Revised Estimates of Mental Illness from the National Survey on Drug Use and Health

The Substance Abuse and Mental Health Services Administration (SAMHSA) has produced national and State estimates of mental illness among adults using its National Survey on Drug Use and Health (NSDUH) since 2008. These estimates are based on the Mental Health Surveillance Study (MHSS), initiated in 2008. The overarching goal of the MHSS was to provide accurate estimates of serious mental illness (SMI) among adults aged 18 or older at both the national and State levels, with a secondary goal of providing accurate estimates of any mental illness (AMI). This report discusses recent improvements to the methods used to produce these estimates and provides revised 2011 estimates. The improvements in the methodology were facilitated by a partnership between SAMHSA and the National Institute of Mental Health (NIMH). This partnership resulted in the collection of a large, nationally representative sample of clinical psychiatric interviews from a subsample of adult NSDUH respondents.¹ The data from this sample were used to refine the methods used to estimate SMI and AMI.^2,3

In accordance with the 1992 Alcohol, Drug Abuse, and Mental Health Administration (ADAMHA) Reorganization Act, SAMHSA developed conceptual and operational definitions of SMI and developed a methodology for producing estimates of SMI.^4,5,6 SAMHSA defined SMI as having at least one mental disorder, other than a developmental or substance-use disorder, in the past 12 months that resulted in serious impairment.⁷ SAMHSA also established a definition for AMI as having at least one mental disorder, other than a developmental or substance-use disorder, in the past 12 months, regardless of the level of impairment. Using these definitions, the MHSS was initiated in 2008 comprising the following:

• short scales measuring psychological distress and functional impairment included in the main NSDUH interview⁸;
• clinical psychiatric diagnostic data collected from a nationally representative subset of NSDUH respondents each year from 2008 to 2012⁹; and
• development of an estimation methodology for SMI and AMI among adults based on a weighted, statistical model.

The clinical diagnostic interview to assess the presence of mental disorders and functional impairment was administered to a subsample of NSDUH adult respondents within 2 to 4 weeks of the NSDUH interview. Each participant was assessed by a trained clinical interviewer (master's or doctoral-level clinician, counselor, or social worker) over the telephone. A subsample was collected because it was not feasible to administer a structured diagnostic clinical interview to assess mental illness to all NSDUH adult respondents (about 45,000 respondents each year). Administration of this type of interview to such a large number of NSDUH respondents would require a large staff of trained mental health clinicians and would require each respondent to participate in both the NSDUH interview and a clinical diagnostic interview.

A prediction model of SMI was developed from the clinical subsample of respondents. The model used the short scales of psychological distress and functional impairment in combination with the clinical diagnostic data. Specifically, the short scales were used as predictor variables of SMI diagnosis in a statistical model developed from the subsample. The model was then used to classify each of the NSDUH adult respondents as having SMI or AMI based on their responses to the distress and impairment scales. The prediction models were developed in two phases.

Phase One: Initial Estimation Method

The initial model was developed using the 2008 MHSS subsample of approximately 750 respondents. The psychological distress and functional impairment indicators were used as predictor variables of SMI diagnosis in a statistical model developed from the subsample. The model was then used to classify each of the NSDUH adult respondents as having SMI or AMI based on their responses to the distress and impairment scales.

Although clinical interview data were collected annually between 2008 and 2012,¹⁰ producing a new model each year based on the small annual clinical samples would have resulted in large changes in the estimates due to sampling error, making it impossible to detect trends in SMI and AMI over time. Furthermore, an evaluation of the 2008 model, using 2009 clinical interview data, found that the model could not be significantly improved with the addition of the 500 cases from the 2009 sample. Therefore, SAMHSA continued using the 2008 model to generate 2009 to 2011 estimates, to provide consistency for trend measurement. SAMHSA planned to reevaluate the model after more clinical data were accumulated from 2009 to 2012.^11,12

Phase Two: Improved Estimation Method

By the end of 2012, SAMHSA had accumulated a nationally representative sample of about 5,000 completed clinical assessments. This larger clinical sample enabled SAMHSA to evaluate the current methods for estimating mental illness to see if they could be improved. The evaluation of the methods included an assessment of the 2008 statistical model and the weights applied to the clinical interview sample used to develop the model.

The 2008 model included two predictors (i.e., short scales of psychological distress and impairment). The exploration of the model with the combined 2008 to 2012 sample indicated that additional predictors would improve the model. For example, incorporating age into the model corrected for the fact that the 2008 model substantially overestimated SMI and AMI among young adults.¹³ A new model was identified that improved the two-predictor 2008 model by adding three more predictors based on the respondent's age, past year thoughts of suicide, and past year depression.

The weights applied to the 2008 clinical sample also contributed to the overestimation of SMI and AMI. As part of the reevaluation, it was determined that respondents with mental health issues were more likely to respond to the clinical interview. This resulted in an overrepresentation of respondents with SMI and AMI in the sample and an overestimation of the prevalence of mental illness in the Nation. Therefore, SAMHSA developed new weights that included better adjustments for sample nonresponse because of the higher response rate among those with mental health issues in the clinical subsample and that also accounted for coverage error.¹⁴

Using combined 2008 to 2012 data, SAMHSA produced estimates using the initial 2008 method and the improved method, and compared these estimates with those from the clinical interview sample. This comparison showed that estimates based on the improved methods are more similar to the clinical data than estimates based on the 2008 methods (Table 1). Furthermore, estimates of SMI and AMI by age groups produced using the improved methods were closer to the estimates produced using the clinical data. Thus, revising the weights and adding new predictors to the model produced more accurate estimations of SMI and AMI.

To date, reports providing 2011 NSDUH data on SMI and AMI have used the 2008 estimation method. This section presents the first look at 2011 estimates produced using the improved method; hereafter, such estimates are referred to as the "revised 2011 estimates" or "revised estimates." Additionally, this section compares the revised estimates to the 2011 estimates produced using the 2008 methods (hereafter referred to as the "prior 2011 estimates" or "prior estimates"). Comparisons are made for the overall population and by age group and gender. Revised and prior estimates by other characteristics (e.g., race/ethnicity, region) are provided in tables at the end of this document (see Supporting/Additional Tables). For all mental illness estimates, overall and by characteristics, the 2012 improved methods reduce the number of individuals who have mental illness compared to the 2008 methods.¹⁵

Serious Mental Illness

Among adults aged 18 or older, the revised 2011 estimate of past year SMI is 3.9 percent (9.0 million adults) (Table A1). By age groups, the revised estimates were 3.8 percent for those aged 18 to 25, 5.0 percent for those aged 26 to 34, 5.1 percent for those aged 35 to 49, and 2.8 percent for those aged 50 or older (Figure 1). The revised estimates of SMI by gender are 4.8 percent among females and 2.9 percent among males (Figure 2).

The revised 2011 estimate of SMI is lower than the prior 2011 estimate (3.9 vs. 5.0 percent) (Table A1). The distribution of the prevalence across age groups differs depending on whether the revised or prior methods are applied (Figure 1). For example, in the revised estimates, the prevalence of SMI was highest among the 26 to 34 and 35 to 49 age groups; whereas in the prior estimates, the prevalence was highest among the 18 to 25 age group. Females were more likely than males to have SMI regardless of whether revised or prior methods are used; however, the revised estimates were lower than the prior estimates among both females and males (Figure 2).

Any Mental Illness

The revised 2011 estimate of past year AMI among adults is 17.8 percent (41.4 million adults) (Table A2). By age groups, the revised estimates were 18.5 percent for those aged 18 to 25, 20.9 percent for those aged 26 to 34, 20.0 percent for those aged 35 to 49, and 15.0 percent for those aged 50 or older (Figure 3). The revised estimates of AMI by gender are 21.1 percent among females and 14.2 percent among males (Figure 4).

Similar to SMI, the revised 2011 estimates of AMI are lower than the prior 2011 estimates (17.8 vs. 19.6 percent) (Table A2). The distributions of AMI across age groups for the revised and prior estimates differ (Figure 3). For example, in the revised estimates, the prevalence of AMI was highest among the 26 to 34 and 35 to 49 age groups; whereas in the prior estimates, the highest prevalence was among the 18 to 25 age group. Females were more likely than males to have AMI regardless of whether revised or prior methods are used (Figure 4). However, the revised estimates were lower than prior estimates for both females and males.

Substance Use Disorder among Persons with Mental Illness

When presented independently, estimates of substance use disorder are not affected by using the improved methods. However, estimates of substance use disorder among those with mental illness are affected by using the improved methods.^16,17 Therefore, revised and prior estimates of substance use disorder among adults with mental illness are described and compared in this section.

Among the 9.0 million adults aged 18 or older with SMI in the past year, the revised 2011 estimate of meeting criteria for substance use disorder is 23.6 percent (2.1 million adults). The revised estimates of having a substance use disorder among adults with SMI by age groups were 40.8 percent among adults aged 18 to 25, 31.4 percent among adults aged 26 to 34, 18.6 percent among adults aged 35 to 49, and 16.0 percent among adults aged 50 or older (Figure 5). The revised estimates of having a substance use disorder among adults with SMI by gender are 20.6 percent among females and 29.0 percent among males (Figure 6).

Among the 41.4 million adults (17.8 percent) with AMI in the past year, the revised 2011 estimate of meeting criteria for substance use disorder is 16.5 percent (6.8 million adults). The revised estimates of having a substance use disorder among adults with AMI by age groups were 32.4 percent among adults aged 18 to 25, 22.4 percent among adults aged 26 to 34, 15.6 percent among adults aged 35 to 49, and 7.4 percent among adults aged 50 or older (Figure 5). The revised estimates of having substance use disorder among adults with AMI by gender are 13.2 percent among females and 21.8 percent among males (Figure 6).

The revised estimates of having a substance use disorder among adults with SMI were generally similar to the prior estimates (23.6 and 22.6 percent, respectively), as were the estimates for having a substance use disorder among adults with AMI (16.5 and 17.5 percent, respectively). The distributions of AMI and SMI across age groups and gender for the revised and prior estimates were fairly similar. For example, both the revised and prior estimates of having a substance use disorder among adults with SMI is highest for adults aged 18 to 25 followed by adults aged 26 to 34, adults aged 35 to 49, and adults aged 50 or older. Similarly, both the revised and prior estimates of having a substance use disorder among adults with AMI is highest for adults aged 18 to 25 followed by adults aged 26 to 34, adults aged 35 to 49, and adults aged 50 or older.

Mental Health Service Use

Revised and prior estimates of mental health service use among adults with mental illness are described and compared in this section (Table 2).¹⁸

Among adults with SMI, the revised 2011 estimate of receipt of mental health services was 64.9 percent. The revised estimates of mental health service use among adults with SMI by age groups were 52.1 percent among adults 18 to 25, 56.8 percent among adults aged 26 to 34, 67.6 percent among adults aged 35 to 49, and 73.2 percent among adults aged 50 or older. The revised estimates of receipt of mental health services among persons with SMI by gender are 68.0 percent among females and 59.3 percent among males.

Among adults with AMI, the revised 2011 estimate of mental health service use was 40.8 percent. The revised estimates of mental health service use among adults with AMI by age groups were 32.9 percent among adults aged 18 to 25, 37.0 percent among adults aged 26 to 34, 43.7 percent among adults aged 35 to 49, and 43.6 percent among adults aged 50 or older. The revised estimates of AMI by gender are 44.9 percent among females and 34.2 percent among males.

The revised estimates of mental health service use among persons with SMI or AMI were generally greater than prior estimates. The revised estimate for mental health service use among adults with SMI was 64.9 percent, and the prior estimate was 59.6 percent. The revised estimate for mental health service use among adults with AMI was 40.8 percent, and the prior estimate was 38.2 percent.

Both the revised and prior estimates indicate that the older age groups were more likely to use mental health services compared with the 18 to 25 age group among adults with SMI or AMI. However, the proportion of adults with SMI or AMI receiving mental health services was higher in the revised estimates than in the prior estimates for all age groups. Both the revised and prior estimates indicate that females with SMI or AMI were more likely than males to use mental health services; however, the percentage with SMI or AMI receiving mental health services was higher in the revised estimates than in the prior estimates for both genders.

This report includes information on the improvement to the methods used to produce annual estimates of SMI and AMI among adults. The improvements in methodology were the result of a planned evaluation of these methods at the end of the clinical data collection in 2012 in which a total sample size of over 5,000 clinical interviews was obtained. This report demonstrates that the improved methods produce model based estimates that are more accurate because they are more consistent with the direct estimates from the clinical interview.¹⁹ Therefore, the 2008 methods will no longer be used by SAMHSA to produce SMI and AMI estimates. The 2012 Mental Health National Findings Report scheduled for release in November 2013 will only include SMI and AMI estimates produced using the improved method. In addition, SAMHSA will release revised 2008 to 2011 detailed tables that include the revised estimates for those years. Beginning in October 2013, any CBHSQ report that includes AMI or SMI estimates will be based on the improved method. However, existing reports will not be revised and will continue to be available on the SAMHSA website with a notification about these changes.

The NSDUH public-use files available at http://www.datafiles.samhsa.gov will be updated to include the revised mental illness variables. The updated data files are expected to be made available in the fall of 2013. Additionally, the 2-year Restricted-Use Data Analysis System (R-DAS) data file that allows for analysis of SMI and AMI at the State level is also expected to be updated by the end of 2013.

NSDUH_148