-up questions if they reported lifetime use of only alcohol or only drugs.
If respondents reported the lifetime use of alcohol, and they answered “no” to all questions about their use of drugs in their lifetime, then they were logically inferred to have received these types of treatment for their use of alcohol only. Similarly, if respondents reported the lifetime use of one or more drugs, and they answered “no” for their lifetime use of alcohol, then they were logically inferred to have received these types of treatment for their use of drugs only.
Edited variables for the receipt of these types of treatment for the use of alcohol only, drugs only, or both alcohol and drugs retained missing values if respondents reported the lifetime use of only alcohol or only drugs but they had missing data for the lifetime use of some substances. The recoded variables for the receipt of these types of treatment for the use of alcohol or the use of drugs reduced the amount of missing data by incorporating data from imputed variables for the lifetime use of alcohol and drugs. Respondents with missing data were excluded from the analyses to produce published estimates for the 2022 NSDUH. See Section 3.3.2 for a discussion of the potential bias in estimates because of missing data.
3.4.5.3.3 Substances for Which People Received Other Substance Use Services
For the other substance use services described in Section 3.4.5.2, respondents who reported lifetime use of alcohol and drugs were asked for these other services whether they received these other services for their use of alcohol only, drugs only, or both alcohol and drugs. Respondents were not asked these follow-up questions if they reported lifetime use of only alcohol or only drugs.
If respondents reported the lifetime use of alcohol, and they answered “no” to all questions about their use of drugs in their lifetime, then they were logically inferred to have received these other services for their use of alcohol only. Similarly, if respondents reported the lifetime use of one or more drugs, and they answered “no” for their lifetime use of alcohol, then they were logically inferred to have received these other services for their use of drugs only.
Edited variables for the receipt of specific other services for the use of alcohol only, drugs only, or both alcohol and drugs retained missing values if respondents reported the lifetime use of only alcohol or only drugs but they had missing data for the lifetime use of some substances. For example, if respondents reported the lifetime use of only alcohol, but they did not know or refused to report whether they had ever used some drugs, then it could not be determined unambiguously that these respondents were lifetime users of only alcohol.
3.4.5.4 Need for Substance Use Treatment
Historically, NSDUH data products have included substance use treatment at a “specialty facility” in the past year as part of the definition for whether people needed substance use treatment. With the changes to the questionnaire in 2022, the term “specialty facility&38221; was dropped from 2022 NSDUH data products.
Consequently, the definition of the need for substance use treatment was revised for 2022. Respondents were classified as needing substance use treatment if they had an SUD in the past year, as defined in Section 3.4.4, or they received substance use treatment in the past year, as defined in Section 3.4.5.1. As noted previously, the “other services” described in Section 3.4.5.2 were not counted as substance use treatment. Because the variables for SUD and receipt of substance use treatment in the past year were imputed, the variable for the need for substance use treatment does not have missing data.
3.4.5.5 Perceived Unmet Need for Substance Use Treatment and Reasons for Not Receiving Treatment
Questions about whether respondents perceived an unmet need for substance use treatment and the reasons for not receiving treatment were also revised for the 2022 NSDUH.
- These questions were asked only if respondents did not report any substance use treatment in the past year (i.e., inpatient or outpatient treatment, MAT, telehealth treatment, or treatment in a prison, jail, or juvenile detention center).
- Respondents were asked whether they sought treatment rather than being asked whether they made an effort to get treatment.
- Respondents who did not report seeking treatment were asked whether they thought they should get treatment rather than being asked whether they thought they needed treatment.
NSDUH respondents were classified as having a perceived unmet need for substance use treatment if they did not receive substance use treatment in the past year and reported either of the following:
- They did not receive other services (i.e., from a support group, from a peer support specialist or recovery coach, in an emergency room, or detoxification or withdrawal support services), but they sought or thought they should get treatment in the past 12 months for their use of alcohol or drugs.
- They received other services, described in the preceding bullet, but they sought or thought they should get additional professional counseling, medication, or other treatment in the past 12 months for their use of alcohol or drugs.
Questions about reasons for people not receiving substance use treatment (i.e., barriers to treatment) were asked of respondents who reported a perceived unmet need for substance use treatment. The list of possible barriers to treatment also was expanded for 2022. For each reason for not receiving treatment, respondents were asked whether that reason was “one of the reasons” or “not one of the reasons” they did not seek or get professional counseling, medication, or other treatment for their use of alcohol or drugs.
Variables for the perceived unmet need for substance use treatment and reasons for not receiving substance use treatment were not imputed and had missing data. Respondents with missing data were excluded from the analyses used to produce published estimates for the 2022 NSDUH. See Section 3.3.2 for a discussion of the potential bias in estimates because of missing data.
3.4.6 Mental Health Treatment
The mental health treatment questions underwent considerable revisions for the 2022 NSDUH (see Section 2.2.2). These revisions were intended to reflect contemporary changes in the delivery of mental health treatment services. The changes also made the content more similar between the alcohol and drug treatment and the mental health services utilization sections of the questionnaire. The following is a summary of key changes to these questions:
- Adolescents aged 12 to 17 and adults aged 18 or older received the same questions about mental health treatment. Before 2022, adolescents and adults received different sets of questions, such that measures differed for adolescents and adults. Consequently, the youth mental health service utilization section was dropped from the 2022 questionnaire.
- The instruction was dropped for respondents not to report mental health treatment associated with their use of alcohol or drugs.
- Terminology was updated, and questions were included about newer treatment options.
- Additional inpatient and outpatient locations were provided for adult respondents to report where they received treatment in the past 12 months. As noted previously, adolescents were asked about the receipt of treatment in these same inpatient and outpatient locations.
- The question about the use of medication that was prescribed to help with people’s mental health applied to all respondents, regardless of age. Before 2022, only adults were asked whether they took prescribed medication to help with their mental health.
- The question about treatment received in prison, jail, or a juvenile detention center applied to all respondents, regardless of age. Before 2022, only adolescents were asked whether they received services in a juvenile detention center, prison, or jail.
- The question about the receipt of treatment over the phone or through video (i.e., telehealth treatment) was revised to emphasize that the treatment was from a therapist or other healthcare professional.
Because of these changes, the definition for the receipt of mental health treatment changed for 2022. Estimates based on these outcomes should not be compared with estimates from 2021 because of these changes.68
Questions for whether respondents perceived an unmet need for mental health treatment and the reasons for respondents not receiving treatment were also revised for the 2022 NSDUH. Specific changes are discussed in Section 3.4.6.3 in the context of measurement of these outcomes.
3.4.6.1 Locations and Types of Services for Mental Health Treatment
Receipt of mental health treatment includes the receipt of treatment in the past year to help people with their mental health, emotions, or behavior that was received in an inpatient location, in an outpatient location, via telehealth, or in a prison, jail, or juvenile detention center. The definition also includes the receipt of prescription medication to help with mental health, emotions, or behavior.
Inpatient treatment was defined as treatment in locations where people stayed overnight or longer and included the following locations:
- a hospital,
- a residential mental health treatment center,
- a residential drug or alcohol rehab or treatment center, or
- some other place the respondent stayed overnight or longer.
Outpatient treatment was defined as treatment people received in locations that did not require them to stay overnight or longer. These measures included the following locations:
- a mental health treatment center as an outpatient;
- a drug or alcohol treatment or rehab center as an outpatient;
- the office of a therapist, psychologist, psychiatrist, or mental health professional;
- a general medical clinic or doctor’s office;
- a hospital as an outpatient;
- a school health or counseling center; or
- some other place as an outpatient.
Variables for the receipt of mental health treatment in the past year in inpatient locations, in outpatient locations, via telehealth, or in a prison, jail, or juvenile detention center were statistically imputed. The variable for the receipt of prescription medication to treat mental health, emotions, or behavior in the past year was also statistically imputed. Consequently, variables in 2022 for the receipt of mental health treatment in the past year do not have missing data.
In addition to presenting estimates for all respondents who received mental health treatment in the past year, NSDUH presents estimates for the receipt of mental health treatment among people with mental health issues. NSDUH produces direct estimates only for the mental disorder of an MDE and an MDE with severe impairment in the past year (see Section 3.4.9). Estimates are presented in national reports and tables for the receipt of mental health treatment in the past year among adolescents aged 12 to 17 who had an MDE or an MDE with severe impairment in the past year. Adolescent respondents with missing data for the MDE outcomes were excluded from the analyses (see Section 3.3.2).
The 2022 Key Substance Use and Mental Health Indicators report (CBHSQ, 2023j) also presents estimates for the receipt of mental health treatment among adults aged 18 or older who had an MDE in the past year. As noted in Table 2.5 and Section 3.4.9, the past year MDE variable for adults was imputed for 2022 and had no missing data.
NSDUH also produces estimates for AMI and SMI in the past year among adults aged 18 or older. However, these estimates are model-based predictions rather than direct estimates (see Section 3.4.8). Nevertheless, national reports and tables for the 2022 NSDUH present estimates of the numbers and percentages of adults with AMI or SMI who received mental health treatment in the past year. Because variables for AMI, SMI, and the receipt of mental health treatment among adults were imputed for 2022, these estimates had no missing data.
3.4.6.2 Receipt of Other Services for Mental Health Treatment
The 2022 NSDUH also collected information on the receipt of other services in the past year to help people with their mental health, emotions, or behavior. These other services include the following:
- services from a support group,
- services from a peer support specialist or recovery coach, or
- services in an emergency room or emergency department.
NSDUH did not classify these other services as “mental health treatment.” However, they were included in a separate aggregate measure created to cover the receipt of mental health treatment or other services. Variables for the receipt of other services also were statistically imputed, so these variables do not have missing data.
3.4.6.3 Perceived Unmet Need for Mental Health Treatment and Reasons for Not Receiving Treatment
Questions about whether respondents perceived an unmet need for mental health treatment and the reasons for not receiving treatment were also revised for the 2022 NSDUH.
- These questions were asked of respondents of all ages, including those aged 12 to 17, not just adults aged 18 or older.
- These questions were asked only if respondents did not report mental health treatment (i.e., inpatient or outpatient treatment, prescription medication, telehealth treatment, or treatment in prison or jail).
- Respondents were asked whether they sought treatment.
- Respondents who did not report seeking treatment were asked whether they thought they should get treatment.
NSDUH respondents were classified as having a perceived unmet need for mental health treatment if they did not receive mental health treatment in the past year and reported either of the following:
- They did not receive other services (i.e., from a support group, from a peer support specialist or recovery coach, or in an emergency room), but they sought or thought they should get treatment in the past 12 months for their mental health, emotions, or behavior.
- They received other services, described in the preceding bullet, but they sought or thought they should get additional professional counseling, medication, or other treatment in the past 12 months for their mental health, emotions, or behavior.
Questions about reasons for people not receiving mental health treatment (i.e., barriers to treatment) were asked of respondents who reported a perceived unmet need for mental health treatment. The list of possible barriers to treatment was expanded for adults in 2022, and the same list was asked of adolescents aged 12 to 17. For each reason for not receiving mental health treatment, respondents were asked whether that reason was “one of the reasons” or “not one of the reasons” they did not seek or get professional counseling, medication, or other treatment for their mental health, emotions, or behavior.
Variables for the perceived unmet need for mental health treatment and reasons for not receiving mental health treatment were not imputed and had missing data. Respondents with missing data were excluded from the analyses to produce published estimates for the 2022 NSDUH. See Section 3.3.2 for a discussion of the potential bias in estimates because of missing data.
3.4.7 Definition of County Type
County type is based on the “Rural-Urban Continuum Codes”69 developed by the U.S. Department of Agriculture. The county type measure used in 2022 was based on the 2013 Rural-Urban Continuum Codes. Because counties are defined for all NSDUH respondents, the county type measures did not have missing data.
To create the 2013 Rural-Urban Continuum Codes, all U.S. counties and county equivalents were first grouped according to their official metropolitan-nonmetropolitan status (i.e., statistical area definitions), as determined by the Office of Management and Budget (OMB) in February 2013. This grouping distinguished metropolitan counties by the population size of their metropolitan area and nonmetropolitan counties by their degree of urbanization and adjacency to a metropolitan area. The OMB determined current metropolitan status by applying population and worker commuting criteria to the results of the 2010 census and the 2006-2010 ACS. No major changes were made in either the metropolitan-nonmetropolitan or urban-rural criteria between 2000 and 2010. However, the decennial census long form was eliminated in 2010, and the OMB used 5-year average commuting flow data from the 2006-2010 ACS rather than a point-in-time estimate to delineate metropolitan and micropolitan areas.
Nonmetropolitan counties in the three urban-sized categories were further subdivided by whether the county was adjacent to one or more metropolitan areas. A nonmetropolitan county was defined as adjacent if it physically adjoined one or more metropolitan areas and had at least 2 percent of its employed labor force commuting to central metropolitan counties. Nonmetropolitan counties not meeting these criteria were classed as nonadjacent. The 2006-2010 ACS commuting flow data were also used to compute adjacency for the 2013 Rural-Urban Continuum Codes.
Metropolitan and nonmetropolitan categories were subdivided into three metropolitan and six nonmetropolitan categories, resulting in a nine-part county codification.
- Large metropolitan statistical areas (MSAs) (large metropolitan) have a total population of 1 million or more.
- Small MSAs (small metropolitan) have a total population of fewer than 1 million. Small MSAs are further subdivided into two categories:
- 250,000 to 1 million population; and
- less than 250,000 population.
- Nonmetropolitan counties were classified according to the aggregate size of their urban population and whether they were adjacent to a metropolitan area. Nonmetropolitan areas include counties in micropolitan statistical areas and counties outside of both metropolitan and micropolitan statistical areas and are classified as follows:
- urban population of 20,000 or more, adjacent to a metropolitan area (“urbanized adjacent”);
- urban population of 20,000 or more, not adjacent to a metropolitan area (“urbanized nonadjacent”);
- urban population of 2,500 to 19,999, adjacent to a metropolitan area (“less urbanized adjacent”);
- urban population of 2,500 to 19,999, not adjacent to a metropolitan area (“less urbanized nonadjacent”);
- completely rural or less than 2,500 urban population, adjacent to a metropolitan area (“completely rural adjacent”); and
- completely rural or less than 2,500 urban population, not adjacent to a metropolitan area (“completely rural nonadjacent”).
For NSDUH, nonmetropolitan counties were categorized as “urbanized,” “less urbanized,” and “completely rural.” The terms “urbanized,” “less urbanized,” and “completely rural” for counties are not based on the relative proportion of the county population in urbanized areas but rather are based on the absolute size of the population in urbanized areas. For example, some counties classified as “less urbanized” had over 50 percent of the county population residing in urbanized areas, but this percentage represented fewer than 20,000 people in the county.
3.4.8 Estimation of Serious and Other Levels of Mental Illness
3.4.8.1 Background
The 1992 Alcohol, Drug Abuse, and Mental Health Administration Reorganization Act created SAMHSA and required the new organization to develop a definition and methodology for estimating SMI among adults for use by states in developing their plans for use of block grant funds distributed by SAMHSA. A technical advisory group convened by SAMHSA was tasked with developing a definition of SMI, which was published in the Federal Register in 1993 (SAMHSA, 1993):
Pursuant to Section 1912(c) of the Public Health Service Act, as amended by Public Law 102-321, “adults with serious mental illness” are defined as the following:
- Individuals aged 18 and over, who currently or at any time during the past year, have had diagnosable mental, behavioral, or emotional disorder of sufficient duration to meet diagnostic criteria specified within DSM-III-R that has resulted in functional impairment, which substantially interferes with or limits one or more major life activities.
- These disorders include any mental disorder (including those of biological etiology) listed in DSM-III-R or their ICD-9-CM equivalent (and subsequent revisions), with the exception of DSM-III-R “V” codes, substance use disorders, and developmental disorders, which are excluded unless they co-occur with other diagnosable serious mental illness.
- All of these disorders have episodic, recurrent, or persistent features; however, they vary in terms of severity or disabling effects. Functional impairment is defined as difficulties that substantially interfere with or limit role functioning in one or more major life activities including basic daily living skills (e.g., eating, bathing, dressing); instrumental living skills (e.g., maintaining a household, managing money, getting around the community, taking prescribed medication); and functioning in social, family, and vocational/educational contexts.
- Adults who would have met functional impairment criteria during the referenced year without benefit of treatment or other support services are considered to have serious mental illness.
In December 2006, a new technical advisory group was convened by SAMHSA’s Office of Applied Studies (which later became CBHSQ) and the Center for Mental Health Services to solicit recommendations for data collection strategies to address SAMHSA’s legislative requirements. Although the technical advisory group recognized the ideal way to estimate SMI in NSDUH would be to administer a clinical diagnostic interview annually to all adult respondents, this approach was not feasible because of constraints on the interview time and the need for trained mental health clinicians to conduct the interviews. Therefore, the approach recommended by the technical advisory group and adopted by SAMHSA for NSDUH was to utilize short scales in the NSDUH interview to separately measure psychological distress and functional impairment. NSDUH used the resulting data collected from these short scales in a statistical model to predict whether a respondent had mental illness.
To create the model, SAMHSA’s CBHSQ initiated a Mental Health Surveillance Study (MHSS) in 2007 as part of NSDUH to develop and implement methods to estimate SMI. Models using the short scales for psychological distress and impairment to predict mental illness status were developed from a subsample of adult respondents who had completed the NSDUH interview and were administered a clinical psychological diagnostic interview soon afterward. For the clinical interview data, people were classified as having SMI if they had a diagnosable mental, behavioral, or emotional disorder in the past 12 months according to the criteria in the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV; APA, 1994) (other than a developmental disorder or SUD) that resulted in substantial functional impairment. This estimation methodology was implemented in the 2008 NSDUH.
3.4.8.2 Historical Summary of the 2012 Model
The model used for 2022 to predict mental illness was developed for the 2012 NSDUH (subsequently referred to as the “2012 model”). This 2012 model was previously updated from a model developed in 2008 (subsequently referred to as the “2008 model”).
The 2008 model was created from a randomly selected subsample of approximately 1,500 adults in 2008 who had completed the NSDUH interview. Respondents were recruited for a follow-up clinical interview consisting of a diagnostic assessment for mental disorders.70 In order to determine the optimal scale for measuring functional impairment in NSDUH, roughly half of the adult respondents were assigned to receive an abbreviated eight-item version of the World Health Organization Disability Assessment Schedule (WHODAS; Novak et al., 2010), and the other half were assigned to receive the Sheehan Disability Scale (SDS; Leon et al., 1997). The WHODAS was chosen as the impairment scale to be administered in the 2009 and subsequent NSDUHs (Office of Applied Studies, 2009). For more information on the 2008 MHSS design and analysis, see Colpe et al. (2009) and Office of Applied Studies (2009). Information about the 2008 model is available in Appendix B of the Results from the 2012 National Survey on Drug Use and Health: Mental Health Findings (CBHSQ, 2013b) and in Section 3.4.7 of the 2020 Methodological Summary and Definitions report (CBHSQ, 2021).
Based on the accumulated MHSS clinical data collected from 2008 to 2012, however, SAMHSA determined the 2008 model had some important shortcomings not detected in the original model fitting because of the small sample of clinical interview respondents in 2008. Specifically, estimates of SMI and AMI among young adults based on the 2008 model were higher than the estimates for this age group based on the clinical interview data. In addition, improvements were needed in the weighting procedures for the MHSS clinical data to account better for undercoverage and nonresponse. Only NSDUH respondents in 2008 who answered their surveys in English were eligible for the clinical follow-up, and people with mental illness appeared to be more likely to participate in the follow-up.
Because of these concerns, SAMHSA fit a more accurate model using combined 2008-2012 clinical data for the 2012 model. To reduce bias and improve prediction, additional mental health-related variables and an age variable were included in the model. In addition, to protect against potential coverage and nonresponse error, alternatives for the weights were applied to the clinical sample data for the model development.
The next sections describe the instruments and items used to measure the variables employed in the 2012 model. Specifically, sections include descriptions of the instrument used to measure mental illness in the clinical interviews, followed by the scales and items in the main NSDUH interviews used as predictor variables in the model (i.e., the Kessler-6 [K6] and WHODAS total scores, age, MDE, and suicidal thoughts); see Sections 3.4.9 and 3.4.13 for more information on the measurement of MDE and serious thoughts of suicide among adults, respectively. Next, procedures for the MHSS clinical interview sampling and weighting and for developing the 2012 model are described. Section 3.4.8.9 discusses SEs for the mental illness estimates based on the 2012 model. Remaining sections discuss miscellaneous issues for the mental illness variables.
3.4.8.3 Clinical Measurement of Mental Illness
Mental illness was measured in the MHSS clinical interviews using an adapted version of the Structured Clinical Interview for the DSM-IV-TR Axis I Disorders, Research Version, Non-patient Edition (SCID-I/NP) (First et al., 2002) and was differentiated by the level of functional impairment based on the Global Assessment of Functioning (GAF) scale (Endicott et al., 1976).71 Past year disorders assessed through the SCID included mood disorders (e.g., MDE, manic episode), anxiety disorders (e.g., panic disorder, generalized anxiety disorder, posttraumatic stress disorder), eating disorders (e.g., anorexia nervosa), intermittent explosive disorder, and adjustment disorder. In addition, the presence of psychotic symptoms was assessed. SUDs were also assessed, although these disorders were not used to produce estimates of mental illness.
- Respondents were classified as having any mental illness (AMI) if they were determined to have any of the mental disorders assessed in the SCID (not including SUDs), regardless of the level of functional impairment.
- Respondents were classified as having low (mild) mental illness if they had any of the mental disorders assessed in the SCID (not including SUDs), but these disorders resulted in no more than mild impairment, based on GAF scores of greater than 59.
- Respondents were classified as having moderate mental illness if they had any of the mental disorders assessed in the SCID (not including SUDs), and these disorders resulted in moderate impairment, based on GAF scores of 51 to 59.
- Respondents were classified as having serious mental illness (SMI) if they had any of the mental disorders assessed in the SCID (not including SUDs), and these disorders resulted in substantial impairment in carrying out major life activities, based on GAF scores of 50 or below. The SMI diagnosis was used as the response variable in both the 2008 and 2012 prediction models.
The SCID and the GAF in combination were considered to be the “gold standard” for measuring mental illness.
3.4.8.4 K6
The K6 in the main NSDUH interview consists of two sets of six questions in the mental health section for adult respondents. These questions ask adult respondents how frequently they experienced symptoms of psychological distress during two different time periods: (1) during the past 30 days, and (2) if applicable, the 1 month in the past year when they were at their worst emotionally. Respondents were asked about the second time period only if they indicated there was a month in the past 12 months when they felt more depressed, anxious, or emotionally stressed than they felt during the past 30 days. All questions had the same response categories:
- 1All of the time
2Most of the time
3Some of the time
4A little of the time
5None of the time
Don’t know/Refused
The six questions in the K6 scale for the past month are as follows:
- NERVE30
- During the past 30 days, how often did you feel nervous?
- HOPE30
- During the past 30 days, how often did you feel hopeless?
- FIDG30
- During the past 30 days, how often did you feel restless or fidgety?
- NOCHR30
- During the past 30 days, how often did you feel so sad or depressed that nothing could cheer you up?
- EFFORT30
- During the past 30 days, how often did you feel that everything was an effort?
- DOWN30
- During the past 30 days, how often did you feel down on yourself, no good or worthless?
In the 2022 NSDUH, all adult respondents with item nonresponse for these psychological distress items had their scores statistically imputed using the modPMN methodology (see Section 2.3.3). Thus, there were no missing values in the 2022 survey for measures of psychological distress (based on the K6 distress scale) used in the mental illness prediction model. Imputation also mitigated potential effects of nonresponse from adult respondents who broke off the interview before or during the mental health section of the interview.
To create the score corresponding to the past month, the imputation-revised values for the six items for the past 30 days (NERVE30, HOPE30, FIDG30, NOCHR30, EFFORT30, and DOWN30) on the K6 scale were recoded from 0 to 4 so that “all of the time” was coded as 4, “most of the time” as 3, “some of the time” as 2, “a little of the time” as 1, and “none of the time” as 0. Summing across the transformed values for responses in these six items resulted in a score with a range from 0 to 24.
If respondents were asked about a month in the past 12 months when they felt more depressed, anxious, or emotionally stressed than they felt during the past 30 days, they were asked comparable K6 items for that particular month in the past 12 months. The imputation and scoring procedures for these K6 items corresponding to the worst month in the past 12 months were the same as those described previously for the past 30 days.
The maximum of the two K6 total scores for the past 30 days or past 12 months was used for MHSS analysis purposes and in the adult respondents’ final data. If respondents were asked K6 items for both the past 30 days and past 12 months, two K6 total scores were calculated for these periods, and the maximum of the two scores was identified. For respondents who were asked the K6 questions only for the past 30 days, the maximum score was the score for the past 30 days.
An alternative K6 total score was also created from the maximum K6 score. Maximum K6 scores less than 8 were recoded as 0. A maximum score of 8 was recoded as 1, a maximum score of 9 was recoded as 2, and so on, until a score of 24 was recoded as 17 in the alternative score variable. This alternative K6 score was used in the 2012 SMI prediction model because SMI prevalence typically was extremely low for respondents with past year K6 scores of less than 8. The prevalence started increasing only when scores were 8 or greater.
3.4.8.5 WHODAS
The WHODAS was modified for use in a general population survey such as NSDUH by making minor changes to question wording and reducing its length (Novak, 2007). A subset of eight items was found to capture the information represented in the full 16-item scale with no significant loss of information.
Respondents were asked the WHODAS questions if they reported having at least some symptoms of psychological distress in the past 30 days or in their worst period in the past 12 months at least a little of the time (i.e., their answers yielded a K6 score greater than zero). Approximately 450 respondents in 2022 were administered the WHODAS but did not provide a response to every applicable question. As for the K6 variables, statistical imputation was used to replace missing data in the WHODAS variables (see Section 2.3.3). Thus, there were no missing values in the 2022 survey for measures of adult SMI and other mental illness measures created from a model using the WHODAS scores. Imputation also mitigated potential effects of nonresponse from adult respondents who broke off the interview before or during the mental health section of the interview.
Respondents were not asked the WHODAS questions if their maximum reported K6 score was zero. Respondents had a maximum unimputed K6 score of zero if they reported that they had all six symptoms of psychological distress “none of the time” in the past 30 days and in their worst period in the past 12 months (if applicable). Respondents also could have a maximum unimputed K6 score of zero because of missing data with no reports of symptoms at least a little of the time. For example, respondents could have answered some of the K6 questions as “don’t know” or “refused” and the remainder as “none of the time.” All K6 questions had missing data if respondents answered all questions as “don’t know” or “refused” or if they broke off the interview before being asked the K6 questions.
Approximately 13,000 adult respondents were skipped out of the WHODAS questions in 2022 because the sum of all reported K6 item scores was zero. However, approximately 1,100 of these respondents were missing at least one of the K6 items and were subsequently imputed to have a maximum K6 score greater than zero. For these respondents whose K6 score was imputed to a value greater than zero, missing data in the skipped WHODAS variables also were statistically imputed.
The imputation-revised values for the eight WHODAS items included in the main NSDUH mental health section of the interview were coded on a 0 to 3 scale, with imputed responses of “no difficulty” coded as 0; “mild difficulty” coded as 1; “moderate difficulty” coded as 2; and “severe difficulty” coded as 3. Some items had an additional category for respondents who did not engage in a particular activity (e.g., they did not leave the house on their own). Respondents who reported they did not engage in an activity were asked a follow-up question to determine whether they did not do so because of emotions, nerves, or mental health. Respondents with an imputation-revised value of “yes” to these follow-up questions were subsequently assigned to the “severe difficulty” category. Respondents with an imputation-revised value of “no” to these follow-up questions were assigned to the “no difficulty” category. Summing across these codes for the eight responses resulted in a total score with a range from 0 to 24.
An alternative WHODAS total score was used in the 2012 SMI prediction model starting from the imputation-revised WHODAS items. Individual item scores of less than 2 were recoded as 0 and item scores of 2 to 3 were recoded as 1. The individual alternative item scores then were summed to yield a total alternative score ranging from 0 to 8. Creation of an alternative version of the WHODAS score assumed a dichotomous measure dividing respondents into two groups (i.e., severely impaired vs. less severely impaired) would fit better than a linear continuous measure in models predicting SMI.
3.4.8.6 Suicidal Thoughts, MDE, and Age
In addition to the K6 and WHODAS scales, the 2012 model included the following measures as predictors of SMI: (1) serious thoughts of suicide in the past year, (2) having a past year MDE, and (3) adjusted age, as defined later in this section. The first two variables were added to the model to decrease the error rate in the predictions (i.e., the sum of the false-negative and false-positive rates relative to the clinical interview results). A recoded age variable reduced the bias in estimates for particular age groups, especially for 18- to 25-year-olds.
All adult respondents in NSDUH were asked the following question in the mental health section about serious thoughts of suicide: “At any time in the past 12 months, that is from [DATEFILL] up to and including today, did you seriously think about trying to kill yourself?”72 Definitions for MDE in the lifetime and past year periods were based on questions in the adult depression section and are discussed in Section 3.4.9. For the modeling, starting in 2021, missing data from adult respondents for whether they had serious thoughts of suicide or for having a past year MDE were statistically imputed. See the 2021 Editing and Imputation Report (CBHSQ, 2023b) for specifics on how the adult suicidality and MDE measures were imputed for the 2021 NSDUH; these procedures did not change for the 2022 NSDUH.
For respondents aged 18 to 30, an adjusted age was created by subtracting 18 from the respondent’s current age, resulting in values ranging from 0 to 12. For a respondent aged 18, for example, the adjusted age was 0 (i.e., 18 minus 18), and for a respondent aged 30, the adjusted age was 12 (i.e., 30 minus 18). For respondents aged 31 or older, the adjusted age was assigned a value of 12.
3.4.8.7 Sampling and Weighting of the 2012 Model
The target annual respondent sample sizes for the MHSS clinical interviews were 1,500 in 2008 (750 of which received the WHODAS and were used in developing the 2008 model), 500 in 2009 and 2010, and 1,500 in 2011 and 2012. Respondent sample sizes were roughly equal across quarters.
A stratified Bernoulli selection process was used in which all eligible NSDUH respondents were given an independent probability of selection into the MHSS based on their strata. In 2008 and the first two quarters in 2009, stratification was based on K6 scores in an attempt to minimize the variance of the estimate for SMI prevalence. In the last two quarters in 2009, stratification attempted to minimize the variance of the AMI prevalence estimate rather than the variance of the SMI estimate. This change reduced the probability a respondent with an extremely large weight would be selected. Starting from 2010, stratification for the MHSS sample incorporated information on functional impairment levels (WHODAS scores) and age in addition to K6 scores. Younger age groups were undersampled for the MHSS clinical sample to reverse the impact of the oversampling of young adults aged 18 to 25 in the main survey (see Section A.1 in Appendix A in the 2012 NSDUH Mental Health Findings report [CBHSQ, 2013b]). This undersampling of younger age groups resulted in a more equally allocated clinical sample by age. More details about the sample design for the MHSS clinical study can be found in the 2012 National Survey on Drug Use and Health Methodological Resource Book, Section 2: Sample Design Report (CBHSQ, 2013a).
Special clinical sample analysis weights were created. Each was the product of the following seven weight components: (1) the NSDUH analysis weight; (2) a coverage adjustment for Hispanics completing the main NSDUH interview in English to account for Hispanics who completed it in Spanish and thus were not eligible for the English-language clinical follow-up interview; (3) the inverse of the selection probability for clinical follow-up; (4) a refusal adjustment to account for NSDUH respondents who were selected for the MHSS but declined to be contacted for the clinical interview; (5) another nonresponse adjustment to account for MHSS nonresponse among NSDUH respondents who had originally agreed to be recontacted for the clinical interview but did not complete the interview; (6) poststratification adjustments to reduce the variance of the resulting estimates by matching the weighted main NSDUH interview sample by age, gender, race/ethnicity, alternative K6 score, alternative WHODAS score, having had serious thoughts of suicide in the past year, and having had an MDE;73 and (7) a yearly scaling factor. The first six weight components were created separately for each year.
The 2012 model was fit by assuming the relationship between SMI and the covariates of the model stayed the same from 2008 through 2012. Because the sample size, sampling allocation, and weight adjustments for the MHSS clinical samples differed across years, gains in statistical efficiency were realized by scaling the weights in each year using the following scaling factors: 12 percent for 2008, 4 percent for 2009, 14 percent for 2010, 35 percent for 2011, and 35 percent for 2012. The scaling factors were determined based on the relative sizes of the estimated variances for estimates of SMI, AMI, and past year MDE made directly from SCID diagnoses.74
3.4.8.8 2012 SMI Model
The 2012 SMI prediction model was fit with data from 4,912 WHODAS MHSS respondents from 2008 through 2012. The response variable Y equaled 1 when an SMI diagnosis was positive based on the clinical interview; otherwise, Y was 0. Letting X be a vector of the characteristics attached to a NSDUH respondent and letting the probability this respondent had SMI be 
, the 2012 SMI prediction model was as follows:
where 
refers to the estimate of the SMI response probability 
.
The beta values for equation (1) are displayed in Table 3.7. The following covariates in equation (1) came from the main NSDUH interview data and were statistically imputed:
= Alternative Past Year K6 Score: Past year maximum K6 score of less than 8 recoded as 0; past year maximum K6 score of 8 to 24 recoded as 1 to 17.
= Alternative WHODAS Score: WHODAS item score of less than 2 recoded as 0; WHODAS item score of 2 to 3 recoded as 1, then summed for a score ranging from 0 to 8.
= Serious Thoughts of Suicide in the Past Year: Coded as 1 if “yes”; coded as 0 otherwise.
= Past Year MDE: Coded as 1 if the criteria for past year MDE were met (see Section 3.4.9);75 coded as 0 otherwise.
= Adjusted Age: Coded as age minus 18 if aged 18 to 30; coded as 12 otherwise.
A cut point probability 
was determined so that if 
for a particular respondent, then the respondent was predicted to be SMI positive; otherwise, the respondent was predicted to be SMI negative. The cut point (0.260573529) was chosen so that the weighted numbers of false positives and false negatives in the MHSS dataset were as close to equal as possible. The predicted SMI status for all adult NSDUH respondents was used to compute prevalence estimates of SMI.
A second cut point probability (0.0192519810) was determined so that any respondent with an SMI probability greater than or equal to the cut point was predicted to be positive for AMI, and the remainder were predicted to be negative for AMI. The second cut point was chosen so that the weighted numbers of AMI false positives and false negatives were as close to equal as possible.
The 2022 national reports and tables show estimates for AMI, SMI, and “AMI excluding SMI.” Adults with AMI excluding SMI currently or at any time in the past year have had a diagnosable mental, behavioral, or emotional disorder resulting in less than substantial impairment in carrying out major life activities (see Appendix A in this report).
3.4.8.9 Standard Errors for Mental Illness Estimates
For the national reports and tables, SEs for mental illness estimates (SMI, AMI, and AMI excluding SMI) were computed using the NSDUH dichotomous variable values without taking into account any variance introduced through using a model based on the clinical subsample data. This ignores the added error resulting from fitting the 2012 SMI model, which can be very large (see the 2012 National Survey on Drug Use and Health: Methodological Resource Book [Section 16a, 2012 Mental Health Surveillance Study: Design and Estimation Report]; CBHSQ, 2014a). These conditional SEs (conditional on the model predictions being correct) are useful when making comparisons across years and across subpopulations (except those involved in modeling) within years because the errors due to model fitting are nearly the same across the estimates being compared.
3.4.8.10 Limitations in Using the Mental Illness Variables in Analyses
There are many advantages to using the cut point methodology described in this section to predict the SMI and AMI status for every adult responding to the NSDUH main survey interview. For some analyses, however, these predicted values should not be used. In particular, these predicted values should not be employed in analyses using the mental illness variables in conjunction with variables used or closely related to variables used in the prediction model. Because variables used in the prediction models would be expected to be correlated with the SMI or AMI probabilities, SMI or AMI should not be estimated for the following groups of adults: among people with past year or lifetime MDE; among people with past year suicidal thoughts, suicide plans, or suicide attempts; or among people with particular K6 or WHODAS scores. For details, see the Estimating Mental Illness among Adults in the United States: Revisions to the 2008 Estimation Procedures report (CBHSQ, 2015).
3.4.9 Major Depressive Episode (Depression)
Two sections related to MDE were included in the 2022 questionnaire: an adult depression and an adolescent depression section. These sections were originally derived from DSM-IV criteria for MDE and remained applicable to the more recent DSM-5 criteria. Consistent with the DSM-5 criteria, NSDUH does not exclude MDEs occurring exclusively in the context of bereavement. In addition, no exclusions were made for MDEs caused by medication, alcohol, illicit drugs, or any medical illness.
Questions on depression permit estimates to be calculated for the occurrence of MDE in the population and receipt of treatment for MDE. Separate sections were administered to adults aged 18 or older and youths aged 12 to 17. The adult questions were adapted from the depression section of the National Comorbidity Survey Replication (NCS-R), and the questions for youths were adapted from the depression section of the National Comorbidity Survey Replication Adolescent Supplement (NCS-A).76 To make the sections developmentally appropriate for youths, there are minor wording differences in a few questions between the adult and youth sections. Revisions to the questions in both sections were made primarily to reduce their length and to modify the NCS questions, which were interviewer-administered, for self-administration in NSDUH.
According to DSM-5, people are classified as having had an MDE77 in their lifetime if they had at least five or more of nine symptoms nearly every day (except where noted) in the same 2-week period, where at least one of the symptoms is a depressed mood or loss of interest or pleasure in daily activities. These symptoms are as follows:
- depressed mood most of the day;
- markedly diminished interest or pleasure in all or almost all activities most of the day;
- significant weight loss when not sick or dieting, or weight gain when not pregnant or growing,78 or decrease or increase in appetite;
- insomnia or hypersomnia;
- psychomotor agitation or retardation at a level observable by others;
- fatigue or loss of energy;
- feelings of worthlessness or excessive or inappropriate guilt;
- diminished ability to think or concentrate or indecisiveness; and
- recurrent thoughts of death or suicidality (i.e., recurrent suicidal ideation without a specific plan, making a specific plan, or making an attempt).
Unlike the other symptoms, recurrent thoughts of death or suicidality did not need to have occurred nearly every day.
Respondents who have had an MDE in their lifetime are asked if, during the past 12 months, they had a period of depression lasting 2 weeks or longer while also having some of the other symptoms mentioned previously for the lifetime period. Respondents reporting experiences consistent with their having had an MDE in the past year are asked questions from the SDS to measure the level of functional impairment in major life activities reported to be caused by the MDE in the past 12 months (Leon et al., 1997).
NSDUH measures the nine symptoms associated with MDE as defined in DSM-5 with the following questions. The questions shown are taken from the adult depression section of the 2022 NSDUH questionnaire. A few of the questions in the youth section were modified slightly to use wording more appropriate for youths aged 12 to 17.
- Depressed mood most of the day
The following questions refer to the worst or most recent period of time when the respondent experienced any or all of the following: sadness, discouragement, or lack of interest in most things.
During that [worst/most recent] period of time …
- … did you feel sad, empty, or depressed most of the day nearly every day?
- … did you feel discouraged about how things were going in your life most of the day nearly every day?
- Markedly diminished interest or pleasure in all or almost all activities most of the day
- … did you lose interest in almost all things like work and hobbies and things you like to do for fun?
- … did you lose the ability to take pleasure in having good things happen to you, like winning something or being praised or complimented?
- Weight
In answering the next questions, think about the [worst/most recent] period of time.
- Did you have a much smaller appetite than usual nearly every day during that time?
- Did you have a much larger appetite than usual nearly every day?
- Did you gain weight without trying to during that [worst/most recent] period of time?
- … because you were growing?
- … because you were pregnant?
- How many pounds did you gain?
- Did you lose weight without trying to?
- … because you were sick or on a diet?
- How many pounds did you lose?
- Insomnia or hypersomnia
- Did you have a lot more trouble than usual falling asleep, staying asleep, or waking too early nearly every night during that [worst/most recent] period of time?
- During that [worst/most recent] period of time, did you sleep a lot more than usual nearly every night?
- Psychomotor agitation or retardation
- Did you talk or move more slowly than is normal for you nearly every day?
- Did anyone else notice that you were talking or moving slowly?
- Were you so restless or jittery nearly every day that you paced up and down or couldn't sit still?
- Did anyone else notice that you were restless?
- Fatigue or loss of energy
- During that [worst/most recent] period of time, did you feel tired or low in energy nearly every day, even when you had not been working very hard?
- Feelings of worthlessness
- Did you feel that you were not as good as other people nearly every day?
- Did you feel totally worthless nearly every day?
- Diminished ability to think or concentrate or indecisiveness
- During that [worst/most recent] time period, did your thoughts come much more slowly than usual or seem confused nearly every day?
- Did you have a lot more trouble concentrating than usual nearly every day?
- Were you unable to make decisions about things you ordinarily have no trouble deciding about?
- Recurrent thoughts of death or recurrent suicidal ideation
- Did you often think about death, either your own, someone else’s, or death in general?
- During that period, did you ever think it would be better if you were dead?
- Did you think about committing suicide?
Starting in 2021, missing data were statistically imputed in variables for whether adult respondents had an MDE in their lifetime and whether adult respondents had an MDE in the past 12 months. MDE variables were not statistically imputed for youths aged 12 to 17. Respondents aged 12 to 17 who had missing data for whether they had an MDE in the past 12 months were excluded from the analyses to produce published estimates for the 2022 NSDUH. See Section 3.3.2 for a discussion of the potential bias in estimates because of missing data.
NSDUH also collects data on impairment using the SDS, which is a measure of impairment due to mental health issues in four major life activities or role domains. These four domains are defined separately for adults aged 18 or older and youths aged 12 to 17 to reflect the different roles associated with the two age groups. See Section 3.4.9.1 for details about the questions for adults and Section 3.4.9.2 for details about the questions for youths. Each role domain consists of four questions, and each item uses an 11-point scale ranging from 0 (no interference for adults and no problems for adolescents) to 10 (very severe interference for adults and very severe problems for adolescents). The impairment score is defined as the single highest severity level of role impairment across the four SDS role domains. Ratings greater than or equal to 7 on the scale were considered severe impairment.
In addition to past year MDE, NSDUH shows estimates for past year MDE with severe impairment. Estimates for severe impairment are calculated separately for youths and adults because the four domains are slightly different for the two groups. In 2022, missing data for MDE with severe impairment were statistically imputed for adults but not for youths aged 12 to 17. Respondents aged 12 to 17 who had missing data for impairment were excluded from the analyses to produce published estimates for MDE with severe impairment in the 2022 NSDUH. See Section 3.3.2 for a discussion of the potential bias in estimates because of missing data.
Because variables for lifetime and past year MDE among adults and MDE with severe impairment in the past year were imputed beginning in 2022, the main analysis weight described in Section 2.3.4 was used to produce 2022 estimates. For youths aged 12 to 17, the main person-level analysis weight was used to produce estimates of MDE and MDE with severe impairment in the past year because the number of break-offs among youths was minimal.
3.4.9.1 Adult Depression Section: Functional Impairment
The questions pertaining to the four domains of functional impairment for adults aged 18 or older are listed below. The scale is shown below for the first domain but applies to all four domains.
- ASDSHOME
- Think about the time in the past 12 months when these problems with your mood were most severe.
- Using the 0 to 10 scale shown below, where 0 means no interference and 10 means very severe interference, select the number that describes how much these problems interfered with your ability to do each of the following activities during that period. You can use any number between 0 and 10 to answer.
- How much did your [depression symptoms] interfere with your ability to do home management tasks, like cleaning, shopping, and working around the house, apartment, or yard?
- ASDSWORK
- During that time in the past 12 months when your [depression symptoms] were most severe, how much did this interfere with your ability to work?
- ASDSREL
- How much did your [depression symptoms] interfere with your ability to form and maintain close relationships with other people during that period of time?
- ASDSSOC
- How much did [depression symptoms] interfere with your ability to have a social life during that period of time?
3.4.9.2 Youth Depression Section: Functional Impairment
The questions pertaining to the four domains of functional impairment for adolescents aged 12 to 17 are listed below. The scale is also shown below for the first domain but applies to all four domains.
- YSDSHOME
- Think about the time in the past 12 months when these problems with your mood were the worst.
- Using the 0 to 10 scale shown below, where 0 means no problems and 10 means very severe problems, select the number that describes how much your [depression symptoms] caused problems with your ability to do each of the following activities during that time. You can use any number between 0 and 10 to answer.
- How much did your [depression symptoms] cause problems with your chores at home?
- YSDSWORK
- During that time in the past 12 months when your [depression symptoms] were worst, how much did this cause problems with your ability to do well at school or work?
- YSDSREL
- How much did your [depression symptoms] cause problems with your ability to get along with your family during that time?
- YSDSSOC
- How much did your [depression symptoms] cause problems with your ability to have a social life during that time?
3.4.10 Perceived Recovery
Adult respondents aged 18 or older in 2022 were asked up to four questions about perceived recovery from substance use problems or mental health issues. These questions were in the emerging issues section of the questionnaire. Respondents first were asked whether they thought that they ever have had a problem with their own drug or alcohol use. If adult respondents answered “yes,” they were asked whether they considered themselves to be in recovery or to have recovered from their own problem with drug or alcohol use. These first two questions on recovery from a substance use problem were followed by a set of two similar questions asking adult respondents whether they have ever had a problem with their own mental health and, if so, whether they considered themselves to be in recovery or to have recovered from their own mental health problem.
These perceived recovery estimates were included in national reports and tables for the 2022 NSDUH. Estimates of perceived recovery were reported in the 2022 Detailed Tables (CBHSQ, 2023k) among (1) adults who reported ever having a substance use problem or mental health issue and (2) all adults, regardless of whether they perceived themselves ever to have had a problem. To generate estimates among the total adult population, adults who reported not having a problem were classified as not being in recovery or having recovered from a problem. Respondents were excluded from substance use or mental health analyses if they had unknown information for whether they ever had a substance use problem or mental health issue, respectively. Respondents were also excluded from analyses if they had unknown information for whether they perceived themselves to be in recovery or to have recovered from their respective problem (e.g., if respondents reported ever having had a substance use problem but did not know or refused to report whether they perceived themselves to be in recovery or to have recovered from their substance use problem). For a discussion on how procedures for handling missing data may bias estimates, see Section 3.3.2.
Consistent with the discussion in Section 3.3.4, data users are reminded that these estimates are based on self-reports. Specifically, these estimates reflect adults’ perceptions of whether they had substance use or mental health problems, but not necessarily the clinical assessments of medical or mental health professionals or the internal consistency of respondents’ answers. In addition, data on adults’ perceptions of whether they had a problem with their substance use or mental health and whether they perceived themselves to have recovered or to be in recovery from these problems were not edited relative to data in other sections of the interview for substance use, SUDs, substance use treatment, mental health issues, or the receipt of mental health treatment (see Section 2.3.2). Therefore, perceptions may seem to be inconsistent with substance use and mental health data from earlier sections of the interview.
The emerging issues section occurred in the 2022 NSDUH interview after the mental health and adult depression sections. Also, the perceived recovery variables were not imputed for 2022. Therefore, 2022 estimates for perceived recovery were created using the break-off analysis weight described in Section 2.3.4.
3.4.11 Nicotine Vaping
Starting with the 2022 NSDUH, questions on vaping of nicotine were included in the nicotine section of the questionnaire instead of in the later emerging issues section. The nicotine section was one of the earliest sections presented to respondents in 2022. Respondents aged 12 or older were asked whether they ever vaped nicotine with an e-cigarette or other vaping device, including devices that may also be called vapes, vape pens, or mods. Respondents were asked to consider any device that heats a liquid containing nicotine into a vapor.
Follow-up questions if respondents reported nicotine vaping in their lifetime were patterned after similar questions for tobacco products and included questions for the first time respondents vaped nicotine (age at first use, and if applicable, the year and month of first use), when respondents last vaped nicotine, and the number of days that respondents vaped nicotine in the past 30 days, if they last vaped nicotine in that period. Associated measures for initiation of nicotine vaping, most recent nicotine vaping, and the frequency of nicotine vaping in the past 30 days were all statistically imputed for 2022 and do not contain missing data.
As noted at the beginning of this section, questions for nicotine vaping changed between 2021 and 2022. These changes affected the comparability of estimates of nicotine vaping between 2021 and 2022. Therefore, estimates for nicotine vaping in 2021 are not shown in the 2022 NSDUH data products. The remainder of this section discusses reasons for the nicotine vaping estimates not being comparable between these 2 years.
In 2021, questions about nicotine vaping in the emerging issues section of the questionnaire first required respondents to answer a “gate” question for whether they vaped any substance in their lifetime; respondents were not asked whether they vaped nicotine in their lifetime unless they reported that they ever vaped any substance. Respondents in 2021 who reported that they vaped nicotine in their lifetime were asked to report only the last time when they did so. In 2022, all respondents were asked whether they ever vaped nicotine without first being asked a gate question. The editing and imputation procedures in 2022 for assigning the period when respondents last vaped nicotine also used data for the first time that respondents vaped nicotine (see Sections 2.3.2 and 2.3.3); these data were not available in 2021 to create the final recency variables for nicotine vaping. For example, an inconsistency occurred if respondents in 2022 reported that they first vaped nicotine at their current age (suggesting use in the past 12 months), but they last vaped it more than 12 months ago. This inconsistency was resolved through statistical procedures. The resolution could result in final assignment of nicotine vaping to have occurred in the past 12 months instead of what respondents originally reported. In 2021, if respondents reported that they last vaped nicotine more than 12 months ago, then that was what their final data would have indicated.
As noted previously, questions about nicotine vaping were also asked much earlier in the 2022 NSDUH interview than they were for 2021. Respondents in 2021 could discontinue the interview (i.e., “break-offs”) before they reached the emerging issues section, especially for adult web respondents, and would have missing data for nicotine vaping, even though they were considered to have provided usable information on their substance use (see Section 2.3.1). Along with the absence of a gate question for lifetime nicotine vaping, a consequence of the earlier placement of the nicotine vaping questions in the 2022 interview was that the unweighted percentage of respondents with missing data for lifetime nicotine vaping was lower in 2022 (less than 0.05 percent) than in 2021 (3.44 percent).
3.4.12 Central Nervous System Stimulant Misuse
CNS stimulants are a group of drugs that include cocaine, methamphetamine, and prescription stimulants. These drugs act in similar ways to stimulate the brain. They produce stimulant effects, such as increased alertness, wakefulness, or energy. They can also produce physical side effects of rapid or irregular heartbeat or increased blood pressure and body temperature (NIDA, 2018b, 2019, 2021a).
An aggregate measure for CNS stimulant misuse was created for inclusion of estimates in the 2022 Detailed Tables and 2022 Key Substance Use and Mental Health Indicators report (CBHSQ, 2023k, 2023j). Because this aggregate measure includes the misuse of prescription stimulants in addition to the use of cocaine or methamphetamine, it was defined as CNS stimulant misuse.
CNS stimulant misuse data for 2022 were available for the past year and past month, and data for any use of CNS stimulants were available for the past year. Because of potential measurement issues for the lifetime misuse of prescription drugs (see Section 4.4.3), lifetime estimates for CNS stimulant misuse and any use were not presented in national reports and tables for the 2022 NSDUH. Measures for CNS stimulant misuse in the past year or past month periods and any use of CNS stimulants in the past year were created according to the most recent time when respondents used or misused these substances. Because the measures were imputed for cocaine use, methamphetamine use, and prescription stimulant misuse for the past year and past month, the aggregate measures for CNS stimulant misuse in those periods had no missing data for 2022. Similarly, the aggregate measure for any CNS stimulant use in the past year had no missing data because the measure for any past year use of prescription stimulants was imputed. Section 3.4.4.2 also describes the creation of measures for CNS stimulant use disorder.
3.4.13 Suicidal Thoughts and Behavior
The 2022 NSDUH included separate sets of questions asking adults aged 18 or older and adolescents aged 12 to 17 whether they had serious thoughts of suicide, made a suicide plan, or attempted suicide in the past 12 months. All adult and adolescent respondents in 2022 were asked whether they made a suicide plan or attempted suicide regardless of whether they reported that they had serious thoughts of suicide in the past 12 months. Respondents who reported that they made a suicide attempt were asked whether they received medical attention or stayed overnight in the hospital because of their suicide attempt. The 2022 Detailed Tables and 2022 Key Substance Use and Mental Health Indicators report present estimates for suicidal thoughts and behavior among adults and adolescents (CBHSQ, 2023k, 2023j).
In addition, follow-up questions were asked of adults or youths reporting suicidal thoughts or behavior. These follow-up questions asked whether the suicidal thoughts or behavior were because of the COVID-19 pandemic. For example, if respondents reported that they had serious thoughts of suicide in the past year, they were asked whether this was because of the COVID-19 pandemic. Estimates for suicidal thoughts and behavior because of the COVID-19 pandemic are presented in the 2022 Detailed Tables but not in the 2022 Key Substance Use and Mental Health Indicators report (CBHSQ, 2023k, 2023j).
3.4.13.1 Suicidal Thoughts and Behavior among Adults
The mental health section of the NSDUH questionnaire included questions about suicidal thoughts and behavior among adults. Beginning in 2021, the variables for suicidal thoughts and behavior among adults were statistically imputed, so these variables had no missing data for 2022. Estimates for suicidal thoughts and behavior among adults in the past year were created for 2022 using the standard analysis weight described in Section 2.3.4.
3.4.13.2 Suicidal Thoughts and Behavior among Adolescents
Starting with the 2022 NSDUH, questions about adolescents’ suicidal thoughts and behaviors in the past 12 months were included in the youth experiences section of the questionnaire instead of in the youth mental health service utilization section, which was removed from the 2022 NSDUH questionnaire. However, the wording of the adolescent suicidal thoughts and behavior questions did not change for 2022. Unlike the questions for adults, the questions about suicidal thoughts and behavior among adolescents included response choices for “I'm not sure” and “I don’t want to answer,” in addition to standard response choices for “yes” and “no.” Adolescent respondents also could choose these response choices for “I'm not sure” and “I don’t want to answer” instead of using function keys (as is the practice elsewhere in the interview) for answers of “don’t know” or “refused,” respectively.
Estimates for suicidal thoughts and behavior among adolescents in national reports and tables for 2022 included estimates for “I'm not sure,” and “I don’t want to answer,” in addition to estimates for “yes” and “no.” Responses of “don’t know” were grouped with “I'm not sure,” and refusals were grouped with “I don’t want to answer.” Thus, measures for suicidal thoughts and behavior among adolescents were not statistically imputed for 2022, but adolescent respondents who did not know or refused to report whether they had suicidal thoughts or behavior in the past year were not excluded from analyses. However, adolescents who broke off the interview before reaching these questions were excluded from the analyses.
The 2022 estimates for suicidal thoughts and behavior among adolescents were created using the main analysis weights, with no adjustment because of break-offs. As discussed in Section 2.3.4, review of the 2022 NSDUH data indicated that a small number of adolescents aged 12 to 17 broke off the interview before they reached the youth experiences section where the questions were located for suicidal thoughts and behavior among adolescents.
3.4.13.3 Suicidal Thoughts and Behavior Because of the COVID-19 Pandemic
As previously noted, adults and adolescents who reported suicidal thoughts and behavior in the past year were asked whether these suicidal thoughts and behaviors were because of the COVID-19 pandemic. Variables for suicidal thoughts and behavior among adults because of the COVID-19 pandemic were imputed for 2022. Therefore, estimates for suicidal thoughts and behavior among adults because of the COVID-19 pandemic were created using the main analysis weight described in Section 2.3.4. Estimates for suicidal thoughts and behavior among adolescents because of COVID-19 were not imputed in 2022. Nevertheless, these estimates for adolescents were created using the main analysis weight because a small number of adolescents aged 12 to 17 broke off the interview before they reached the COVID-19 section of the interview.
3.4.14 Perceived Effects of the COVID-19 Pandemic
Because of widespread interest in the effects of the COVID-19 pandemic on substance use and mental health, questions were included in the 2022 NSDUH that asked respondents how the pandemic affected these and related aspects of their lives. Questions were included on how the COVID-19 pandemic affected respondents’ mental health, their substance use, their financial situation, their housing, and access to mental health treatment, substance use treatment, or medical care. Appendix A includes definitions for specific measures related to the perceived effects of the COVID-19 pandemic. The 2022 Detailed Tables presents estimates for these topics (CBHSQ, 2023k).
A limitation of 2022 NSDUH estimates for the perceived effects of the COVID-19 pandemic is the length of time that respondents would need to recall events in answering the questions. For questions on the perceived effects of the COVID-19 pandemic on mental health and substance use, respondents were asked to compare how their mental health and substance use have changed since the beginning of the COVID-19 pandemic. Questions that asked about things that happened “because of the COVID-19 pandemic” also required respondents to recall events over the entire period since the start of the COVID-19 pandemic. If respondents were interviewed after March 2022, they would need to think back over more than a 2-year period. Therefore, the accuracy of respondents’ self-reported perceptions of the effects of the COVID-19 pandemic could erode over time.
Questions about these topics related to the COVID-19 pandemic occurred in the 2022 NSDUH questionnaire after the mental health and adult depression sections. Also, the variables for perceived effects of the COVID-19 pandemic were not imputed for 2022. Therefore, estimates for perceived effects of the COVID-19 pandemic were created using the break-off analysis weight described in Section 2.3.4.
In addition, measures for suicidal thoughts and behavior among adults and adolescents because of the COVID-19 pandemic are discussed in Section 3.4.13. That section also discusses the handling of missing data for these suicidality measures.
3.4.15 Modes of Marijuana Use
The marijuana section of the 2022 NSDUH questionnaire included new questions to assess the variety of ways that people consume marijuana. Respondents who reported using marijuana in the past year or past month were asked to report whether they used marijuana in any of the following ways in these periods:
- smoking;
- vaping;79
- dabbing waxes, shatter, or concentrates;
- eating or drinking;
- putting drops, strips, lozenges, or sprays in their mouth or under their tongue;
- applying lotion, cream, or patches to their skin;
- taking pills; or
- using it in some other way.
Respondents could report that they used marijuana in more than one way in the past year or past month. For example, respondents could report that they smoked marijuana and vaped it in the past year.
3.4.15.1 Data Inconsistencies
The following patterns of inconsistent data occurred in 2022 for modes of marijuana use:
- Respondents did not report a particular mode of use (e.g., vaping) in the past year or past month but reported “some other way” and specified that mode for using marijuana in that period.
- Respondents did not report a particular mode of use in the past year but reported it as a mode of use for the past month.
- Respondents had missing data for the mode of marijuana use in the past month (i.e., they did not know or refused to report how they used it). However, they reported using marijuana in only one way in the past year.
If respondents did not report a particular mode of use in the past year or past month but specified it as some other way they used marijuana in that period, then these respondents were inferred to have used marijuana in that way. For example, if respondents did not report that they smoked marijuana in the past year but reported smoking as some other way they used marijuana in the past year (e.g., bongs, water pipes), then respondents were logically inferred to have smoked marijuana in the past year.
Percentages were low for respondents who reported past year or past month marijuana use and specified a particular mode as some other way they used marijuana without having originally reported that mode. The highest percentage was for eating marijuana in the past year: 0.4 percent of past year marijuana users specified eating as some other way they used marijuana (e.g., gummies) without having reported eating as a mode of marijuana use in the past year. Eating as some other way also was the most commonly reported inconsistency among past month marijuana users: 0.2 percent of past month users reported eating as some other way they used marijuana in the past month without having chosen eating or drinking from the list of available responses.
If respondents did not report a particular mode of use in the past year but reported it as a mode of use for the past month, then these respondents were inferred to have used marijuana in that specific way in the past year. Among respondents in 2022 who reported that they used marijuana in the past year, 1.3 percent reported vaping marijuana in the past month, but they did not report vaping as a way they used marijuana in the past year. For consuming marijuana in edible form, 1.1 percent of past year marijuana users reported eating or drinking marijuana in the past month but not the past year. For smoking and dabbing, 0.7 to 0.8 percent of respondents who were past year marijuana users reported using marijuana in these ways in the past month but not in the past year. About 0.3 percent or less of past year marijuana users reported using marijuana in other ways in the past month but not in the past year.
If respondents used marijuana in the past month, reported using marijuana in only one way in the past year, and had missing data for how they used it in the past month, then the one way they used marijuana in the past year would have been the only way they could have used it in the past month. About 0.1 to 0.3 percent of respondents who reported past month marijuana use but had missing data for their mode of use in the past month were logically inferred to have used marijuana in a particular way in that period based on the mode of use they reported for the past year. All variables for modes of marijuana use in the past year or past month took into account the logical assignments discussed in the preceding paragraphs.
In addition to mode questions in the marijuana section, questions about marijuana vaping remained in the emerging issues section of the 2022 questionnaire. However, data for marijuana vaping from the emerging issues section were not used in creating the published 2022 estimates for marijuana use in the lifetime, past year, or past month periods or for marijuana vaping in the past year or past month. Consequently, although inconsistencies may remain in the 2022 data for marijuana use and marijuana vaping between the marijuana and emerging issues sections, they will not be reflected in published estimates for 2022.
3.4.15.2 Measures for Modes of Marijuana Use
Variables for vaping marijuana in the past year and past month were statistically imputed for 2022 and do not have missing data. Variables for other modes of marijuana use were not imputed for 2022; missing data from these variables were excluded from analyses. See Section 3.3.2 for discussion of potential bias in estimates because of missing data.
Measures for the following specific modes of marijuana use in the past year or past month were created to produce estimates in the 2022 Detailed Tables (CBHSQ, 2023k): smoking; vaping; dabbing waxes, shatter, or concentrates; and eating or drinking. Modes for applying lotion, cream, or patches to the skin; putting drops, strips, lozenges, or sprays in the mouth or under the tongue; taking pills; and using marijuana in some other way were included in an aggregate “other” category. Respondents who reported using marijuana in some other way could specify other valid ways of using it.
Estimates are presented in the 2022 Detailed Tables for these modes of use among past year users and past month marijuana users for the total population aged 12 or older and by age group. The 2022 Detailed Tables also present estimates by additional demographic characteristics for marijuana vaping in the past month.
For the 2022 Key Substance Use and Mental Health Indicators report (CBHSQ, 2023j), estimates are presented for all eight specific modes of marijuana use described at the beginning of this section, including use in some other way. Estimates are presented in the report only for the past year period.
Estimates for marijuana vaping in the past year and past month were also created for 2021 from the marijuana vaping questions in the emerging issues section. However, these estimates for marijuana vaping in 2021 are not comparable with estimates for marijuana vaping in 2022 from the marijuana section because of the changes to the marijuana section that were described previously, including changes to how respondents could report that they vaped marijuana in the past year or past month. Also, estimates for vaping marijuana in the lifetime period were not created for 2022 because respondents in 2022 could report marijuana vaping only for the past year or past month.
, to reduce bias due to nonresponse, to poststratify to known population control totals, and to control for extreme weights when necessary.
, where 
as follows:
are prespecified centering constants, such that 
and 
. The variables 
, 
are unit-specific bounds, and 
is the column vector of model parameters corresponding to the covariates in vector 
denotes control totals that could be either nonrandom, as is generally the case with poststratification, or random, as is generally the case for nonresponse adjustment.
, defined as
. This general approach was used at several stages of the weight adjustment process, including (1) adjustment of household weights for nonresponse at the screener level, (2) poststratification of household weights to meet population controls for various household-level demographics by state,
. 
, is applied when the domain size estimates, 
, are among those forced to match their respective U.S. Census Bureau or American Community Survey (ACS) population estimates through the weight calibration process.
when 
, or
when 
, or
, with nominal sample size, 
, or
); a random number, 
, is independently chosen from the uniform distribution on the interval (0,1); and an imputed value of 1 is assigned for the element to be imputed with sorted index, 
, if and only if the predicted mean is greater than 
.
for these domains is considered negligible. Therefore, the same race domains are considered fixed for every year.
: 
for 
. The Taylor-series linearization of the numerator 
is 
, which approximately equals 
by Taylor-series linearization, which in turn equals 
. The same principles apply for the computational formula when 
, except that 
is replaced with 
.
. The W-shaped plot in