9  Literature Review Summaries

9.1 Opiate Use Patterns Review:

  • The purpose of this review is to inform how we ask people about opioid lapses (e.g., creation of time windows to report a lapse), routes of administrations, and types of opioids used for non-medical reasons Opiate Use Patterns First, opiate use patterns appear to be not as clear cut as one would think and there is little research on rates of drug administration. One study specifically examines opiate consumption at a heroin maintenance program in Geneva (Perneger et al., 2000). At this program heroin users are prescribed heroin and oral opiates as needed without any limit to the amount prescribed (wow!). On about 50% of total patient-days (days researchers documented consumption) patients used a combination of intravenous

    1. heroin and oral opiates (methadone and/or oral morphine; Perneger et al., 2000). This brings me to my first observation: participants may not be using just one opiate type. Perneger et al.
    1. found that the average daily dose of IV heroin was about 500 mg taken in three injections throughout the day. This dose was slightly lower when taken in conjunction with oral opiates, which lowered IV users’ number of injections by one or two injections on average. One thing to note about this research is that it is on heavy users. So, my second thought is: participants who are relapsing after periods of abstinence may have different consumption patterns than regular users. This deviation could go either way – resulting in higher (i.e., they could binge one night on the entirety of drugs they pick up) or lower consumption (i.e., they have a slight lapse and only use one time). Evidence from the previously mentioned study supports the latter. Perneger et al.
    2. found that when higher amounts of oral methadone was administered, fewer amounts and number of injections of IV heroin was documented. One patient, representing a typical pattern seen in the study, was able to decrease their daily injections from three injections to two to one by increasing their oral methadone dosage each time (Perneger et al., 2000). It is not clear, however, if this same pattern would be seen with other opioid maintenance medications. Unfortunately, this information is specific to IV usage and heroin. A 2004 study on opioid prescription pill users in a methadone maintenance program found that oxycodone and codeine were the most commonly abused pills (Brands et al., 2004). The same study reports that the average number of pills taken per day prior to treatment was about 20 – 22 tablets per day (Brands et al., 2004). Table 1 (taken from a different study) illustrates the average daily dose (mg), heaviest dose (mg) in a day, and route of administration used for various opiate prescription pills (Sproule et al., 2009).

Table 1. From Sproule et al. (2009). Users of controlled-extended release oxycodone take on average higher daily doses than other opiate prescription pills. Other issues to consider are route of administration and variation in duration of the high for different opiates. According to an older research report put out by the National Institute on Drug Abuse, a typical heroin user will inject up to four times a day (National Institue on Drug Abuse, 2000). There does not appear to be the same available data for other routes of administration, but with the lower bioavailability of the drug seen in routes like smoking (Rook et al., 2006), it is likely they are used more frequently. Sniffing/snorting heroin is the second most common route of administration followed by smoking (National Institue on Drug Abuse, 2000). However, one newer study suggests sniffing is the most common route of administration, even surpassing injection (Ihongbe & Masho, 2016). With opioid prescription pills, chewing is the most common route of administration (Gasior et al., 2016; Sproule et al., 2009). However, some research suggests choice in route of administration is influenced by external factors. One study reported that route of administration varied as a function of both type of prescription opioid and whether the user was from a rural or urban community (Young et al., 2010). For example, rural users were more likely to snort hydrocodone tablets whereas urban users were more likely to swallow/chew the tablets (Young et al., 2010). One final consideration is the comorbidity of drug use common in opiate addiction (Brands et al., 2004; Monga et al., 2007). In one sample of 114 opiate users, 97% reported using other non-opiate drugs in the past 30 days, with alcohol (70%), cannabis (64%), and benzodiazepines (60.5%) being the most prevalent (Fischer et al., 1999). A larger study on 1075 opioid users reported that 54% of the sample regularly used benzodiazepines and 59% regularly used stimulants (Marsden et al., 2000). I do not think this is a concern currently for us; however, I think it is worth noting as it is possible that certain drugs may have synergistic effects with opioids resulting in more or less frequent drug administrations. Ultimately, this still doesn’t answer our question about the exact dosing frequency of opioid users who are not injecting heroin. To help guide us in making decisions for RISK2, I have also attached a table exhibiting the half-lives of various opiates (Table 2) and a table of duration of action for the same opiates (Table 3). From these tables you will see that the average duration of action for each of these opiates is around 3 – 6 hours. As a result, I would recommend that a time window of three hours is sufficient for measuring opioid lapses if needed. Furthermore, due to lack of research on opiate administration/use patterns it might be worth it to discuss whether we want to collect this information. For example, when a participant reports their lapse, they could also report the type of opiate (i.e., prescription pill, heroin, other) and the route of administration used (i.e., snorting/sniffing, smoking, IV, other).

Table 2. From Trescot et al. (2008). Not listed above is heroin (diamorphine/diacetylmorphine) which has a half-life of .1 - .25 hours (https://www.aruplab.com/files/resources/pain-management/DrugAnalytesPlasmaUrine.pdf)!

Drug Duration of action

heroin 4 – 5 hours morphine 3 – 7 hours hydromorphone 4 – 5 hours oxycodone 3 – 6 hours (up to 12 hours with extended release) hydrocodone 4 – 8 hours fentanyl 1 hour codeine 4 – 6 hours buprenorphine up to 24 hours methadone prolonged use: 1 – 2 days https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6264977/ Table 3. Duration of action for each of the common opiates. –

9.2 EMA Pain Item Review

  • The purpose of this review is to determine the convergent and criterion validity for a daily EMA pain item, and to inform how we ask participants about their daily pain

A review of convergent and criterion validity for a daily EMA pain item Note: The following bibliography primarily focusses on research using a single EMA measure of pain intensity (i.e., daily “worst pain” question from BPI ). I could not find an example from a study on patients without chronic pain so the following draws on a variety of different types participant samples actively excluding cancer samples. Kratz, A. L., Murphy, S. L., & Braley, T. J. (2017). Ecological Momentary Assessment of Pain, Fatigue, Depressive, and Cognitive Symptoms Reveals Significant Daily Variability in Multiple Sclerosis. Archives of physical medicine and rehabilitation, 98(11), 2142–2150. https://doi.org/10.1016/j.apmr.2017.07.002 Sample: Multiple Sclerosis participants (N = 107) Measure modality: EMA and end-of-day diaries Measure: Pain intensity (1 question on EMA, 3 questions on end-of-day diary) 1. EMA: “What is your level of pain right now?” – 0 (“no pain”) to 10 (“worst pain imaginable”) 2. End-of-day: Asked participants to rate their daily “worst”, “average”, and “current” pain – 1 (“no pain”) to 5 (“very severe”) Measure source: EMA from Brief Pain Inventory (BPI) (Daut, Cleeland & Flanery, 1983); diaries from Patient Reported Outcomes Measurement Information System (PROMIS; Revicki et al., 2009 ). Validity type: Convergent Validity stats: Shared between-person variance for daily EMA and end-of-day pain intensity constructs, beta = 1.91, SE = 0.05, p < .0001, R2 = .798 Other notes: Also reported within (1.99, 41%) and between person (2.89, 59%) variance in EMA single pain item. Important to note that they used BPI “right now” instead of “worst pain” question on EMA but did ask about worst pain on end-of-day diary.

Kuerbis, A., Reid, M. C., Lake, J. E., Glasner-Edwards, S., Jenkins, J., Liao, D., Candelario, J., & Moore, A. A. (2019). Daily factors driving daily substance use and chronic pain among older adults with HIV: An exploratory study using ecological momentary assessment. Alcohol (Fayetteville, N.Y.), 77, 31–39. https://doi.org/10.1016/j.alcohol.2018.10.003 Sample: HIV positive participants (N = 55). Measure modality: EMA Measure: Pain intensity (2-part question) 1. “In the past 24 hours did you experience any pain?” – yes or no (if yes Q2 shown) 2. “Rate the pain you experienced at its WORST in the last 24 hours” – 0 (“no pain at all”) to 10 (“pain as bad as you can imagine”) Measure source: Questions taken from the Brief Pain Inventory (BPI) (Daut, Cleeland & Flanery, 1983). Validity type: Possibly criterion Validity stats: Multilevel models of predictors of daily worst pain include: 1. more daily number of drinks, B = 0.48, SE = 0.12, p < .001 2. less daily happiness, B = -0.21, SE = 0.09, p = .03 3. poorer sleep quality overall (person average not daily), B = -0.29, SE = 0.09, p < .01 4. lower confidence to cope with the pain without medication (person average not daily), B = -0.34, SE = 0.10, p < .01 5. no exercise, B = -0.67, SE = 0.25, p = .02

Lapane, K. L., Quilliam, B. J., Benson, C., Chow, W., & Kim, M. (2014). One, two, or three? Constructs of the brief pain inventory among patients with non-cancer pain in the outpatient setting. Journal of pain and symptom management, 47(2), 325–333. https://doi.org/10.1016/j.jpainsymman.2013.03.023

Sample: Various non-cancer short and long-term chronic pain patients obtained from oxycodone users registry (N = 777). Measure modality: Daily (not clear how administered) Measure: Pain intensity (3 item) 1. pain ‘right now’ 2. average pain in the past 24 hours 3. pain at its worst in the past 24 hours Response options were 0 (“no pain”) to 10 (“worst possible pain”) Measure source: Questions taken from the Brief Pain Inventory (BPI) (Daut, Cleeland & Flanery, 1983). Validity type:
Validity stats: Other notes: Study finds support for a two-factor model of pain intensity and interference. In introduction authors mention that “worst pain” item on BPI is commonly used as a single item measure. They do not expand on this thought though. I think this study is important because it seems to confirm that the research I found in my last draft can extend to non-cancer patients with varying degrees of pain. This is helpful because most of the research using the BPI (and single “worst pain” item) are on cancer patients.

Keller, S., Bann, C. M., Dodd, S. L., Schein, J., Mendoza, T. R., & Cleeland, C. S. (2004). Validity of the brief pain inventory for use in documenting the outcomes of patients with noncancer pain. The Clinical journal of pain, 20(5), 309–318. https://doi.org/10.1097/00002508-200409000-00005 - “The reliability of BPI data collected from non-cancer pain patients was comparable to that reported in the literature for cancer patients and sufficient for group-level analyses (coefficient alphas were greater than 0.70). The factor structure of the BPI was replicated in this sample and the relationship of the BPI to generic measures of pain was strong. The BPI exhibited similar relationships to general and condition-specific measures of health as did a generic pain scale (SF-36 Bodily Pain). Finally, the BPI discriminated among levels of condition severity and was sensitive to change in condition over time in arthritis and LBP [lower back pain] patients.” Above is from abstract, have not gotten access to this article yet. It was cited in the above study and seems to work to again support the BPI for use in non-cancer pain patients.

Atkinson, T. M., Mendoza, T. R., Sit, L., Passik, S., Scher, H. I., Cleeland, C., & Basch, E. (2010). The Brief Pain Inventory and Its “Pain At Its Worst in the Last 24 Hours” Item: Clinical Trial Endpoint Considerations. Pain Medicine, 11(3), 337–346. https://doi.org/10.1111/j.1526-4637.2009.00774.x - “Throughout BPI validation, the item”pain at its worst in the last 24 hours” has consistently shown the highest degree of internal consistency, with Cronbach’s alpha coefficients ranging from 0.77 to 0.90.” (See Table 2) - “When examining test–retest reliability, the”pain at its worst in the last 24 hours” item had acceptable reliability during validation of the BPI in German and Taiwanese subjects (0.80 and 0.96, respectively).” - “The test–retest reliability of the”pain at its worst in the last 24 hours” item is highest when administered over a short time span (i.e., hourly or daily), suggesting these may be the optimal intervals during future drug trials.” - “During instrument development,”pain at its worst in the last month” was tested as an item rather than “pain at its worst in the last 24 hours.” The “pain at its worst in the last month” item was found to be highly related to pain interference items, consistent with the pattern of relationship between “pain at its worst in the last 24 hours” and pain interference items in subsequent studies.” - Note: these above findings are in the context of validating the BPI worst pain item as a measure of pain reduction in clinical treatment. Not sure how this may affect what we are looking for.

9.3 Progress Tracking Review

  • The purpose of this review is to summarize a literature search related to progress tracking and messaging, and to inform how we will allow and encourage participants to track their progress in the study with respect to study task completion Progress tracking and messaging literature search (The purpose of this review is to summarize a literature search related to progress tracking and messaging, and to inform how we will allow and encourage participants to track their progress in the study with respect to study task completion) Progress tracking: • We can disassociate points from $ bonuses. Merely framing something as a game (e.g., using visual signals of gaming) seems to be motivating (e.g., Lieberoth, 2014). It might make sense for us to keep the current payment scheme but additionally reinforce completion of individual surveys with a more detailed points tracker (which when clicked can still show amount earned). • People should accumulate points for incremental progress. Every time people complete a survey / do something we want them to do, they should get points even if there is no financial earning. • We should provide positive messages and/or images upon completion of every survey. We need to positively reinforce completion of surveys. For example, when people complete a survey maybe we show a star on the daily survey button, when people complete all daily surveys in a week, the tracker bar could turn green. • We should take inspiration from existing apps (e.g., fitbit, strava). Private companies have poured a ton of money into testing progress trackers so it will be wise for us to look at how other apps function and use similar design features or language. • I didn’t find great papers on ‘streak’ feedback but I know this is popular on apps like Duolingo, so we might consider it (although it could backfire if people aren’t consistent… and maybe we don’t care that much about streaks) On messaging (e.g., if we incorporate text under the tracker or for notifications/reminders) • We should incorporate meaning/value information in messages. Occasionally reminding people why we want them to complete surveys will probably be motivationally beneficial (e.g., ‘Filling out surveys will help improve scientific understanding of relapse risk’; general support for this comes from theories of motivation (SDT, expectancy-value) but also supported by evidence in ‘gamification’ literature like Mekler et al., 2013). • We should vary message content based on progress. Shifting the reference point strategically can increase motivation (Wallace & Etkin, 2018). For example, we should be vague about goals when people are doing badly, so if people have completed 0 surveys we say something like “Complete more surveys to earn a bonus this week” rather than something like “Complete at least 5 surveys to earn a bonus this week” (or whatever our compensation scheme is). When people are doing well, we should be specific and focus on what remains versus how much progress they have already made… so e.g., when 6/7 daily surveys are done, rather than saying something like ‘You’ve completed six of seven surveys required to earn a bonus.’ we say ‘Complete one more daily survey to earn a bonus’ • If our in-app notifications are attention-grabbing, we should allow people to control the frequency of and/or disable them. If notifications annoy people, and if people habitually ignore notifications, they will not work.

9.4 Risky Dates Review

  • The purpose of this review is to summarize the literature surrounding meaningful dates and its relationship to substance use. It was also used to inform our decision to not ask participants to report potentially “risky dates”)

Event -Based Substance Use Often referred to in the literature as event-based or event-level substance use, there appears to be a general consensus that specific meaningful dates, events, and contexts are associated with increased likelihood of substance use on or immediately after said date, event, or context. Somewhat unsurprisingly, the majority of this event-based substance use research has focused on drinking patterns among college-aged populations (Stanesby, Labhart, Dietze, Write & Kuntsche, 2019). Consequently, the events that have received the most attention are those that have been identified as having a unique relevance to college-aged populations (21st birthdays, Spring Break, Frat Parties, Collegiate Sporting Events)(Patrick & Azar, 2018). Despite this methodological skew towards college-aged populations, I did come across a handful of event-based studies that included clinical, older, and more diverse populations. The events investigated in these studies included Holidays, Sporting Events, and Disaster/Terror Events.

Holidays Predictably, people tend to drink more on certain holidays. In fact, a report by Mothers Against Drunk Driving put out a report on the top ten holidays with the highest number of alcohol related accidents 1. New Years Holiday, 2. Labor Day, 3. Fourth of july 4. Super Bowl Sunday 5. Christmas 6. Memorial Day 7. Fourth of July (2002) 8. Thanksgiving, 9. Halloween 10. St. Patrick’s Day (Lapham, Forman, Alexander, Illeperuma & Bohn, 2009). However, increases in high-intensity drinking may depend on the holiday in question. One study observed that on family-oriented holidays such as Thanksgiving and Christmas, the number of young people who consumed alcohol increased but the average number of drinks consumed per person (counting only those who drank) actually decreased (Goldman,Greenbaum, Darkes, Brandon, Del Boca, 2011). In contrast, on holiday weeks that included a Halloween-like holiday, New Year’s Eve, and the Fourth of July, the average number of drinks consumed per drinker increased significantly compared with non-holiday weeks (Goldman et al., 2011). Though it is unclear whether these trends would translate into increased risk of opioid lapse given the considerable difference in social use of the two substances, if RISK 2 participants are polysubstance users, it is possible that increased risk of holiday drinking would increase the risk of opioid lapses. Interestingly, there is evidence that using medication to assist with substance use disorder can be helpful in reducing holiday substance use. One study investigating the effect of Naltrexone combined with psychosocial support on holiday drinking in treatment seeking populations with AUD found that taking Naltrexone was associated with near-complete abstinence during the holidays (Lapham et al., 2009).

Sporting Events Sporting events are also associated with risky drinking behavior. Among college football fans, particularly men, drinking on days of high-profile football games is as heavy as alcohol consumption on other well known drinking days, including New Year’s Eve and Halloween weekend (Neal & Fromme, 2007). Another study that looked at three years worth of drinking data found that heavy drinking male participants drank significantly more of Super Bowl Sundays compared to other heavy drinking days (Saturdays) across all three years, while women’s drinking was higher in only one of those three years (Dearing, Twaragowski, Smith, 2014). These findings suggest sporting events present a significant risk for risky substance use, especially among men. Disaster/Terror Events and Substance Use Given that we will have negative anniversaries as a category on the risky dates question, I looked at articles that covered the effect of disaster/terror events on substance use behaviors. Most of the research I found on this topic covered substance use in participants exposed to terror attacks, specifically 9/11 and the 1995 Oklahoma City Bombing (Vlahov et al., 2002; Pfefferbaum et al.,2002; Welch et al., 2014; DiMaggio & Galea, 2009). These studies generally found that those individuals who reported high or very high exposure to the event reported increased smoking and increased alcohol intake. Critically, these effects on participant substance use behavior were observed in those individuals with high or very high exposure to the event,up to a decade after exposure, regardless of PTSD diagnoses (Welch et al., 2014; Welch, Zweig, McAteer & Brackbill., 2017). Similar increases in substance use behavior have been observed in Jewish settlers who were directly injured during the intifada, particularly male settlers who had “sustained traumatic events with a great risk to personal integrity” (Keinan-Boker, Kohn, Billig & Levav, 2011). It is important to note that these studies were unable to establish any sort of causal relationship between the terror attacks and increased substance use. They do, however, suggest that traumatic exposure to a specific event is an important risk factor for risky substance use, especially if the individual is directly impacted by the event. 

While these studies demonstrate a significant relationship between increased substance use and certain meaningful dates and events, it is worth noting that the majority of these studies did not explicitly discuss relapse. Additionally, I did not find any research papers that had actually measured the relationship between sobriety anniversary and relapse. In the course of my search, I found a few non-research article sources, namely blog posts and online recovery resources (See References), that identify sobriety anniversaries, holidays, and recovery milestones as possible triggers for relapse. This dearth in the literature, especially as it relates to opioid use disorder, underscores the value of collecting this data as part of RISK 2. However, weighed the lack of clarity as to what might constitute a “risky” date to individuals in recovery, we have decided not to ask about risky dates in RISK 2.

9.5 Motivational Constructs Review

  • The purpose of this report was to consider how the RISK 2 measures are directly and indirectly assessing motivation in the context of motivational construct literature

Motivational Constructs in Risk 2 Across the daily, monthly, and intake survey, we have one item that directly measures participants’ motivation “How motivated are you to completely avoid using opioids for non-medical reasons?”. However, given the “growing consensus that motivation involves multiple constructs”, it is worthwhile to consider measuring additional “theoretically relevant predictors of motivation” (DiClemente, Doyle, & Donovan, 2009). Norozi et al. (2017) sought to investigate potential predictors of motivation by focusing on three cognitive constructs, perceived avoidance self-efficacy, perceived social support, drug quitting outcome expectancies. Of these three cognitive constructs, two are represented in the current RISK 2 surveys. Perceived avoidance self-efficacy refers to an individual’s belief in their ability to avoid drug use, particularly in high risk situations (Norozi et al. 2017). Several studies have shown that perceived avoidance self-efficacy is highly related to treatment motivation and commitment to abstinence (Germeroth et al. 2019; Chauchard et al. 2013; Laudet & Stanick, 2010). Because perceived avoidance self-efficacy is believed to fluctuate over the course of treatment, perceived avoidance self-efficacy is a better predictor of overall treatment outcomes when it is measured at multiple points during and after treatment (Ilgen etal, 2005; McKay et al 2013). In RISK 2, we ask participants to report their perceived avoidance self-efficacy in the Daily Survey using a single item (“How confident are you in your ability to completely avoid using opiates for non-medical reasons?”). In the Monthly Update and the Intake Survey, we have a set questions concerning Abstinence Confidence/Efficacy: “How satisfied are you with your progress toward achieving your recovery goals?”, How motivated are you to completely avoid using opioids for non-medical reasons?, How confident are you in your ability to avoid using opioids for non-medical reasons?, Do you intend to completely avoid using any other drugs? The other construct assessed in both Norozi et al. and the current RISK 2 measures is perceived social support. Norozi et al. used the Multidimensional Scale of Perceived Social Support (MPSS), a 12-item measure that asks participants to rate the general social support adequacy from family, friends, and significant other. Several studies have found that strong social support is associated with greater treatment retention, behavioral improvement during treatment, and better treatment outcomes (Broome et al. 2002, 2010). However, there is also evidence to suggest that the relationship between social networks and motivation may undermine recovery efforts if the participant maintains substance-using network members (Tracey et al. 2009). In RISK 2, Social Support will be assessed in the Intake Survey and Monthly Update using the Social Connectedness Scale, Relationship Assessment Scale, and individual items from the Social Networks Survey (In the past month, how satisfied are you with… The support you get from your friends, Your personal relationships). Additionally, the Contacts Report has four items that directly ask about social support (This person… is a person I can tell private things to, helps me when I need it, would support me through bad times, makes me feel good about myself even when mess up) as well as an item that asks participant to report whether interacting with the contact put their recovery at risk (How risky are your interactions with this person to your recovery goals?). The final cognitive construct Norozi et al. considered was outcome expectancies. Several studies have demonstrated that addicts’ perceptions of the costs and benefits of quitting can predict treatment motivation (Battjes et al., 2003; Chauchard et al., 2013; Jackson et al.,2003; Laudet & Stanick, 2010). In Norozi et al., the researchers created their own Drug Quitting Consequences Questionnaire, a 23-item measure “that assesses the costs and benefits of quitting substance use”. Following multivariate analyses, outcome expectancies emerged as a significant predictor of treatment motivation. However, a more recent study by the same group of researchers using the same scale found drug quitting outcomes was not a significant predictor of actual drug abstinence (Eslami et al, 2018). This inconsistency between predictors of treatment motivation and predictors of treatment outcomes is seen throughout the literature, with the predictive ability of outcome expectancies on either treatment motivation or treatment outcomes varying widely depending on factors such as the type of outcome expectancies measure, subject demographics, the stage of treatment and the behavior being assessed. The current RISK 2 measures do not directly assess outcome expectancies. While I did come across one scale that compressively assesses drug quitting outcome expectancies, The Perceived Risks and Benefits Questionnaire (PRBQ), it is both long (39 items) and highly specific to nicotine addiction (See Figure 1). However, it might be worthwhile to consider creating our own questions based on the different subscales within the PRBQ (i.e. Health Benefits, General Well Being, Self-Esteem) which could be added to the intake survey and monthly survey. I am imagining these items as being similar in theme to the items in the WHO Quality of Life Scale we are currently using, but with a more future-thinking orientation. Example Instructions: Use the scale (1= Very unlikely…7=Very Likely) to rate how likely each item would be if you were to achieve your recovery goals - I will lower my chance of developing future health problems - My overall well-being will be higher - I will have a greater sense of self esteem Ultimately, there could be some value in gathering information on how participants outcome expectancies might change over the course of RISK 2. Gathering such information would not only draw attention to the potential role of outcome expectancies in predicting opioid lapses, but also allow us to identify how outcome expectancies interact with constructs like self-efficacy and social support. However, for the narrower purposes of RISK 2 I did not find sufficiently compelling evidence that the value of measuring outcome expectancies would justify the added burden of a drug quitting outcome expectancies scale.