1997 National Household Survey on Drug Abuse
Overview of NHSDA Main Findings Report
In addition, two summary measures of drug use were constructed to examine the overall extent of involvement in drug use: "any illicit drug use" and "nonmedical use of any psychotherapeutic drugs." The summary measure "any illicit drug use" includes use of illegal drugs (such as marijuana, cocaine, inhalants, hallucinogens, and heroin) and the nonmedical use of prescription-type psychotherapeutic drugs. Thus, throughout this report, the terms "any illicit drug use" and "illicit drugs" refer to both the use of illegal drugs and the nonmedical use of psychotherapeutic drugs. The summary measure "nonmedical use of any psychotherapeutic drugs" includes use of prescription-type psychotherapeutic drugs, such as stimulants, sedatives, tranquilizers, and analgesics, without a doctor's prescription or in amounts or for purposes other than prescribed.
In 1994, two separate versions of the NHSDA questionnaire were administered as part of a split-sample experiment: 1994-A, which was essentially the same questionnaire as had been used in previous NHSDAs, and 1994-B, which featured certain improvements in questionnaire design and editing procedures over the previous version. The 1995, 1996, and 1997 survey questionnaires and editing procedures were similar to those implemented in the 1994-B NHSDA. In this 1994-B version, changes were implemented to improve the measurement of trends, reduce data processing time, and improve the overall quality of data essential to policymakers at all levels of government. Because the new methodology affected the levels of substance use reported by respondents and, therefore, the estimates of prevalence, many previously published estimates in Main Findings reports for the survey years prior to 1994 are not comparable to the estimates for 1994 and later NHSDAs. To account for these methodological effects, an adjustment procedure was developed and applied to data presented in Chapter 2 of this report for years prior to 1994. (See Appendix E for a description of the adjustment method.) Readers need to be aware that all 1979-1993 data shown in this report are different from those previously published in NHSDA Main Findings reports for 1979-1993.
Throughout the survey series, the NHSDA questionnaire has been sufficiently flexible to permit examination of special topics. In 1982, the NHSDA devoted considerable attention to medical as well as nonmedical use of stimulants, sedatives, tranquilizers, and analgesics. In the 1979 and 1982 surveys, the survey obtained supplementary information about the prevalence of heroin use by questioning respondents about their friends' use of heroin. Since 1985, the survey has included additional questions about cigarettes and other tobacco products, such as smokeless tobacco. Questions about the ways cocaine is used were added in 1985. Additional questions about "crack" cocaine and sharing needles for injection of drugs have been included since 1988. Beginning in 1985, questions on the perceived consequences of use of various drugs were added. In 1990, questions were added about health insurance and total annual family income. Since 1991, the survey has obtained information about employment, including drug testing (a more detailed set of questions on these two topics was used in 1991, 1992, and 1993).
The new questionnaire, 1994-B, included questions on mental health, access to health care, and drug testing in the workplace. In 1995, questions on the need for treatment for drug or alcohol use, and criminal record, were included under special topics. The 1994-B mental health and access to care questions were used again from 1995 through 1997, and questions on the need for treatment for drug or alcohol use, and criminal record, were included under special topics. In 1996, the questions on the risk/availability of drugs were re-introduced and continued to be asked in 1997. The 1997 questionnaire also included new items about use of cigars, people present when respondents used marijuana or cocaine for the first time (if applicable), reasons for using marijuana or cocaine the first time, reasons for using marijuana or cocaine in the past year, reasons for discontinuing use of these two drugs (for lifetime but not past year users), and reasons why respondents never used these two drugs. In addition, a new series of questions was introduced in 1997 that was asked only of youthful respondents aged 12 to 17. These questions covered a variety of topics relevant to the lives and perceptions of young respondents that may be associated with substance use behaviors.
To increase the reliability of estimates of drug use, the NHSDA has oversampled population groups of special interest. People under age 35 have traditionally been oversampled. Non-Hispanic blacks and Hispanics have been oversampled since 1985. In 1979 and in 1994, the NHSDA oversampled respondents from rural areas. In 1990, the Washington, DC, Metropolitan Statistical Area (MSA) was oversampled. In 1991, 1992, and 1993, six MSAs were oversampled: Chicago, Denver, Los Angeles, Miami, New York, and Washington, DC. (The six-city oversample was discontinued in 1994.) In 1991-1993, urban areas of the six MSAs that were low in socioeconomic status (SES) also were oversampled so that separate estimates might be reported for low-SES urban areas and for all other areas of each oversampled MSA. Current smoking status was first used in the selection process in 1993, and the 1994 and 1995 NHSDAs continued to oversample smokers 18 to 34 years old (see Appendix E of OAS, 1996a). In the 1996 NHSDA, the oversampling of cigarette smokers aged 18 to 34 was discontinued. The oversampling of 18- to 34-year-old past month cigarette smokers was abandoned because the precision gains experienced from oversampling smokers did not justify the additional screening costs.
The principal demographic correlates of drug use
examined in this and previous NHSDA reports are as follows:
The variable composition of the "other" employment category is a source of possible confusion because the composition of this category varies substantially across the four age groups. Students make up the largest fraction in the youngest adult age group (18 to 25); retired and disabled persons make up the largest fraction in the oldest adult age group (35 or older). Thus, the differential composition of the "other" category should be considered in comparisons of employment groups. Cautionary notes are provided in the text where appropriate.
In tables reporting drug use by gender, race/ethnicity, population density, region, educational attainment, or employment status, results are generally presented separately for each of the four age groups. No further cross-classification of demographic characteristics, however, is provided. For example, tables showing drug use by gender, and tables showing drug use by employment status, also control for age, but none of the tables examines drug use while simultaneously controlling for gender, employment status, and age. Because of this limitation, there is a risk that demographic comparisons in this report may be misinterpreted when demographic characteristics are highly associated within an age group. For example, within each age group, women are less likely than men to be currently employed full-time, so associations between drug use and gender might be at least partially attributable to the effects of current employment. Multivariate analyses would permit a more thorough examination of the unique effects of each demographic characteristic; this type of analysis can be conducted using the NHSDA's public use files. The public use files from 1979 to 1996 are currently available on the World Wide Web (WWW), and 1997's will soon be available, within the Substance Abuse and Mental Health Data Archive (SAMHDA) at the University of Michigan: http://www.icpsr.umich.edu/SAMHDA/nsduh.html.
Empirical associations between demographic variables and drug use do not imply causal relationships. In particular, except for stratification by age, this report does not attempt to control for potentially confounding variables that might help to account for the observed associations. This point is particularly salient with respect to associations between race/ethnicity and drug use. Race/ethnicity is highly associated with SES, educational attainment, geographic location, and many other features of the broader social environment.4 The tables presented in this report, however, are particularly useful for the purpose of identifying demographic subgroups with relatively high and low levels of drug use, regardless of the underlying causes of drug use.
This report presents results of tests of statistical significance for comparisons between demographic groups for 1997 and for comparisons of estimates between 1996 and 1997. Differences in levels of drug use between groups and between years were tested for statistical significance using Z tests. The Z test takes into account the sizes of the subsamples being compared and the degree of variation among sample members. An observed sample difference is designated as "statistically significant" if the probability of a sample difference equal to or larger than the observed sample difference is less than or equal to .05. Unless otherwise noted, differences discussed in the text were statistically significant at the .05 level or lower.
Some substantively large and interesting differences also are discussed, even though these are not statistically significant at the .05 level. A difference between subgroups or between years can fall short of statistical significance even when the rate for one subgroup or year is twice or more the rate for another subgroup or year. This seeming incongruity can occur when the rates for both groups are small, when the sample sizes are small, or both. For example, if the rate of use of one comparison group is 1% and the rate of use of the other is 1.5%, the difference may be statistically insignificant. Although the rate of use of the second group is 50% higher than the rate of use of the first group, the NHSDA sample may not be large enough to reliably demonstrate or detect a difference of this magnitude.
Estimates of individual rates, percentages, and proportions considered to be unreliable were omitted from all tables and were noted by asterisks (*). An estimate was considered to be unreliable if the relative standard error (RSE) was greater than 17.5% of the log transformation of the estimated proportion. Because of the relatively large sample sizes for most population subgroups, low precision usually occurs only for prevalence estimates that are very close to zero or 100%. In addition, very small estimates that round to zero (i.e., are <.05%) and are not already suppressed due to low precision are indicated by an asterisk (*) in table cells.
Estimates are not available for some survey years because of differences in the survey instruments. These data points are footnoted in the tables and marked with a double hyphen (--). As mentioned above, the 1979-1993 estimates presented in this report are not comparable to previously published NHSDA estimates for 1979-1993 that were not adjusted. These estimates were adjusted to account for the new methodology that was begun in the 1994 NHSDA (see Appendix E for more detailed discussion of adjustment procedure).
Appendix A of this report contains definitions of key terms, including drug prevalence measures, demographic characteristics, and statistical terms for the 1972 to 1997 surveys. Appendices B, C, and D include additional information on the quality of the data, sample selection, sampling errors, confidence intervals, significance testing, and weighting procedures. Appendix E explains the adjustment procedure that was applied to the 1979-1993 NHSDA estimates in order to facilitate long-term trend analysis. Appendix F contains the drug answer sheets and special topic answer sheets from the 1997 NHSDA questionnaire.
3 In the interest of readability for this report, "white" is used to indicate "white, non-Hispanic," and "black" to indicate "black, non-Hispanic."
4 For a demonstration of the extent to which racial/ethnic differences in drug use prevalence may be influenced by differences in other sociodemographic characteristics measured in the NHSDA, see Flewelling, Ennett, Rachal, and Theisen (1993).
This page was last updated on December 30, 2008.