2. Data and Methods - Initiation of Marijuana Use: Trends, Patterns, and Implications Report

The National Household Survey on Drug Abuse (NHSDA) is the primary source of statistical information on the use of licit and illicit drugs by the U.S. population aged 12 or older. Conducted by the Federal Government since 1971, the survey collects data by administering questionnaires to a representative sample of the population through face-to-face interviews at their place of residence. The survey is sponsored by the Substance Abuse and Mental Health Services Administration (SAMHSA). Data collection is carried out by RTI of Research Triangle Park, North Carolina, under a contract with SAMHSA's Office of Applied Studies (OAS).

The target population covers residents of households, noninstitutional group quarters (e.g., shelters, rooming houses, dormitories), and civilians living on military bases. Persons excluded from the survey include homeless people who do not use shelters, active military personnel, and residents of institutional group quarters, such as prisons and long-term hospitals. The survey is conducted from January through December each year.

Prior to 1999, the NHSDA was administered in about an hour and used paper-and-pencil interviewing (PAPI) methods. The NHSDA PAPI instrumentation consisted of a questionnaire booklet completed by the interviewer and a set of individual answer sheets completed by the respondent. All substance use questions and other sensitive questions appeared on the answer sheets so that the interviewer was not aware of the respondent's answers. Less sensitive questions, such as demographics, occupational status, and household size and composition, were asked aloud by the interviewer and recorded in the questionnaire booklet.

Beginning in 1999, the NHSDA underwent a major redesign. The new features of the survey design produce a significant impact on the NHSDA estimates for substance use. In addition to the following summary, see the report titled Development of Computer-Assisted Interviewing Procedures for the 1999 National Household Survey on Drug Abuse (OAS, 2001a).

First, the method of data collection was changed from a paper questionnaire administration to a computer-assisted administration. The 1999 NHSDA marked the first survey year in which the national sample was interviewed using a computer-assisted interviewing (CAI) methodology. The survey used a combination of computer-assisted personal interviewing (CAPI) conducted by the interviewer and audio computer-assisted self-interviewing (ACASI). For the most part, questions previously administered by the interviewer are now administered by the interviewer using CAPI, and questions previously administered using answer sheets are nowadministered using ACASI. The CAI method has many advantages over PAPI, including more efficient collection and processing of the data and improved data quality.

Use of ACASI is designed to provide the respondent with a highly private and confidential means of responding to questions and should increase the level of honest reporting of illicit drug use and other sensitive behaviors. The interview averages about an hour. In brief, the interview begins in CAPI mode with the field interviewer (FI) reading the questions from the computer screen and entering the respondent's replies into the computer. The interview then transitions to the ACASI mode for the sensitive questions. In this mode, the respondent can read the questions silently on the computer screen and/or listen to the questions read through headphones and enter his or her responses directly into the computer. At the conclusion of the ACASI section, the interview returns to the CAPI mode with the interviewer completing the questionnaire. No personal identifying information is captured in the CAI record for the respondent.

Second, the sample design was changed from a strictly national design to a State-based sampling plan. Beginning in 1999, the NHSDA sample employed a 50-State design with an independent, multistage area probability sample for each of the 50 States and the District of Columbia. The eight States with the largest population (which together account for 48 percent of the total U.S. population aged 12 or older) were designated as large sample States (i.e., California, Florida, Illinois, Michigan, New York, Ohio, Pennsylvania, and Texas). For these States, the design provided a sample large enough to support direct State estimates. For the remaining 42 States and the District of Columbia, smaller, but adequate samples were selected to support State estimates using small area estimation (SAE) techniques. The design also oversampled youths and young adults so that each State's sample was approximately equally distributed among three major age groups: 12 to 17 years, 18 to 25 years, and 26 years or older.

The NHSDA also tripled its sample size in 1999, which makes it possible to produce marijuana use estimates separately for every State and the District of Columbia and for smaller population subgroups. It also allows more detailed analyses of national patterns of use, predictors of recent initiation, and consequences of early first use of marijuana. The precision of the estimates at the national level has been improved substantially. The CAI methodology has made data collection and processing more efficient and improved the quality of the data. However, because of the major differences between the CAI and PAPI methods, it is not appropriate to compare the 1999 and 2000 CAI estimates of substance use with earlier NHSDA estimates in order to assess changes over time in substance use. In addition, the sample expansion had unexpected effects on some aspects of the data collection. In-depth analyses of these methodological issues are described in another SAMHSA report (Gfroerer, Eyerman, & Chromy, in press).

Third, beginning in 1999, the NHSDA questionnaire allows for collecting year and month of first use for recent initiates (i.e., new drug users). In addition, the questionnaire call record provides the date of the interview. Exposure time to substances can be determined in terms of days and converted to an annual measure. Having data about exact dates of birth and first use allows person time of exposure to a drug during the targeted period to be determined. In prior years, before exact date data were available for computing incidence of drug use, the calculation of the person time exposure for incidence rates of drug use was based on an approximation, rather than an exact computation for each person. Thus, because of the changes in methodology since the 1999 NHSDA, the estimates from the 1999 and 2000 surveys are not completely comparable with data obtained from prior surveys. Nonetheless, because all incidence estimates in this report, including pre-1999 estimates, are based on the 1999 and 2000 NHSDAs, they are comparable. A more complete discussion of the differences between the old and new incidence estimates is presented elsewhere (Chromy, in press; Gfroerer et al., in press).

Regardless of the survey year, the NHSDAs are all based on retrospective reports by survey respondents, and they may be subject to similar kinds of recall and reporting biases. Some sources of biases are related to the NHSDA designs and retrospective self-reports.

First, some degree of underreporting on drug use-related behaviors might have occurred because of the social acceptability of drug use behaviors and respondents' fear of disclosure. Prior studies showed that underreporting of drug use among youths in their homes may be substantial (Gfroerer, 1993; Gfroerer, Wright, & Kopstein, 1997). Self-report data also are influenced by memory and recall errors, including recall decay (tendency to forget events occurring long ago) and forward telescoping (tendency to report that an event occurred more recently than it actually did). These memory errors would both tend to result in estimates for earlier years (i.e., 1960s and 1970s) that are downwardly biased (because of recall decay) and estimates for later years that are upwardly biased (because of telescoping).

Second, the NHSDA target population focuses on civilian, noninstitutionalized household residents. Although it includes almost 98 percent of the U.S. population aged 12 or older, some population subgroups who may have different drug-using patterns are excluded, such as active military personnel, people living in institutional group quarters, and homeless persons not living in identifiable shelters. Thus, the generalizability of the findings to the excluded subgroups is limited. Further, the estimates for drug use should be considered conservative.

Third, there is a potential bias associated with differential mortality because some individuals who were exposed to the risk of first drug use in the historical periods shown in thetables died before the 1999 NHSDA was conducted. This bias is probably very small for analyses of recent marijuana initiation.

Fourth, marijuana incidence trends based on NHSDA data may be biased. Johnson, Gerstein, and Rasinski (1998) concluded that the marijuana incidence trend from the NHSDA may be biased because the reporting of initiation declines as the length of time between initiation and the survey increases. However, their analysis did not address very recent estimates, which could be biased because they reflect recent drug use and because they are heavily based on the reports of adolescents. Appendix A presents estimates for cocaine, heroin, and marijuana use based on single years of NHSDA data in order to better understand the size of the biases and to assess the reliability of estimates for recent years. This analysis shows that marijuana initiation rates appear to have small biases.

A total of 66,706 respondents aged 12 or older completed the 1999 survey, and a total of 71,764 respondents completed the 2000 survey (Tables 2.1 and 2.2). The analysis samples for this report vary depending on the specific objective of the chapter. The full sample of 1999 and 2000 CAI data (N = 138,470) was used to estimate trends in the incidence of marijuana use (Chapter 3), as well as State-level incidence rates (Chapter 4). The characteristics of recent initiates (Section 5.2) were examined in a sample consisting of individuals who started to initiate marijuana use in 1998 and individuals who had never used marijuana prior to 1998 in the 1999 NHSDA, as well as 1998 and 1999 marijuana initiates and individuals who had never used marijuana prior to 1998 in the 2000 NHSDA (n = 99,752). The analysis of seasonality (Section 5.3) was based on all marijuana users who were asked the month of first use question and reported it (i.e., no imputed data) (n = 2,085). These users reflect persons interviewed during 1999 and 2000 who initiated marijuana either at their current age or at 1 year less than their current age. Although this captures some initiation occurring in 1997 and 2000, it primarily represents 1998 and 1999 initiation.

The sample for the analysis of predictors of recent initiation (Section 5.4) was restricted to 1998 marijuana initiates and individuals who had never used marijuana prior to 1998 in the 1999 NHSDA, as well as 1999 marijuana initiates and individuals who had never used marijuana prior to 1999 in the 2000 NHSDA (n = 97,530). Using this sample allows the analysis to reflect the population that had never used marijuana at a point in time and model the transition from nonuse to use over a 1-year time period.

For the analysis of the relationship between early use and later drug use patterns (Chapter 6), the sample included all lifetime marijuana users aged 26 or older in 1999 and 2000 CAI data(n = 16,652). Because the 1999 questionnaire did not fully represent the criteria in the 4^th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (American Psychiatric Association [APA], 1994), the analysis of substance dependence and/or abuse was conducted on data from the 2000 survey (n = 8,927). A subset of the analysis investigated whether the risk of substance dependence and/or abuse was greater for early initiates than late initiates among lifetime marijuana users who used it in the past year. The analysis sample was based on lifetime marijuana users aged 26 or older who also used marijuana in the past year (n = 1,447).

This section describes the NHSDA measures and definitions of the following study variables: first marijuana use; social and demographic variables; use of cigarettes, alcohol, and other drugs; heavy use of illicit drugs; and dependence on or abuse of alcohol and other drugs.

Estimates of first use, incidence, or initiation of marijuana use were based on the following questions: age at first use, year and month of first use for recent initiates, the respondent's date of birth, and the interview date. By using this information, along with editing and imputation when necessary, an exact date of first use was determined for each marijuana user. Recent marijuana initiates were defined as persons who reported that their first use of marijuana occurred during 1998 or 1999. Age at first marijuana use was defined as self-reported age at first use of marijuana and was grouped into four categories (i.e., aged 14 or younger, 15 to 17, 18 to 20, and 21 or older).

A range of social and demographic variables was included in the 1999 and 2000 NHSDAs. Age of the respondent was defined as "age at time of interview." In the predictor analysis (Section 5.4), age was defined as "age on January 1, 1998" in the 1999 survey and as "age on January 1, 1999" in the 2000 survey. Race/ethnicity was coded into the following categories: (a) non-Hispanic whites (referred to as "whites"); (b) non-Hispanic blacks (referred to as "blacks"); (c) Hispanics; (d) non-Hispanic American Indians/Alaska Natives; (e) non-Hispanic Asians/Pacific Islanders/Native Hawaiians; and (f) non-Hispanic persons reporting more than one race. Level of education (for 18 to 25 year olds only) was categorized into four groups: school dropouts, high school students, high school graduates, and college students or graduates. A school dropout was defined as an individual aged 18 to 25 who had completed less than the 12^th grade, reported not being currently enrolled in school, and did not have a high school degree or take an equivalency exam (e.g., a general equivalency diploma [GED]). Employment status (for 18 to 25 year olds only) was categorized into employed persons (full-time or part-time) and not employed persons (unemployed individuals, students, or others). Not employed persons in this age group primarily consist of students. Marital status was categorized into two groups: never married and ever married (for 18 to 25 year olds only).

Population density was grouped into three categories: large metropolitan, small metropolitan, and nonmetropolitan. Large metropolitan areas had a population of 1 million or more; small metropolitan areas had a population of less than 1 million; and nonmetropolitan areas were areas outside metropolitan statistical areas (MSAs). Residential region was categorized into four regions: Northeast, Midwest, South, and West. The Northeast region has nine States: Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, and Vermont. The Midwest region has 12 States: Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin. The South region has 16 States, as well as the District of Columbia: Alabama, Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia. And the West region has 13 States: Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming.

The definitions for the use of cigarettes, alcohol, and other drugs include past year use and lifetime use. Lifetime use referred to a respondent reporting any use of the substance at least once in his or her lifetime. Past year use referred to a respondent reporting any use of the substance at least once during the 12 months preceding the interview date. Use of psychotherapeutic drugs was defined as any nonmedical use of prescription-type pain relievers, sedatives, tranquilizers, or stimulants (i.e., when it was not prescribed for the respondent, or used only for the experience or feeling it caused). Pain relievers include painkillers like Darvon, Demerol, Percodan, and Tylenol with codeine. Sedatives are sometimes referred to as "downers" and include barbiturates, sleeping pills, and Seconal. Tranquilizers include antianxiety drugs, such as Librium, Valium, Ativan, and Meprobamate. Stimulants are often called "uppers" or "speed" and include amphetamines and Preludin.

Heavy marijuana use was defined as using marijuana daily or almost daily in the past year (i.e., at least 300 days). Heavy use of other illicit drugs referred to using one or more of the following drugs on at least 50 days in the past year: cocaine, hallucinogens, heroin, inhalants, pain relievers, sedatives, tranquilizers, or stimulants, regardless of heavy marijuana use.

The 2000 NHSDA included a series of questions to assess substance dependence and abuse based on DSM-IV criteria (APA, 1994). The seven dependence criteria are (1) tolerance; (2) withdrawal or avoidance of withdrawal; (3) persistent desire or unsuccessful attempts to cut down or stop substance use; (4) spending a lot of time using the substance, obtaining the substance, or recovering from its effects; (5) reducing or giving up occupational, social, or recreational activities in favor of substance use; (6) impaired control over substance use; and (7) continuing to use the substance despite physical or psychological problems. A respondent was considered to be dependent on a substance when he or she reported having at least three of the dependence criteria.

The four substance abuse criteria are (1) having serious problems due to substance use at home, work, or school; (2) the use of that substance putting the respondent in physical danger; (3) substance use causing the respondent to be in trouble with the law; and (4) continuing to use the substance despite having substance-use-related problems with family and friends. A respondent was classified with abuse when he or she reported having at least one of the four abuse criteria.

The following types of substance dependence or abuse were studied in relation to age at first use of marijuana:

SUrvey DAta ANalysis (SUDAAN) software (Shah, Barnwell, & Bieler, 1996) was used for the analyses to take into account the complex survey design of the NHSDA. The incidence rate of marijuana was defined as the rate of new marijuana users in a given year (i.e., the number of new marijuana users divided by the person time of exposure) (Appendix A). By applying sample weights to the incidence of first use, estimates of the number of new users for each year were made. The incidence of first use was classified by year of occurrence. For age-specific incidence rates, the period of exposure was defined for each respondent and age group for the time that the respondent was in the age group during the calendar year. For the analysis that used aggregated 1999 and 2000 data, sample weights were adjusted to obtain a simple average weight over 2 years (i.e., averaging the weights by dividing them by two).

The average annual numbers of marijuana initiates and rates by State, as reported in Chapter 4, were obtained using small area estimation (SAE) methods applied to the pooled 1999-2000 survey data and are, therefore, different from incidence estimates reported in the other chapters. A detailed discussion of the SAE methodology can be found in State Estimates of Substance Use from the 2000 NHSDA (Wright, in press).

In brief, NHSDA State estimates of each substance use measure are produced by combining an estimate of the measure based on the State sample data with the estimate of the measure based on a national regression model applied to local-area county and Census block group/tract-level estimates from the State. The parameters of the regression model are estimated from the entire national sample. Because the 42 smaller (in terms of population) States and the District of Columbia have smaller samples than the eight large States, estimates for the smaller States rely more heavily on the national model. The model for each substance use measure typically utilizes from 50 to 100 independent variables in the estimation. These variables include basic demographic characteristics of respondents (e.g., age, race/ethnicity, and gender), demographic and socioeconomic characteristics of the Census tract or block group (e.g., average family income and percentage of single-mother households), and county-level substance abuse and other indicators (e.g., rate of substance abuse treatment, drug arrest rate, and drug- and alcohol-related mortality rate). Population counts by State and age group are applied to the estimated rates to obtain the estimated number of persons with the substance use characteristic.

Corresponding to each SAE estimate is a 95 percent prediction interval (PI) that indicates the precision of the estimate. The PI accounts for variation due to sampling as well as variation due to the model and is derived from the process that generates the State SAE. There is a 95 percent probability that the true value lies within the interval.

The incidence estimates discussed in this report are based on the combination of two separate measures: (1) the number of marijuana initiates during the past 24 months, and (2) the number of persons who have never used marijuana. Each of these measures is generated independently using SAE, by State and age group. The following formula was used to generate the average annual rate of first use of marijuana for each State:

Average annual incidence rate =
0.5 * {Number of initiates in past 24 months /
[(Number of initiates in past 24 months * 0.5) + Number of persons who never used]}.

Logistic regression procedures were used to (1) determine the characteristics and suspected predictors of recent initiation and (2) examine the relationship between early onset of marijuana use and later drug use patterns (e.g., heavy illicit drug use, alcohol and/or illicit drug dependence or abuse). Odds ratio (OR) estimates derived from logistic regression procedures denote the estimated magnitude of an association between a binary outcome and a covariate. In this report, the p value equal to or less than .05 is considered statistically significant. The OR estimate greater than 1 indicates a positive association between the outcome of interest and the covariate; a value of less than 1 reflects an inverse association.

Table 2.2 Survey Sample Sizes for All Respondents Aged 12 or Older, by Age Group and Geographic Characteristics: 1999 and 2000
Geographic Characteristic	Total		Age Group (Years)
	Total		12-17		18-25		26 or Older
	1999	2000	1999	2000	1999	2000	1999	2000
Total	66,706	71,764	25,357	25,717	21,933	22,613	19,416	23,434
Geographic Division
Northeast	11,830	14,394	4,475	5,102	3,656	4,310	3,699	4,982
Midwest	18,103	19,355	6,530	6,655	6,165	6,236	5,408	6,464
South	21,018	22,041	7,731	7,856	7,189	7,189	6,098	6,996
West	15,755	15,974	6,621	6,104	4,923	4,878	4,211	4,992
County Type
Large metro	25,901	28,744	10,116	10,576	8,121	8,759	7,664	9,409
Small metro	22,612	24,579	8,316	8,505	7,859	8,108	6,437	7,966
Nonmetro	18,193	18,441	6,925	6,636	5,953	5,746	5,315	6,059
Source: SAMHSA, Office of Applied Studies, National Household Survey on Drug Abuse, 1999 and 2000.

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.

* Adobe™ PDF and MS Office™ formatted files require software viewer programs to properly read them. Click here to download these FREE programs now

Table 2.1 Survey Sample Sizes for All Respondents Aged 12 or Older, by Age Group and Demographic Characteristics: 1999 and 2000
Demographic Characteristic	Total		Age Group (Years)
	Total		12-17		18-25		26 or Older
	1999	2000	1999	2000	1999	2000	1999	2000
Total	66,706	71,764	25,357	25,717	21,933	22,613	19,416	23,434
Gender
Male	32,092	34,386	12,798	12,977	10,411	10,716	8,883	10,693
Female	34,614	37,378	12,559	12,740	11,522	11,897	10,533	12,741
Hispanic Origin and Race
Not Hispanic
White only	46,054	49,415	16,90	17,047	14,697	14,991	14,456	17,377
Black only	7,982	8,494	3,297	3,367	2,729	2,711	1,956	2,416
American Indian or Alaska Native only	739	769	273	288	278	270	188	211
Native Hawaiian or other Pacific Islander	232	261	92	92	84	108	56	61
Asian only	2,146	2,393	795	784	765	856	586	753
More than one race	1,072	1,039	483	468	380	352	209	219
Hispanic	8,481	9,393	3,516	3,671	3,000	3,325	1,965	2,397
Adult Education¹
< High school	7,458	8,376	N/A	N/A	4,347	4,771	3,111	3,605
High school graduate	14,845	16,026	N/A	N/A	8,218	8,234	6,627	7,792
Some college	11,692	12,577	N/A	N/A	6,990	6,954	4,702	5,623
College graduate	7,354	9,068	N/A	N/A	2,378	2,654	4,976	6,414
Current Employment¹
Full-time	23,723	26,826	N/A	N/A	11,433	11,984	12,290	14,842
Part-time	7,220	7,567	N/A	N/A	5,184	5,113	2,036	2,454
Unemployed	1,705	1,706	N/A	N/A	1,266	1,237	439	469
Other²	8,701	9,948	N/A	N/A	4,050	4,279	4,651	5,669
N/A: Not applicable. ¹ Data on adult education and current employment not shown for youths aged 12 to 17. Estimates for both adult education and current employment are for persons aged 18 or older. ² Retired, disabled, homemaker, student, or "other." Source: SAMHSA, Office of Applied Studies, National Household Survey on Drug Abuse, 1999 and 2000.