Higher education among the “forgotten half”: The association between substance use and college completion among 2-year college students versus matched 4-year college students from a nationally representative longitudinal study

Background: Recent policy changes have increased marijuana availability to college students. Past research has not evaluated the association between substance use and college completion among the forgotten half of disadvantaged young adults who access higher education through 2-year colleges. Methods: This study uses a subsample of college-enrolled participants from the National Longitudinal Study of Adolescent and Adult Health to evaluate whether substance use in 2001 among students attending 2-year college and 4-year college predicts educational attainment in 2008. We used Mahalanobis nearest-neighbor and exact matching within propensity score calipers to identify a comparison group of 4-year college students (n=888) similar to the 2-year college students (n=1398) on 15 baseline measures including grades, test scores, and substance use. We used multivariate regression in the matched sample using a Poisson working model to estimate the relative risk of earning no post-secondary degree. Results: Compared with matched 4-year college students, 2-year college students were more likely to use methamphetamines, cocaine, marijuana, and more likely to report problematic substance use, and less likely to use alcohol. Two-year college students who used methamphetamines in the past year (IRR=1.51, 95% CI (1.12, 2.04), p=0.007) or past month (IRR=1.69, 95% CI (1.09, 2.61), p=0.02), or completed alcohol abuse treatment (IRR=1.58, 95% CI (1.21, 2.07), p<0.001) were less likely to complete college. Four-year college students who reported that drugs interfered with school or work in the past year (IRR=1.84 (1.28, 2.64), p=0.001), used cocaine in the past year (IRR=1.47 (1.04, 2.08), p=0.03), and used marijuana in the past year (IRR=1.30 (1.07, 1.57), p=0.007), past month (IRR=1.31 (1.07, 1.61), p=0.01), or ≥5

Conclusions: Substance use interventions should target 2-year and 4-year college students. Two-year colleges that better accommodate students who complete substance use treatment may improve these students' completion. Students who use marijuana or cocaine or whose drug use impairs functioning may benefit from an incremental approach of completing a 2-year degree prior to transfer for a 4-year degree, rather than enrolling directly in a 4-year program.

Background
Alcohol and illegal drug use is common during adolescence and emerging adulthood, but often declines with age. Many young adults (18)(19)(20)(21)(22)(23)(24)(25) who meet criteria for problem drinking decrease alcohol consumption over 7 years (1). For other youth, alcohol and illegal drug use can interfere with or be a marker for difficulties in the transition to adulthood including the completion of schooling. The existing longitudinal research on whether substance use predicts long-term educational attainments focuses on 4-year college completion. For instance, the Coronary Artery Risk Development in Young Adults study, a random sample from 4 cities, found past-month marijuana users and heavy drinkers were less likely to have graduated college within 10 years (2) and that high levels of alcohol consumption at ages 18-30 predicted lower chances of college graduation and lower occupational prestige 15 years later (3).
As a result of community college's lower entry barriers, community college students are highly similar to high school graduates who do not matriculate in post-secondary education, with just slightly higher pre-college socioeconomic status, grades, and test scores (4). Prior to entering college, 4-year college students have vastly higher grades and test scores, socioeconomic status, better health, and fewer health risk behaviors than adolescents who did not matriculate in 4-year colleges (4). Non-representative studies have found that community college students are at least as likely to binge drink as 4-year 4 college students, adjusting for socioeconomic factors (5)(6)(7)(8), although the frequencies of binge drinking vary greatly among studies, from 25% (6) to 46% (9).
Little evidence exists about associations between the incidence of substance use in community college or its association with educational attainment among community college students (10), although some pilot studies have evaluated text-messaging antialcohol interventions with this population (11,12). Disadvantaged youth primarily access higher education through community college, which some regard as a promising vehicle for social mobility (13). Community colleges are affordable and have reduced traditional barriers to admission to disadvantaged youth who would be otherwise unlikely to attend college (14). Community college was formerly a negligible portion of the post-secondary educational system, but currently community college students comprise 46% of undergraduates (in the most recent available data from 2016) (15,  locations, low tuition, and free tuition in some states (e.g., Tennessee). Completion rather than access remains the chief challenge for community college students. Among community college students, about 15% finish an associates degree or certificate within 3 years, and 45% leave school with no credential (16). Half of former community college students cite "personal reasons" as their explanation for leaving school, about twice as many as cite "family" or "finances." (17). Personal reasons could include many risk factors not measured by standard education surveys, such as current and past drug use (18). This study evaluates factors that are not even included in national education surveys.
Recent changes to the marijuana policy environment have increased the availability of marijuana to college students. Evidence from long-term longitudinal studies in the current cohort of college students is needed to show the associations between marijuana and 5 educational attainment; the extent to which marijuana will substitute for use of other drugs; and how illegal drug use will affect educational attainment. This study supplements the literature on community college completion by using a nationally representative health dataset from a prior cohort of college students to identify associations between drug and alcohol use, matriculation in 2-year or 4-year college, and educational attainment. We evaluate whether students attending 2-year and 4-year college who also use alcohol and various drugs are less likely to have attained any post-secondary degree seven years later.

Data
These hypotheses were tested in the National Longitudinal Study of Adolescent and Adult Health (Add Health), a nationally representative sample of students grades 7-12 in 1995, followed in 1996, 2001, and 2008 and surveyed in in-home interviews (19). Respondents and parents were interviewed in their homes in 1995 (baseline) and 1996, and respondents were followed in 2001 and 2008. Sensitive questions were asked by audio computer-assisted self-interview. This study used the baseline, 2001, and 2008 waves.
The 1995 baseline survey was used instead of the 1996 wave 2 survey because the sample size was larger; a subset of baseline respondents were not invited to participate in wave 2, although they were asked to participate in waves 3 and 4.
We measured pre-college variables from the wave 1 in-home interview in 1995, when the the past year was the response to the question "During the past 12 months, on how many days did you drink alcohol?" Number of episodes of binge drinking in the past year was defined as number of times in the past year that the respondent consumed "5 or more drinks in a row." Number of alcoholic drinks typically consumed in one episode of drinking was the response to the question "Think of all the times you have had a drink during the past 12 months. How many drinks did you usually have each time? A 'drink' is a glass of wine, a can of beer, a wine cooler, a shot glass of liquor, or a mixed drink." Number of episodes of binge drinking in the past 2 weeks was defined as "five or more drinks on a single occasion, for example, in the same evening" for males and "four or more drinks on a single occasion, for example, in the same evening" for females. Number of times drunk was the response to the question "During the past 12 months, on how many days have you been drunk or very high on alcohol?" Number of episodes of marijuana use in the past month was the response to the question "During the past 30 days, how many times have you used marijuana?" The binary variables were self-reported marijuana use in the past month and past year, cocaine use in the past month and past year, methamphetamine use in the past month and past year, injection drugs in the past month and past year (fewer than 10 cases), "other illegal drugs" in the past month and past year, whether drugs interfered with life in the past year, and obtaining treatment for alcohol abuse. "Other" illegal drugs were not listed in the survey, but it could include any illegal drug other than cocaine and marijuana, such as MDMA/ecstasy or LSD.

College enrollment in 2001
College enrollment in 2

Multivariate analysis
A multivariate model was used to estimate how much more common each substance use condition was among community college students than 4-year students. This model was used first in the raw data and then after matching. We used multivariate analyses in both raw and matched samples to demonstrate that matching does not change the direction of association. Multivariate regressions used a Poisson working model with robust standard errors to yield consistent and unbiased estimators that are also easily interpretable (21)(22)(23)(24). The multivariate regressions yielded incidence rate ratios (IRRs), which can be interpreted as relative risks; they are not odds ratios.
The control variables in the regression model included gender, age, race/ethnicity (Latino, Asian, African-American), and 1995 marijuana use, friends' smoking, out-of-school suspension history, parent income, test score, grade point average (GPA), GPA missing, school attachment, ever pregnant, and college expectancies.

Matched sampling 9
Differences in substance use between 2-year and 4-year college students could be due to their college setting, but they could also be due to adolescents with specific pre-college drug habits choosing 2-year or 4-year colleges. This paper matched on baseline factors that may be important in students' self-selection into 2-year versus 4-year colleges, measured in 2001. We implemented matching using the MatchIt library in the R statistical package (25).
All factors used for matching were measured at baseline (1995), except age, which is computed at wave 3 (2001) from self-reported birth date and the date of the interview. We refined the matching model until there were no significant differences between the two groups of college students on key variables. We matched exactly on black race, GPA being unreported, and a binary measure of college expectancies, used 1:1 nearest-neighbor Mahalanobis matching with replacement within propensity score calipers of 0.25 standard deviations of estimated propensity score. The Mahalanobis metric estimated the distance between participants on age, grade point average (GPA), parent-reported household income, and standardized test score. The propensity score model used for estimating the propensity score calipers included binary indicators for male gender, ever being suspended, pregnant, using marijuana, using alcohol unsupervised, mother college graduation status, as well as number friends who smoke and an index for school attachment problems.

Post-matching multivariate analysis
After matching, statistical analysis used a Poisson working model and controlled for demographics (gender, age, Latino, Asian, African-American race/ethnicity); deviance (ever used marijuana, number of friends who smoke, having ever had an out-of-school suspension, ever pregnant); and socioeconomic status (parent-reported household income, test score, GPA, GPA missing, school attachment, college expectancies). As in the pre-matching multivariate analysis, the analysis used a Poisson working model, yielding estimates of incidence rate ratios, which can be interpreted as relative risks; they are not odds ratios.

Substance use of 2-year versus 4-year college students
Community college students were more likely than 4-year college students to have used methamphetamines (past year and past month), cocaine (past year), and injection drugs (past year and past month), and used marijuana more times in the past month, according to bivariate analysis (Table 1). Four-year college students reported more frequent alcohol use, binges, and episodes of intoxication in the past year, and more alcohol binges in the past 2 weeks, but community college students were more likely to have been in alcohol abuse treatment, and reported drinking more alcoholic drinks per night. college students who reported that they usually drink 1-3 alcoholic drinks were more likely to attain a post-secondary degree than students who reported that they usually do not drink alcohol (IRR 0.83, 95% CI (0.68, 1.00), p = 0.05).

Discussion
With recent public policy changes, marijuana is increasingly available to college students.
The results in this cohort of college students find that students at 4-year colleges who use marijuana at various levels are less likely to have attained a post-secondary degree within 7 years, but the same is not true for students at 2-year colleges. Use of a variety of other drugs predict lower likelihood However, 2-year college students engage in more drug use risk behaviors than matched 4year college students, with greater use of methamphetamines, cocaine, marijuana, and greater life impairment from substance use, but lower likelihood of binge-drinking alcohol, even after matching on pre-college substance use, test scores, grades, socioeconomic, and demographic factors that may have affected students' self-selection into 2-year versus 4-year colleges. Substance use appears to reduce the chances of graduation from 4-year college more than from 2-year college, even after using matched sampling to adjust for observed factors that predict selection into 2-year versus 4-year college.

Marijuana use predicts lower likelihood of graduation from 4-year college but not 2-year
college, which is particularly important with increased availability of marijuana and changed social norms due to decriminalization or legalization.
Despite serving students from more socioeconomically disadvantaged backgrounds who may have greater health needs and lower health care access, 2-year colleges generally do not have as many resources or services to help students with substance use (13).
Past research in Add Health finds that high school students diagnosed with depression are more likely to select into 2-year college than 4-year college (26). The greater substance use among 2-year college students could be partially explained by some 2-year college students with depression self-medicating with substance use.
The finding that alcohol abuse treatment predicts greater risk of non-completion for 2year college students is consistent with past research that non-childbirth hospital admissions predicts 44% lower risks of completing an associates degree (26). Two-year colleges need to better accommodate students' health needs and help students resume college after medical treatment.
Students who attend 4-year college engage in more alcohol use, especially binge drinking.
Four-year college students may have adopted these alcohol use patterns at college. A longitudinal study of beginning college students at the University of Texas found social motives (agreeing that e.g., attending social events is important) predicts greater alcohol use, and that academic motives are protective against alcohol use for females during the second year of college (27).
Moderate alcohol use may be protective for 4-year college students' college graduation relative to no use because in some cases, moderate alcohol use may signal greater social 14 integration at college, consistent with Tinto's theory that students who become wellintegrated, both academically and socially, are more likely to persist in college than students who are not well-integrated (28). Alcohol use is central to the social life at many residential 4-year colleges, and students who are able to navigate these social spheres while engaging in moderate alcohol use may have greater social integration at these colleges, and thus maintain the social connections needed to graduate college. This evidence is consistent with past research finding that moderate alcohol use predicts higher wages (29,30), including a study in the Add Health data that explains the association between moderate alcohol use and wages by sociability (31)      Substance use factors that predict non-completion (attaining no post-secondary degree by 2008), in community college students (blue) and 4 year college students (green) in matched sample (n=2286).