#hope-s #alcohol potential #Biases #systemic #problem # Reducing Drinking Among People Experiencing Homelessness: Protocol for the Development and Testing of a Just-in-Time Adaptive Intervention https://www.researchprotocols.org/2020/4/e15610/ JMIR RESEARCH PROTOCOLS Michael S Businelle1*, PhD; Scott T Walters2*, PhD; Eun-Young Mun2, PhD; Thomas R Kirchner3, PhD; Emily T Hébert1, DrPH; Xiaoyin Li2, PhD ---- **Thoughts and Questions I have:** - I felt quite troubled by this. Although the intention seems good.. to provide interventions for people with drinking problem. - Not expert here, i wonder why do people develop alcohol dependency? - Drinking cause them to be homeless? Or they drink because they are homeless? - Do we see substance addiction as a symptom of society / relationship / systemic issues or a flaw, deficit in the person? - Suppose that this system work, and can detect drinking and rank the person as high or low risk, what will be the implication? - Is that a little like China Credit System? Will this be abused? If this person have "high risk of drinking", then that person will be stigmatised and will that cause a spiral/revolving issues? - The Algorithm is built on the assumption that those theories are valid... how do we know that there are no biases built in? How is it decided that individual have what level of risk? - "==Different levels of risk across individuals== will also be included to examine how intraday risk gets intensified or ameliorated by ==personal trait-level variables== (eg, sex, psychosocial resources, stress/adversity, and negative mood). We expect that analyses of SCT and ecological constructs (eg, affect, expectancies, self-efficacy, and proximity to drinking areas), gathered during random EMAs and breadcrumb trail geolocation, will identify patterns that predict drinking in near real time" ----- [[15-02-2022]] Objective: The aim of this study is to (1) identify variables (eg, affect, stress, geolocation, and cravings) that predict drinking among homeless adults (phase I), (2) develop a mobile intervention that utilizes an algorithm to identify moments of risk for drinking and deliver treatment messages that are tailored to the individual’s current needs in real time (phase II), and (3) pilot test the intervention app (phase III). Methods: In phase I, adults experiencing homelessness with an AUD (N=80) will complete baseline, equipment, 2-week, and 4-week follow-up visits in person. Participants will be prompted to complete five daily ecological momentary assessments #EMA on a study-provided smartphone for 28 days. The smartphone app will collect GPS coordinates every 5 min for the entire 28-day study period. Participants will wear a transdermal alcohol sensor that will objectively measure alcohol use. In phase II, we will use phase I data to develop an algorithm that identifies moments of heightened risk for drinking and develop treatment messages that address risk factors for drinking. Phase III will pilot test the intervention in 40 adults experiencing homelessness with AUD. ---- **Researcher want to examine interaction between psychosocial variables and alcohol use. Based on a social cognitive theory constructs (about the patient's self...) and social-ecological model constructs (self to others?). Then construct these as a traits.. i.e how "Self" state correlate with behavior (drinking)** *(?Are we seeing alcoholism as a problem with the person, or symptoms of a relations)* - During phase I, we will use smartphones and passive sensing to continuously monitor geolocation and to measure psychosocial variables (eg, negative affect, stress, and urge to drink) and alcohol use in a sample of 80 adults experiencing homelessness enrolled in shelter-based treatment programs. EMAs will be used to examine the moment-to-moment relationship between social cognitive theory (SCT) constructs (eg, affect, abstinence motivation and self-efficacy, alcohol use expectancies, and cravings) [36,37], social-ecological model constructs (eg, current proximity to previous drinking areas or alcohol outlets, social setting, and social support) [38,39], and drinking. We will also assess these constructs as trait-like variables at baseline to examine how trait and state processes interact to influence drinking behaviors. **Then, design algorithms that trigger interventions to reduce those traits, to reduce behavior.** - In phase II, we will use this information to develop optimized risk algorithms and develop tailored treatment messages that can be provided before anticipated alcohol use given personal, situational, and environmental triggers (eg, presence of drinking others, location, elevated positive or negative moods, and high stress). **Lastly test out the system. Sending EMIs (messages only) at the end of EMAs.** - In phase III, we will pilot test the smartphone app for utility, satisfaction, and preliminary effectiveness in another sample of 40 homeless adults enrolled in shelter-based treatments. Algorithm-driven treatment messages will be automatically delivered at the end of EMAs. Phase III participants will complete a qualitative interview that will examine their opinions of the app design and intervention content and ways to improve the app user interface. In phases I and III, self-reported alcohol use will be validated via a transdermal alcohol sensor (ie, Secure Continuous Remote Alcohol Monitor [SCRAM], Alcohol Monitoring Systems, Inc) worn by participants. #Motivational-interviewing Motivational- (eg, derived from motivational interviewing) and self-efficacy- (eg, derived from SCT) themed messages are commonly used in technology-based alcohol interventions [40]. Interventions for AUDs have often drawn from these underlying theories, but mobile interventions have the additional strength of fostering self-regulation through triggering goal salience and re-evaluation of short- versus long-term goals [41]. Recent work has indicated that smartphone apps (eg, the ACHESS app) that incorporate preloaded videos, interactive features, and weekly check-ins can reduce heavy drinking days in alcohol dependent adults [42] and college students [43]. [[202202151536 The use of Geo-location to trigger EMI, to remind user about their training, to reduce relapses]] Others have begun to use geolocation data to alert individuals with SUDs about potentially high-risk environments [44-46]. For example, some SMS text messaging interventions for AUD have focused on encouraging self-regulation and planning before drinking episodes [47,48]. For those who are enrolled in treatment, the messages can reinforce treatment concepts. For those who are not enrolled in treatment, messages may serve as a primary intervention (or at least a reminder of past concepts) to short-circuit alcohol use before it occurs. **Sounds like the app is trying to be like the "angel" on the patient's shoulder. Reminding them to keep it up when faced with multiple temptations** Our central hypothesis is that alcohol use is strongly affected by moment-to-moment risk and protective factors, and we can use EMAs to identify and automatically intervene during moments when people are at high risk for drinking. Our hypothesis is based on preliminary findings from our own studies among homeless [33,49], justice involved [50], and socioeconomically disadvantaged safety net hospital patients [31,32]. If effective, this smartphone treatment app could significantly improve treatment engagement, drinking outcomes, and quality of life among adults experiencing homelessness with AUDs. # What are the EMA used? Researcher collect lots of data. ![[Screenshot 2022-02-15 at 3.45.25 PM.png]] - EMA items completed on the phone (see Table 3) will assess SCT constructs (eg, affect, abstinence motivation and self-efficacy, expectancies, and cravings) and social-ecological model constructs (eg, proximity to previous drinking areas, social setting, and social support) to identify key variables and time- and location-dependent fluctuations in variables, which will be used to predict study outcomes. Most of these items have been used in our previous studies and studies from other labs [22,68]. ==Three types of EMAs will be used: daily diary, random sampling, and event sampling.== Daily diary and random sampling EMAs will be initiated by the phone. The phone will audibly and visually cue these EMAs for 30 seconds. If the participant has not responded after 5 prompts, the assessment will be recorded as missed. Event sampling is initiated by participants if/when they consume their first drink in a day. On average, random and event sampling assessments take 2 min to complete, and daily diary assessments take less than 5 min to complete. - Daily Diary - Daily Diary EMAs will be completed each day 30 min after the participant’s self-reported wake time; questions will ask about the previous day (ie, “yesterday”) and current (ie, “right now”) experiences. Alcohol consumption will be assessed with the item “Did you drink any alcohol yesterday?” If the participant answers “yes,” he/she will be prompted to indicate the number of standard drinks that were consumed. EMA reports have generally been seen as valid measures of drinking, even when participants are intoxicated [69,70]. See Figure 2. Additional items will assess sleeping arrangements from the prior night (example answer options: friend or family member’s house or apartment, homeless shelter, jail, car, outside on the street), quality of sleep the previous night, social support and types of social interactions, stressors, other substance use, and substance abuse treatment attendance (see Table 3). **To collect Mood state..** Random Sampling- Participants will be prompted at random times to complete EMAs 4 times each day, scheduled to occur during the participant’s normal waking hours. Participants will rate their affect by indicating the extent to which they agree or disagree with each of 13 statements at the moment: I feel irritable, happy, content, angry, sad, worried, miserable,restless, stressed, hostile, calm, bored, anddepressed (most items are from the circumplex model of affect [71]). In addition, participants will describe their current environment (eg, shelter, work, outside, or bar) and social setting (eg, alone, with others, or with others who are drinking). Alcohol urges (ie, “I have an urge to drink alcohol”; answer options range from strongly disagree to strongly agree), alcohol availability (ie, “Alcohol is available The use of Transdermal Alcohol Monitor to track user alcohol use. https://www.scramsystems.com/monitoring/scram-continuous-alcohol-monitoring/ --- Phase III App The phase III intervention app will have multiple components including (1) an on-demand “Tips” function/button, (2) a “Helpful Websites” function/button, (3) a “Call Staff” function/button, and (4) ==an algorithm that will use recent EMA responses and geolocation to assess current risk for alcohol use and automatically push relevant tailored messages to participants==. The phone will record date/time when each of the components is accessed. See Figure 4 for the anticipated phase III home screen. ![[Screenshot 2022-02-15 at 3.51.23 PM.png]] # How do they calculate the risk? How does the algorithm work? Any potential biases in the algorithm? Risk Algorithm The algorithm used to guide the just-in-time treatment messages will be similar to the algorithm that was developed for the Smart-T smoking cessation app [31]. Specifically, the algorithm will estimate risk for alcohol use using variables identified in phase I. In Smart-T, we attempted to develop risk algorithms that could predict smoking at 8, 12, and 24 hours before the lapse, but these algorithms were far less sensitive than the 4-hour lapse prediction algorithm. The resulting Smart-T algorithm combined six EMA variables (ie, urge, stress, cigarette availability, alcohol use, motivation to quit, and proximity to others smoking) to successfully predict 80% of all smoking lapses within 4 hours of lapse occurrence (false positive rate=17%) [31,32]. What kind of messages will be pushed out to the user? Based on risk level Level 1 - focus on increasing motivation for abstinence, avoiding people/places/things that may trigger alcohol use, benefits of sobriety, advice on ways to escape high-risk situations, and advice to seek support from others [50]. These messages will complement the treatment themes for those who are in an alcohol treatment program Level 2 - if the algorithm determines that there is heightened risk for imminent (eg, within the next 4-8 hours) alcohol use. These messages will focus on in-the-moment distraction, reframing, immediate help-seeking, planning, and other tools to reduce craving. The highest rated indicator/trigger of alcohol use in that moment will be the topic of the level 2 tailored treatment messages. For example, if a participant reports low motivation for sobriety and average ratings on the other variables, they will receive a message that aims to boost motivation. An example may read: “You said that family was an important reason for staying sober. You’re looking forward to a better life!” Likewise, if exposure to drinkers is an identified alcohol use trigger and a participant reports that he/she is near individuals who are consuming alcohol, he/she may receive a tailored suggestion on how to escape that high-risk situation, such as “You said that removing yourself from a situation was often helpful in managing cravings. Some people decide to get out of the situation, before they are tempted to drink.” Level 3- Participants will receive level 3 messages when they report recent drinking. Level 3 messages will focus on reframing the drinking episode as a learning experience and considering strategies for handling the situation differently in the future. We will draw from best-practice recommendations around message content and tone [83]. Our past interventions have contained hundreds of possible message combinations, depending on a person’s baseline profile and current responses. ----