This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
Prescription drug misuse and abuse is an established public health challenge, and young adults are particularly affected. There is a striking lack of real-time, naturalistic data collection assessing intentions to misuse and other precipitating factors at the time of actual misuse, leaving the conditions under which individuals are most likely to misuse prescription medications unknown. Ecological momentary assessment (EMA) apps and protocols designed to capture this information would accelerate and expand the knowledge base and could directly contribute to prevention and treatment efforts.
The objectives of this study are to describe the development and administration of a mobile app and the EMA protocol designed to collect real-time factors associated with college students’ prescription drug misuse intentions and behaviors in daily life; present completion rates, compliance, acceptability, and reactivity associated with the EMA protocol for participants who endorsed recent prescription drug misuse at screening (ie, risk group; n=300) and those who did not (ie, nonrisk group; n=55); and establish initial construct validity by linking the reports of misuse behaviors in daily life collected via the EMA app to prescription drug misuse reported on a standard survey.
An EMA data collection app and protocol were designed specifically to capture hypothesized contextual factors along with prescription drug misuse intentions and behaviors in daily life. Using this protocol, young adult college students (N=352) completed signal- and event-contingent reports over a 28-day period. When the intention to misuse a prescription drug was endorsed, a brief follow-up prompt was sent 15 min later to collect participants’ indications of whether or not misuse had occurred.
Risk-group participants were significantly more likely than nonrisk counterparts to endorse any prescription drug misuse intentions in daily life (
The novel EMA app and protocol provide an effective way to assess real-time factors associated with prescription drug misuse intentions and behaviors in daily life. The resulting investigations offer the potential to provide highly translatable information for research and prevention efforts.
Prescription drug misuse is an established public health concern in the United States and beyond [
Prescription drug misuse itself is potentially harmful and poses additional costs to individuals and society for its established links with illicit drug use, alcohol abuse, mental health problems, risky sexual behaviors, and overdose-related deaths [
The status quo in the literature on college students’ serious medication practices has been largely cross-sectional and based on retrospective surveys of prescription drug misuse over an extended time frame. For example, the available research predominantly involves participants indicating whether they have misused any prescription drugs over the past 12 months or their lifetime [
Preliminary findings have demonstrated the utility of collecting young adults’ momentary prescription drug misuse reports in real-world settings. We obtained data from 49 mixed-sex dating couples reporting 3 times per day for 10 days [
Building on our preliminary work, the objectives in this study are to (1) describe the development and administration of a mobile app that was designed to collect real-time factors associated with college students’ prescription drug misuse in daily life; (2) present completion rates, compliance, acceptability, and reactivity associated with the EMA method for both risk group (n=300; endorsed recent prescription drug misuse of one or more medications in the past 3 months) and nonrisk group (n=55; did not endorse recent prescription drug misuse) participants; and (3) establish initial construct validity by linking the reports of misuse behaviors in daily life collected via the EMA app to prescription drug misuse reported on a standard survey.
Participants attended two laboratory sessions that were scheduled an average of 35 days apart, with the EMA protocol implemented in between these laboratory sessions. Before this study, university institutional review board approval and the National Institutes of Health Certificate of Confidentiality were obtained. Participants completed questionnaires in each laboratory session, including an EMA feedback form to describe their reporting experiences.
Between September 2017 and September 2019, participants at a large university in the Midwestern United States were continuously enrolled into an ongoing longitudinal study on daily behaviors and health in college life. This study is drawn from the baseline phase of a broader study [
The web-based screening obtained prospective participants’ indications that they were in a private location and agreed to complete the screening questions, which asked about potentially sensitive day-to-day behaviors. Participants were instructed to think back over the past 3 months and indicate whether they used the medications listed in any way a doctor did not intend, such as use without a prescription, increased amounts, more often, or longer than directed. The screener presented 4 prescription medication classes adapted from the 2015 National Survey on Drug Use and Health (NSDUH) [
Given the main objective of capturing prescription drug misuse in daily life during this study, we oversampled for participants who endorsed recent prescription drug misuse (ie, risk group). This study included 300 participants from the risk group and 55 participants who did not endorse recent prescription drug misuse in the screener (ie, the nonrisk comparison group). To minimize any differences in recruitment between groups, the nonrisk participants were enrolled simultaneously with the risk participants throughout the recruitment period. Participants completed informed consent procedures at the start of their first laboratory session. The typical EMA reporting period was scheduled for 28 days following the first laboratory session. The length of the reporting period was adjusted for some participants due to scheduling conflicts (eg, campus recess periods) or device issues. During the second laboratory session, participants returned their devices and completed additional measures. Participants received their choice of electronic or check payments; compensation included US $75 for session 1, US $84 for reporting in daily life (prorated for partial completion), US $55 for session 2, and a US $36 bonus for maintaining compliance across the planned reporting period.
The app was developed in collaboration with the university’s technology division and was installed on an Apple sixth-generation iPod Touch device. Before the initiation of this study, the research team met with developers several times to provide feedback on the app features and interface design. The survey was administered through an app presented across multiple screens, with question completion automatically advancing the participant to the next screen. The questions allowed for different user inputs (eg, drop-down, radio button, and checkbox). Throughout the course of this study, the developers provided technical assistance related to software updates and other questions that arose. Developers also provided training to the research team on how to program the app and device for individual participants.
Given the sensitive nature of the data, security was a primary concern. We maximized the security of the data in several ways. First, the app was installed locally on laboratory-owned devices, which were provided to all participants for the duration of their reporting periods. Second, the data collected from the app were stored directly on the devices; the resulting data files were downloaded directly from the device to a secure server when participants returned to the laboratory (ie, never transmitted via the internet). Third, all data were recorded using a study identifier that contained no personally identifying information. Fourth, devices were also placed in restricted mode, which prevented participants from accessing a completed report or using other device functions. In addition, wireless transfer of data to and from the device was prevented. Finally, responses were all numeric and stored on the device with nondescript variable names. The list of variables was kept separate and accessible only by laboratory members. In the unlikely event someone gained access to the completed surveys, there would be no indication of what the data represented.
Consistent with numerous EMA protocols on addictions [
During the first session, participants were thoroughly trained to complete the app reporting procedures. A comprehensive definition of the focal behavior was provided (ie, using a medication without a prescription of your own; using it in greater amounts, more often, or longer than you were told to take it, or using it in any other way a doctor did not direct you to use it); the term
During the first laboratory session, participants reported on their gender, race, and ethnicity. In addition, the participants reported fraternity or sorority affiliation and medication prescriptions they had received in their lifetime.
EMA reports captured prescription drug misuse intentions. Participants were asked “Are you about to take a medication listed here in any way a doctor did not direct you to use it?” (0=no; 1=yes). Four medication classes and examples were provided (pain relievers, tranquilizers, stimulants, and sedatives or barbiturates; 0=no; 1=yes). Reports also collected contextual information about the current location of participants and who they were with as well as hypothesized predictors of prescription drug misuse, including other substance use, mood, pain, fatigue, and stressful events. To strengthen the predictive design, reports assessed current feelings and situations; substance use covered behaviors in the past 15 min. Items were selected from brief, validated scales used in previous research with college-based populations when possible. The number of items on each report could vary because questions used conditional answer choices (eg, skip logic) whenever possible. Follow-up reports presented the following question for each of the 4 medication classes: “Have you recently taken a medication listed here, in any way a doctor did not direct you to use it?” (0=no; 1=yes). Additional questions about current mood states and recent physical activities were included in the follow-up reports.
During their second laboratory session, participants reported on their experiences with the app and device procedures. Participants rated the extent to which “the reporting occurred during a period that reflects my typical daily life” (0=not typical; 3=very typical) and the extent to which “the iPod touch device was user-friendly” (0=not friendly; 3=very friendly)
During the baseline visit, participants completed a series of questions about the use of prescription medications following the 2015 NSDUH [
We employed independent samples
Participants in the risk and nonrisk groups did not differ with respect to the demographic characteristics of age or class standing (
Sample characteristics at baseline and tests of risk (n=300) and nonrisk (n=55) group differences.
Characteristics | Risk participants | Nonrisk participants | Statistical comparison | ||||||||
|
|
|
Chi-square ( |
|
|||||||
Age (years), mean (SD) | 19.5 (0.71) | 19.36 (0.68) | 1.38 (353) | N/Aa | .17 | ||||||
Class standing (freshman), n % | 169 (56.3) | 35 (64) | N/A | 1.0 (1) | .31 | ||||||
Sex (female), n % | 207 (69) | 30 (55) | N/A | 4.4 (1) | .04 | ||||||
Race and ethnicity (non-Hispanic, White), n %b | 240 (80.3) | 32 (58) | N/A | 12.7 (1) | <.001 | ||||||
Fraternity/sorority member (affiliated), n % | 107 (35.7) | 6 (11) | N/A | 13.1 (1) | <.001 | ||||||
|
207 (69) | 22 (40) | N/A | 17.1 (1) | <.001 | ||||||
|
Pain reliever (endorsed) | 151 (50.3) | 18 (33) | N/A | 5.8 (1) | .02 | |||||
|
Tranquilizer (endorsed) | 66 (22) | 5 (9) | N/A | 4.8 (1) | .03 | |||||
|
Stimulant (endorsed) | 53 (17.7) | 2 (4) | N/A | N/A | .007c | |||||
|
Sedative or barbiturate (endorsed) | 23 (7.7) | 1 (2) | N/A | N/A | .15c |
aN/A: not applicable.
bResponse missing for 1 risk participant.
cFisher exact test.
A total of 28,701 EMA reports were completed by 352 of the 355 participants; data were not available from 3 risk-group participants (2 did not return their data collection device and data from 1 participant’s device was not retrievable due to a password error). An important feature of the EMA protocol was collecting self-initiated reports; therefore, the app did not restrict time between reporting and allowed participants to complete several surveys within a short time frame. Upon examination of the collected data, the study team deemed it reasonable for 2 reports to be completed within 5 min of each other. Any reports beyond the second that were completed within 5 min (ie, the third or greater report) were removed. This resulted in the removal of 1.1% (329/28,701) of the obtained reports that we did not plan to include in any analysis stemming from the broader project.
Some report submissions did not adhere to the EMA protocol. Given our goal of documenting the utility of the protocol, we focus only on the reports that were obtained during the scheduled reporting periods and only the follow-ups that were completed within the instructed time frame. Although the typical reporting period was scheduled for the 28 days following the first laboratory session, some participants continued reporting until they returned for the second laboratory session. We removed a total of 2147 reports that were completed outside of the participants’ designated reporting days. The resulting 91.4% (26,225/28,701) of obtained reports were associated with 439 completed follow-ups; 64.0% (281/439) of the follow-ups were completed within 15 min of being sent (following the associated report) and were retained in this analysis.
In line with our sampling strategy, risk-group participants (143/297, 48.1%) were significantly more likely than the nonrisk counterparts (4/55, 7%) to endorse any prescription drug misuse intentions in daily life across the four medication classes (Fisher exact
Overall, participants demonstrated consistent engagement with the EMA procedures and returned an average of 74.5 reports (SD 23.82; range 10-122).
Ecological momentary assessment completion by risk (n=297) and nonrisk (n=55) group status.
Variables | Risk participants | Nonrisk participants | Statistical comparison, |
|
Reports, mean (SD) | 73.65 (24.07) | 79.11 (22.01) | –1.57 (350) | .12 |
Reporting days, mean (SD) | 26.50 (4.05) | 27.07 (3.05) | –0.99 (350) | .32 |
Completion ratea | 0.69 (0.19) | 0.73 (0.18) | –1.48 (350) | .14 |
aCalculated as the number of completed reports divided by the expected number of reports based on assigned reporting days.
Participants in the risk versus nonrisk groups reported similar experiences with procedures and uniformly positive experiences. As shown in
Ecological momentary assessment experiences by risk (n=298) and nonrisk (n=55) group status.
Variables | Risk participants | Nonrisk participants | Statistical comparison, |
||
Reflecting typical daily lifea, mean (SD) | 2.13 (0.74) | 2.33 (0.70) | –1.83 (351) | .07 | |
User-friendly devicea, mean (SD) | 2.87 (0.37) | 2.84 (0.37) | 0.61 (351) | .54 | |
|
118 (39.6) | 28 (51) | 2.5 (1)b | .12 | |
|
Helpful contacta, mean (SD) | 2.94 (0.27) | 2.86 (0.36) | 1.16 (34.76) | .25 |
aRated on scales from 0 to 3.
bThis result is from χ2, not a
We examined reactivity by testing whether prescription drug misuse intentions and behavior were more or less likely to be endorsed by participants over time, in terms of their reporting length (the number of days) and reporting amount (the number of reports). Specifically, reactivity was documented with prescription drug misuse intentions decreasing as a function of the number of reporting days (
There was a robust association between participants engaging in prescription drug misuse in the past year at the baseline assessment and during the daily life procedure: Reporting misuse behavior (of one or more medication classes) in daily life on the EMA follow-up was significantly more likely among participants who had indicated any past-year prescription drug misuse (90/255, 35.3%) than among those who did not (1/97, 1%; Fisher exact
This study presented the development and implementation of an app installed on study-owned devices to collect participants’ momentary experiences at the time of their prescription drug misuse intentions and behaviors. This study documented participants’ EMA completion rates and provided evidence for compliance and acceptability among participants in both risk and nonrisk sampling groups. Low levels of reactivity have been reported (described later). In addition, we presented the initial construct validity of our EMA approach by linking reports of misuse behavior in daily life to prescription drug misuse reported on a standard survey. Together, the results of this study provide early support for the feasibility and value of collecting young adults’ reports of prescription drug misuse intentions and behavior in daily life.
Although EMA still relies on self-report, this method offers improvement over more traditional assessments by minimizing reliance on memory and collecting situational factors at the time of the behavior of interest. Accordingly, these designs can provide specific momentary information to identify
Our sample was highly committed to the overall study and was particularly engaged with reporting in daily life. From the start of recruitment, interest in the research was high. We maintained continual interest from potential participants. Positive engagement was further reflected by nearly all participants returning to the second laboratory session with their device and saying they would recommend the experience to others. We obtained average completion rates that were similar to a pooled compliance rate on the basis of many EMA protocols that examined different substances [
The results indicated some evidence of reactivity to the momentary reporting procedure, as has been found in reporting other substances using EMA approaches (eg, alcohol) [
In testing adherence to the method here, we conservatively focused on reports that matched protocol training time frames. The reports we plan to include in future studies with this data set will be preregistered on the Open Science Framework and will depend on particular research questions. When we investigate momentary experiences as predictors of prescription misuse behaviors in daily life, we plan to restrict analysis of the focal misuse behavior outcomes to those that were reported on follow-ups completed within 15 min of being sent (consistent with protocol training). Alternatively, studies that are designed to test background factors, or other person-level characteristics, associated with risk of engaging in prescription drug misuse in daily life would not need to be bound by these temporal restrictions. Indeed, for certain research aims, having more instances of the behavior available would result in more representative findings and more powerful statistical tests. These trade-offs should be considered and justified by the purpose of this study.
We note that our EMA reports were identical regardless of whether participants were responding to a prompt or self-initiating a report. Thus, we could not determine the completion rates for signal-based versus event-based reporting. We were also unable to establish how many participants completed event-based assessments (or the associated follow-ups). Another protocol development decision was that we did not have participants indicate retrospectively whether prescription drug misuse had occurred since their last report, that is, a coverage strategy [
Our sample was representative of the college campus from which it was drawn but is not necessarily reflective of students at other types of higher education institutions or of nontraditional students. Additional work will be needed to understand the applicability of this method to individuals across different contexts and locations. A practical limitation of using a separate device for EMA data collection is that the approach may be more burdensome for participants and requires greater investment in study materials. We expect that these considerations were offset by the anonymity and protection of the participant data provided by a separate device. Although reported behavior instances might be lower if the device was not readily available to the participant, the responses obtained may be more forthright due to security and privacy assurances. Nevertheless, in light of the recent successful apps on young adults’ own smartphones to assess drinking in daily life [
Establishing real-time mood states, pain symptoms, stressors, and other substance behaviors as predictors of college students’ prescription drug misuse is critical for obtaining a more precise understanding of the behavior in daily life situations. In line with the social ecological perspective [
ecological momentary assessment
National Survey on Drug Use and Health
This work was supported by the National Institute on Drug Abuse of the National Institutes of Health under Award Number R01DA042093. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
None declared.