Mobile Health Intervention to Address Chronic Pain Among Those Who Engage in Hazardous Drinking: A Pilot Study

Abstract

Background:

Hazardous drinking has been associated with chronic pain in community and medical samples. The purpose of this study was to develop a novel, integrated mobile health intervention that improves pain management and reduces hazardous drinking that may be implemented in primary care settings.

Methods:

Forty-eight participants with moderate or greater chronic pain and hazardous drinking were recruited from primary care clinics and through social media sites. Following baseline assessment, participants were randomized to a counselor-supported smartphone app intervention (INTV) or a counselor delivered treatment-as-usual control condition (CTL).

Results:

Results supported the feasibility and acceptability of the smartphone app intervention. Participants found it easy to use, reported high levels of satisfaction, and showed high levels of engagement with the app. Between-group effect size estimates at follow-up showed small effects for the intervention on pain ratings. However, using clinically meaningful change thresholds of 30% and 50% improvement in pain scores, 38% and 25% respectively of those in the INTV condition showed reductions compared to 20% and 12.5% respectively in the CTL condition. Effect size estimates did not indicate intervention superiority on alcohol outcomes as participants in both conditions showed considerable reductions in drinking over the course of the study.

Conclusions:

Results supported the view that a mobile health intervention delivered via smartphone with electronic coaching is a feasible and acceptable method of addressing chronic pain among those who engage in heavy drinking. Future work should test the efficacy of this approach in a fully powered trial.

Introduction

Chronic pain represents one of the main sources of disability and diminished quality of life among adults in both community (Rikard et al., 2023) and medical samples (Cohen et al., 2021). Indeed, pain is consistently identified as one of the main reasons why patients seek medical care, with some clinic settings reporting over 50% of patients experiencing pain (Cordell et al., 2002). Although chronic pain is typically addressed in primary care (Breuer et al., 2010), pain management represents a challenge for health care providers as they are limited in the pharmacological approaches that may be used to ameliorate persistent pain associated with many chronic conditions (Dowell et al., 2016; Varrassi et al., 2010). Although psychological interventions have been shown to have efficacy for pain (Ehde et al., 2014), they have generally not been readily integrated into primary care systems, often due to inadequate resources to deliver empirically supported treatments (Morley, 2011) and difficulty engaging patients in behavioral interventions that require multi-session visits (Oosterhaven et al., 2019; Turk and Rudy, 1991).

Compounding these challenges is the fact that chronic pain often co-occurs with other health conditions and unhealthy behaviors that influence pain severity, interference, and recovery, and complicate pain management (Haibach et al., 2014; Otis et al., 2006; Uebelacker et al., 2016). Among the more impactful of these is hazardous drinking (also known as excessive or risky alcohol consumption), defined as the consumption of 4+ drinks per occasion for women and 5+ for men or weekly drinking of more than 7 standard drinks for women and more than 14 for men (National Center for Chronic Disease Prevention and Health Promotion, 2022), which is estimated to occur in up to 20% or more of patients in primary care samples (see Reid et al., 1999; Sterling et al., 2020). Hazardous drinking has direct and indirect influences on pain (Yeung et al., 2020) including increased sensitivity to pain (Sullivan et al., 2008), stress reactivity (Apkarian et al., 2013), and non-adherence to pain management recommendations (Timmerman et al., 2016). Chronic pain has a reciprocal influence on alcohol outcomes as it has shown to have negative impacts on hazardous drinking and alcohol treatment (Caldeiro et al., 2008; Ditre et al., 2019; Witkiewitz et al., 2015; Zale et al., 2015). Despite the interacting influences of hazardous drinking and chronic pain on health outcomes and on each other, no behavioral intervention has been developed to address these common co-occurring conditions in a manner that is efficacious, easily utilized by patients, and may be readily integrated into primary care.

Digital health approaches, such as web and app-based interventions, provide a potentially valuable approach to address these conditions in a manner that increases availability and accessibility for patients and reduces resource needs for health care settings. Web-based cognitive and behavioral approaches to pain management have shown to have small, but significant effects on pain outcomes (Gandy et al., 2022). Although the vast majority of the mobile apps available for pain management have not been empirically evaluated, recent empirical studies have suggested a similar benefit (Koumpouros and Georgoulas, 2023; Thurnheer et al., 2018).

Web-programs and smartphone apps have been extensively used to address hazardous/harmful drinking and have also demonstrated small but significant effects both as stand-alone approaches and as part of integrated telehealth strategies across a range of hazardous/harmful drinking populations (e.g., Kaner et al., 2017; Kiluk et al., 2019). However, previous studies have not addressed the co-occurring influences of hazardous drinking and pain in an app-based format. Though integrated interventions to address pain and substance use related to pain coping have been developed (e.g., Ilgen et al., 2020, 2016), they have generally been in-person approaches that target substance users in treatment rather than alcohol across the spectrum of hazardous/harmful drinking.

The current program was also designed to address one of the more challenging features of traditional digital-only programs for pain management, that of drop-out (Macea et al., 2010). Treatment drop-out is a concern for in-person cognitive and behavioral interventions for pain (Macea et al., 2010; Turk and Rudy, 1991), which is exacerbated when individuals are left to navigate digital programs without ongoing guidance or support as is common in empirical studies (Macea et al., 2010). It has been suggested that the absence of “supportive accountability” (Mohr et al., 2011) provided by ongoing, low intensity involvement of health care providers, may account, in part, for the poor uptake of digital interventions in real world settings (Graham et al., 2019). As Mohr et al. (2018) have suggested, to actualize the benefits of enhanced access through digital interventions, it is important design digital approaches from the outset that provide features to foster uptake and utilization of apps, which includes minimizing participant burden, making apps use easy, and providing ongoing support.

The current study sought to test the feasibility and acceptability of a smartphone app-based, integrated intervention for chronic pain among those who engage in hazardous drinking. The intervention, developed based on qualitative interviews with primary care patients (Palfai et al., 2021), included low intensity counselor support through instant messaging given evidence that provider contact enhances engagement with mobile health interventions (Mohr et al., 2011). The primary objectives of this pilot study were to: (1) determine the feasibility and acceptability of the intervention through assessment of patient experiences and engagement in the mobile intervention, and (2) provide preliminary information about the impact of the intervention on indices of pain and hazardous drinking.

Methods

Participants

Forty-eight participants were recruited from primary care clinics in a large, urban hospital and through social media. Participants were 18 years of age or older, fluent in English, engaged in primary care (i.e., had a primary care physician), experienced chronic non-cancer related pain, and reported hazardous drinking. The chronic pain inclusion criterion was pain of moderate severity or greater in the past week that lasted for at least 3 months. Hazardous drinking was determined as described above. Participants currently using pharmacological approaches for pain or alcohol use were permitted in the study if medication doses were stable for at least two months. Exclusion criteria were: history of bipolar disorder/schizophrenia/complicated alcohol withdrawal, psychological treatment for pain or alcohol use within the past 3 months, and planned surgery for a pain-related condition in the subsequent six months. Participants were also excluded if they did not have a smartphone.

Recruitment from primary care took place within clinic waiting rooms and through letters sent to patients. Interested patients were then screened by phone. Study advertisements were also placed through social media and were linked to a preliminary, web-based screening form. Individuals who met initial eligibility criteria based on the web-based screening were contacted by a member of the research team for phone screening to determine eligibility. Study procedures were approved through university IRB protocol #4947.

Assessments

Sample Characteristics

Participants completed a series of measures to assess demographic characteristics including race/ethnicity, marital status, education, and employment. Participants reported substance use frequency in the past 30-days (Rosen et al., 2000; Turk et al., 2003). Motivation-to-change drinking was assessed using the staging algorithm of the Readiness-to-Change Questionnaire (Budd and Rollnick, 1996).

Pain Outcomes

The primary pain outcome for the trial was the Pain, Enjoyment of Life, and General Activity (PEG; Krebs et al., 2009) scale which has been widely utilized as a pain outcome measure for clinical trials (e.g., Wandner et al., 2022). On this 3-item measure, participants rate the severity of their pain and the degree to which their pain interfered with enjoyment in life and general activity “in the past 7 days” using 0–10 scales. The PEG has excellent reliability and validity and has been widely established with medical populations (e.g., Keane et al., 2019; Kroenke et al., 2009).

Secondary pain outcomes were average past week pain severity rated on an 11-point scale and pain severity and pain interference composite scores from the Brief Pain Inventory-Short Form (BPI), (Dworkin et al., 2005).

Alcohol Outcomes

The primary alcohol outcomes were frequency of heavy drinking episodes (women > 3 per occasion, men > 4 per occasion) in the past 30-days and mean number of drinks per week as assessed by the 30-day alcohol Timeline Follow-Back calendar method (Alcohol TLFB-30) (Sobell et al., 1979). Alcohol-related consequences were assessed with the Short Inventory of Problems-Revised (Kiluk et al., 2013).

Additional Outcome Assessments

Participants completed additional assessments to examine the transdiagnostic effects of the intervention including depressive symptoms with the PHQ-8 (Kroenke et al., 2009), days of cannabis use in the past 30 days (Rosen et al., 2000; Turk et al., 2003) and physical functioning from the PROMIS-29 v2.0 (Hays et al., 2018).

Health Behavior Change Processes

To provide initial data regarding hypothesized mechanisms underlying the intervention, we utilized pain and alcohol specific measures of the Goal Systems Assessment Battery (Karoly and Ruehlman, 1995). Participants rated their self-regulatory processes related to the behavior change goals of “pain management” and “moderating/limiting alcohol use” at baseline and at the 8-week post-intervention outcome. For each of these two change goals, they completed 4-point 16-item Likert-scale measures (“not at all” to “very much”) to assess self-regulatory domains of goal value, self-monitoring, self-efficacy, and planning. Internal reliability for the pain version (value = 0.85, self-efficacy = 0.94, self-monitoring = 0.82, planning = 0.75) and alcohol version (value = 0.93, self-efficacy = 0.86, self-monitoring = 0.80, planning = 0.77).

Conditions

Participants were randomized to one of two intervention conditions following baseline based on a block randomization schedule stratified by sex at birth and level of heavy drinking (+/− 5 heavy drinking episodes).

MCBMAP Intervention (INTV).

The intervention utilized a self-regulation framework (Karoly, 2012) to integrate evidence-based approaches for hazardous drinking and pain including Motivational Interviewing (Miller and Rollnick, 2013) and cognitive-behavioral and self-management training (Morgenstern et al., 2007; Otis, 2007; Ruehlman et al., 2012; Sobell and Sobell, 1996). Participants first met with the health coach, who introduced the program, reviewed participants’ experience of pain and prior treatment, provided psychoeducation on the interaction between pain and alcohol use, and presented the program rationale. Participants were then provided access to the smartphone app. Intervention content was delivered through a series of app-based video modules that were supplemented with a brief (15 minute) weekly health coaching session delivered through instant messaging. Each of the eight intervention weeks were conducted through the app and included viewing one or two short video modules, completing two daily check-in and activity surveys. In addition, participants had access to a Skills Library which contained a brief summary of each of the skills learned.

Control (CTL).

Those in the CTL condition completed a single videoconferencing session with the health coach which consisted of a brief review of the participant’s pain and treatment history, psychoeducation about the interactive effects of pain and alcohol, discussion of resources available for pain and alcohol use, and the option to review information regarding available local and web-based behavioral health and wellness resources to address pain and alcohol use. These procedures are consistent with a treatment-as-usual-strategy to address these conditions in the clinic setting with available resources.

Training, Supervision and Fidelity Assessment.

The health coaches were advanced level PhD clinical psychology students who were trained for content related to both conditions. Supervision was provided biweekly and involved individual discussion of ongoing cases. Intervention fidelity assessment was conducted through a checklist of session components (specific to condition) conducted by research assistants based on audio-recordings and instant messaging transcripts.

Indices of Feasibility and Acceptability

A variety of indices were used to evaluate the acceptability and feasibility of the intervention and trial procedures. Study attrition and responses to a modified version of the Client Satisfaction Questionnaire-8 (CSQ-8), an 8-item validated measure of perceived value of treatment services (Kelly et al., 2018; Larsen et al., 1979) was used. Items range from 1 “poor” to 4 “excellent” with the total score ranging from 8 to 32, with higher scores indicating higher treatment satisfaction. In addition, we used study specific items based on the Perceptions of Treatment Questionnaire (PTQ-17), a 17-item questionnaire used in previous treatment development studies (Pincus et al., 2010) that provides descriptive information about perceived comprehensibility and utility of the intervention using 0 “not-at-all” to 8 “very much” Likert-type scales.

To gather usability and acceptability data on the intervention itself, we examined rates of completion for the video modules, adherence to health coach instant messaging sessions, and completion of smartphone daily activities to assess intervention acceptability. We also assessed whether participants engaged with the optional skills library. Participants also completed 3-items selected from the Systems Usability Scale (Brooke, 1996) to provide information about the experience of program use through subjective ratings.

Procedures

Participants were screened by phone and, if eligible, scheduled for an initial videoconferencing visit. During the initial visit, participants were consented, completed a 50-minute baseline assessment, and were then randomized to the Control or Intervention condition. Participants then met with a study health coach for approximately 20-minutes. Those randomized to Intervention concluded the initial videoconferencing visit by downloading the study smartphone application and learning the procedures for using the app.

In addition to completing weekly lesson video-modules through the app to learn skills and strategies, participants completed two daily surveys to practice skills and evaluate outcomes. Weekly health coaching sessions were scheduled to be approximately 15 minutes each. Participants were reminded via text message the day before and the day of the instant messaging health coaching session. Those who experienced any difficulties using the technology were able to consult a “troubleshooting” form or call the study assistant to guide them through the smartphone procedures.

Post-intervention and follow-up assessments were conducted approximately 8 and 16-weeks following baseline by a research assistant who was blind to study condition. Study visits were conducted using videoconferencing and used an interviewing assessment method supplemented with on-screen displayed measures.¹ Study data were collected and managed using REDCap electronic data capture tools (Harris et al., 2009).² Participants were compensated for their time and effort related to assessments at baseline, 8 weeks, and 16 weeks in the amount of $50 for each session.

Data Analyses Approach

Feasibility of methods and procedures were evaluated with descriptive data on the percent of patients screened, percent of eligible patients enrolled, and percent of patients who completed assessments. For those in the Intervention condition, adherence rates for the percent of modules completed and the percent of coaching sessions attended were assessed. Patient satisfaction with the intervention and treatment adherence, mean and median treatment satisfaction ratings were calculated based on measures described above.

Analyses of intervention impact were based on the full sample of participants who completed baseline and were randomized to condition. Lost to follow-up primary and secondary outcomes was defined as missing the 16-week assessment. For the primary analyses, multiple imputation was used to address missing outcome data and the full randomized sample was included. Completer analyses were conducted as sensitivity analyses for the primary outcomes and for all remaining outcomes. We used a parallel approach (i.e., f² calculated from regression) to provide effect size estimates for the influence of the intervention on secondary outcomes. To provide information about whether the intervention was associated with change in self-regulatory mechanisms, we estimated the effect of intervention on self-regulatory processes by comparing CTL to INTV post-intervention (8-weeks post baseline), controlling for corresponding baseline values. Analyses were conducted using SPSS 29.0.1.

Results

Study participants were enrolled from August 26, 2021 to November 14, 2022 with the final follow-up assessment completed on March 27, 2023. Screening, eligibility, and randomization rates are described in the Consort diagram (Figure 1). Forty-eight participants were randomized. Two participants withdrew post-randomization and 5 additional participants were lost to follow-up resulting in a total of 85% who provided post-intervention (8 week) follow-up data and 85% who provided 16-week data which served as the primary outcome point. There were no significant differences by group. As shown in Table 1, baseline characteristics were similar across groups. The sample consisted of 58.3% female at birth, majority white (83.3% White, 14.6% Black, 2.1% multi-racial), 6.3% Hispanic, with a mean age of 52.79 (SD = 15.23). The mean number of heavy drinking days was 10.02 (SD = 10.41) with a median of 5.50 (interquartile range [IQR], 1.00–15.75) while the mean number of drinks per week was 20.35 (SD = 19.24) with a median of 14.00 (IQR, 6.82–14.85). Overall, the sample exhibited moderate-severe pain as indicated by mean PEG ratings of 6.06 (SD = 1.78) with a median of 6.17 (IQR, 4.67–7.33). There were no significant differences between 16-week completers and non-completers on outcomes.

Figure 1.

Consort Diagram

Table 1.

Baseline Descriptive Statistics by Intervention Arm (N = 48)

	Intervention (n = 24)	Control (n = 24)	Overall (N = 48)
	M (SD) or %	M (SD) or %	M (SD) or %
Age	53.17 (13.15)	52.42 (17.35)	52.79 (15.23)
Sex at birth (% female)	14 (58.3%)	14 (58.3%)	28 (58.3%)
Race and Ethnicity
White	20 (83.3%)	20 (83.3%)	40 (83.3%)
Black	3 (12.5%)	4 (16.7%)	7 (14.6%)
Asian	0	0	0
Native American	0	0	0
More than one	1 (4.2%)	0	1 (2.1%)
Hispanic	0	3 (12.5%)	3 (6.3%)
Patient Health Questionnaire-8 (% > 10)	11 (45.8%)	13 (54.2%)	24 (50%)
General Anxiety Disorder-7 (% > 8)	15 (62.5%)	14 (58.3%)	29 (60.4%)
PROMIS Physical Functioning	12.88 (3.51)	13.50 (4.54)	13.19 (4.03)
Cannabis Use (%)	5 (20.8%)	10 (41.7%)	15 (31.3%)
ASI Days with Cannabis Use (out of 30)	4.00 (9.06)	7.88 (11.35)	5.94 (10.35)
Smoking (%)	8 (33.3%)	3 (12.5%)	11 (22.9%)
Readiness to Change– Alcohol
Precontemplation	9 (37.5%)	7 (29.2%)	16 (33.3%)
Contemplation	8 (33.3%)	13 (54.2%)	21 (43.8%)
Action	7 (29.2%)	4 (16.7%)	11 (22.9%)

Legend. PROMIS = Patient Reported Outcomes Measurement Information System-29; ASI = Addiction Severity Index.

Feasibility and acceptability outcomes

Participants in the INTV and CTL conditions both reported moderate to high levels of satisfaction with the program that they received. Those in the Intervention condition reported a mean modified-CSQ rating of 26.00 (SD = 5.8), while those in the Control condition reported a mean of 21.76 (SD = 6.2). Individual ratings from the PTQ on the benefits of each program showed that participants in both groups experienced moderate benefit for pain and alcohol use behavior. As shown in Table 2, rating on perceived benefits for “coping with pain” were: INTV M = 5.14, (2.40), Median = 6; Control M = 3.95 (SD = 2.98), Median = 5; “help with drinking”: INTV M = 3.62, (3.04), Median = 2; Control M = 4.05 (SD = 2.80), Median = 4; “coping with negative mood”: INTV M = 4.33, (3.04), Median = 5; Control M = 3.26 (SD = 2.77), Median = 3.

Table 2.

Acceptability and Usability Indices of Intervention

	Intervention (n = 20)				Control (n = 21)
Measure	Mean	SD	Med.		Mean	SD	Med.
Cope with pain1	5.14	(2.40)	6.00		3.95	(2.98)	5.00
Cope with negative mood1	4.33	(2.85)	5.00		3.26	(2.77)	3.00
Help with drinking1	3.62	(3.04)	2.00		4.05	(2.80)	4.00
Intervention only components
	Mean	SD	Med.	IQR
Check-ins (out of 8)	5.87	(2.50)	6.00	[5, 8]
Video modules (out of 12)	9.92	(2.91)	11.00	[7.75, 12]
Surveys completed (out of 89)	50.1	(27.2)	56.00	[22.5, 73.25]
Morning surveys completed (out of 42)	23.75	(12.93)	26.50	[11, 34.75]
Evening surveys completed (out of 42)	22.71	(12.84)	24.50	[9.75, 33]
Skill reviews competed (out of 5)	3.63	(1.84)	4.50	[3, 5]
Skill library use (%)	83.33
Ease rating (1–5)2	4.10	(1.00)	4.00	[3.5, 5]
Most learn quick (1–5)2	3.81	(0.93)	4.00	[3, 4.5]
Confidence in use (1–5)2	4.38	(4.38)	5.00	[4, 5]
Usability rating (1–5)3	4.13	(0.88)	4.50	[3.67, 4.67]
Module video/exercise	5.95	(2.36)	7.00	[4, 8]
Number of Check-ins	4.71	(2.96)	6.00	[2, 7]
Library helpfulness rating4	5.48	(2.82)	7.00	[3, 8]

Legend. Med. = Median

¹Response to questions 1, 2, and 3 (listed in order) of the Perceptions of Treatment Questionnaire (How much do you think the program helped you…).

²Response to questions 5 (I thought the program was easy to use), 6 (I think that most people would learn to use this program very quickly), and 9 (I felt confident using the program) of the System Usability Scale.

³Average of Ease, Most learn quick, and Confidence in use ratings.

⁴Response to question 6 of the Perceptions of Treatment Questionnaire (How helpful was the skills library to you?).

To provide data on the feasibility and acceptability of the intervention, we relied on behavioral and self-reported indices. Of the 12 video modules available over the course of the program, participants viewed a mean of 9.92 (SD = 2.91) with a median number of 11 sessions (IQR =7.75, 12). There were 8 instant messaging sessions available for participants, of which a mean of 5.87 (SD = 2.5, median = 7 (IQR = 5, 8) were completed. Participants completed surveys during each day to set plans for skill use and report on the effects of skill practice. Over the course of the 8-week program, up to 89 surveys were delivered. There was a wide range of surveys completed by participants (4–87) with a mean of 50.1 (SD = 27.2), median = 56 (IQR = 11, 34.75), though this includes those who discontinued and two who temporarily lost phone access during the trial. Among those who provided at least one survey response in the last week, the mean response rate was 59.78% (SD = 20.76).

Ratings of the intervention also provided support for patient satisfaction with the smartphone intervention. Mean rating of the “helpfulness of the videos/exercises” was 5.95 (SD = 2.36), with a median of 7 (IQR 4, 8). The adjunctive skills library was well-received by participants as 20/24 of them made use of it during the trial and overall rated it highly in terms of its helpfulness (M = 5.48 (2.82), median = 7 [IQR 3, 8]). The weekly check-in with the coach was also identified as a useful component of the program as indicated by ratings (M = 4.71 (SD = 2.96, median 6.0 [IQR 2, 7]). Finally, participant ratings of usability of the app itself on 3 dimensions using a Likert scale (‘1” strongly disagree to “5” strongly agree) supported the value of the program. Mean ratings across items measuring ease of use, ease of learning, and confidence in use was 4.13 (SD = 0.88), median = 4.33 (IQR 3.67, 4.67).

Pain and Hazardous Drinking Outcomes

We calculated effect size estimates and rates of behavior change by condition to provide initial information about the potential impact of the integrated intervention. For each primary analysis, we examined the condition differences at 16-week post-intervention outcomes, controlling for corresponding baseline value. The conditions (Control vs. Intervention) were coded as an indicator variable with 0 representing the control group reference condition. We used these dummy codes in the specific comparisons of interest namely: CTL vs. INTV. Mean ratings of pain and alcohol outcomes by condition and timepoint are presented in Table 3. As noted above, analyses for the primary outcomes were conducted using multiple imputation for missing outcome values.

Table 3.

Descriptive statistics at baseline and 16 weeks for primary and secondary outcomes.

	Baseline (n = 48)						16-wk Outcome (n = 41)
	Intervention (n = 24)			Control (n = 24)			Intervention (n = 21)			Control (n = 20)
Outcomes	Mean (SD)	Med.	IQR	Mean (SD)	Med.	IQR	Mean (SD)	Med.	IQR	Mean (SD)	Med.	IQR
Primary
PEG	6.36 (1.69)	6.50	[5.08, 7.33]	5.76 (1.83)	5.50	[4.33, 7.33]	4.60 (2.56)	5.00	2.33, 6.50]	4.82 (1.79)	4.83	[3.12, 6.33]
BPI Severity	5.40 (1.71)	5.50	[4.13, 6.44]	4.66 (1.46)	4.75	[3.81, 5.69]	3.98 (1.93)	3.50	[2.53, 5.25]	4.16 (1.21)	4.00	[3.75, 4.69]
BPI Interference	6.21 (1.88)	6.07	[5.18, 7.79]	5.24 (2.15)	4.93	[3.57, 7.11]	4.37 (2.63)	4.57	[2.14, 6.00]	4.68 (2.36)	4.36	[2.86, 6.57]
HDE	11.25 (11.78)	5.00	[1.00, 25.50]	8.79 (8.93)	7.00	[1.25, 12.75]	7.90 (11.32)	2.00	[1.00, 10.50]	4.30 (5.57)	2.00	[0.00, 7.75]
Weekly Drinks	22.33 (22.27)	13.18	[6.01, 29.51]	18.36 (15.89)	14.58	[9.27, 22.28]	16.89 (20.60)	9.33	[2.45, 28.58]	10.56 (9.02)	7.82	[4.49, 15.17]
Secondary
Short Inventory of Problems	10.50 (11.72)	6.50	[2.00, 16.50]	8.88 (10.14)	6.50	[0.50, 12.00]	6.48 (8.26)	3.00	[0.00, 10.50]	6.55 (7.29)	5.50	[0.25, 10.25]
Patient Health Questionnaire-8	10.13 (5.84)	8.50	[6.00, 15.75]	10.13 (4.89)	10.50	[6.00, 14.75]	6.71 5.07)	5.00	[3.00, 9.50]	9.20 (5.29)	9.50	[4.00, 13.75]
PROMIS Phys. Functioning	12.88 (3.51)	12.00	[10.00, 16.50]	13.50 (4.54)	14.00	[9.25, 17.75]	13.52 (4.54)	14.00	[9.50, 18.50]	14.35 (4.91)	14.50	[11.25, 19.00]
ASI Days with Cannabis Use (out of 30)	4.00 (9.06)	0.00	[0.00, 0.00]	7.88 (11.35)	0.00	[0.00, 20.00]	4.43 (9.83)	0.00	[0.00, 2.50]	6.05 (10.94)	0.00	[0.00, 9.75]

Legend. Med. = Median; PEG = Pain, Enjoyment of Life, and General Activity scale; BPI = Brief Pain Inventory; HDE = heavy drinking episodes as assessed by 30-day Timeline Followback; PROMIS Phys. Functioning = Patient Reported Outcomes Measurement Information System-29 Physical Functioning Subscale; ASI = Addiction Severity Index.

To provide a preliminary effect size estimate for the influence of the intervention on pain, we examined PEG ratings at the 16- week outcome, controlling for PEG scores at baseline. Regression analyses revealed a small effect (f² = .04) of intervention on PEG ratings as both conditions showed reductions over time (using standards based on Cohen’s guidelines of f² = .02 as a small effect, f² = .15 as a medium effect, and f² = .35 as a large effect). Given the limitations of between-group effect size estimates with small samples (e.g., Leon et al., 2011), we also sought to provide preliminary data on intervention effects based on clinical improvement metrics where applicable, notably for pain. Following recommendations from the IMMPACT trial (Turk et al., 2008), we examined the percent of participants who exhibited moderate clinical improvement based on reductions of 30% and substantial clinical improvement based on 50% reduction in PEG scores. Using these thresholds, a total of 38% and 25% of participants in the INTV condition showed moderate and substantial change respectively compared to only 20% of those in the CTL condition who showed at least moderate change and only 12.5% who showed a substantial reduction.³

Because the primary alcohol outcomes (i.e., number of heavy drinking episodes, number of drinks per week) were skewed, transformations to drinking data were required to provide an effect size estimate to parallel pain outcomes (i.e., f²). To reduce the effects of extreme values, one drinking quantity outcome greater than 3 SD was recoded to one greater than the next largest value (Tabachnick and Fidell, 2013). Alcohol outcomes were square root transformed in each analysis. Due to sex differences in consumption, analyses with alcohol outcomes were conducted with sex at birth as a covariate along with the corresponding baseline alcohol variable. Pooled estimates from multiple imputations were used for each of these primary outcomes. Analyses showed minimal effects of intervention on alcohol use with a minimal effect favoring CTL (f² = .02) on the heavy drinking outcome and no difference on quantity of consumption (f² = .00) at the 16-week outcome.

Secondary and Exploratory Outcomes

Secondary outcomes showed a similar pattern of small effects in favor of the intervention condition with changes over time in both conditions. Rating of average pain in the past week showed an intervention effect of f² =.03, pain severity from the BPI severity composite showed an effect of f² =.05, while pain interference showed an effect of f² = .06. Estimated effect size for the intervention on alcohol related consequences was minimal f² = .01. Given the common overlapping conditions of those with pain, we also examined effect size estimates for depression symptoms as measured by the PHQ-8 which showed an intervention effect of f² = .05. There was no evidence of beneficial effects of intervention on physical functioning as measured by the PROMIS (f² = .00) or frequency of cannabis use in the past 30 days (aIRR = 1.99, CI, 0.63, 6.32). In sum, although effects were in the small range, the smartphone intervention appeared to show benefit for pain and depressive symptoms but did not show a clear benefit for alcohol use.

Self-regulatory processes. Measures of self-regulatory processes (i.e., GSAB) for pain and alcohol were examined at the post-intervention (8-week) timepoint. Covariate adjusted means are presented in Table 4 along with effect size estimates for each subscale outcome for pain and drinking goals, controlling for corresponding ratings at baseline. These analyses suggested beneficial effects of intervention on self-efficacy and self-monitoring related to pain management and on self-efficacy for the goal of “reducing/moderating alcohol use”. Overall, self-regulatory variables showed change in the direction of the intervention effect.

Table 4:

Covariate Adjusted Means for Health Behavior Change Process Measures¹

	Control ( n = 21)		Intervention ( n = 20)
	Mean	SE	Mean	SE	Effect (f2)
GSAB - Pain
Value	14.70	0.34	14.91	0.35	0.01
Self-Efficacy	8.92	0.60	11.68	0.62*	0.27
Self-Monitoring	9.36	0.61	11.12	0.63	0.11
Planning	10.83	0.52	10.88	0.56	0.00
GSAB – Alcohol
Value	11.40	0.67	12.03	0.69	0.01
Self-Efficacy	10.85	0.62	12.56	0.64	0.10
Self-Monitoring	9.07	0.72	10.33	0.73	0.04
Planning	7.21	0.77	8.63	0.79	0.04

Legend. GSAB = Goal Systems Assessment Battery

¹At 8-week outcome point.

Intervention Adherence and Fidelity.

To provide an estimate of intervention adherence, 20% of the initial videoconferencing sessions and instant messaging sessions were coded. To ensure reliability, 20% of these sessions were assessed by a second coder. The intervention instant messaging following smartphone session 1, sessions 2–7, and session 8 each had a set of content codes. Adherence for the CTL condition was based on a smaller set of codes for the single videoconferencing session. These analyses showed that the intervention and control condition was delivered as designed with intervention segments showing 95% of content successfully delivered and 100% of the control condition.⁴

Discussion

The current study sought to: (1) assess the acceptability and utility of an app-based pain management intervention with electronic coaching for those who engage in hazardous drinking, (2) provide initial evidence for the impact of the intervention on pain and alcohol outcomes, and (3) test the feasibility of study procedures to conduct a full-scale, randomized controlled trial to examine intervention efficacy.

Overall, the intervention was easy to implement and well received by participants. This intervention provided a number of unique features including video-based intervention modules to address pain in the context of unhealthy drinking, a readily accessible skill library of strategies to manage pain and alcohol use, and an instant messaging coaching platform. Participants showed high levels of satisfaction with the program and high engagement. In particular, participants exhibited high levels of adherence with the core skills-based components as reflected in number of videos viewed and use of the skills library. The use of a coaching component through instant messaging represents a novel method to enhance engagement with a mobile app-based intervention (Mohr et al., 2011; Gandy et al., 2022) and was well-received by participants.

Responses also indicated areas that may benefit from closer attention in subsequent work. Namely, participants showed a wide range of responses to the daily surveys to support homework and monitor the impact of their skill use. Efforts to increase engagement in daily activities by making the value of these intervention components more salient in instructions and providing additional reminders may provide a means of increasing engagement. This must be balanced, however, with the time and resources that participants are able to invest in behavior change in the context of other daily life tasks (e.g., employment requirements) and the desire to maximize participant autonomy as they navigate through the intervention (Palfai et al., 2021).

Although results from this small sample study should be interpreted with caution, analyses estimated generally small effects of the intervention on pain outcomes in contrast to the treatment as usual control condition. These between-group effects were modest; however, it should be noted that the intervention condition showed evidence of clinically meaningful change (Dworkin et al., 2008) at almost twice the rate of those in the control condition. Moreover, it should be noted that participants in both conditions showed reductions over time in pain outcomes. These within-effects were moderate to large and consistent with previous trials on pain interventions. For example, baseline to follow-up change reductions in PEG scores and interference scores in the INTV condition averaged 1.7, which represented a within effect of d= .68. This pre-post change is consistent with findings from previous studies of in-person CBT interventions for chronic pain (Merlin et al., 2018; Murphy et al., 2022). Similarly, the small estimated between-condition effects on pain outcomes are also consistent with previous meta-analyses that have tested web-based interventions compared to active control conditions (versus assessment only; Gandy et al., 2022).

There was little evidence that the intervention reduced alcohol use compared to the treatment-as-usual control group. This may have reflected the fact that the emphasis in the app was pain, while alcohol use was addressed as a key lifestyle factor that contributes to pain. Future work may benefit from an increase in alcohol emphasis in the intervention app modules. It should be noted, however, that there was considerable reduction in alcohol use across conditions. Indeed, those in the Control condition showed a 70% reduction in median heavy drinking episodes, while those in the Intervention showed a 60% median reduction in heavy drinking episodes. It may be that the treatment-as-usual control condition, which included a review of alcohol use patterns, discussion of its interaction with pain, and provided information about available treatment resources, was sufficient to promote change in alcohol use among this population. It suggests that it may be worth exploring other control conditions (e.g., assessment only; resource information only) when evaluating the impact of the intervention in a full-scale trial. Of course, changes across conditions may be due to other factors including the act of self-reporting alcohol use (McCambridge and Kypri, 2011) and regression to the mean. Trials of brief interventions for alcohol use have yielded wide variability in differences between intervention and control (Jenkins et al., 2009), suggesting that further clarification is needed to identify suitable control conditions in alcohol-related trials.

Similarly, there was no evidence that changes in alcohol use were linked with corresponding increases in other substances (Vuchinich and Tucker, 1988). It should be noted that those in this pilot sample reported minimal use of substances other than cannabis, and the use of cannabis did not increase with the decrease of alcohol use. Thus, we did not find evidence for substance substitution in this sample. Whether this would occur with chronic pain samples that include participants with higher rates of non-prescription opioid use is an open question. Although intervention components included in the integrated intervention seek to address transdiagnostic processes that would be of benefit to reduce other substances (e.g., behavioral activation, stress coping), there may be benefit to incorporate modules/components to address potential co-morbid substance use.

Examination of the health behavior processes from the GSAB measures suggested that the intervention may foster improvements in self-regulatory processes for both pain management and drinking reduction. In particular, the effect size estimates suggested that the intervention may influence self-efficacy for reducing/moderating alcohol use and pain management. Additional potential benefits of the intervention on pain-related self-monitoring were also suggested by results. Future work with larger sample sizes will provide the opportunity to test the potential mediational role of these self-regulatory variables for the intervention.

Limitations.

There are important limitations to consider that should be addressed in future work. First, the recruitment of patients who met alcohol and pain study criteria directly from primary care proved to be a challenge. For the primary care clinics in which this intervention was initially tested, fewer participants were identified as eligible than anticipated, and the recruitment strategies of engaging potentially relevant patients in the clinic (based on medical record review and scheduled visits) was not particularly effective and very labor intensive. Changes to the approach to identify primary care patients who may benefit from this approach through social media proved to be a much more viable approach and may be consistent with alternative ways of making this intervention widely accessible (e.g., online availability through referrals, patient generated inquiries, etc.). Second, it will be important to develop more effective strategies to increase engagement in daily activities completed through the app. Finally, although the app was developed with a primary care sample that was largely African American (Palfai et al., 2021), the sample recruited from social media in this study tended to be predominantly white. It will be important to ensure that race and ethnicity is broadly represented in future work.

Contributions and future directions.

The current study provides a number of contributions to the literature on pain management. First, this study supports the feasibility and acceptability of an easily utilized, digital intervention to address chronic pain among hazardous drinkers with low participant burden. Despite the frequent co-occurrence of chronic pain among hazardous drinkers, an integrated, mobile intervention to address pain and its interaction with hazardous drinking has been absent from the literature. It also provided support for the feasibility of utilizing a low intensity coaching protocol to maintain high levels of engagement in the use of the app features. These features of low burden and supportive accountability through low intensity coaching (Mohr et al., 2017) are important to consider for implementation in real world contexts such as primary care settings.

It is important to emphasize that evidence from this current study alone does not suggest modifications to clinical practice. Future work with fully powered samples is needed to test the efficacy of this mobile app approach. Results from this study provide guidance on the implementation of a larger scale study and insight on the mechanisms that may underlie the benefits of this approach.

Conclusions

In sum, this study supports the feasibility and acceptability of a novel mobile health intervention to address chronic pain and heavy drinking for primary care patients. However, the use of a screening and brief intervention approach in the primary care setting itself proved to be challenging and suggests that the optimal way of engaging hazardous drinking patients with chronic pain may be through referral to a centralized behavioral health resource. A fully remote procedure for engaging patients in the app-based intervention lends itself to wider applications across other medical settings.

Highlights.

• Integrated cognitive-behavioral program for pain and hazardous drinking

• Pilot study of counselor-supported app for chronic pain among hazardous drinkers

• Novel mobile health app-based program showed high acceptability and feasibility

• Intervention associated with improved self-efficacy for pain and alcohol management

List of abbreviations

mhealth

mobile health

BPI

Brief Pain Inventory

CBT

Cognitive -Behavioral Therapy

CSQ-8

Client Satisfaction Questionnaire

CTL

Control condition

GSAB

Goal Systems Assessment Battery

HDE

Heavy Drinking Episode

IMMPACT

Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials

INTV

Intervention condition

PEG

Pain, Enjoyment of Life, and General Activity Scale

PHQ-8

Patient Health Questionnaire-8

PROMIS-29

Patient-Reported Outcomes Measurement Information System

PTQ-17

Patient Treatment Questionnaire

TLFB-30

Time Line Follow-back, 30 days