Prevalence of Substance Use Among Racial & Ethnic Subgroups in the U.S.

Chapter 2 Measures and Methods

2.1 Measurement of Substance Use, Dependence, and Need for Treatment

Each NHSDA interview takes about one hour to complete and incorporates procedures designed to maximize honest reporting of substance use. Respondents aged 12 to 17 are allowed to participate in the survey only after the interviewer has received parental permission. In each of the 1991, 1992, and 1993 surveys, more than 40 percent of the adolescent interviews were conducted in complete privacy, with only the interviewer and respondent present, while less than 20 percent were conducted with at least one other person constantly present. (For details, see SAMHSA, 1993, Table B.10; SAMHSA, 1995b, Table B.10; and SAMHSA, 1995c, Table B.10.) Self-administered answer sheets are used by respondents for all alcohol and illicit drug use questions, so that responses are not revealed to interviewers or to any other persons who may have been present at the interview. [ Exceptions to the use of self-administered answer sheets are the questions about tobacco use. In 1991-1993, these questions were interviewer-administered. This procedure was revised in the 1994 NHSDA so that youths now respond using self-administered answer sheets. Results indicate that estimates of youth tobacco prevalence prior to 1994, such as those in this report, are substantial underestimates.] Respondents are given a guarantee of confidentiality and told that only persons connected with the research will ever see their answers. After the questionnaire is completed, the answer sheets are placed in an envelope that is then sealed and mailed back to the data collection contractor with no name or address information included.

The analyses in this report are based on NHSDA variables that are derived from questionnaire items on substances used in the past year and problems associated with use. The following NHSDA questionnaire items, called "recency" items, were used to determine past-year drug use:

1) Cigarettes: "When was the most recent time you smoked a cigarette?"

2) Alcohol: "When was the most recent time you had an alcoholic drink, that is, beer, wine, or liquor or a mixed alcoholic drink?"

3) Marijuana: "When was the most recent time you used marijuana or hash?"

4) Cocaine: "When was the most recent time you used cocaine, in any form?"

5) Illicit drugs other than marijuana or cocaine: "When was the most recent time you used [name of drug]?"

Respondents who reported that their most recent use of the specified substance occurred within the twelve-month period preceding the interview were classified as having used the substance during the past year. The recency question varies depending on the type of substance. For example, when asking about sedatives, tranquilizers, stimulants, and analgesics, the question limits use to "nonmedical use." These forms of drug use make up the illicit drugs along with marijuana, cocaine (including crack), inhalants, hallucinogens, and heroin. In other words, any use of any of these drugs in the past year is identified as use of illicit drugs.

To measure heavy past-month alcohol use, we used responses to the following question:

6) Heavy past-month alcohol use: "On about how many days did you have five or more drinks of beer, wine, or liquor on the same occasion during the past 30 days?"

Individuals who responded five or more days were classified as heavy past-month alcohol users.

To measure heavy past-month cigarette use, we used responses to the following question:

7) Heavy past-month cigarette use: "How many cigarettes have you smoked per day, on the average, during the past 30 days?"

Individuals who responded "about a pack a day (16-25 cigarettes)" or more than a pack a day were classified as heavy past-month cigarette smokers.

We also want to identify patterns of problem substance use. Therefore, two kinds of general indicators of problem use were developed, based on responses to questions about drug problems: a measure of dependence on alcohol and a measure of need for illicit drug abuse treatment. First, each type of substance use identified above has a corresponding dependence indicator that is created from respondents' self-reports of behavioral consequences and dependence symptoms they may have experienced as a result of using the substance during the twelve months preceding the interview. Respondents were asked to identify specific substances that caused them to experience each of seven consequences/symptoms during the year preceding the interview. (The questions were based on questionnaire items and diagnostic criteria presented in the Diagnostic and Statistical Manual of Mental Disorders, Third Edition [DSM-III-R; American Psychiatric Association, 1987].) The seven consequences/symptoms were: 1) feeling depressed or losing interest, 2) getting involved in arguments or fights, 3) feeling alone or isolated, 4) feeling nervous or anxious, 5) developing health problems, 6) experiencing difficulty in thinking clearly, and 7) feeling irritable or upset. For each substance, we used the respondents' reports of these seven consequences/symptoms, together with their reports of the most recent time the substance was used, and of their experiences of "trying to cut down" and using more of the substance "to get the same effect," to define five "criteria" of dependence.

A. The substance caused one or more of the seven consequences/symptoms, and the substance was last used during the past month.

B. The substance caused the respondent to get less work done during the past year, and the respondent last used the substance during the past year.

C. The respondent was unable to reduce his or her use of the substance ("cut down") during the past year.

D. The respondent needed more of the substance to get the same effect during the past year.

E. The respondent felt sick when he or she tried to cut down during the past year.

We classified respondents as "dependent on alcohol" during the past year if two or more of the five criteria of dependence were satisfied for alcohol.

Second, following Epstein and Gfroerer (1996), we constructed a measure of the need for illicit drug abuse treatment. A respondent was classified as in need of illicit drug abuse treatment if he or she 1) was dependent on an illicit drug during the past year, 2) reported injection substance use or heroin use during the past year, 3) used sedatives, stimulants, tranquilizers, analgesics nonmedically, inhalants, hallucinogens, or cocaine weekly during the past year, or 4) received treatment for illicit drug use during the past year.

Table 2.1 presents the estimated number and percentage of total persons, and males and females aged 12 and older, who were classified as positive according to this measure. This information is presented for each of the nine measures of substance use, alcohol dependence, or need for treatment, which are defined in this section and analyzed in Chapters 3, 4, and 5. For example, we estimate that there were about 63.4 million persons who used cigarettes in the year preceding the survey, about 31 percent of the NHSDA target population aged 12 and older. Of the 63.4 million past-year cigarette users, 33.1 million were male, accounting for about 34% of the male population aged 12 and older, and 30.3 million were female, accounting for about 28% of the female population aged 12 and older. Standard errors for Table 2.1, and subsequent tables in this report, are presented in the Appendix.

2.2 Measurement of Racial and Ethnic Identification

Several studies have investigated the associations between race/ethnicity and substance use among adolescents and among adults. As mentioned above, the principal advantage of this study is a much larger sample size than previous studies, approximately 87,000 respondents aged 12 and older in the combined 1991-93 NHSDAs, including more than 22,000 adolescent respondents aged 12 to 17, more than 21,000 young adult respondents aged 18 to 25, and more than 44,000 adult respondents aged 26 and older. The large sample size makes possible 1) a more refined approach to measuring race/ethnicity, and 2) more extensive breakdowns by demographic and other social factors in analyzing the associations between race/ethnicity and substance use.

The racial/ethnic classification used in this report is based on responses to three interviewer-administered questions:

1) "Are you of Hispanic or Spanish origin or descent?"

YES ___

NO ___

2) (If yes to preceding question) "Which of these Hispanic-origin groups best describes you?"

Puerto Rican ___

Mexican ___

Cuban, or ___

Some other group (SPECIFY) ______________________

3) Which of the groups on this card best describes you? Just give me the number.

WHITE ___ 01

BLACK ___ 02

INDIAN (American), ALEUT, ESKIMO ____ 03

ASIAN OR PACIFIC ISLANDER (Including Asian Indian) ___ 04

OTHER (Specify) ______________________ 05

As stipulated in Statistical Policy Directive No. 15, issued by the Office of Management and Budget (U.S. Department of Commerce, 1978), persons of Hispanic origin (first and second questions) may be of any race (third question). Both in the eleven-category racial/ethnic classification used in Chapters 3 and 4 and in the seven-category racial/ethnic classification used in Chapter 5, an individual's Hispanic origin takes precedence over any racial identification. For example, an individual who was identified as "Hispanic-Mexican" based on the first two questions and as "Indian (American), Aleut, Eskimo" based on the third question was classified as "Hispanic-Mexican" rather than as "Indian (American), Aleut, Eskimo." Similarly, the classifications "black" and "white" as used in this report refer to non-Hispanic blacks and to non-Hispanic whites, respectively.

Write-in responses to the second question, "some other group (SPECIFY)," were used to extend our list of Hispanic-origin groups from four to seven, including "Hispanic-Caribbean" (other than Puerto Rico and Cuban), "Hispanic-Central American" (other than Mexico), "Hispanic-South American," and "Hispanic-Other" (other than Caribbean, Central American, Cuban, Mexican, Puerto Rican, and South American).

Individuals who gave write-in responses to the third question, "OTHER (Specify)," and who were not Hispanic were usually recoded to one of the four explicit non-Hispanic racial categories—Native American, Asian/Pacific Islander, non-Hispanic black, and non-Hispanic white. For example, individuals who reported that they were "African American" were recoded as "black," and individuals who reported that they were "Caucasian" were recoded as "white." A small number of respondents, less than 1%, were statistically imputed to a Hispanic origin or racial classification based on the interviewer-observed race of the respondent and on data on race and Hispanic origin that were collected during household screening. No information is collected on respondents’ citizenship or country of birth, and thus, no inferences can be made about either. Details of the recoding of "other Hispanic" and "other race" responses and of the statistical imputation procedures are available from SAMHSA.

Table 2.2 presents the sample sizes (combined 1991-93 NHSDAs) of the eleven racial/ethnic categories that are analyzed in Chapters 3 and 4, and the estimated number and percentage of NHSDA target population persons aged 12 and older in each racial/ethnic category. The eleven categories are grouped into five major categories: Native American (including American Indian, Eskimo, Aleut), Asian/Pacific Islander, Hispanic, non-Hispanic black, and non-Hispanic white, and there are seven subcategories of Hispanic: Caribbean, Central American, Cuban, Mexican, Puerto Rican, South American, and other Hispanic. Also shown in Table 2.2 are the corresponding population estimates and population percentages of the eleven subgroups based on the 1990 U.S. Census. The population estimates based on the 1990 U.S. Census are larger, because the Census data are not restricted to individuals aged 12 and older, but the estimated population percentages based on the 1990 U.S. Census are closely similar to the corresponding percentages based on the combined 1991-93 NHSDAs.

To be sure, the similarity between NHSDA and Census estimates in Table 2.2 results in part from the procedures that are used to weight the NHSDA data. As discussed in SAMHSA (1993, 1995b, c), the NHSDA sampling weights are post-stratified to ensure that NHSDA estimates of the percentages of U.S. population in demographic subgroups defined by age, sex, Hispanic origin, and race are consistent with Census Bureau projections based on the 1990 Census. The post-stratification procedure explains why NHSDA and Census estimates of the percentages of Asian/Pacific Islanders, Hispanics, Non-Hispanic blacks, and Non-Hispanic whites are similar in Table 2.2. However, the post-stratification procedure does not fully explain the similarity between NHSDA and Census estimates of the percentages of Native Americans and Hispanic subgroups, because this procedure ensures consistency only for subgroups defined by two categories of Hispanic origin (Hispanic, non-Hispanic) and four categories of race (Asian/Pacific Islander, black, white, other).

Table 2.3 presents the sample sizes of the eleven racial/ethnic subgroups by age group (12 to 17, 18 to 25, 26 to 34, 35 and older), region (Northeast, North Central, South, West), and population density (metropolitan statistical area greater than 1 million, other MSA, and not in MSA). The sample sizes are especially small for Native Americans (n = 416), Hispanic-Caribbean (n = 611), Hispanic-Central American (n = 1,962), Hispanic-South American (n = 1,075), and other Hispanic (n = 1,181). The small sample sizes of these five subgroups explain why the eleven-category classification of Chapters 3 and 4 had to be reduced to only seven categories in the detailed analyses of Chapter 5. The racial/ethnic classification of Chapter 5 combines four Hispanic subgroups of Chapters 3 and 4—Caribbean, Central American, South American, and Other Hispanic—into an expanded "Other Hispanic" category, and Chapter 5 does not report results for Native Americans at all.

For brevity in this report, we refer to the Hispanic-Cuban subgroup as "Cubans," to the Hispanic-Mexico subgroup as "Mexicans," to the Hispanic-Puerto Rico subgroup as "Puerto Ricans," to the Hispanic-Caribbean subgroup as "Caribbeans," to the Hispanic-Central America subgroup as "Central Americans," and to the Hispanic-South America subgroup as "South Americans." Except in referring to the results of previous research, this report does not use the terms "Asian-American," "Cuban-American," "Mexican-American," "Puerto Rican-American," and so on. Yet it is worth emphasizing that all of the conclusions of this report pertain strictly to residents of the United States and do not pertain to residents of the countries and geographic areas whose names are used to designate U.S. racial/ethnic subgroups.

2.3 Sociodemographic Control Variables

All analyses in this report are presented separately for adolescents (individuals aged 12 to 17) and older respondents. Chapter 4 broadly describes the associations between race/ethnicity and substance use by controlling only for the gender and detailed age of the respondent (12 to 17, 18 to 25, 26 to 34, 35 and older). With a few exceptions, the analyses of Chapter 4 suggest that racial/ethnic differences in substance use are similar for males and females. Chapter 5 presents a more detailed analysis of the associations between race/ethnicity and substance use, by controlling for a dozen sociodemographic variables in addition to age of respondent (12 to 17 versus 18 and older). The sociodemographic control variables in Chapter 5 depend on whether the respondent is aged 12 to 17 or 18 or older. For respondents aged 12 to 17, the control variables are region, population density, language of interview (Spanish vs. English), family income, health insurance coverage, receipt of welfare by a member of the household, school enrollment, and family structure (living with two biological parents vs. all other living arrangements). For respondents aged 18 and older, the control variables are region, population density, language of interview (Spanish vs. English), family income, health insurance coverage, receipt of welfare by the household, educational attainment, marital status, employment status, and own children (none vs. one or more).

Based on preliminary analyses, we collapsed family income into three income groups according to whether the family income was less than $20,000, between $20,000 and $40,000, or greater than $40,000. Based on research showing a complex, nonlinear relationship between educational attainment and substance use (Zhu et al., 1996), we distinguished four categories of educational attainment: 0-8 years of schooling, 9-11 years, 12 years, and more than 12 years.

Table 2.4 defines the levels of all sociodemographic control variables that are used in this report and also provides the operational definitions of the levels.

2.4 Statistical Methods

Percentage estimates in the text tables of Chapters 3-5 have been rounded to the nearest one-tenth of one percent and screened to determine whether they meet a required level of statistical precision. Differences between racial/ethnic categories and other demographic subgroups that are reported in the text are statistically significant at the .05 level.

Appendix A discusses details of the statistical precision requirements, sample weighting, and statistical estimation and testing techniques. The tables of Appendix A present standard errors for all statistical estimates that are presented in the text of this report. Like the estimates themselves, the standard errors fully take into account the complex sample design of the NHSDA.

In interpreting the statistical estimates of this report, it is useful to remember that associations between variables are more likely to be statistically significant when the sample size is large. Thus, if differences in substance use prevalence by racial/ethnicity are found to be significant in the total population aged 12 and older but not significant within either the 12 to 17 or the 18 and older age groups, this does not mean that there are no real differences by race/ethnicity within age groups. It only means that such differences cannot be inferred given samples of the size used in this report. In general, associations involving substance use or dependence measures with low overall prevalence are more difficult to identify as statistically significant than associations involving measures with high overall prevalence.

Although statistical significance is an important criterion in interpreting the findings, it should not be the sole one. Some findings of this report involve broad patterns of differential substance use prevalence and dependence by race/ethnicity, patterns that show up in the estimates reported for most or all measures of substance use prevalence and dependence. Such persistent and recurring patterns in the data are often worth noting even when such trends and patterns are difficult to evaluate precisely on the basis of statistical significance tests.

Table 2.1 Measures of Substance Use, Dependence on Alcohol, and Need for Illicit Drug Abuse

Treatment. Respondents Aged 12 and Older, by Gender, 1991-1993.

¹The size of the target population for the combined 1991-93 data sets equals the average of the populations represented by the 1991, 1992, and 1993 NHSDAs. The target population comprises 205,257,000 persons, with 98,447,000 males and 106,810,000 females.

²Illicit drugs include marijuana, nonmedical use of psychotherapeutics (sedatives, tranquilizers, analgesics, and stimulants), cocaine (including crack), inhalants, hallucinogens, and heroin.

Source: Office of Applied Studies, SAMHSA, National Household Survey on Drug Abuse, 1991-93.

Table 2.2 Estimated Population Sizes of Racial/Ethnic Subgroups:

NHSDA 1991-1993 and U.S. Census 1990

1. 1990 U.S. Census data. Data on Hispanic origin by country are based on a sample tabulation of responses to the long form question on ancestry. From U.S. Bureau of the Census, http://venus.census.gov/cdrom/lookup/862430630.

2. Native American (n =826) and Asian/Pacific Islander (n =305) respondents who also identified themselves as Hispanics were classified as Hispanics.

3. The Caribbean estimate from the 1990 Census is based on a sample tabulation of persons who identified themselves as Dominican in responses to the long form ancestry item.

4. The "other" Hispanic category is comprised mostly of respondents who reported a Hispanic group of "Spanish (Spain)" or "Spanish American." This group also includes respondents who indicated that they were Portuguese, Filipino, "mixed," or did not report a specific Hispanic group.

* NHSDA data were post-stratified to 1980 Census tables by age, gender, and race/ethnicity.

Table 2.3 Sample Sizes of Racial/Ethnic Subgroups by Age Group, Region, and Population Density: NHSDA 1991-1993

Source: Office of Applied Studies, SAMHSA, National Household Survey on Drug Abuse, 1991-93.

Table 2.4 Description of Sociodemographic Variables

 

Table 2.4 Description of Sociodemographic Variables (continued)
Variable	Levels	Definition
Health Insurance Coverage	Yes No	Health insurance in the month preceding interview includes private health insurance, Medicaid, Medicare, CHAMPUS, CHAMPVA, VA, or military health care.
Welfare Receipt	Yes No	The respondent or another member of the respondent’s family received any kind of welfare in the month preceding the interview.
Number of People in Household	Mean	Number of people living in the respondent’s household (including respondent) at the time of interview.

Return to Table of Contents (TOC)

This page was last updated on May 19, 2008.

SAMHSA, an agency in the Department of Health and Human Services, is the Federal Government's lead agency for improving the quality and availability of substance abuse prevention, addiction treatment, and mental health services in the United States.