Using a theoretical approach to predict college students’ non-medical use of prescription drugs – a survival analysis

Introduction

The non-medical use of prescription drugs (NMUPD) is a growing problem among college students. NMUPD has been characterized as the use of prescription drugs in ways other than directed by health care providers or use without a legal prescription.¹ Young adults between the ages of 18–25 are at high risk for substance use and misuse.² Recent data shows that the non-medical use of prescription drugs for the past year (12.2%), and the past month (4.8%) was highest among young adults 18–25 years of age.³ In addition, NMUPD has been associated with over-the-counter (OTC) medication misuse, risky behaviors, depressive symptoms, poor sleep quality, high-intensity drinking, self-injury, adverse childhood experiences, extreme weight control behavior, and sexual violence.^4,5–10

College students are at a particularly high risk for NMUPD. Researchers have found that young adults who were enrolled in college were more likely to engage in NMUPD than those who did not attend college.^11–13 College students may use several classes of prescription drugs for non-medical purposes, including prescription stimulants, analgesics, and sedatives. The prevalence of lifetime NMUPD across different medication classes range between 4%−22% for college students.^14–17 Research suggests that prescription stimulants and analgesics (e.g., opioids) are among the most common classes of medications used for non-medical purposes by college students.^7,18,19 McCabe et al²⁰ found that 60.8% of undergraduate college students who reported NMUPD during some point in their lifetime used stimulants, 46.2% used sedatives, 45% used pain medications, and 38.6% used sleeping medications. Stone and Merlo²¹ reported that 12.4% of students misused a prescription stimulant at least once during their lifetime. Benotsch et al⁷ reported the most common class of drugs being used for non-medical purposes were prescription analgesics followed by stimulants.

A handful of studies have examined college students’ NMUPD from college entrance onward. Arria et al²² found that the lifetime use of prescription stimulants doubled during the first year of college and quadrupled from pre-college to second year. The increase in stimulant use surpassed the use of prescription analgesics and hallucinogens. Garnier et al²³ found that by the fourth year, 61.8% of college students were offered prescription stimulants at least once, and 31.0% reported lifetime use. In a cross-sectional survey conducted from 2003 to 2013, McCabe et al¹⁷ found a significant increase in lifetime prevalence of non- medical use of prescription stimulants. McCabe et al¹⁶ examined the 12-month prevalence of NMUPD among college students from 1993 to 2001, and found that the 12-month prevalence increased from 4.41% in 1993 to 9.97% in 2001. However, as NMUPD has increased in recent years by point prevalence, longitudinal studies of college student cohorts and NMUPD reflecting current trends are scarce. Furthermore, the transition to college is a challenging time in which students are adjusting to new academic tasks and social networks, which can lead to experimentation with substances and other maladaptive behaviors.^24,25

Several studies have used theoretical frameworks to gain an understanding of mechanisms that influence college students’ NMUPD.^26,27 The Theory of Reasoned Action (TRA) was among the first theories to propose specific individual and social characteristics can influence substance use. According to the TRA, attitudes and subjective norms are important determinants of performing specific behavior.^28,29 Attitudes are based on individuals’ beliefs about outcomes or attributes of performing the behavior; positive attitudes will increase the likelihood of performing the behavior.^28,29 Subjective norms are based on beliefs about whether important referent individuals approve or disapprove of performing the behavior; greater subjective norms will increase the likelihood of performing the behavior.^28,29 The Theory of Planned Behavior (TPB) is an extension of the TRA, and includes the construct of perceived behavioral control.³⁰ The TPB has been used to explain NMUPD among college students. Although previous studies have found consistent results for perceived behavioral control, the impact of attitudes and subjective norms on college students’ NMUPD is less clear.^27,31,32 Ponnet et al³¹ found that subjective norms were the strongest predictor of college students’ intentions to use prescription drugs for non-medical purposes, followed by attitudes and perceived behavioral control. Judson and Langdon³² found that attitudes, social norms, and perceived control were associated with lifetime NMUPD. Gallucci et al²⁷ found significant associations between perceived control and current NMUPD. However, the researchers found no such relationships between current NMUPD, attitudes and subjective norms.²⁷ This difference could be attributed to the assessment of NMUPD at different stages (intentions, past 30 days, and lifetime).²⁷ Given these previous findings, this study will focus on attitudes and social norms and thus apply the TRA.

Scant research has examined the potential for conceptual constructs, measured during the transition period between high school and college, to predict future NMUPD during college using a longitudinal approach. In addition to understanding prevalence over time, the identification of factors that predict NMUPD among college students can inform and bolster prevention efforts. Gaining a better understanding of how NMUPD evolves across students’ college careers and factors that contribute to students’ NMUPD can shed light on appropriate time points and potential targets for intervention. The objectives of this study are to examine students’ NMUPD during college and examine factors (attitudes and subjective norms) that predict NMUPD. Based on TRA, we hypothesize that there are positive relationships between attitudes towards and subjective norms regarding NMUPD and the use of prescription drugs for non-medical reasons.

Methods

Study design

A longitudinal cohort study design was used to examine the non-medical use of prescription drugs across time, and NMUPD-related attitudes and subjective norms among college students. Mixed methods of data collection were used to obtain information for this study. This study was conducted with the entering classes at two large US universities; located in the West and Midwest regions of the country. This study took place between May 2011 and September 2015 and received approval from the two relevant Institutional Review Boards.

Participants

Potential participants were randomly selected from the registrar’s lists of incoming first-year students from both universities for recruitment. Participants were eligible if they were between the ages of 17 and 19 years and enrolled as full-time first-year students for Fall 2011 at one of these two universities. Participants were recruited through several steps. Potential participants were sent a pre-announcement postcard followed by emails, phone calls or Facebook messages. Next, potential participants were provided study information and informed consent was conducted over the phone. Individuals were excluded if they had already arrived on campus for early-enrollment programs, were an age other than 17–19, or were non-English speaking. Participants were compensated for each interview (see Data collection below) in this longitudinal study; over time the incentives increased to promote ongoing engagement. In addition, a Certificate of Confidentiality was used in this study to protect the privacy of participants given the topic area and longitudinal design.

Data collection

Trained research assistants conducted up to five annual telephone interviews to collect study data. Baseline interviews were conducted with all participants at the time of enrollment before arriving at college (Time 0). The second through fifth interviews took place during the summers after the subsequent years in college: after year 1 (Time 1), after year 2 (Time 2), after year 3 (Time 3), and after year 4 (Time 4). Research assistants were trained to record participant responses directly into a customized FileMaker database.

Measures

The interview included questions to assess participants’ NMUPD, attitudes, and social norms. Research assistants informed participants that the interview questions discussed using prescription drugs for non-medical purposes. Participants were also informed that using a prescription drug for non-medical purposes include “using a prescription drug that was your own or someone else’s for anything other than its intended use, such as to get high or to get through finals or something like that.”

Non-medical use of prescription drugs

Participants reported lifetime non-medical use of prescription drugs. Lifetime NMUPD was assessed with the following question: “Have you ever used a prescription drug that was not your prescription, or used your own prescription drug in a non-medical way?” If a participant responded “yes,” then the interviewer asked the respondent to list all drugs that were used prior to the data collection time point (open-ended response). A research assistant categorized reported prescription drugs into four categories: sleep medication (e.g., Ambien, Halcion, Restoril), sedative/anti-anxiety medication (e.g., Ativan, Xanax, Valium, Klonopin), stimulants (e.g., Adderall, Ritalin, Dexedrine, Concerta, Vyvanse), and pain medication (e.g., Vicodin, OxyContin, Tylenol 3 with Codeine, Lortab, Norco). A random 10% of the categorized prescription drugs were assessed for the accuracy of the placement into categories.

Attitudes

At the time of enrollment before arriving at college (Time 0), attitudes about the non-medical use of prescription drugs were measured with the question, “On a scale between 0 and 6, with 0 as very negative, 3 as neutral, and 6 as very positive, what would you say your own attitude towards prescription drugs for non-medical use is?”³³. Participants’ responses to this question were scored on a 7-point Likert-type scale, ranging from 0=very negative to 6=very positive.

Subjective norms

At Time 0, subjective norms related to the non-medical use of prescription drugs were measured in reference to participants’ friends, as the proximity of this reference group renders it more influential than distal reference groups such as typical college students.³⁴ Specifically, descriptive subjective norms that indicate what a person thinks other people actually do were assessed.³⁵ Participants provided an estimated percentage in response to: “What percentage of your friends engage in non-medical use of prescription drugs?” Higher values indicate greater social norms for the non-medical use of prescription drugs.

Analyses

Descriptive statistics were calculated to characterize participants’ NMUPD across all time periods and study variables. Survival analysis was used to assess NMUPD across four years of college (i.e., event of NMUPD). Survival analysis allows participants to be followed across time from an initial point (before college – Time 0) to the discrete event (i.e., first report of NMUPD).³⁶ However in survival analysis studies, only some individuals will experience the discrete event; survival times will be not be known for a subset of the study group. This phenomena is called censoring and can be due to the following issues: (1) the individual did not experience the event by the end of the study; (2) the individual was lost to follow up; or (3) the individual experienced another event that makes follow-up impossible.³⁷ Therefore, censoring in survival analysis (time-to-event) occurs when there is incomplete information about the individual regarding the discrete event. For this study, censor and time variables were created. The censor variable was assigned a “1” if a participant never reported using prescription medications for non-medical reasons (no event observed) and a “0” if a participant reported first use during the study (event observed). The time variable is the time of the event. If a participant reported using a prescription medication for non-medical reasons during an interview, then the event occurs and this value is the time during which the participant first used. If the participant did not report NMUPD during an interview, then the value is the school year in which data was last collected from that individual. Participants’ results were considered right-censored if they did not indicate NMUPD during the time period. Kaplan-Meier analysis was conducted to estimate the proportion of the cohort who reported NMUPD for any observation point in the study period. Kaplan-Meier analyses also were used to examine differences in NMUPD across gender (male, female), race/ethnicity (White, Non-White), and university (West, Midwest). There were no significant differences between the two universities in terms of NMUPD, temporal patterns of reporting NMUPD, or predictors of NMUPD. Thus, data from both universities were pooled. Cox proportional hazards regression analysis was used to model the time to NMUPD based on attitudes and social norms; models included gender, race/ethnicity, and university. Analyses were conducted with SAS version 9.3.

Results

A total of 338 students participated in the study and completed interviews at Time 0 (54.6% response rate); 259 participants completed interviews at Time 4 (76.6% retention rate). There were no significant differences between completers and dropouts regarding gender, race, or university. In all, there were 1459 separate phone interviews, an average of 4.32 per student, over the time period from Time 0 to Time 4. The majority of participants were Caucasian/White (74.8%) and female (56.1%) (Table 1). Overall, about 35% of the students (n=119) reported using prescription drugs for non-medical reasons during the study period. Mean attitudes toward NMUPD (at Time 0) lay between very negative and negative (0.72, SD=1.09), and the mean subjective norm (at Time 0) was very low (8.29%, SD=12.36), indicative of believing that small percentage of friends use prescription drugs for non-medical purposes. Across all years in college, the most commonly used medications for non-medical purposes were prescription stimulants and pain medications (see Figure 1); note that some participants indicated using multiple categories of medications.

Table 1.

Participant information for incoming college students from two US universities (N=338)

Variables	N (%)
Gender
Female	190 (56.1%)
Male	148 (43.9%)
University
Midwest	199 (58.8%)
West	139 (41.2%)
Race/Ethnicity
Caucasian/White	253 (74.8%)
Non-White	85 (25.2%)

Figure 1. Non-medical use of prescription drugs across college years.

Notes: Categories are not mutually-exclusive.

Table 2 details the censor and time variables for the survival analysis. The censor = 0 row represents participants’ first report of NMUPD during the study, categorized by whether first-use occurred prior to entering college (Time 0) or after year 1 (Time 1), after year 2 (Time 2), after year 3 (Time 3), or after year 4 (Time 4) years. The censor = 1 row represents those who did not report NMUPD during the study. However, some of these individuals only participated in a portion of the study (i.e., drop outs). For example, there were 17 participants who took the first survey (and did not report NMUPD at that time), so no data beyond baseline are available.

Table 2.

Censor and time information for survival analysis

	Time
Censor	0	1	2	3	4	Total
0	39	30	27	13	10	119
1	17	10	12	19	161	219
Total	56	40	39	32	171	338

Note: The censor variable was assigned a “1” if a participant never reported using prescription medications for non-medical reasons (no event observed) and a “0” if a participant reported first use during the study (event observed).

Based on the Kaplan-Meier analysis, the probability of having used prescription drugs for non-medical purposes before entering college is 39/338=.115. The next year, 282 of those who had not previously reported NMUPD responded, making the proportion of those who first reported NMUPD during their first year 30/282 = 0.106. The rest follows similarly: 27/242 = 0.112 (during year 2); 13/203 = .064 (during year 3); and 10/171 = .058 (during year 4). Thus, the Kaplan-Meier estimator of making it through college without using prescription medications for non-medical reasons is Q = (299/338)*(252/282)*(215/242)*(190/203)*(161/171) = .619. Note that the denominator at each year deletes those who had previously used as well as those who had dropped out before the current year. For example, the denominator of 282 for Time 1 (during year 1) is 338 - (39 previous users) - (17 who dropped out after Time 0). In the first three years, the annual rate of NMUPD (among those eligible) is about 11%. However, this annual rate drops in the last two years to approximately 6%, as those participants who are left are more resistant to NMUPD (i.e., a decreasing hazard rate). The change in hazard rate is reflected in Kaplan Meier function in Figure 2.

Figure 2. Kaplan Meier survival function for all students.

The percentage of NMUPD by females was 27.32%, and the percentage of NMUPD by males was 46.62%. The Log-Rank (Chi-square = 15.97, p<0.05) results indicate a significant difference between male the female participants. No significant differences were found across race/ethnicity or university. In the Cox proportional hazard regression model, male gender was positively associated (HR= 1.89, p<0.01) with NMUPD. Even after controlling for gender, race/ethnicity, and university, the time to NMUPD during college was significantly associated with more positive attitudes toward NMUPD (HR=1.73, p<0.01) and with reporting larger numbers of friends using prescription medications in a non-medical way (HR=1.01, p<0.01).

Discussion

College students’ NMUPD is a critical and growing issue, given the increases in use in the recent past.¹⁷ The identification of critical times during which college students initiate the use of any type of prescription medication for non-medical purposes as well as factors that predict this non-medical use can shed light on when preventative efforts should take place and strategies to improve the success of those efforts. In order to support preventative efforts, this research examined students’ NMUPD across their progression through college and factors that predicted the time to initiation. Our findings show that of the college students who use prescription drugs for non-medical purposes, the majority do so by their second year in college. Study results also indicate that attitudes were the strongest predictor of time to the initiation of NMUPD; subjective norms and gender (male) also were significant predictors of increased risk. These findings can be used to inform the timing and targeting of campus efforts to reduce NMUPD.

The results of the current study suggest that prevention and intervention strategies to curb NMUPD should start early, perhaps in early adolescence, as the risk is greatest during the early years of college. Specifically, 35% of college students in our study indicated using prescription drugs for non-medical purposes, and the majority of those initiated non-medical use within the first two years of college. Arria et al²² found that lifetime prevalence prescription drugs increased during the first two years of college, but not theresafter. McCabe et al³⁸ found that NMUPD was highest at age 18, followed by a consistent decrease in use across subsequent years. Researchers have called for the development of early intervention strategies to assess risk, and prevent progression to serious substance misuse and dependence problems.³⁹ Prevention efforts could be included in campus orientation sessions where substance misuse awareness and screening programs occur. In addition, family physicians and pediatricians could engage in preventative strategies such as screening NMUPD at earlier time periods, counseling to encourage healthy behaviors or reduce risky behaviors, and using prescription drug monitoring programs.⁴⁰

Study findings also revealed that attitudes and subjective norms that students come to college with were significant predictors of NMUPD during college. Students with more positive attitudes towards NMUPD and those who had more friends that used NMUPD proceeded more quickly to NMUPD during their college years. Results from this longitudinal study are in-line with previous cross-sectional research that examined associations between attitudes, social norms, and NMUPD. Donaldson et al⁴¹ found that attitudes were significantly associated with behavioral intentions to use prescription drugs non- medically. Using a cross-sectional study design, researchers also found that students who used prescription drugs non-medically had perceived greater subjective norms about NMUPD in comparison to those who did not use prescription drugs non-medically.³² In addition, the number of close friends who use prescription drugs for non-medical purposes is a predictor of NMUPD among students.^42,43 One plausible reason for the associations between attitudes, subjective norms, and NMUPD may be that friends play a substantial role in both the acquisition and social use of prescription drugs, as part of social convention or the “in-thing” to do.⁴⁴ A recent meta-analyses found that friends and family were the most prominent source of drug diversion for NMUPD, and individuals at risk of diversion were those with substance use issues and those with fraternity or sorority affiliations.⁴⁵ Another plausible explanation is that students may believe that friends are using prescription drugs to bolster academic success and pressure from peers to perform academically motivates NMUPD. Although the notion that NMUPD improves academic performance is unsubstantiated, students may have adopted such behaviors in hopes of obtaining such favorable outcomes.⁴⁶ For college counselors, directly addressing the lack of evidence for the positive impact of NMUPD on academic performance early in the college experience may also deter NMUPD.

Results in the present study also suggest that NMUPD may be particularly problematic among male college students. The percentage of males using prescription drugs for non-medical purposes was significantly higher than females. Previous studies also found that male college students were more likely to use prescription pain medications, sedative or anxiety medications, and stimulant medications for non-medical purposes in comparison to female students.^17,20,47 The gender difference may be due to the higher occurrence of medical conditions in males that require medications for treatment such as attention deficit hyperactivity disorder, which increase access to prescription medications.⁴⁸ While our study did not assess these types of co-morbidities, college health services may wish to specifically address availability of assessments and assistance, including prescription medications, to support academic achievement.

This study contains limitations that warrant mentioning. First, although 76% of participants completed the entire duration of the study, attrition could create noise and lessen the generalizability of the study. However, post-hoc comparisons between participants who dropped out and those who completed the study revealed no significant differences for gender, race/ethnicity, or university. Second, the NMUPD was assessed by research assistants conducting an interview, and may be subject to social desirability bias. Although the prevalence of NMUPD found in this study is similar to rates found in previous studies, some participants may have perceived stigma associated with using prescription drugs for non-medical purposes and sought to not disclose their non-medical use for fear of embarrassment. Future studies could use alternative methods to reduce the impact of social desirability bias such as self-administered questionnaires and the use of proxy subjects.⁴⁹ Next, attitudes and subjective norms were assessed with single-item measures resulting in an inability to calculate internal consistency reliability statistics and vulnerability to random measurement errors. However, research has shown that multiple-item and single-item measures of psychological constructs exhibit comparable psychometric qualities.^50,51 Also, NMUPD was assessed once a year during the study interviews. Participants may have had a difficult time remembering occurrences of using prescription drugs for non-medical purposes during the past one year period of time (i.e., recall bias), resulting in under-estimation of NMUPD. Future research could assess NMUPD multiple times during a year to address memory issues; however respondent burden should be considered as well. Finally, this study did not account for other factors such as mental health, socio-economic status, use of other substances, and Greek affiliation which could impact NMUPD.

College students are among those who are at great risk for substance use and misuse. Of the study participants who reported NMUPD during their college careers, a majority initiated non-medical use before the end of their second year. In addition, attitudes and subjective norms regarding NMUPD assessed just before entering college were significant predictors of using prescription drugs for non-medical reasons. These results suggest that engaging students during their early college careers, or perhaps even before they step foot on a college campus, may be a pivotal time to address the issue of non-medical use of prescription drugs. Future studies should evaluate prevention efforts that target students’ attitudes about the non-medical use of prescription drugs and peer networks in an effort to prevent misuse and negative outcomes.

Table 3.

Cox proportional hazard model predicting time to non-medical use of prescription drugs

Variables	Hazard Ratios	95% CI	P-value
Attitudes	1.726	(1.503, 1.976)	<0.001
Subjective norms	1.012	(1.003, 1.022)	0.008
Race/Ethnicity (non-White)	0.762	(0.470, 1.218)	0.263
Gender (male)	1.894	(1.297, 2.740)	<0.001
University (west)	1.107	(0.740, 1.683)	0.627