This report presents the first comprehensive analysis of the expanded set of risk and protective factors included in the 1999 National Household Survey on Drug Abuse (NHSDA). Chapter 2 presents the prevalence levels of the risk and protective factor measures using single items and the average scale scores and distributions of the youth scores for the risk and protective factor measures using multiple items. These nationally representative scores by age, race/ethnicity, and gender may provide useful benchmarks for other smaller studies.
Chapter 3 presents the association of these variables with past year marijuana use, revealing the same types of strong associations with marijuana that were exhibited in the report on the 1997 NHSDA (Lane et al., 2001). These relations also are presented separately by racial/ethnic groups and gender and subsequently explored for additional demographic variables, including household income, number of parents in the household, county type, and geographic region. For most of the risk and protective factors, controlling for these demographic variables did not alter the expected relationship between these factors and substance use.
Chapter 4 presents multiple logistic regression models that assess the ability of the enhanced set of risk and protective variables in the 1999 survey to predict youth past year marijuana use (using "prediction" to refer to an association between independent variables and marijuana use in a cross-sectional survey). The results indicate that the explanatory power of the 1999 model was similar to the 1997 model. Among the four domains of risk and protective factors, the peer/individual domain explained the largest amount of variation, with the strongest predictors being participation in antisocial behavior, friends' marijuana use, low perceived risk of marijuana use, and positive attitude toward marijuana use. The finding that the peer/individual domain contained the strongest predictors of youth marijuana use was consistent with the results of the 1997 report.
An implication for future research of the relative explanatory power of the different domains is that constructs from the peer/individual domain should be given a stronger representation (relative to the other domains) in the annual fielding of the survey. However, one should be cautious when using the amount of explained variation as the sole basis for interpreting the relative importance of these domains. For example, Kandel (1996) noted two possible reasons that the influence of peers can be overstated relative to the influence of families in studies such as this. First, parents often play an important role in youths' selection of friends, but cross-sectional surveys, such as the NHSDA, confound these peer selection and family socialization effects. Second, these types of studies rely on perceptions of peer behaviors rather than peer self-reports. The effect of a youth's projection of his or her own attitudes to those of his or her peers may be significant; however, in the context of the combined influence of theindividual and his or her peers, the impact will not be large because the individual variables themselves account for most of the explained variation. For the purpose of balance, some constructs should be included from each of the domains, perhaps based on their explanatory power.
The enhancement of the questions relating to school appeared to improve the explanatory power of the school domain, relative to 1997. It was informative to note that although the total explanatory power of the risk and protective factors remained quite high for both the 1997 NHSDA and the 1999 NHSDA, the enhancement of the Youth Experiences module in the 1999 NHSDA did not increase this explanatory power. In fact, it was slightly lower in 1999 compared with 1997.
Chapter 4 also includes a section on hierarchical modeling, the goal of which was to indicate how this type of modeling can result in richer models, and how those models might lead to a better understanding of substance use (e.g., marijuana use) among youths. These models indicate that most of the total variation in past year use of marijuana among youths aged 12 to 17 occurred at the person level (79 percent), while another 15 percent was present at the family level and 6 percent at the neighborhood level. In the example, it is noted that relatively large percentages of the variation at the community level (58 percent) and family level (40 percent) were explained by the hierarchical model, but a relatively small percentage was explained at the person level (18 percent)implying that those models could be improved at the person level. The example suggests that it would be helpful to know what percentage of person-level marijuana use is explained by the final model of Chapter 4 (Table 4.6). The example, which included only one person-level variable, also suggests that more research is needed to determine how much of the total person-level variation in marijuana use would be explainable if the full set of risk and protective factors were included.
Chapter 5 includes an investigation into whether the decrease in the prevalence rate of youth marijuana use between 1997 and 1999 could be attributed to changes in the risk and protective factors. One limitation was that these analyses were necessarily restricted to risk and protective factors that were measured using the same questions in both the 1997 and 1999 NHSDAs; this requirement limited these analyses to 11 comparable questions. Nevertheless, that set of factors was able to "explain" about 50 percent of the total variation in past year marijuana use in each of those 2 years. To compare the models in the 2 years, the changes in factors were grouped into two sets: changes in the distributions of the risk and protective factors and changes in the individual associations between the risk and protective factors and youth marijuana use. The 11 variables were first examined individually, and for some variables there had indeed been changes in both the distributions and the individual associations with marijuana use between 1997 and 1999. However, the direction of the changes was inconsistent, which made it difficult to reach overall conclusions from those analyses.
In an attempt to obtain a more comprehensive picture of the impact of the risk and protective factors on the change in past year marijuana use among youths, a traditional model was used in which the 1997 and 1999 data were combined, with the goal of assessing how the addition of a "year" variable and the interaction terms of year by each risk and protective factor variable would affect the model. Although the year and a number of the risk and protective variables showed significant main effects, very few of the individual interaction terms were significant. Nevertheless, collectively the addition of the year-by-factor interaction terms resulted in a small, but statistically significant, increase in the explained variation in past year marijuana use. However, that type of analysis did not directly address the impact of changes in the distribution of risk or protective factorsrelative to the impact of changes in the associations between the risk or protective factors and marijuana useon the decrease in youth marijuana use between 1997 and 1999.
In the final analyses of Chapter 5, a new technique is used to partition the change in youth marijuana use between 1997 and 1999 into different components. This partitioning was estimated from models based on the set of 11 common explanatory variables. In the first partitioning, it was confirmed that changes in demographics had little effect on the change in the prevalence of youth marijuana use between 1997 and 1999. The second partitioning was aimed at disaggregation of the effects of the risk and protective factors into changes in the distributions (i.e., the prevalence) of the factors and changes in the strength of the association of those factors with marijuana use. The implications of this disaggregation are tentative at present because the methodology is new, the estimated variance of the estimates is not known, and the number of risk and protective factors that were common to both years was small. It appears, however, that more of the decrease in prevalence rates of marijuana between 1997 and 1999 was due to changes in the strength of the associations between the risk and protective factors and marijuana use than to changes in the distributions of these factors. The pattern of changes in the associations of both risk and protective factors with marijuana use between 1997 and 1999 (see Section 5.5.1) further suggests that the relationship between risk factors and using marijuana was weakened during that period while the relationship between protective factors and not using marijuana was strengthened.
Future analyses that compare multiple years of data will benefit from having larger samples and a larger set of common risk and protective factors between years. Because it appears that these factors can change in their importance and relationship to drug prevalence rates over timein addition to changes in the percentages of youths who evidence each of the factorsit is important to identify those factors that are most related to substance use and to track them over time.
Other research currently under way may help to better understand trends in youth substance use. For example, a report is being prepared that focuses on transition probabilities, including moving from drug nonuse to use and from current drug use to continuation or quitting. That report deals with transitions for both youths and adults and on how these transition probabilities change over time. A second example is a report that is being prepared on the increase in youth marijuana use during the period from 1992 to 1997. The latter report analyzes a number of reasons that were suggested for the increase and tries to draw conclusions about the most likely causes. Included among those analyses are the effects of risk and protective factors and the role of increases in the number of new users of a substance over a number of years.
One goal of future research using the NHSDA will be to monitor a stable set of risk and protective factors of youth substance use and to analyze changes in those factors to determine whether they could account for changes in the prevalence of youth substance use.
This page was last updated on July 17, 2008.