An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Substance Abuse and Mental Health Services Administration . Comparing and Evaluating Youth Substance Use Estimates from the National Survey on Drug Use and Health and Other Surveys [Internet]. Rockville (MD): Substance Abuse and Mental Health Services Administration (US); 2012 Dec.
As noted in Chapter 1 and Appendix A, school-based surveys such as Monitoring the Future (MTF) and the Youth Risk Behavior Survey (YRBS) do not include youths who have dropped out of school. MTF also does not include data from youths who were absent on the day of survey administration at the sampled schools. YRBS makes follow-up attempts to obtain data from youths who were absent on the day of survey administration but nevertheless does not obtain complete coverage of these youths. School dropouts and those who often are absent from school are at increased risk of substance use (Bachman & O'Malley, 1981; Gfroerer et al., 1997; Johnston & O'Malley, 1985). Researchers for MTF and YRBS indicate that the surveys allow inferences to be made for secondary school students (for MTF, this includes 8th graders) or high school students (for YRBS) (Brener et al., 2004b; Johnston, O'Malley, Bachman, & Schulenberg, 2009). However, policymakers and others may use these school-based survey data to make inferences about substance use among youths in the United States as a whole. Depending on the effects of the exclusion of dropouts and frequent absentees from school-based surveys, data from these surveys may not generalize to the population of adolescents as a whole.
Therefore, this chapter uses combined data from the 2002 to 2008 National Surveys on Drug Use and Health (NSDUHs) to examine effects of school dropouts and absentees on substance use estimates. The chapter also examines relationships between the seasonality of NSDUH interviews (i.e., whether they were conducted in January through June or in July through December of a given survey year) and estimates of substance use among youths.
NSDUH defined persons aged 12 to 25 as being enrolled in school if they reported that (1) they were enrolled in or attending school or (2) they did not report being enrolled in school because they were on vacation or break from school, and they planned to return to school when their vacation was over. Persons who were defined as enrolled in school were asked to report the grade that they were currently in or the grade that they would be in when they returned to school after their vacation.
Conversely, NSDUH defined persons aged 12 to 25 as not being enrolled in school if they reported that (1) they were not enrolled in or attending school and were not on vacation from school or (2) they were on vacation from school and did not plan to return to school when their vacation was over. Persons aged 12 to 25 who were not enrolled in school were asked whether they had received a high school diploma or GED (general educational development) certificate of high school completion. A school dropout was defined to have completed less than the 12th grade, not to be currently enrolled in school, and not to have received either a high school diploma or a GED certificate.
Analyses in this section focus on estimates for youths who were enrolled in a given grade in school and youths who had dropped out of school but would have been at that corresponding grade level if they had not dropped out. Estimates for youths at a given grade level were made for the following groups based on interviews conducted in January through December: (1) nondropouts only; (2) dropouts only (i.e., who were assumed to be at that grade); and (3) both dropouts and nondropouts. Use of the full 12 months of data from a given survey year improved the precision of estimates based only on dropouts. Therefore, estimates excluding dropouts at a given grade may differ from corresponding estimates presented in Chapter 2 for youths who were in school at that grade.
Because dropouts by definition were not asked to report their current grade in school, available NSDUH data were used to assign assumed grade levels to these dropouts. Specifically, NSDUH includes information on dropouts' current ages, the highest school grade they completed, and their age when they left school. Appendix B gives additional details about the assumptions and procedures for assigning grade levels to dropouts.
Dropouts whose ages were within narrowly defined ranges relative to both their highest grade and their expected next highest grade if they had remained in school (see Table B.2 in Appendix B) were considered to have dropped out of school relatively recently. These youths were assigned to be at the next highest grade level in school based on their highest completed grade and current age. For example, if a dropout completed the 10th grade and was aged 15 to 18, the inference was made that the youth would have been at the 11th grade level if he or she were still in school.
If youths had dropped out of school less recently or if they were older than the typical age range for youths who had completed a given grade, their current age, highest grade completed, and age when they left school were used to assign a grade level. The difference between the current age and age when a dropout left school was added to the current grade level, and then the assumed grade level was further incremented by 1 year. For example, if a dropout's last grade was the 9th grade, the base assumption was that the dropout's next grade level would be the 10th grade. Thus, if there was no difference between the dropout's current age and his or her age when the dropout left school, the dropout was assumed to be at the 10th grade level; this procedure allowed for assignment of grade levels to youths who left school at their current age but were older than most youths in that grade when they dropped out. Extending this example, if there was a 1-year difference between the two ages, the dropout was assumed to be at the 11th grade level, and if there was a 2-year difference, the dropout was assumed to be at the 12th grade level. If dropouts were aged 17 or 18 and they did not report an age when they had left school, it was assumed that they would be at the 12th grade level. In addition, the current grade was capped at the 12th grade level if this calculation procedure yielded an assumed grade level past the 12th grade. Dropouts for whom these procedures did not yield an assumed grade level within the 8th through 12th grades (as if they had remained in school) were excluded from further analysis.
Based on these procedures, 97 percent or more of the youths at the 8th, 9th, or 10th grade levels were identified as being in school (Figure 3.1). In particular, 99.7 percent of youths at the 8th grade level were classified as being in school, and only 0.3 percent were classified as dropouts who would be at this grade level if they were still in school. Dropouts would comprise more than one in five of the older adolescents assigned to the 12th grade level.
Student in School and Dropout Composition at the 9th through 12th Grade Levels: 2002 to 2008 NSDUH.
Among students in the 12th grade, 52.7 percent were aged 17, 35.9 percent were aged 18, and about 6 percent were aged 19 (4.8 percent) or 20 (1.3 percent) (Figure 3.2). Among dropouts who were assigned to the 12th grade level based on these procedures (i.e., the highest proportion of dropouts assigned to a given grade level), less than 5 percent were aged 17 or younger. More than half were aged 19 (37.0 percent) or 20 (37.7 percent), and 21.3 percent were aged 18.
Age Distributions among Students in School and Dropouts at the 12th Grade Level: 2002 to 2008 NSDUH.
Two sets of substance use estimates at a given grade were compared when estimates had acceptable precision (see Appendix B): (1) estimates excluding dropouts (shown in tables as "Without Dropouts") versus those for dropouts only; and (2) estimates including both dropouts and youths in school (shown in tables as "With Dropouts") versus those excluding dropouts.
Elsewhere in this section, dropouts who were assumed to be at a given grade level may simply be referred to as being at that grade level for the sake of brevity. For the first set of comparisons, estimates were mutually exclusive. For the second set of comparisons, estimates were not mutually exclusive; statistical testing took this into account.
Dropouts who were assigned to a given grade level typically were more likely to be substance users than their counterparts who were in school (i.e., nondropouts). Highlights are presented for comparison of selected substance use estimates among dropouts and nondropouts. Notable exceptions to patterns of higher estimates among dropouts than among nondropouts also are presented.
In addition, for the subgroup consisting only of school dropouts, substance use estimates did not have acceptable precision for dropouts who were assigned to the 8th grade level. Most or all substance use estimates specifically for dropouts also did not have acceptable precision for females who were assigned to the 9th grade level or for several racial/ethnic groups at the 9th through 12th grade levels. However, estimates for dropouts who were assigned to the 12th grade level often had adequate precision for whites, blacks, and Hispanics.
Past Year Alcohol Use among Students and Dropouts at the 9th through 12th Grade Levels: 2002 to 2008 NSDUH.
Past Year Marijuana Use among Students and Dropouts at the 9th through 12th Grade Levels: 2002 to 2008 NSDUH.
This section discusses the effects on prevalence estimates for youths in the 8th through 12th grade levels when data from dropouts at those grade levels were included along with data from youths who were in school. These estimates that included data from dropouts at a given grade level (subsequently referred to as "with dropouts") are compared with estimates at that grade that excluded dropouts (subsequently referred to as "without dropouts"). Especially at lower grade levels—in which there are very few dropouts, and students in school comprise the large majority of youths at those grade levels (Figure 3.1)—the estimates with and without dropouts were based on almost the same set of respondents. Consequently, statistical test results comparing estimates with and without dropouts could show significant differences despite small differences in the actual percentages.
Although dropouts who were assigned to a given grade level often were more likely to be past year or current substance users than their counterparts who were in school, inclusion of data from dropouts had little effect on estimated percentages of children at the 8th or 9th grade levels who were substance users, even if differences were statistically significant. In contrast, inclusion of data from dropouts showed more notable increases for some estimated percentages of substance use at the 12th grade level. These general findings are consistent with what Gfroerer and colleagues (1997) hypothesized—that inclusion (or exclusion) of data from dropouts would have the smallest effect on estimates for 8th graders and the greatest effect on estimates for 12th graders. These findings also are consistent with U.S. Department of Education statistics on dropout rates. In the 2007-2008 school year, for example, the dropout rate (defined as the number of dropouts from a given grade divided by the number of students in that grade at the beginning of the school year) among all reporting States was 3.0 percent for the 9th grade, 3.6 percent for the 10th grade. 4.0 percent for the 11th grade, and 6.1 percent for the 12th grade (Stillwell, 2010).
Even if dropouts do not appreciably affect estimates of the percentages of youths in a given grade level, data from dropouts may be important for estimating the sizes of different populations of substance users (i.e., estimated numbers of users) for youths in at least some age categories. In particular, information about the number of users of different substances at each grade level can be useful to policymakers and planners for developing more effective substance abuse prevention and dropout prevention policies and programs. This type of information also can be useful to practitioners in identifying specific needs for prevention, intervention, or treatment services for adolescents.
Highlights are presented in this section for comparison of selected substance use estimates with and without dropouts. When an estimate that includes data from dropouts who were assigned to a given grade is significantly different from the corresponding estimate without dropouts (i.e., for students who were in school at that grade), the magnitude of the difference is presented; this is done even for statistically significant differences that were relatively small in magnitude. The purpose of presenting this information is to quantify the impact on adolescent substance use estimates when school dropouts are included or excluded.
At the 8th and 9th grade levels, inclusion of dropouts had little or no effect on percentages of youths who were past year or current users of cigarettes (Tables C.3.1B and C.3.2B), smokeless tobacco (only current use compared; Table C.3.3B), cigars (only current use compared; Table C.3.4B), alcohol (Tables C.3.5B and C.3.6B), marijuana (Tables C.3.8B and C.3.9B), or cocaine (Tables C.3.10B and C.3.11B). Estimates with and without dropouts were not significantly different for the following estimates: current use of smokeless tobacco, current use of marijuana, past year and current use of cocaine, past year use of inhalants, and lifetime use of Ecstasy at the 8th grade level, as well as current use of smokeless tobacco and lifetime use of heroin at the 9th grade level. However, differences between estimates with and without dropouts that were significantly different from a statistical standpoint at the 8th and 9th grade levels were small or showed no apparent differences in the estimates.
Consistent with the estimated percentages at the 8th grade level, inclusion of dropouts had little effect on estimates of the numbers of youths at this grade level who were past year or current users of cigarettes (Tables C.3.1A and C.3.2A), current users of smokeless tobacco (Table C.3.3A), current users of cigars (Table C.3.4A), past year or current users of alcohol (Tables C.3.5A and C.3.6A), current binge alcohol users (Table C.3.7A), past year or current users of marijuana (Tables C.3.8A and C.3.9A), or past year or current users of cocaine (Tables C.3.10A and C.3.11A), despite all differences in estimated numbers of users being statistically significant except for past year cocaine use and current use of smokeless tobacco and cocaine. Similar minimal effects on estimated numbers of youths at the 9th grade level who were past year or current users of these substances were observed when data from dropouts were included, although these differences were statistically significant.
Inclusion of dropouts had little effect on percentages of youths at the 10th and 11th grade levels who were current users of smokeless tobacco (Table C.3.3B) or cigars (Table C.3.4B), past year or current alcohol users (Tables C.3.5B and C.3.6B), current binge alcohol users (Table C.3.7B), past year or current marijuana users (Tables C.3.8B and C.3.9B), or past year or current users of cocaine (Tables C.3.10B and C.3.11B); except for current smokeless tobacco use, however, differences in these estimated percentages with and without dropouts were statistically significant. Inclusion of data for dropouts had a more appreciable effect on estimated numbers of users at the 11th grade level.
Annual Average Percentages of Cigarette, Alcohol, Binge Alcohol, and Marijuana Users Who Were Students in School or Dropouts at the 12th Grade Level: 2002 to 2008 NSDUH.
Annual Average Percentages of Cocaine and Ecstasy Users Who Were Students in School or Dropouts at the 12th Grade Level: 2002 to 2008 NSDUH.
As noted previously, school-based surveys are likely to miss absentees, and students who often are absent from school may be at increased risk of substance use. Therefore, this section uses NSDUH data to examine the effects of including or not including absentees in substance use estimates among adolescents who were enrolled in school. NSDUH respondents aged 12 to 25 who reported that they were enrolled in school (see Section 3.1) were asked whether they were full-time or part-time students. Those who reported that they were full-time students first were asked to report the number of days that they missed school in the past 30 days because they were sick. These respondents also had the option to report that school was not in session during the past 30 days. If respondents did not report that school was not in session, they also were asked to report the number of days they missed school in the past 30 days because they skipped or "cut" school.
Valid values for reports of the number of days that full-time students were absent because they were sick or because they skipped school in the past 30 days ranged from 0 to 30. These continuous values were recoded into the following categories for days missed from school: (1) 0 days; (2) 1 to 2 days; (3) 3 to 5 days; and (4) 6 or more days.
In principle, days that full-time students missed from school because they were sick or because they skipped school could overlap. For analyses discussed in this section, however, a "composite" variable for overall days missed from school in the past 30 days was created that assumed that the number of days that youths were absent for these different reasons did not overlap; review of the data indicated that students tended to be absent for one reason or another (but not both) or to report small numbers of days that they were absent for each reason (see Section B.3.2 and Table B.5 in Appendix B of this report). To create a measure of overall absences, therefore, the numbers of days that students reported they missed school because they were sick or because they skipped school were treated as additive. The variable for the total number of days that students were absent was top-coded at 30 days for sums greater than 30. The same four categories that were described previously also were created for this overall absence variable.
By definition, this variable for the overall number of days that students were absent from school in the past 30 days for any reason did not include data from respondents who reported that they were part-time students. In creating this overall variable for the number of absences, data also were not included from respondents if (1) they reported that school was not in session in the past 30 days or (2) they had missing data (e.g., responses of "don't know" or "refused") either for the number of days they missed school because they were sick or the number of days they missed school because they skipped school (or both). Section B.3.2 in Appendix B contains additional information about how the number of days that youths were absent from school was defined for these analyses.
Estimates of substance use were made for 8th through 12th grade full-time students and by gender at these grades according to the categories for the number of days that they were absent from school in the past 30 days. Estimates were made for absences because of sickness, absences because youths skipped school, and absences for either reason in the past 30 days; these latter absences are subsequently referred to as absences for "any" reason. Estimates also were made at each grade for these absence measures regardless of substance use. For consistency with NSDUH estimates presented in Chapter 2 for youths who were school, analyses presented in this section focused on youths who were interviewed in January through June of their survey year. Even with the combined 2002 through 2008 NSDUH data and even when data were run on the full sample (rather than using data only from respondents who were interviewed in January through June), many estimates by race/ethnicity did not meet the precision criteria for publication (see Section B.2.3 in Appendix B). Therefore, absence data are not presented by race/ethnicity.
The second main analytic approach for investigating the effect of absences on substance use estimates involved evaluating how the inclusion or exclusion of absentees in surveys of students could affect substance use estimates. This was done by comparing standard NSDUH estimates for full-time students with adjusted estimates that accounted for how frequently respondents reported missing school in the past month; the more days that youths reported being absent from school, the greater the likelihood that they would have been absent on the day of a school-based survey. To accomplish this evaluation, the NSDUH analysis weights of respondents who had defined absence data were multiplied by the proportion of days that respondents were inferred to have been in school in the past 30 days. These proportions were derived from the number of days that respondents reported being absent from school in that period. Thus, youths' relative weighted contribution to the overall estimate was decreased according to the number of days they reported being absent; the more days that they were absent, the less they contributed to the total estimate.
Although youths could report that they missed school for a total of 30 days in the past 30 days because they were sick or because they skipped school (or both), school was not likely to be in session for every day in that period. Therefore, the number of actual days that school was assumed to be in session in the past 30 days was set at 20 days; this was set as the base for adjustment of the weights. If the total number of days that youths were absent from school because they were sick or because they skipped school was calculated to be greater than 20 days, the total was reset to a maximum of 20 days for this analysis.
Under this approach, an adjustment factor was calculated for the analysis weights for youths aged 12 to 20 who were in the 8th through 12th grades by first subtracting the calculated total number of days that youths were absent from school from 20. This difference then was divided by 20 (i.e., the assumed total number of school days in the past month) to yield a proportion. A youth's analysis weight than was multiplied by this adjustment factor:
where WGTAdj = adjusted weight for absences, WGTOrig = respondent's original analysis weight, and DAYSAbs = number of days absent from school in the past 30 days (capped at 20 days).
For example, if a youth was not absent for any of the 20 school days in the past month, the adjustment factor was equal to 1, and no adjustment was made to that respondent's analysis weight. At the other extreme, if a youth was absent on 20 out of 20 school days, the adjustment factor was equal to zero, and the youth then had an adjusted weight of zero. At the middle of the range, the weight of a youth who was absent on 10 out of 20 days would have his or her weight reduced by half. Section B.3.2 in Appendix B contains additional information about these adjustment procedures.
New estimates of substance use were generated using these adjusted analysis weights; these are referred to in this section and in tables as "adjusted" estimates. Based on the weight adjustment procedures, the population of inference for these adjusted estimates would be full-time students aged 12 to 20 in the 8th through 12th grades who were likely to be in school on a given school day. These adjusted substance use estimates were compared with estimates in which the analysis weights had not been adjusted for absences; these are referred to in this section and in tables as "unadjusted" estimates.
For consistency with other school-based data in this report, estimates are presented for persons aged 12 to 20 who were interviewed in January through June. Tables present unadjusted and adjusted prevalence estimates. Data were restricted to full-time students, and youths with missing data for the number of days they were absent were excluded from the analysis. Consequently, the unadjusted estimates presented in this section may differ from corresponding estimates at a given grade or for a subgroup within a grade that were presented in Chapter 2 and in Section 3.1. Although these estimates may differ from corresponding estimates elsewhere in the report, these exclusions allowed for statistical testing of differences between the unadjusted and adjusted estimates because the corresponding estimates were based on the same samples.
As noted previously, NSDUH respondents were asked about the number of days they missed school in the past 30 days because they were sick and the number of days they missed school because they skipped school. Therefore, discussion of substance use according to the number of days that students were absent from school focuses on estimates of past year use (which are proximal to the past 30 days) or current use.
Students in a given grade who missed school on 6 or more days in the past 30 days because they were sick often were more likely to be past year or current substance users than their counterparts who did not miss any days of school because they were sick. This association between substance use and missing a high number of school days because of sickness was especially consistent across different substances for 8th and 9th graders but was less consistent by the 12th grade. Highlights are presented in this section for selected estimates of past year and current substance use according to the number of days that youths missed school because they were sick.
Past year or current substance use also was associated with youths missing school in the past 30 days because they skipped or "cut" school or they just did not want to be there. Among 9th through 12th graders overall, for example, students who missed school on 6 or more days in the past 30 days because they skipped school often were more likely to be past year or current substance users than their counterparts who did not miss any days of school for this reason. This section presents highlights for selected estimates of past year and current substance use according to the number of days that youths missed school because they skipped school.
Consistent with findings for absences because students were sick or because they skipped school, students in a given grade who missed school on 6 or more days in the past 30 days for either reason often were more likely to be past year or current substance users than their counterparts who did not miss any days of school for either reason. Highlights are presented in this section for selected estimates of past year and current substance use according to the total number of days that youths missed school for either reason.
As noted in the preceding sections, youths who were absent from school for several days in the past 30 days usually were more likely to be past year or current substance users than their counterparts who did not miss any days of school. The more days that youths are absent from school, the less likely that school-based surveys such as MTF or YRBS will be to capture information about the substance use characteristics of these youths. However, the effects of excluding absentees on school-based survey estimates will depend on the relative numbers of youths who are likely to be absent from school on a given day and the size of any differences in substance use prevalence between youths who are likely to be in school and those who are likely to be absent. As was the case with estimates for dropouts and nondropouts in the lower grades, for example, even large differences in prevalence between youths who are likely to be in school and those who are likely to be absent could have minimal effects on substance use estimates if the size of the subpopulation of youths that is likely to be in school on a given day is considerably larger than the size of the subpopulation that is likely to be absent.
This section presents the results from an analysis that used adjusted analysis weights to account for the total number of days that full-time students were absent in the past 30 days (i.e., because of sickness or skipping school) in order to replicate the consequences of conducting a school-based study on a single day and not including students who were absent on that day. As discussed previously, these adjustment procedures reduce the contributions of data from youths according to their frequency of being absent from school in estimating of the prevalence of substance use among the school-based population. Given the higher prevalence estimates for youths who are frequently absent from school than for those who are less likely to be absent, therefore, a given substance use estimate for youths based on these adjustment procedures would be expected to be lower than the corresponding unadjusted estimate (which would include data from youths who were interviewed in their homes but frequently were absent from school).
Because youths' analysis weights could be adjusted only if they had fully defined absence data, youths with missing data for the number of days absent were excluded from analysis. This exclusion was necessary to allow statistical testing between the estimates from the unadjusted and adjusted analysis weights to be based on the same samples. In many instances, however, this exclusion also yielded almost a 100 percent overlap between the unadjusted and adjusted data but with different weights.3 As for the previous comparisons between estimates that included dropouts and those that did not include dropouts, high correlations between unadjusted and adjusted estimates resulted in most differences being statistically significant, even if differences were small. Differences between unadjusted and adjusted estimates were not statistically significant only for a limited number of smokeless tobacco, cocaine, and Ecstasy use estimates among males or females in grades earlier than the 12th grade. Therefore, this section discusses the relative differences between unadjusted and adjusted estimates without emphasizing statistical significance of differences.
Highlights of the results are presented in the remainder of the section. Selected annual average estimates are presented in separate sections for (1) 8th and 9th graders, (2) 10th and 11th graders, and (3) 12th graders. Consistent with the findings reported previously for dropouts, the adjustment procedures had less of an effect on estimates for youths in the 8th and 9th grades and had a greater effect on estimates for youths in the 12th grade.
Johnston and colleagues (2009) hypothesized that if rates of absenteeism were comparable across survey years, exclusion of data from absentees would not affect trends in substance use. This assumption would hold true if the trends for frequent absentees were similar to the rest of the student population or if the proportion of absentees was so small that differences in trends for absentees did not affect overall trends. To test this assumption, estimates of lifetime, past year, and current use of cigarettes, alcohol, marijuana, cocaine, and inhalants were run for 12th graders in each year from 2002 to 2008. In each year, estimates were run that were unadjusted or adjusted for absences. For consistency with estimates discussed in preceding sections and with the period in which school-based surveys are administered, estimates in this section focus on interviews conducted in January through June of a given survey year.
Adjusting the weights for 12th graders who were more likely to be absent on a given day within a 30-day period from January to June did not affect differences in estimates between most consecutive survey years. If an unadjusted estimate was significantly different from the estimate in the next survey year, the difference between the same 2 years also was significant for the adjusted estimate. If an unadjusted estimate did not differ significantly from the estimate in the next survey year, then the adjusted estimates often were not significantly different between the same 2 years.
For some estimates, however, the unadjusted estimates for 12th graders who were interviewed in January through June did not differ significantly between consecutive survey years, but the adjusted estimates were significantly different. In addition, one set of unadjusted estimates showed significant differences between a pair of survey years, but the corresponding adjusted estimates did not.
Trends in Lifetime Alcohol and Cigarette Use among 12th Graders: Percentages Unadjusted and Adjusted for Absences, 2002 to 2008 NSDUH.
Trends in Lifetime Marijuana and Cocaine Use among 12th Graders: Percentages Unadjusted and Adjusted for Absences, 2002 to 2008 NSDUH.
A decrease in the lifetime prevalence of cigarette use relative to the preceding year could be explained by differences in substance use among cohorts of youths in different grades. Specifically, the lifetime prevalence among 12th graders could decrease between consecutive years as cigarette users in the 12th grade cohort in the earlier year move out of the 12th grade and are replaced with a new cohort of 12th graders with a higher proportion of lifetime nonusers of cigarettes.
Although Section 3.2.2 did not discuss estimates of lifetime substance use, 12th graders who frequently were absent from school were more likely than those who did not miss any school to be lifetime cigarette users. For example, an annual average of 61.9 percent of 12th graders who were absent on 6 or more days because of illness or skipping school were lifetime cigarette users. In comparison, an annual average of 40.9 percent of 12th graders who were not absent on any days were lifetime cigarette users.
Consequently, the greater likelihood that frequent absentees in the 12th grade would be lifetime cigarette users could have contributed to the decrease in the lifetime prevalence from 2004 to 2005 in the adjusted estimates despite the change not being significant for the unadjusted data. Because frequent absentees contributed less to adjusted prevalence estimates than students who missed few or no days of school did, the adjustment procedures yielded a 5.2 percentage point decrease in the lifetime prevalence of cigarette use from 2004 to 2005 (Table C.3.34 and Figure 3.7); the unadjusted estimates showed a 4.8 percentage point decrease over this same period.
For cocaine, if an increase in the lifetime prevalence of use relative to the preceding year can be explained by increases in the initiation of use of that substance in the past year, then a significant increase in the past year prevalence of cocaine use also would be expected. For past year cocaine use in 2002 and 2003, the unadjusted estimates were 4.2 percent in 2002 and 4.7 percent in 2003 (Table C.3.35 and Figure 3.9). The corresponding adjusted estimates were 3.5 percent in 2002 and 4.3 percent in 2003. Although the unadjusted and adjusted estimates of past year cocaine use among 12th graders were in the direction of an increase from 2002 to 2003, neither showed a significant difference in estimates between these 2 years. The lack of a significant difference in past year prevalence of cocaine use between 2002 and 2003 is consistent with the lack of a significant difference in lifetime prevalence between 2002 and 2003 for the unadjusted estimates but the significant increase in lifetime prevalence for the adjusted estimates is inconsistent with this pattern for the past year estimates.
Trends in Past Year Cocaine Use among 12th Graders: Percentages Unadjusted and Adjusted for Absences, 2002 to 2008 NSDUH.
As for lifetime cigarette use, 12th graders who were absent on a total of 6 or more days because of illness or skipping school were more likely to be lifetime cocaine users than those who were not absent on any days (annual averages of 11.1 vs. 3.6 percent). As noted in Section 3.2.2, 8.3 percent of 12th graders who missed school on 6 or more days for any reason and 2.2 percent of those who did not miss any days of school were past year cocaine users.
These data suggest an explanation for the finding of an increase in the adjusted lifetime cocaine prevalence among 12th graders from 2002 to 2003 despite the lack of significant differences in the past year prevalence from 2002 to 2003 for the adjusted and unadjusted estimates. Specifically, adjusting the weights of 12th graders for absences yielded a 2.5 percentage point increase in the lifetime prevalence of cocaine use from 2002 to 2003 (Table C.3.34 and Figure 3.8). In contrast, the lower prevalence of past year cocaine use in the 12th grade yielded only a 0.8 percentage point increase in the adjusted prevalence of past year use from 2002 to 2003 (Table C.3.35 and Figure 3.9). Thus, the samples of 12th graders in 2002 and 2003 may have had sufficient statistical power to detect the 2.5 percentage point increase in lifetime prevalence from 2002 to 2003 for the adjusted estimates but did not have sufficient power to detect a 0.8 percentage point increase for the adjusted past year prevalence estimates.
The remainder of the section presents highlights for the effects of absentees on trends in past year and current use among 12th graders.
Trends in Past Year Alcohol and Cigarette Use among 12th Graders: Percentages Unadjusted and Adjusted for Absences, 2002 to 2008 NSDUH.
Trends in Past Year Marijuana and Inhalant Use among 12th Graders: Percentages Unadjusted and Adjusted for Absences, 2002 to 2008 NSDUH.
Trends in Current Cigarette Use among 12th Graders: Percentages Unadjusted and Adjusted for Absences, 2002 to 2008 NSDUH.
The Monitoring the Future (MTF) and the Youth Risk Behavior Survey (YRBS) survey students in the spring of the school year. If youths who have used different substances vary in their likelihood of using them during different times within a year (i.e., seasonal variation), then estimates of current use among youths could vary depending on whether data collection occurred during or shortly after a period when they were more likely to use a substance or they were interviewed after a month or more had elapsed since their most recent use.
However, estimates for substance use behaviors in which certain youths engage fairly regularly might show less variation depending on when they were interviewed. For example, an older adolescent who has been a daily cigarette user for more than a year might be likely to report current cigarette use regardless of whether he or she was interviewed in the first or second half of a survey year or even regardless of the specific month in which he or she was interviewed. Any seasonal variation that exists also could have a minimal effect on year-to-year trends in estimates of current use, as long as the variation is constant across years, or seasonal variations are small. Similarly, small seasonal variations could have a minimal effect on estimates of current use at a single point in time.
Survey questions about use in the past 12 months, in principle, could compensate for any seasonal variation in adolescents' substance use over the course of a year by including times of the year when adolescents might be more likely to use various substances in addition to times of the year when they may be less likely to use substances. However, accurate recall of substance use in the past 12 months could decay if the data collection period is relatively distant from when the most recent use occurred in the past 12 months. Especially for respondents who are infrequent users of a substance, recall of use in the past year is likely to be easier for use that occurred more recently relative to the interview date. In addition, if youths in a given half of the year are more likely than those in the other half to have used a substance last in the period close to the boundary between the past year and lifetime periods (e.g., in the past 11 months), they may be more likely to misreport their use as not having occurred in the past year.
Therefore, a limited set of estimates were run for 12th graders who were interviewed in the first half of the year (January through June) and for those who were interviewed in the second half of the year (July through December). These estimates were run as part of comparisons of annual trends among 12th graders. Estimates for the first and second halves of the year are presented for individual survey years rather than as annual average estimates based on combined 2002 through 2008 NSDUH data. In addition, estimates for 12th graders that have not been adjusted for absences and estimates that have been adjusted for absences are shown for the first and second halves of each survey year.
This section presents two sets of findings for 12th graders. First, trend data are compared across years, holding the period of data collection constant (i.e., January through June or July through December in each year). For example, estimates in 2002 based on surveys in January through June are compared with estimates in 2003 for that same data collection period. Thus, these comparisons examine whether trend data are comparable, as long as data were consistently collected in the first or second half of each year.
In addition, estimates for 12th graders surveyed in the first or second halves of a given calendar year would be based on different cohorts of youths. For example, 12th graders who were interviewed in July through December would have been 11th graders if they had been interviewed in January through June; that is, they would represent the cohort of youths who were 11th graders during the first half of the year. In addition, youths in NSDUH who are interviewed during their summer break are asked to report the grade they will be in when they return. Therefore, the second set of findings in this section compares estimates for 12th graders in a given year who were interviewed in July through December with estimates for 12th graders who were interviewed in January through June of the next calendar year. For example, estimates for 12th graders who were interviewed in July through December of 2002 were compared with estimates for 12th graders who were interviewed in January through June of 2003 because these groups represent members of the same cohort of 12th graders.
With few exceptions, trends in lifetime use of cigarettes, alcohol, marijuana, and inhalants among 12th graders showed no significant differences between consecutive years regardless of whether estimates were based on interviews conducted in January through June of a given year or in July through December (Table C.3.34). This pattern held both for estimates that were not adjusted for absences and those that were adjusted.
Trends for lifetime use of cocaine among 12th graders showed more variability depending on whether estimates were based on data from the first or second halves of the year or whether estimates were unadjusted or adjusted for absences (Table C.3.34).
Past year use of cocaine and inhalants did not show any significant changes between consecutive years regardless of the data collection period (Table C.3.35). Trends in past year use of cigarettes, alcohol, and marijuana among 12th graders showed some variation according to whether data were collected in the first or second halves of the year. For the most part, however, trends were comparable for past year use of these substances regardless of the data collection period.
For current use of marijuana, there were no significant differences between consecutive years regardless of whether data were collected in the first or second halves of the year (Table C.3.36). Except for a decrease from 2007 to 2008 for unadjusted (but not adjusted) estimate for July through December, most data for current alcohol use showed no significant changes between consecutive years. Trends for current use of cigarettes, cocaine, and inhalants among 12th graders showed more variation according to whether data were collected in the first or second halves of the year.
Most estimates of lifetime use within a cohort of 12th graders did not differ according to whether data were collected in the first or second half of the year (Table C.3.34). The exceptions were for comparison of lifetime marijuana and cocaine use based on data from 12th graders who were interviewed in July through December 2003 and those who were interviewed in January through June 2004. The unadjusted and adjusted lifetime prevalence estimates were greater for January through June 2004 than for July through December 2003.
Except for the cohort of 12th graders in 2003 and 2004 and for the cohort of 12th graders in 2006 and 2007 for one substance (inhalants), estimates of past year use within a cohort of 12th graders did not differ according to whether data were collected in the first or second half of the year (Table C.3.35).
The cohort of 12th graders in the second half of 2003 and the first half of 2004 also had different estimates of current use of cigarettes, alcohol, and inhalants for both unadjusted and adjusted estimates (Table C.3.36). As for lifetime and past year use of cigarettes and alcohol, estimates from the first half of 2004 were greater than the estimates from the second half of 2003. However, estimates of current use of inhalants were lower in the second half of 2004 than in the first half of 2003.
Percentage of 12th Graders within a Given School Year Who were Past Month Alcohol Users Based on the Period of Data Collection in NSDUH, 2002 to 2008.
Although there was a general—but often not statistically significant—pattern in the direction of higher estimates in January through June than in July through December, the lack of significant differences between the halves of the year for 12th graders might be attributable to smaller sample sizes when the estimates were limited to single years of data and further subdivided into the two halves of the year. Therefore, annual average estimates for 12th graders in the two halves of the year were run for the combined 2002 to 2008 data to improve the precision of estimates. Estimation of annual averages was limited to data that were not adjusted for absences. A total of six cohorts were included in the annual averages. These annual averages did not include data from 12th graders at the two extremes (i.e., January through June 2002 and July through December 2008) because these two groups did not include complete cohorts; for example, the corresponding group in the same cohort as 12th graders who were interviewed in July to December 2008 would consist of 12th graders in January to June 2009.
As noted in the introduction to this chapter, school-based surveys such as MTF and YRBS do not include data from dropouts and absentees (or do not include data from chronic absentees, in the case of YRBS), two groups that are likely to have a higher prevalence of substance use. Results of the analyses in this chapter are consistent with other literature (e.g., Gfroerer et al., 1997; Stillwell, 2010) that suggests data from dropouts will have a relatively small effect on the overall percentages of 8th and 9th graders who are estimated to be substance users. Examination of data on numbers of users also suggests that inclusion of dropouts will have small effects on the numbers of youths in these two grades who are estimated to be substance users.
Consistent with what Gfroerer and colleagues (1997) hypothesized, inclusion of data from dropouts had greater effects on the percentages of persons at the 12th grade level who were estimated to be substance users than for youths in the 8th and 9th grades. For the most part, however, estimated percentages of persons at the 12th grade level or of subgroups of persons at this level with data from dropouts showed fairly modest increases relative to data without dropouts, often by a factor of about 10 percent. For example, if 20 percent of youths in a given grade were estimated to be users of a particular substance when data from dropouts were not included, an increase by a factor of 10 percent would yield an estimate of 22 percent. Even if such an increase were statistically significant, a change from 20 to 22 percent, for example, likely would be of limited practical significance for making decisions about substance use prevention policy or substance use prevention programming.
However, inclusion of data from dropouts had a greater effect on the percentages of persons at the 12th grade level or certain subgroups at this grade level who were current cigarette users, current alcohol users, or current binge alcohol users. Specifically, data from dropouts raised the overall percentages of 12th graders who were estimated to be current cigarette users by a factor of 30 percent. Inclusion of data from dropouts raised the percentage of 12th grade blacks who were current cigarette users by about 50 percent: from 14.0 percent without dropouts to 21.1 percent with dropouts. Similarly, the estimated percentage of 12th grade blacks who were current alcohol users increased by about 20 percent when data from dropouts were included; the estimated percentage of 12th grade blacks who were current binge alcohol users increased by about 40 percent when data from dropouts were included.
Inclusion of data from dropouts also had a more substantial effect on percentages of 12th graders who were estimated to be users of substances other than tobacco, alcohol, or marijuana. Specifically, inclusion of data from dropouts increased the estimated percentage of 12th graders who were lifetime users of cocaine by about 50 percent overall and by about 30 to 50 percent for certain subgroups of 12th graders. In addition, inclusion of data from dropouts increased the percentage of 12th graders who were estimated to be lifetime Ecstasy users by about 40 percent overall and by about 40 to 60 percent for subgroups of 12th graders. Given the small percentages of 12th graders who were lifetime heroin users (0.6 percent based on data without dropouts and 1.7 percent including data from dropouts), the estimated percentage of persons at the 12th grade level who were lifetime heroin users appeared to be particularly sensitive to the effects of excluding or including data from dropouts.
Furthermore, data from dropouts significantly affected estimates of the numbers of substance users in the 11th and 12th grade levels. Information on numbers of substance users is important for estimating the size of the adolescent population needing early intervention or treatment services. Data on estimated numbers of users among youths in a given grade and among dropouts assigned to that grade suggest that in-school surveys would capture the majority of youths in the 11th grade category who were current users of cigarettes, current alcohol users, current binge alcohol users, past year and marijuana users, and lifetime cocaine users. Conversely, these data for 11th graders suggest that about 10 to 13 percent of these users of cigarettes, alcohol, or marijuana would be missed for youths at that grade level through school-based surveys. In addition, exclusion of data from dropouts would appear to miss more than 20 percent of the 11th graders who were lifetime cocaine users.
Data from dropouts had an even greater effect on numbers of 12th graders who were estimated to be users of different substances. Specifically, dropouts would appear to comprise about 40 percent of the current cigarette users, about 25 percent of the current alcohol users, about 30 percent of the current binge alcohol users, and about 30 percent of the current marijuana users at the 12th grade. Moreover, inclusion of data from dropouts would appear to nearly double the estimated numbers of 12th graders who were lifetime users of cocaine and Ecstasy relative to the estimated numbers without dropouts.
As noted in the introduction to Section 3.1, however, assumptions needed to be made about the grades that dropouts would be in if they were still in school. Therefore, the conclusions to be drawn will depend on how valid these assumptions are for youths at different grade levels.
Findings from this analysis also may underscore the need for a certain degree of caution about the kinds of inferences to be made about the population of older adolescents as a whole from school-based surveys. In particular, MTF estimates the percentages of youths in different grades who have used various substances in the lifetime, past year, or past month periods but does not make estimates of the numbers of youths in these grades who are substance users (see Section A.4 in Appendix A for additional discussion of this issue). For policymakers and service providers who are interested in knowing the numbers of adolescents who are users of various substances, as opposed to the percentages of adolescents who are substance users, exclusion of dropouts could appreciably affect some estimates of these numbers of adolescent substance users, especially for older adolescents.
Analysis of data on absentees also suggests that exclusion of absentees from school-based surveys will slightly affect estimates of substance use among adolescents because in an average month, most youths are absent from school no more than 1 or 2 days, if at all. Consistent with prior research (Bachman & O'Malley, 1981; Johnston & O'Malley, 1985), however, analyses of NSDUH data found that frequent absentees were more likely to be past year or current substance users than their peers who were not absent from school.
Examination of trend data for 12th graders in NSDUH also raises some questions about whether the inclusion or exclusion of absentees in annual estimates will not appreciably affect substance use trends. Although most trend data were consistent with an assumption of no appreciable effect on trends when absentees are excluded, estimates based on weights that were adjusted for absences sometimes yielded significant differences in prevalence between consecutive survey years when differences were not significant based on unadjusted data for 12th graders.
Trends in lifetime, past year, and current use for 12th graders in NSDUH who were interviewed in the first 6 months of the year and those who were interviewed in the second 6 months of the year were similar for most substances. However, situations were observed in which estimates differed between consecutive years for data collected in one half of the year but not for the other half. As noted previously, trends in lifetime cocaine use between consecutive years that were not adjusted for absences were stable based on data collected in January through June but showed some variation for unadjusted data from July through December and for data that were adjusted for absences. Given that the lifetime prevalence of cocaine use between consecutive years was stable from 2004 onward for estimates based on data from each half of the year and for unadjusted and adjusted data, the changes from 2002 to 2003 may be anomalous, particularly for estimates based on data from the second half of the year. Nevertheless, these findings suggest that trends in lifetime use of less prevalent substances such as cocaine could vary depending on the data collection period.
Estimates within a cohort of 12th graders typically did not differ according to whether data were collected in the July through December or January through June period. The notable exception was that some estimates of lifetime, past year, and current use for the cohort of 12th graders in 2002 to 2003 were greater for January through June 2003 than for July through December 2002. Unadjusted and adjusted estimates of current alcohol use among 12th graders in 2006 to 2007 also were greater based on data from the first half of 2007 than for the second half of 2006. Although the higher prevalence among 12th graders in the second half than in the first half of a school year may be related to 12th graders being older in the latter half of the school year, most estimates did not differ significantly depending on whether data were collected in the first or second half of the year. Consequently, something atypical may have occurred for the cohort of 12th graders in 2002 to 2003.
In the annual average data for 2002 to 2008 with the larger sample sizes for July through December and January through June, only one situation was observed (for current use of inhalants) in which the annual average estimate in July through December was greater than the corresponding estimate in January through June. When other annual average estimates were significantly different between the two halves of the year, the estimate in January through June was greater than the estimate in July through December. For many estimates, however, there was no significant difference in the annual average estimates between the two halves of the year.
Recent initiation of use could account for higher annual average estimates of lifetime use of alcohol, marijuana, and cocaine among 12th graders in the second half of a school year (i.e., January to June) than in the first half of the year (i.e., July to December). Recent initiation could explain the higher lifetime rate of alcohol use in January to June than in July to December, because the annual average estimate of past year use also was greater in January to June than in July to December. That is, initiation of use in the second half of the school year would be expected to yield higher annual average past year estimates in January to June than in July to December, because these recent initiates presumably would not have reported past year use if they had been surveyed in July to December. Furthermore, assuming a greater potential stigma for use of marijuana and cocaine than for alcohol, 12th graders who initiated marijuana or cocaine use in January to June might be willing to report that they used these drugs at least once in their lifetime but to underreport more recent use.
Although these findings are not completely clear-cut, whether students are surveyed in the fall or spring could make a difference for some prevalence estimates. Factors that could affect estimates for 12th graders between the first and second halves of the year include aging of the cohort, seasonal differences in substance use behaviors between the fall and spring semesters of the 12th grade, initiation of use after survey data collection in the fall, and students dropping out of school in or before the spring. With regard to aging of the cohort, rates of substance use steadily rise as youths progress from ages 12 to 18. In 2008, for example, the rate of past month marijuana use in NSDUH (regardless of school enrollment status) was 7.7 percent among 15 year olds, 10.9 percent among 16 year olds, 14.5 percent among 17 year olds, and 17.8 percent among 18 year olds (OAS, 2009b). In addition, 12th graders who have reached their 18th birthday in the spring could be more likely to report cigarette use in the spring than in the fall if they still were aged 17 in the fall. All but the effect of dropouts could increase the estimates in the spring. As indicated by the analysis of data for dropouts and nondropouts in this chapter, missing the 12th graders who were in school in the fall but dropped out of school by the spring would be expected to depress the estimates in the spring.
An additional consideration for whether students are surveyed in the fall or the spring concerns whether statistically significant differences between the fall or spring would be of practical significance. For example, the annual average estimate of past year alcohol use based on combined 2002 to 2008 NSDUH data was greater in January to June than in July to December (62.6 vs. 59.0 percent). From policy and programming perspectives, however, estimates based on either half of the school year indicate that the majority of 12th graders had used alcohol at least once in the past year.
Samples for unadjusted and adjusted estimates may not overlap completely because respondents whose adjusted weight was 0 (i.e., absent the entire period) were excluded from the analysis of adjusted estimates but still were included in the unadjusted data.
All material appearing in this report is in the public domain and may be reproduced or copied without permission from SAMHSA. Citation of the source is appreciated. However, this publication may not be reproduced or distributed for a fee without the specific, written authorization of the Office of Communications, SAMHSA, HHS.
Your browsing activity is empty.
Activity recording is turned off.
See more...
