A Randomized Feasibility Study of a Positive Psychology Journaling Intervention to Support Recovery from Substance-Use Disorders

Abstract

Background.

Individuals in early recovery face significant biopsychosocial stressors causing a preponderance of negative affect. Novel interventions are needed to improve mood and well-being to support recovery. Positive Recovery Journaling (PRJ) combines elements of positive psychology, behavioral activation, and journaling to emphasize what is going right and to encourage small, positive steps that align with an individual’s values to make life in recovery more rewarding and therefore more reinforcing. Our objective was to determine PRJ’s feasibility, acceptability, and impact on a set of strengths-based, multidimensional aspects of recovery, including satisfaction with life, happiness with recovery, and commitment to sobriety.

Methods.

The study randomized adults in substance-use disorder treatment (N = 81) to PRJ or control. Those in PRJ were asked to practice PRJ daily and complete online surveys for four weeks; those in the control group completed online surveys for four weeks. We used multi-level modelling to determine intercept and slope for feasibility and acceptability outcomes as well as to compare differences in recovery indicators between treatment and control at baseline and Weeks 2, 4, and 8. We conducted intention-to-treat and per-protocol analyses for each recovery indicator.

Results.

Participants were 53% female, and 26% Black, Indigenous, People of Color (BIPOC) and mean age of 39 years. PRJ participants attended 71% of groups and completed 56% of the daily PRJ entries. Treatment and control groups rated their study tasks (PRJ for the treatment group, surveys for the control group) as equally easy; however, the PRJ group rated PRJ as significantly more satisfying, helpful, and pleasant. Treatment and control were not significantly different on any recovery indicator. In post hoc analyses, we found that for those with < 90 days sobriety at baseline (51%), PRJ had a statistically significant beneficial effect for satisfaction with life, happiness with recovery, and numerous secondary recovery indicators.

Discussion.

Results suggest a positive impact of PRJ on numerous recovery indices for those in earliest recovery. Integrating PRJ into support services among those with < 90 days sobriety could reinforce what is going well in recovery to encourage its continued maintenance and thereby improve treatment outcomes.

Individuals in early recovery from substance-use disorders face considerable biopsychosocial challenges that threaten sobriety. At the biological level, addiction causes physical illnesses that persist into recovery (Eddie et al., 2019), and changes in the brain cause anhedonia and increased stress reactivity (Ahmed et al., 2002; Koob, 2008). At the psychosocial level, individuals face the stressful, negative consequences of their active addiction, such as degraded personal relationships, problems at work or school, and social stigma (Klingemann & Gmel, 2001; Miller et al., 1995). Such conditions lead to negative emotional states that are exacerbated by difficulties regulating negative emotions (Berking et al., 2011; Emery et al., 2022; Stellern et al., 2023). In addition, this population has demonstrated diminished interpersonal and environmental coping skills, inspiring treatment interventions to address such deficits (Maisto et al., 2000; Sliedrecht et al., 2019). Circumstances such as physical illness, anhedonia, stress, negative psychosocial consequences, and difficulty coping can, it seems, account for observable decreases in self-esteem and happiness during the first year of recovery (Kelly et al., 2018). For all these reasons, during early recovery, negative emotions predominate, setting the stage for recurrence of substance use (Marlatt, 1996; Ramo & Brown, 2008; Sliedrecht et al., 2019).

To address this constellation of challenges, researchers have developed and tested interventions to improve mood and encourage pleasant and meaningful activities to make recovery more reinforcing. Two such approaches are positive psychological interventions (Krentzman, 2013; Seligman & Csikszentmihalyi, 2000) and behavioral activation (Fazzino et al., 2019; Martínez-Vispo et al., 2018). Examples of positive psychological approaches are gratitude practices and acts of kindness to improve mood and to keep the focus on what is going well in recovery (e.g., Akhtar & Boniwell, 2010; Carrico et al., 2018; Hoeppner et al., 2019; Kahler et al., 2015; Krentzman et al., 2015). Behavioral activation encourages individuals to engage in important, enjoyable, and values-congruent activities and to move concretely toward meaningful goals (e.g., Daughters et al., 2008; Daughters et al., 2018; Magidson et al., 2011; Meshesha et al., 2020). In addition, individuals have experienced benefits through journaling, including through expressive writing, where participants convey negative emotions (Pennebaker & Beall, 1986), and interactive journaling, which encourages participants to change their behavior via written responses to short, didactic prompts (Miller, 2014; Proctor et al., 2012).

1.1. Positive Recovery Journaling (PRJ) and Its Theoretical Foundation

PRJ is a written daily practice, developed by the first author, which uses pre-printed prompts in a bound workbook to guide a review of the preceding twenty-four hours and a plan for the upcoming twenty-four hours. The overall intention of PRJ is to draw out what went well that day in recovery and encourage the planning and achievement of another balanced, successful day rooted in recovery activities and the individual’s personal values. To accomplish this, PRJ combines aspects of positive psychology, behavioral activation, and journaling that, together, leverage reinforcement theory (Higgins et al., 2004) to bolster life in recovery—a perspective associated with the successful maintenance of behavior change (Rothman, 2000; Rothman et al., 2016) and congruent with the idea that if recovery becomes more reinforcing, individuals will be motivated to maintain it (Gutierrez et al., 2022; Laudet, 2011; McKay, 2017). PRJ does not require advanced literacy as it uses written lists to reduce obstacles to the benefits of journaling. PRJ is intended to be adjunctive to other recovery activities, including traditional treatment, recovery housing, peer-recovery support services, or a combination of these and other activities. (For a PRJ composite entry, see Figure 1.)

Figure 1.

Positive Recovery Journaling Example

Positive Recovery Journaling (PRJ) example. Note. This is a de-identified composite of participant responses. Grand mean of the number of bullet points for each heading across the sample was 8.2 for Good things that happened; 2.5 for Bad things that happened; 10.8 for Things that I am grateful for; 9.1 for Wishes for others; 1.1 for Work/Education; 1.3 for Home/Housing; 1.7 for Joy; 1.9 for Health; 1.6 for Recovery; 1.3 for Spirituality; 0.8 for Community; 1.5 for Social; 0.9 for Financial; and 0.7 for Amends/Repair. The number of bullet points depicted in this figure reflects these grand means rounded to the nearest integer. Overall count of PRJ items per PRJ entry were M = 37.4 (SD = 13.7, Median = 35.8).

1.2. Early Work on Positive Recovery Journaling

In earlier work, we conducted a mixed-methods uncontrolled trial in which we taught PRJ to fifteen women in residential treatment for substance-use disorders (Krentzman, Hoeppner, et al., 2022). This study showed that participants found PRJ to be feasible and acceptable. We defined feasibility in our previous study and in the current study in accordance with Proctor et al. (2011) as the degree to which PRJ was doable, achievable, easy to complete, and able to be used while receiving treatment in the host treatment setting. Evidence of feasibility from the previous study were participants’ high ratings of the practice as easy, (M = 8.6, SD = 1.3), low ratings of the practice as difficult (M = 1.6, SD = 2.1), and the high levels of effort they were willing to put into the practice (M = 8.7, SD = 1.5) on single-item 0–10 Likert scales where higher scores indicated stronger agreement. We defined acceptability in this previous study and in the current study as the degree to which PRJ was perceived by participants as “agreeable, palatable, or satisfactory” (Proctor et al., 2011, p. 67). Evidence of feasibility in our earlier study were participants’ high ratings of the practice as pleasant (M = 7.9, SD = 1.9) and satisfying (M = 8.0, SD = 1.9), on single-item 0–10 Likert scales. In addition, participants stated that PRJ helped them to realize that more good things than bad things were happening in their lives in recovery and that they had more to be grateful for than they had realized. Participants explained that PRJ helped them to both reassess negative experiences and recognize progress in recovery. We were surprised to learn that participants described the right-hand side of the PRJ daily journaling page (which encourages users to plan a balanced day, See Figure 1) as a simple planner, which helped them remember and accomplish important tasks, such as taking vitamins or making medical appointments, leading to feelings of pride and accomplishment. While practicing PRJ, participants showed significant increases in satisfaction with life, well-being, reward from the environment, happiness with recovery, confidence staying sober, as well as decreases in depression. Although the within-group findings were promising, without a control group we were unable to determine whether these improvements were caused by PRJ.

1.3. The Current Study: Overall Aim and Objectives

The objective of this study was to conduct a small randomized controlled trial to: (a) replicate previous findings as to feasibility (i.e., PRJ as doable, achievable, easy) and acceptability (i.e., PRJ as satisfying, helpful, pleasant); (b) investigate the impact of PRJ on satisfaction with life, happiness with recovery, and commitment to sobriety compared to a control group; and (c) determine the impact of PRJ on secondary outcomes with respect to affect (negative affect, positive affect, serenity, depression, and anxiety); well-being (flourishing and gratitude); and addiction and recovery-related indicators (urges for drugs/alcohol, abstinence self-efficacy, quality of life in recovery, and satisfaction with the host treatment setting) compared to a control group. Study results would inform whether a future full-powered, definitive trial of PRJ would be warranted.

Our decision to cast a broad net when assessing positive change in recovery drew on the current state of recovery science. Both formal and informal conceptualizations of recovery have indicated that a comprehensive assessment of recovery should include aspects of well-being and aspects of pathology rather than focusing exclusively on substance consumption and symptoms (Neale et al., 2014, 2016; Witkiewitz & Tucker, 2020). Yet research on recovery has yet to discern which wellbeing indicators are optimal for the assessment of recovery (Hagman et al., 2022), leading us to include diverse indicators. Discovering which indicators respond to a journaling intervention might provide general guidance about which aspects of well-being are useful to assess in future recovery research.

We hypothesized that compared to control, PRJ would increase general satisfaction with life, which would lead to increases in happiness with recovery and these improvements would buttress participants’ commitment to sobriety as a marker of successful maintenance of behavior change.

2.0. Methods

2.1. Trial Design

We used an unblinded parallel-group randomized controlled design for this study. (See Figure 2 for a diagram of study timeline and procedures.) We intended to conduct in-person implementation but revised the protocol for remote delivery during the COVID-19 pandemic.

Figure 2.

Study Timeline and Procedures

After screening, baseline survey completion, and randomization, we structured the study as follows: At the beginning of Week 1, treatment group members received a journal via U.S. mail. During Weeks 1 and 2, the treatment group learned PRJ over eight 60-minute group sessions, and completed surveys after each group session, while the control group only completed surveys. During Weeks 3 and 4, there were no group sessions: treatment group members journaled independently and completed surveys after journaling, while control group members continued to complete surveys only. Week 4 ended with a survey and a qualitative interview for all participants in both groups. During Weeks 5 through 8 there was no study activity. At the conclusion of Week 8, we conducted a follow-up survey and interview. After the Week 8 follow-up, the control group received a copy of the PRJ journal and written instructions for how to use it. To discourage treatment diffusion, after randomization and during group sessions we asked participants not to discuss the details of the study with other clients at the treatment center. We asked at the qualitative interview: “What was it like to not be able to talk about the study to other [treatment center] clients?” All participants assured us they had not shared information. All study activities (i.e., screening, group work, and surveys) took place remotely via video conferencing and web-based questionnaires because of the COVID-19 pandemic.

2.2. Participants

2.2.1. Settings and Locations

The study recruited participants from two intensive outpatient programs and one residential program, each of which was operated by the same not-for-profit abstinence-based substance-use disorder treatment program in the upper Midwestern United States. The three settings employed similar treatment elements: educational lectures, group and individual therapy, trauma-informed treatment, gender-sensitive programming, and recovery management. The intensive outpatient programs served individuals of any gender identity; clients received twenty hours of treatment weekly, which was offered remotely during this study due to public health requirements attendant to the COVID-19 pandemic. The residential treatment program served individuals identifying as female, non-binary, or transgender; clients received twenty-one hours of treatment weekly.

2.2.2. Eligibility Criteria

To meet study inclusion criteria, at baseline, participants had to: (1) be at least eighteen years of age; (2) meet criteria for a substance use disorder in the past year; (3) be sufficiently literate in English to comprehend survey questions and write short lists; (4) have approximately two weeks of sobriety; (5) have been enrolled in the host setting for approximately two weeks; (6) agree to be audio-recorded; (7) have an email address that they checked regularly; (8) have access to a device with a camera that could access a stable internet connection; (9) have a planned discharge date from the host setting scheduled after conclusion of study activities; and (10) be available on the days and times of the PRJ group. The study excluded participants if they reported significant (as defined by the Global Appraisal of Individual Needs Short Screener [Dennis et al., 2013]) symptoms of a co-occurring psychiatric disorder in the week before the baseline interview or if they were unable to give informed, voluntary consent to participate.

2.2.3. How Participants Were Identified and Consented

Counsellors identified prospective candidates and study staff made remote recruitment presentations in the sites’ group therapy sessions. Study staff informed prospective participants that the purpose of the research was “to study if a specific kind of journaling … might support recovery from substance-use disorders compared to other kinds of check-ins on your emotions and recovery.” The study pre-screened interested individuals to determine whether they met inclusion criteria, and then they attended a remote appointment for more extensive screening, during which if they were eligible, they provided written informed consent.

2.2.4. Recruitment

Data collection took place on a rolling basis during 2020 and 2021. We enrolled six cohorts of between ten and sixteen individuals (N = 81) during the months of July (n = 10), August (n = 12), October (n = 16), and November (n = 15) 2020, as well as January (n = 13) and March (n = 15) 2021, ending the study when time and financial resources were exhausted.

2.2.5. Subject Remuneration

Participants in both groups received compensation for completing surveys in the form of gift cards to a general merchandise store: $15 for baseline, $5 for each of 21 daily surveys, $10 for the interview at Week 4, $10 for the interview at Week 8, and a $20 bonus for completing all interviews (at baseline and Weeks 4 and 8) and 90% of all other study activities. In total, each person could receive up to $160 in gift cards.

2.3. Intervention

Here and in supplemental materials we have included all the information suggested by the TIDierR checklist (Hoffmann et al., 2014), which provides guidelines about the elements of an intervention that should be presented in a publication. In addition, the PRJ intervention is described in detail by Krentzman and colleagues (2022). PRJ invites information about the preceding twenty-four hours using a bound notebook with pre-printed column headings under which participants write lists in response to two gratitude exercises (Emmons & McCullough, 2003; Emmons & Stern, 2013; Seligman et al., 2005) and one act-of-kindness exercise (Curry et al., 2018; Fredrickson et al., 2008). Individuals are encouraged to list what did not go well in the past day, to encourage emotional self-expression aligned with expressive writing interventions (Pennebaker & Beall, 1986). The plan for the upcoming twenty-four hours uses column headings to invite practices inspired by behavioral activation (Daughters et al., 2008; Daughters et al., 2018; Lejuez et al., 2011), including planning diverse activities and completing a values-clarification exercise (Kirschenbaum, 2013) while using SMART (specific, measurable, assignable, realistic, time-related) goal articulation to support successful completion (Doran, 1981). We administered PRJ via a set of materials including a 166-page PRJ journal printed with prompts, videos that described how to use each aspect of the journal, PowerPoint slides used to guide each treatment group session, and written PRJ instructions for dissemination to the control group upon study completion.

The first author, who has an MSW and a PhD in Social Work, as well as expertise in group leadership, Rogerian counselling, and Motivational Interviewing, taught PRJ to each treatment group during the eight group sessions. A standardized curriculum guided group sessions where each aspect of the journal was described and practiced. Homework involved PRJ journaling on days when group did not meet. The Supplemental Table describes the contents of each group session; contact the first author for a copy of the treatment manual. The intervention was not individualized, titrated, or modified over the course of the study.

2.4. Outcome Measures for Each Study Objective and the Statistical Methods of Estimation

2.4.1. Data Collection Methods

To input data for this study, all participants used their own smartphones or other devices to access web-based surveys. We asked individuals in the treatment group to upload an electronic photograph of that day’s PRJ entry as part of their survey. Survey questionnaire items were otherwise the same for both groups with the exception of minor modifications described below.

2.4.2. Objectives 1a–1e: Measures of Feasibility and Acceptability

There are two aspects of PRJ participation: the “PRJ entry” and a “PRJ item.” A PRJ entry is a daily journaling page with the left and right sides completed (see Figure 1). A PRJ item is a single bullet-pointed element written underneath the pre-printed column headings (e.g., “woke up sober,” “good food,” “go for a walk”). We determined intervention adherence and counts of PRJ entries and PRJ items by examining time and date stamped photographs of PRJ entries uploaded to daily surveys.

We determined the feasibility and acceptability of PRJ by (1a) calculating the percentage of eligible participants who enrolled in the study, (1b) calculating the percentage of those in the treatment group who attended at least one group session and then (for that population) tabulating group participation based on attendance records; (1c) determining the percentage of those in the treatment group who completed at least one PRJ entry and then (for that population) calculating the average number of entries completed per person; and (1d) tabulating the average number of items written in each PRJ entry per person. We calculated (1e) group differences in retention based on participation rates at Week 8 and determined (1f) group differences in participant-rated effort, difficulty, ease, satisfaction, pleasantness, and helpfulness of each groups’ study activities by administering six questions at the conclusion of each daily study activity over twenty-eight days as follows. To determine the degree of effort expended on study activities, for the control group, we asked, “In answering these questions today in this survey, I would say I put forth this much effort . . .”, whereas, for the treatment group, we asked, “For the Positive Recovery Journaling entry I made today, I would say I put forth this much effort . . .”, with options ranging from 0 (No effort) to 10 (Very much effort). To determine the degree to which study activities were experienced as easy, difficult, helpful, satisfying, or pleasant, for the control group, we asked, “I would say that answering the questions in this survey was . . .” whereas, for the treatment group, we asked, “I would say that this Positive Recovery Journaling entry was . . .”, with the corresponding options ranging from 0 (Not easy at all) to 10 (Extremely easy), 0 (Not difficult at all) to 10 (Extremely difficult), 0 (Not satisfying at all) to 10 (Extremely satisfying), 0 (Not pleasant at all) to 10 (Extremely pleasant), and 0 (Not helpful at all) to 10 (Extremely helpful).

2.4.3. Objective 2: Primary Recovery Indicators

We assessed recovery indicators at baseline and at Weeks 2, 4, and 8, except for satisfaction with the host treatment provider which we did not assess at baseline. Cronbach Alphas are reported in Tables 3 and 5.

Table 3, Figures 5, 6, and 7.

Satisfaction with Life

Note. The internal consistency of the Satisfaction with Life Scale ranged from α=.81 at baseline to α=.91 at Week 4. Scale range 5–35.

Table 5, Figures 11, 12, and 13.

Commitment to Sobriety

Note. The internal consistency of the Commitment to Sobriety Scale ranged from α=.88 at Week 2 to .97 at baseline. Scale range: 5–30.

2.4.3.1. Satisfaction With Life.

We assessed this construct with the Satisfaction With Life Scale (Diener et al., 1985), a five-item instrument (e.g., “The conditions of my life are excellent”) that uses a seven-point response format from 1 (strongly disagree) to 7 (strongly agree). Responses are summed, with higher scores indicating greater satisfaction with life. The instrument is a well-established measure (Pavot & Diener, 2008) shown in our previous research to significantly increase over time while participants practiced PRJ (Krentzman et al., 2022).

2.4.3.2. Happiness With Recovery.

To assess happiness with recovery specifically, we employed a single item: “In general, I am happy with my recovery,” with response options ranging from 0 (strongly disagree) to 10 (strongly agree). In our previous study, this item increased significantly over time while participants learned and practiced PRJ (Krentzman et al., 2022).

2.4.3.3. Commitment to Sobriety.

We assessed this construct via the Commitment to Sobriety Scale (Kelly & Greene, 2014), a five-item instrument (e.g., “I will do whatever it takes to recover from my addiction”) that uses a six-point response format from 1 (strongly disagree) to 6 (strongly agree). Items are summed with higher scores indicating greater commitment to sobriety. We employed this scale as an indicator of motivation to maintain abstinence from drugs and alcohol.

2.4.4. Objective 3: Secondary Recovery Indicators

2.4.4.1. Affect.

Improvement in affect is often the primary target in positive psychological interventions (Bolier et al., 2013; Carr et al., 2021). Therefore, we assessed present-moment affect with three subscales of the Positive and Negative Affect Schedule: negative affect (10 items, e.g., scared, irritable), positive affect (10 items, e.g., active, interested), and serenity (3 items, e.g., calm, at ease); (1 = Very slightly or not at all, 5 = extremely; Watson et al., 1988; Watson & Clark, 1994). We assessed affect over the past two weeks with the Hospital Anxiety and Depression Scale’s two 7-item subscales, one measuring depression (e.g., “I still enjoy the things I used to enjoy…” 0 = definitely as much, 3 = hardly at all) and one measuring anxiety (e.g., “I get sudden feelings of panic…” 0 = not at all, 3 = very often indeed; Zigmond & Snaith, 1983).

2.4.4.2. Well-being.

Flourishing and gratitude are commonly associated with positive psychological interventions (Dickens, 2017, 2019; Keyes & Haidt, 2002) and with recovery from substance-use disorders (Krentzman, 2019; Parker et al., 2018). To assess these dimensions of well-being, we employed the 7-item Flourishing Questionnaire (e.g., “I am a good person and I live a good life” 1 = strongly disagree, 7 = strongly agree [Diener et al., 2010]) and the Gratitude Questionnaire–Six Item Form (e.g., “I have so much in life to be thankful for” 1 = strongly disagree, 7 = strongly agree [McCullough et al., 2002]).

2.4.4.3. Addiction and Recovery.

Given that formal definitions state that addiction recovery connotes both decreases in pathology and increases in well-being (Betty Ford Institute Consensus Panel, 2007, 2009; Hagman et al., 2022; Neale et al., 2016; Witkiewitz & Tucker, 2020), we assessed several constructs to capture additional dimensions of recovery. The study measured Urges for Drugs or Alcohol with two items using the prompts, “Please rate the strongest urge to drink you experienced in the past 7 days” and “Please rate the strongest urge to use drugs you experienced in the past 7 days.” Response options ranged from 0 (No urge whatsoever) to 10 (Strongest urge I have ever felt). We used the stronger of the two responses to reflect urge. The study measured Abstinence Self-efficacy with a single item (Hoeppner et al., 2011): “How confident are you that you will be able to stay clean and sober in the next 90 days, or 3 months?” with response options ranging from 0 (Not at all confident) to 10 (Very confident). Quality of Life was assessed by asking the participant to reflect on their quality of life now, in recovery, as compared to then, during active addiction. For this, we modified the EUROHIS-QOL 8-Item Index (Schmidt et al., 2006) to align more strongly with the theory of the maintenance of behavior change (Rothman, 2000; Rothman et al., 2016). Our prompt asked participants to “Think back to your most recent period of active addiction. Compared to that time in your life, how would you rate your life now?” Response options ranged from 1 (… much better then) to 5 (… much better now) for the eight quality-of-life domains assessed in the original instrument (i.e., quality of life, health, energy for everyday life, able to perform daily activities, satisfied with myself, satisfied with my personal relationships, money to meet my needs, satisfied with my living place). The study assessed satisfaction with the services provided by the host treatment setting using the 8-item Treatment Satisfaction Scale (e.g., “If you were to seek help again, would you come back to [host treatment center]?” 1 = No, definitely not, 4 = Yes, definitely; Larsen et al., 1979). We used this instrument because previous research had shown that a behavioral activation intervention to treat substance use disorders improved participants’ satisfaction with host treatment (Daughters et al., 2008). No changes were made to measurements after the trial began.

2.5. Sample Size

2.5.2. Rationale for Numbers in the Pilot Trial

The study did not conduct power analyses for well-being outcomes.

2.6. Randomization

2.6.2. Type of Methods Used

At the end of the first baseline interview for each new cohort, we used REDCap’s computer-generated randomization procedure to assign this first participant in the cohort to the treatment or control group. After randomization of the first person, research staff assigned the next participant to the opposite group to ensure equal numbers in each group, and then used the REDCap randomization feature for the third person in the cohort, etc. This led to slightly different numbers of individuals in the treatment (n = 42) and control (n = 39) groups.

2.7. Implementation

Research staff members consented participants, performed the random allocation procedure, and provided individuals with instructions for their assigned arm of the study. Research staff members who conducted this work were the first author, four social work masters students and two social work PhD students.

2.8. Blinding

We did not employ blinding in this study; participants were aware that they were assigned to the “journaling now” or the “journaling later” group. Staff were not blinded to participant condition.

2.9. Statistical Methods

To summarize feasibility and acceptability metrics (Objectives 1a–d), we calculated means, standard deviations, medians, and percentages; to determine retention by study arm (Objective 1e), we conducted a Chi Square analysis. To analyze daily acceptability ratings (Objective 1f), we fitted separate linear mixed models for each outcome with time (in days), treatment group, and their interaction as the predictors, and time as a repeated measure within individuals. Time was modelled as a linear continuous variable with random intercepts and slopes per participant (i.e., latent growth curve models); these models produced estimates of intercept and slope for treatment and control and for the combined sample.

For primary and secondary recovery indicators (Objectives 2 and 3), we fitted a linear mixed model with four waves of time (baseline and Weeks 2, 4, and 8) as a categorical, fixed predictor, which modelled repeated measurements within persons using an unstructured covariance matrix. We used a contrast statement to examine cross-sectional treatment differences in least square means at each time point. We estimated all models using the proc mixed command in SAS 9.4, relying on maximum likelihood estimation.

2.9.1. Numbers Analyzed

We performed intention-to-treat and per-protocol analyses as follows. For intention-to-treat analyses, each participant remained in the study arm into which they were originally randomized. For per-protocol analyses, the control group was as originally randomized (n = 39), but to represent individuals who had received an acceptable level of PRJ learning, the treatment group was comprised of only those participants who had attended five or more group sessions (n = 27; see Figure 3 for details).

Figure 3.

CONSORT Flowchart of Participants

2.9.2. Differences at Baseline Between Those Retained and Those Not Retained at Week 8

We conducted chi square and t-test analyses to explore differences in baseline characteristics (all factors reported in Table 1) and baseline levels of the three primary recovery indicators (satisfaction with life, happiness with recovery, and commitment to sobriety) between those retained and not retained at Week 8. Those retained had significantly more years of education (M 13.6, SD 2.3 versus M 12.2, SD 1.2; t(77.6) 3.44, p < .001) and significantly more days of sobriety (M 170.5, SD 222.8 versus M 90.6, SD 94.6; t(71.7) 2.23, p = 0.029), with no other differences. Based on these results, we adjusted for baseline years of education in all models and addressed differences in length of sobriety in post hoc analyses (see Section 3.4.1, Post Hoc Hypotheses, below).

Table 1.

Baseline Demographic and Clinical Characteristics of the Sample

Baseline Characteristic	Full Sample (N = 81)	Treatment (n = 42)	Control (n = 39)
Age in years, M (SD)	39.0 (11.0)	38.6 (11.7)	39.4 (10.4)
Gender identity, n (%)
Female	43 (53.1%)	24 (57.1%)	19 (48.7%)
Male	37 (45.7%)	18 (42.9%)	19 (48.7%)
Gender nonconforminga	1 (1.2%)	0 (0.0%)	1 (2.6%)
Race, n (%)
Black or African American	4 (4.9%)	3 (7.1%)	1 (2.6%)
White	60 (74.1%)	28 (66.7%)	32 (82.1%)
Latine	2 (2.5%)	1 (2.4%)	1 (2.6%)
Native American or Alaskan Native	5 (6.2%)	3 (7.1%)	2 (5.1%)
Asian or Pacific Islander	1 (1.2%)	0 (0.0%)	1 (2.6%)
Biracial/Multiracial	9 (11.1%)	7 (16.7%)	2 (5.1%)
Years of education, M (SD)	13.1 (2.1)	13.0 (2.4)	13.1 (1.7)
Past year household income ≤ $15,000, n (%)	51 (63.0%)	26 (61.9%)	25 (64.1%)
Relationship status, n (%)
Never married/single	38 (46.9%)	19 (45.2%)	19 (48.7%)
Married, remarried, in a relationship	21 (25.9%)	13 (31.0%)	8 (20.5%)
Separated, divorced, widowed	22 (27.2%)	10 (23.8%)	12 (30.8%)
Substance most addicted to, n (%)
Alcohol	37 (45.7%)	18 (42.9%)	19 (48.7%)
Opiates	21 (25.9%)	9 (21.4%)	12 (30.8%)
Amphetamines	19 (23.5%)	11 (26.2%)	8 (20.5%)
Cannabis	2 (2.5%)	2 (4.8%)	0 (0.0%)
Cocaine/crack	2 (2.5%)	2 (4.8%)	0 (0.0%)
Length of sobriety in days, M (SD)b	139.9 (187.8)	135.4 (224.3)	144.9 (141.3)
Number of previous treatment episodes, M (SD)	6.0 (3.4)	6.3 (3.6)	5.7 (3.1)
Negative consequences of substance use M (SD) c	15.6 (11.4)	15.9 (10.9)	15.2 (12.0)
Receiving medication-assisted treatment, n (%)	33 (40.7%)	19 (45.2%)	14 (35.9%)
Reports a legal issue, n (%)	35 (43.2%)	19 (45.2%)	16 (41.0%)
Reports a history of trauma, n (%)	77 (95.1%)	40 (95.2%)	37 (94.9%)
Alcoholics Anonymous Affiliation Scale M (SD) d	4.9 (2.3)	4.7 (2.3)	5.2 (2.2)

Note. To determine whether randomization succeeded, we conducted t test and chi square analyses to compare the treatment and control groups on all baseline demographic and clinical characteristics in this table; there were no statistically significant differences between the groups.

^aExpressed in participant’s own words.

^bLength of sobriety ranged from 9 to 498 days with the addition of one person with 649 days and one with 1,426 days; the median value of this variable was 87.5 days (IQR 55.7–148.6).

^cWe used the 10-item Short Inventory of Problems–Alcohol and Drugs Scale (e.g., “I have failed to do what is expected of me because of my drinking or drug use” 0 = never, 3 = daily or almost daily; Hagman et al., 2009) to assess negative consequences of substance use; α at baseline = .97.

^dWe used the 9-item Alcoholics Anonymous Affiliation Scale (e.g., “Have you ever considered yourself a member of [primary mutual-aid group]?” yes/no, Humphreys et al., 1998) to assess affiliation to participant’s self-identified primary mutual-aid recovery group; α at baseline = .80.

2.9.3. Covariates

All models adjusted for years of education and recruitment site as fixed covariates. We could not also adjust for cohort because of small cell sizes.

2.10. Important Changes to Methods After Pilot Trial Commencement

Iterative modifications to the study were necessary given the remote delivery of study activities during COVID-19. To enhance recruitment efforts, we added the residential treatment center and a second outpatient treatment center as sites. We began making recruitment presentations at site therapy groups rather than relying solely on counsellor referrals. To improve retention and to strengthen the intervention, we created short videos to summarize key elements of group sessions so that participants in the treatment group could view them to review key concepts or catch up if they missed a session.

3.0. Results

3.2. Baseline data

Table 1 presents the baseline characteristics of the total sample and the treatment and control groups.

3.3. Outcomes and Estimation

Objective 1: Feasibility and Acceptability. The majority who were eligible for the study enrolled in it (94.2%: those enrolled (N = 81) divided by those assessed (n = 107) minus those who did not meet criteria (no stable internet, significant co-morbid psychiatric conditions, and active substance use, n = 14) and those who could not make the scheduled group sessions (n = 7; i.e., 81/86)). Most treatment group members (n = 39, 92.9%) attended at least one of the eight group sessions. Among those who came to at least one group session, attendance ranged from one to eight sessions (M = 5.7, SD = 2.4, Median = 6.0), indicating that participants attended on average 71.3% of group sessions. Most participants in the treatment group (n = 37, 88.1%) completed at least one PRJ entry (the remaining 5 randomized to the treatment group attended 0 to 2 group sessions and submitted no PRJ entries). Among those who completed one or more entries, the number of PRJ entries completed ranged from one to twenty-eight (which was the maximum number possible; M = 15.8, SD = 8.7, Median = 19.0), indicating that on average participants completed 56.4% of the assigned daily PRJ entries. Participants who completed at least one PRJ entry wrote M items = 37.4 (SD = 13.7, Median = 35.8) on each entry.

Survey completion rates at Week 8 were 52.4% (n = 22) for the treatment group and 71.8% (n = 28) for the control group (X² (1, N = 81) = 3.2, p = .073). Both treatment and control groups reported expending high levels of effort in their daily study activities, rating their respective activities as very easy, satisfying, helpful, pleasant, with means and standard error bars ranging in the top half of the scale for these outcomes (i.e., scores > five on a 0–10 point scale; Figure 4), and not difficult, with means and standard error bars ranging in the bottom half of the scale for this outcome (between scores of zero and five, see Figure 4). In terms of intercept, treatment and control groups did not differ in their rating of the difficulty or ease of their assigned tasks (i.e., PRJ for the treatment group, completing daily surveys for the control group); however, the PRJ group rated their tasks as significantly more satisfying, helpful, and pleasant (ps < .01, Table 2) than the control group. Self-reported effort decreased significantly over time for both groups (by 0.04 points per day). The interaction between treatment group and time was significant for easy; the “ease” of study activity decreased significantly for the control group only (by 0.04 points per day). For all of the other feasibility and acceptability outcomes depicted in Table 2, we were unable to detect any significant change over the course of the study (Table 2, Figure 4.)

Figure 4.

Feasibility and Acceptability Indicators Over Time by Treatment Group

Note. X-axis represents time in days. Error bars indicate the standard error of the mean. Scale range of all outcomes is 0–10. Days with < 4 respondents are not graphed but are included in the analysis.

Table 2.

Differences Between Treatment and Control Groups on Feasibility and Acceptability Indicators

Intercept	Estimate of the least square mean difference between treatment and control at Day 1 (treatment mean–control mean) [95% CI]	t	p	Estimated effect size2
Satisfying	1.4 [0.7, 2.0]	4.33	<.0001	0.78
Helpful	0.9 [0.3, 1.5]	3.19	0.002	0.48
Pleasant	1.4 [0.8, 2.1]	4.67	<.0001	0.88
Effort	0.3 [−0.6, 1.2]	0.58	0.562	0.15
Easy1	0.3 [−0.4, 1.0]	0.91	0.364	0.18
Difficult	0.1 [−0.8, 1.0]	0.22	0.824	0.05
Slope	Time coefficient	t	p	Estimated effect size3
Satisfying	0.003 [−0.012, 0.018]	0.39	0.699	0.05
Helpful	−0.002 [−0.014, 0.009]	−0.42	0.677	−0.03
Pleasant	−0.001 [−0.017, 0.015]	−0.09	0.926	−0.01
Effort	−0.041 [−0.060, −0.022]	−4.26	<.0001	−0.55
Easy (treatment)	0.008 [−0.023, 0.039]	0.54	0.588	−0.14
Easy (control)	−0.036 [−0.063, −0.008]	−2.60	0.012	−0.69
Difficult	0.008 [−0.014, 0.029]	0.72	0.473	0.10

Note. All scales range from 0 (lowest rating of each outcome) to 10 (highest rating of each outcome).

¹Our first set of models included the interaction between time and treatment condition, but the interaction was not significant for any outcome (ps > .3143) with the exception of Easy, for which the interaction was significant (p = .037). Therefore, we removed the interaction term from all of the models except for Easy to produce the intercept and slope estimates for this table. For the model with the outcome Easy, we report the difference in least square means with the interaction term included in the model, which results in two separate time coefficients for treatment and for control for the slope portion of the table, indicating that tasks became less easy for control but showed no change for treatment.

²We calculated effect sizes for differences between treatment and control for their intercepts at Day 1 (the first day when they completed study activities) based on model estimated group intercepts which adjust for site and education. Thus, the effect size is calculated as d _INTERCEPT = (a₁ - a₂) ⁄ SD, where a₁ is the model estimated intercept for treatment and a₂ is the model estimated intercept for control; and SD was the pooled standard deviation of the outcome measure at Day 1.

³We based the effect-size calculation of the time effect on the equation for the effect size for the standardized mean difference at the end of treatment calculated from linear slope difference between treatment and control groups from a growth modelling analysis (GMA; [Feingold, 2009]). Thus, we calculated within-group effect sizes for the linear effect over time as d _GMA = (b * duration) ⁄ SD, where b was the unstandardized coefficient for the effect of time (rate of change of a continuous outcome per unit of time), duration = 28, the number of days in this phase of the study, and SD was the pooled standard deviation of the outcome measure at Day 1.

Objectives 2 and 3: Impact on Primary and Secondary Recovery Indicators. We were unable to detect any significant differences between the PRJ and control groups for any primary or secondary recovery outcomes in either the intention-to-treat or per-protocol analyses. We ran the intention-to-treat analyses with and without covariates. Differences were negligible; therefore, all models reported herein adjust for covariates. (See Tables 3, 4, and 5 and their associated figures for details for primary recovery indicators. See Supplemental Graphs for graphs of the secondary recovery indicators.)

Table 4, Figures 8, 9, and 10.

Happy With Recovery

Note. Assessed with a single item. Scale range: 0–10.

3.4. Ancillary Analyses

3.4.1. Post Hoc Hypotheses

We investigated two post hoc hypotheses to determine whether PRJ might be more effective for two groups of more vulnerable individuals: those in earliest recovery (< 90 days sobriety at baseline, range = nine to 89 days; n = 41) and those with greater depression (depression scores > the median of four at baseline, range > 4 = 5 to 13, n = 36). Our length-of-sobriety hypothesis derived from research that considers <90 days of sobriety to be a vulnerable time attended by highest risk of recurrence of use (Hagman, 2022; Sinha, 2011); it is a phase that the National Institute of Alcohol Abuse and Alcoholism has given the moniker, “initial” recovery, a stage that predates even “early” recovery (Hagman, 2022). The hypothesis also was derived from the findings of our uncontrolled study of PRJ (Krentzman, Hoeppner, et al., 2022) where statistically significant psychosocial improvement co-occurred with PRJ; that sample had lower length of sobriety at baseline (M 47.8 days, SD 31.6) than the current sample (M 139.9 days, SD 187.8), suggesting perhaps that PRJ might be more impactful for those with lower length of sobriety.

We arrived at our hypothesis about depression based on PRJ’s behavioral activation component: behavioral activation was developed to treat depression (Lejuez, et al., 2011) and has been shown to be beneficial for individuals with co-occurring depression and substance use disorder (Daughters, et al., 2008; Magidson et al., 2011). Accordingly, we thought that PRJ might be more impactful for those with greater depression. We conducted subset analyses for this early evaluation rather than full moderation analyses (i.e., including those with > 90 days of sobriety and those with low levels of depression), as those would be underpowered in this small sample.

Therefore, we investigated whether recovery indicators would differ between treatment and control groups among two separate subsets of the sample: those with less sobriety and those with greater depression. First, we graphed the means with standard errors by study arm over time for each outcome for each of these two subsets of the data. We interpreted a lack of overlap in the standard error of the mean bars to indicate differences worth exploring in future work. For those with < 90 days sobriety, mean levels showed beneficial impact of PRJ, with gaps in at least one pair of standard error bars for ten of the fourteen recovery indicators.

We tested whether any of these differences were statistically significant using the per-protocol sample and the analytic model described above while limiting the analysis to individuals who at baseline had < 90 days. For primary recovery indicators for this subset of the data, PRJ showed significantly greater satisfaction with life at Weeks 2, 4, and 8 and significantly more happiness with recovery at Week 4. (See Tables 3 and 4.) For secondary recovery indicators, PRJ showed significantly lower negative affect at Week 8, significantly less depression at Week 4, significantly greater quality of life at Week 4, and significantly greater satisfaction with host treatment at Week 8. We did not adjust for multiple comparisons in these exploratory post hoc analyses. (See Supplemental Graphs for graphs of the secondary recovery indicators.)

Among those with higher depression scores, we observed gaps between the standard error of the mean bars for only two outcomes: the control group had greater quality of life at baseline and lower satisfaction with host treatment at Week 8 compared with the treatment group. Because standard error bars overlapped in the graphs of all other comparisons, we did not further analyze treatment differences in this subgroup.

3.4.4. Harms and Other Important Unintended Consequences

No harms were observed or reported. Interviews conducted at Week 4 revealed that most control group members described having benefitted from the daily surveys, including feeling validated in their current recovery efforts and being inspired to initiate new recovery behaviors. Details of these qualitative analyses are available in prior work (Krentzman & Gass, 2024; Krentzman, Gass, & Lowery, 2023). We extracted substance use and treatment retention data from the treatment record and intended to examine these as outcomes, but we failed to consider that the control group would also have been exposed to the journaling intervention before discharge from treatment, invalidating an investigation of the impact of journaling on these outcomes.

4.0. Discussion

This is the first randomized controlled study to test PRJ, a novel intervention combining aspects of positive psychology and behavioral activation that—with its roots in reinforcement theory (Higgins et al., 2004)—was designed to increase satisfaction and happiness with life in recovery. The current study built on previous work (Krentzman, Hoeppner, et al., 2022) and provided additional evidence for the feasibility and acceptability of PRJ even when, as was the case in this research, the practice was taught and administered remotely. Our work demonstrated that participants were able to complete PRJ activities and rated them as being just as easy as the control group’s activities, albeit significantly more helpful, pleasant, and satisfying. Post hoc analyses suggested that, for individuals who at baseline had < 90 days of sobriety, PRJ could improve satisfaction with life, happiness with recovery, negative mood, quality of life, and satisfaction with the host treatment setting.

Despite these positive findings, it is important to reiterate that in the full-sample intention-to-treat and per-protocol analyses, we observed no significant differences between treatment and control groups with respect to any recovery indicators. While we saw a treatment effect of PRJ for a subsample, it is unclear why we did not observe a stronger effect of PRJ for the entire sample.

Positive psychological and gratitude interventions have yielded only small or in some cases medium effect sizes (e.g., Bolier et al., 2013), with some studies specifically pointing out methodological problems that further reduced the impact of these interventions (Dickens, 2019; White et al., 2019). In addition, meta-analyses of interventions related to PRJ (e.g., loving-kindness meditations [Zhou et al., 2022]) and behavioral activation approaches (e.g., Stein et al., 2021) have shown treatment effects only when the control condition was inactive; treatment effects for these interventions disappeared when the control condition was active. Consistent with this finding, in a qualitative analysis of control group members who participated in the current study (Krentzman & Gass, 2024; Krentzman, Gass, & Lowery, 2023), we found that most control group members reported at least some benefit from their assigned study tasks (daily surveys), suggesting that measurement reactivity experienced by the control group could have made any effect of PRJ harder to detect.

Our finding of positive effects for individuals with < 90 days of sobriety presents a novel contribution to the extant yet limited body of research investigating the impact of positive psychological interventions on addictive behavior. Length of sobriety can be a useful construct for a number of reasons, including its impact on retention. Pott and colleagues (2022) found that more days of abstinence at baseline were correlated with completion of behavioral activation treatment (Pott et al., 2022), a finding consistent with the results of the current study, wherein we found that greater length of sobriety was associated with retention at Week 8. However, in positive psychological recovery research and studies of the impact of behavioral activation on addiction recovery, length of sobriety is inconsistently reported and there is no real measurement standard; Krentzman and colleagues (2015) report length of sobriety in days and Hoeppner and colleagues (2019) report length of sobriety as a categorical variable. Moreover, while scholars have suggested that “pleasure-oriented interventions” might vary based on recovery stage (Boden et al., 2017, p. 121), to our knowledge no prior research has investigated length of recovery as a moderator of the impact of positive psychology or behavioral activation interventions on outcomes in substance use disorder research. In a review of reinforcement-based interventions, Fazzino et al. (2019) reported no tests considering length of sobriety as a moderator. The findings of the current study suggest that length of sobriety should be included in future research along these lines.

This study took place during COVID-19, a period with documented challenges for those seeking recovery and those in recovery, including an upheaval in society in general and in substance use treatment provision, with centers, including the two outpatient sites in this study, switching rapidly from in-person to online care. Therefore, it is important to note that this was not an efficacy study conducted under ideal conditions but a depiction of PRJ amid real-world, distressed circumstances.

4.1. Limitations

Several limitations should be considered when interpreting this study’s findings.

We noted poorer retention of individuals in the treatment group at Week 8 (52.4% compared to 71.8% for the control group), although this difference was not significant. We speculate that retention was weaker for the PRJ group because we asked them to do more work: for example, inviting them to attend eight PRJ groups at noon after completing four hours of treatment from their host setting and also asking them to complete PRJs daily for twenty-eight days before then completing surveys, while the control group only completed surveys. To complete study activities, the PRJ group needed a complete set of materials and resources: their PRJ journal, a pen, their electronic device, and a block of time to complete study activities. By contrast, the control group only needed their electronic device to complete their study activities, making it more convenient for them. In actual practice where PRJ groups would be incorporated within substance use disorder treatment settings, many of these barriers would be removed: PRJ would be completed in a group setting incorporated into the overall treatment schedule, there would be no need to take a photograph of the PRJ entry, and no electronic survey to complete.

In addition, we observed occasions when challenges based on remote administration were followed by a PRJ member dropping out. Such challenges—again, faced only by the PRJ group—included a journal arriving late via US mail, remote conference software glitches, and internet connectivity problems. An additional limitation is that the treatment group received more attention and contact from study staff via the eight-session journaling groups. Future research should provide control group members with a contact-matched control condition.

The study was not blinded. Based on our qualitative data, we know that some in the control group seemed to think that completing questionnaires was in fact “journaling” while others in the control group knew very well that they were in the “journaling later” group and let us know that they preferred to have been assigned to the “journaling now” or treatment group. Whether or to what extent this led to compensatory rivalry we do not know; however, our qualitative data show that the majority of control group members reported at least some benefit from completing daily questionnaires (Krentzman & Gass, 2024; Krentzman, Gass, & Lowery, 2023). Replications of this research should administer assessment questionnaires less frequently to diminish measurement reactivity.

While we took precautions to limit individuals in the journaling group from sharing the intervention with control group members, and while all participants in exit interviews assured us that they had not talked to peers about their assigned study activities, we were adequately reassured, but could not be entirely certain, that this did not occur.

Our results should be generalized with caution. Most participants were white, which reflected the racial and ethnic composition of the recruitment sites. With the limited diversity of this sample, PRJ appeared acceptable and feasible, but PRJ’s acceptability and feasibility to BIPOC clients should be determined with consideration for culturally specific adaptations to both the means of teaching PRJ and the journal’s contents. In the same sense, the generalizability of our study could be limited by the geographic and social context of the Upper Midwest amid the COVID-19 pandemic. The study’s inclusion and exclusion criteria selected individuals with stable internet, approximately two weeks of sobriety, adequate literacy to write short lists, and other factors which limit generalization to individuals with substance use disorders who are more vulnerable based on these clinical and socioeconomic factors. While the host treatment centers endorsed abstinence as the consumption-related aim of treatment, we did not ask participants what their own consumption goals were and so are uncertain of the impact of journaling based on such goals.

4.2. Conclusion

Early recovery is a period of high risk for recurrence to substance use: it is a time when individuals experience a preponderance of negative affect and extremely challenging biopsychosocial circumstances. PRJ is designed to bring positive elements of life in early recovery to the foreground while encouraging additional positive experiences. In a post hoc analysis, our study found evidence of a beneficial effect of PRJ with a vulnerable subgroup, those with less than 90 days of sobriety. Our results emphasize the importance of protocol modifications before proceeding to a definitive trial of PRJ. Participants in both groups should be assessed less frequently to reduce control group reactivity. Future PRJ protocols should either recruit only those in the earliest phase of recovery or should consider length of recovery as an important moderator in a priori hypotheses. Such work would produce a truer estimate of the effect of PRJ, clarify for whom PRJ might be most useful, and identify the well-being indicators through which PRJ might encourage successful, lasting recovery.

Highlights.

• Positive Recovery Journaling (PRJ) is based in positive psychology

• PRJ is designed to make life in recovery more rewarding and reinforcing

• PRJ was rated more satisfying, helpful, and pleasant than the comparison condition

• Treatment and comparison were not different on any other recovery outcome

• Those in earliest recovery benefitted the most from PRJ