Preliminary Estimates of Recovery Capital Growth in Early and Middle Stages: Social Work Practice Implications

Abstract

As the recovery capital framework emerges, we still know little about growth rates in early- and middle-stage recovery. This study aimed to: 1) hypothesize recovery capital growth trajectories via graphical representations employing fitted lines and curves; and 2) generate preliminary recovery capital growth estimates through statistical models that predict monthly and yearly growth in recovery capital scores. We used the Multidimensional Inventory of Recovery Capital (MIRC), a new, psychometrically validated measure, to assess recovery capital on a sample of (n=482) people, as well collecting key control variable data. Regarding Aim 1, graphical plots showed that growth was observed for all forms of capital; however, not for human and cultural capital in the first 12 months. Regarding Aim 2, controlled regression models predict an overall increase of approximately one unit of total recovery capital per year. Social workers and other recovery professionals may use these results to assess and treatment plan.

Introduction

Although millions of people in the United States and globally continue to struggle with problems related to alcohol and other drug use, many do resolve their problems, a phenomenon known as recovery. While past definitions of recovery from substance misuse focused on abstinence, newer scholarship emphasizes the multifaceted nature of recovery (Ashford et al., 2019; Eddie et al., 2022; Hagman et al., 2022). Recovery is a dynamic process characterized by the pursuit of wellness and gains in well-being, and during which a person may be abstinent from all substances or continue to use some substances moderately (Ashford et al., 2019; Hagman et al., 2022; Kelly et al., 2018b).

Holistic definitions of recovery highlight the importance of building internal and external resources as part of the recovery process, drawing on the concept of recovery capital (Hagman et al., 2022). Based on their research on natural recovery (e.g., individuals recovering without the use of formal treatment programs or self-help groups), Cloud and Granfield (2008) conceptualized recovery capital as a network of positive and negative inputs across four domains– social, physical, human, and cultural capital. Social capital relates primarily to social support and utilization of social networks, whereas physical capital pertains to tangible or financial resources, such as housing and income. Human capital is comprised of the individual characteristics and skills someone has, including physical and mental health, coping skills, and attitudes. Finally, cultural capital captures the milieu of environmental conditions with respect to both the physical community as well as cultural values and beliefs in which one is embedded. Across domains, Cloud and Granfield (2008) theorized that greater positive capital and lesser negative capital (for example, negative social influences) facilitate a person’s recovery.

Following Cloud and Granfield’s (2008) foundational work, a burgeoning literature has explored recovery capital and its features and functions in a range of populations (Best & Hennessy, 2022; Hennessy, 2017; Pouille et al., 2021). There has been a particular interest among researchers in developing and using measures of recovery capital, in order to quantify this construct and to study its association with various recovery outcomes (Bunaciu et al., 2023). Although several recovery capital measures exist, many are impacted by significant psychometric shortcomings. These including misalignment with recovery capital theory (e.g., not measuring the four domains conceptualized by Cloud and Granfield) and being evaluated with primarily homogenous samples in which individuals who are White, middle-class, and/or recruited from treatment programs or self-help groups are over-represented (Arndt, 2017; Bowen et al., 2020).

A new recovery capital measure, the Multidimensional Inventory of Recovery Capital (MIRC), was recently developed in an effort to address these shortcomings. The 28-item MIRC assesses both positive and negative recovery capital across four subscales (seven items each), based on Cloud and Granfield’s (2008) domains of social, physical, human, and cultural capital (Bowen et al., 2023). Further, the MIRC was developed and tested with a sample of people in recovery from alcohol problems that was diverse with regard to several characteristics, including race, ethnicity, gender, age, socioeconomic status, and recovery experience (Bowen et al., 2022; Bowen et al., 2023). The measure has demonstrated moderate to strong reliability and validity for the full measure and its four subscales, including moderate to good internal consistency for each subscale (α=0.65–0.81) and for the full measure (α=0.91), while also demonstrating concurrent validity to related World Health Organization measures of social support and human capital among other constructs (Bowen et al., 2023).

One of the applications of the MIRC is for social workers, other recovery professionals, and researchers to understand how recovery capital may be expected to change based on a person’s length of time in recovery. Time in recovery can be conceptualized in broad stages, e.g., initial recovery (less than 3 months), early recovery (three to 12 months), sustained recovery (one to five years), and stable recovery (more than five years) (Betty Ford Institute Consensus Panel, 2009; Hagman et al., 2022). Research suggests that these stages of recovery may vary meaningfully in the amount of support individuals require, and by extension, the amount of recovery capital that individuals are likely to possess (Dennis et al., 2007; Roxburgh et al., 2023). In a large study that included people who reported having resolved a substance use problem for periods of time ranging from less than one year through 40 years, Kelly et al. (2018b) noted that, “the relationship between recovery capital and time since problem resolution was best characterized by a quadratic term, which similarly suggests rapid accrual of recovery capital during the early years after problem resolution followed by a continued, but attenuated accrual” (p. 774). While informative, Kelly and colleagues’ study utilized the 10-item Brief Assessment of Recovery Capital (BARC-10), which does not align with the categories of social, physical, human, and cultural capital (Vilsaint et al., 2017). Thus, it is not known if and how these domains may change over the course of recovery.

Following Best and Hennessy’s (2022) call for research on how recovery capital varies with the length of time in recovery and whether this is consistent across different domains, this study aimed to use the MIRC to enhance this understanding. In the absence of any extant longitudinal data to leverage specific information on individual growth trends, we used cross-sectional data from respondents with varying amounts of time in recovery to generate plausible estimates about recovery capital growth. Specifically, this study had two aims: 1) to hypothesize recovery capital growth trajectories through graphical representations using fitted lines and curves; and 2) to generate preliminary recovery capital growth estimates through statistical models which predict monthly and yearly growth in recovery capital scores. Both Aim 1 and Aim 2 were carried out for Total Recovery Capital (TRC), Social Capital (SC), Physical Capital (PC), Human Capital (HC), and Cultural Capital (CC), using MIRC total and subscale scores.

Methods

Sampling

The sample for this study stems from a parent study developing the MIRC (see Bowen et al., 2023). Participants were eligible for the study based on three demographic criteria (i.e., age 18 or greater; resident of the United States; and English communication ability) as well as one substance use criterion, self-identification as having resolved a prior problem with alcohol (either alcohol alone, or both alcohol and one or more drugs) for a period of at least 30 days prior to survey response. Similar criteria have been used in other recovery research studies (e.g., Gilbert et al., 2021; Kelly et al., 2018a). The eligibility criterion focusing on alcohol was informed by funder priorities, as well as recognition of the unique recovery context surrounding alcohol as a legal, widely available, and largely culturally sanctioned substance.

Data were collected in 2022 through Innovate MR, an online sampling provider that maintains a panel of more than 1 million respondents. Online panels are increasingly used as a means of data collection in addiction research. They offer the advantage of recruiting participants from a broad range of backgrounds and with a variety of experiences with treatment and services, including participants in natural recovery, in contrast to the historical reliance in addiction research on recruiting participants from treatment programs or self-help groups (Jones et al., 2022; Subbaraman et al., 2015). Participants completed an eligibility screening based on inclusion criteria, with additional questions to conceal the study aims, deter false positives, and implement quality control procedures such as attention checking, duplication prevention, and straight lining prevention. All data were subject to quality review by the collection firm, individual review of the research team, and statistical analysis. Eligible participants were then administered a survey which included the MIRC, demographic items, and other recovery-related information, such as length of time in recovery.

Variables

This study makes use of eleven analytic variables, including five outcome variables (i.e., TRC, SC, PC, HC, CC), one key predictor variable (length of time in recovery), and three standard control variables (i.e., age, sex, race/ethnicity), as well as two variables for sensitivity check models (i.e., income and education). The recovery capital outcome variables were measured using the MIRC, taking the total score for the full measure (TRC) and the scores for the four subscales (SC, PC, HC, CC). Length of time in recovery was self-reported with five categorical response options (i.e., 1–3 months, 3–6 months, 6–12 months, 1–5 years, and 5 years or more). The inconsistent interval between these categories hampers sensible interpretability when treated as categorical. Therefore, this variable was recoded as continuous, measured in months of recovery, taking the minimum value of each categorical range for the primary variable and taking maximal values with the final category equal to 120 months as a specification choice sensitivity check.

Sex was a binary variable (male or female), based on self-report of assigned sex at birth. Age was a continuous variable derived from birth year. Race was a categorical variable with response options including White, Black, American Indian, Asian American, Hawaiian or Pacific Islander, more than one race, and other race. Ethnicity was a binary response regarding Hispanic/Latino ethnicity. From these, a binary variable representing White (non-Hispanic White-only) and People of Color was generated. Participants self-reported their household income in a nine-level categorical variable, which was converted to a binary indicator variable of low-income status (having a household income less than 200% of the federal poverty level). Finally, education was a five-level categorical variable including less than high school, high school, associate’s degree, bachelor’s degree, and graduate degree.

Analysis

Our analytic strategy included two methods to investigate the research questions. First, in assessing Aim 1, we produced graphical displays of the uncontrolled aggregate raw data on growth by plotting observations on total recovery capital and each subscale across each observational time point. For each subscale, we display both a linear (line of best fit) and curvilinear (lowess smoothed growth curve) representation, providing alternative central tendency representations of growth across time. Then, to assess Aim 2, we estimated ordinary least squares multiple linear regression equations to produce predictions about the return to recovery capital of an additional month or year of time in recovery. These equations were estimated for the sample of persons in recovery, with time in recovery varying from approximately 1 month to greater than 60 months. All models employed controls for respondent sex, race, and age. We elected to run sensitivity check models for inclusion of income and education controls as well; however, we theorized that controls for education and income may be inappropriate given that these variables may themselves be key mechanisms for the growth of recovery capital. We also ran sensitivity checks for alternate specifications of the time in recovery variable.

Results

Quota sampling of participants (n=482) was conducted to approximate the over-18 demographics of the United States with respect to sex and race, as shown in Table 1. Mean age for the sample was 43.3 years. One quarter of respondents (26%) held a bachelor’s degree or greater and 41% were below 200% of the federal poverty threshold. One quarter of respondents had been in recovery for less than one year, while 41% were in the one-to-five-year range, and 34% for five years or more. The average total score on the MIRC was 77.4 (range=28–112; SD=13.12). The SC subscale had the highest average score (M=20.84, SD=3.49), while the HC subscale was the lowest (M=17.57, SD=4.06).

Table 1.

Sample demographics and MIRC scores (N=482)

Sample Characteristic	N (%)	Mean (sd)
Age		43.3
Sex
Female	238 (49.4)
Male	244 (50.6)
Race
White-only	299 (62.0)
People of color	183 (38.0)
Income
< 200% poverty level	197 (40.9)
> 200% poverty level	285 (59.1)
Education
< High school	29 (6.0)
High school	247 (51.2)
Associate’s degree	82 (17.0)
Bachelor’s degree	92 (19.1)
Graduate degree	32 (6.6)
Recovery Duration
1–3 months	39 (8.1)
3–6 months	34 (7.1)
6–12 months	48 (10.0)
12–60 months	197 (40.9)
60+ months	164 (34.0)
Total Recovery Capital		77.40 (13.12)
Social Capital		20.84 (3.49)
Physical Capital		20.60 (4.47)
Human Capital		17.57 (4.06)
Cultural Capital		18.39 (3.84)

With regard to Aim 1, for TRC, SC, and PC, growth was observed in the early recovery stage (prior to 12 months). Both linear best-fit and lowess smoothed curve plots show upward trends in overall growth for each of these domains across the first 12 months, which then continues during the following years (see Figures 1, 2, and 3). Each of these variables also appear to grow somewhat more quickly across the first 12 months where the estimated growth pattern is greater, before the rate of growth declines, though stays positive, over the longer term. For HC and CC, however, the linear picture shows an early upward trend, but lowess plots reveal that this positive trend was driven by gains accumulated after one year of recovery, while early recovery trends show what is effectively no change in the initial 12-month period (see Figures 4 and 5).

Figure 1.

Lowess Smoothed Curve and Best Fit Line for Total Recovery Capital by Months in Recovery

Figure 2.

Lowess Smoothed Curve and Best Fit Line for Social Capital by Months in Recovery

Figure 3.

Lowess Smoothed Curve and Best Fit Line for Physical Capital by Months in Recovery

Figure 4.

Lowess Smoothed Curve and Best Fit Line for Human Capital by Months in Recovery

Figure 5.

Lowess Smoothed Curve and Best fit Line for Cultural Capital by Months in Recovery

In uncontrolled bivariate regression models assessing Aim 2, time in recovery was significantly related to TRC and each recovery capital domain. Notably, the bivariate relationship between length of time in recovery and TRC was fairly strong for a single predictor (r=.25, p<.000). In fact, on average, one month of time in recovery was associated with an additional 0.14 units of TRC, or a 1.68 point increase on the total MIRC score over one year. Controlled regression models estimated more modest growth. For SC, PC, HC, and CC, growth appeared somewhat meager, whereas greater growth was seen in TRC over time (see Table 2). On average, controlling for age, sex, and race, one additional month in recovery was associated with an overall increase of .09 units of TRC (p<.000), meaning each year, an average person may expect to see a 1.08 unit increase in TRC. The corresponding per month and per annum average expected growth values for SC (.02, .24), PC (.02, .24), HC (.03, .36), and CC (.02, .24) are limited, but clearly trending positively across longer time frames (p<.000).

Table 2.

Age-, sex-, and race-controlled regression model coefficients, standard errors (SE), and p-values for total recovery capital and subscales using the MIRC

	Coefficient	SE	P-value
Total Recovery Capital	.090	.024	<.000
Social Capital	.022	.007	<.001
Physical Capital	.018	.009	<.035
Human Capital	.030	.008	<.000
Cultural Capital	.020	.007	<.006

Sensitivity analyses were performed to test for potentially problematic model dependency based on specification of the time in recovery variable as well as control for income and education (see Tables 3 and 4). Alternatively constructing the continuous variable as a maximal rather than minimal value of time in recovery during the continuous transformation of the original categorical variable returned artificially protracted point estimates for equivalent total growth, but made no impact on the essential interpretation of early- to middle-stage growth for TRC and any individual recovery capital domain. Likewise, introduction of income and education as control variables introduced negligible changes in the estimated per month growth for any recovery capital domain.

Table 3.

Sensitivity check regression models for alternate specification of time in recovery

	Coefficient	SE	P-value
Total Recovery Capital	.048	.014	<.001
Social Capital	.014	.004	<.000
Physical Capital	.011	.005	<.029
Human Capital	.015	.004	<.001
Cultural Capital	.009	.004	<.033

Table 4.

Sensitivity check regression models with added income and education controls

	Coefficient	SE	P-value
Total Recovery Capital	.089	.024	<.000
Social Capital	.022	.007	<.001
Physical Capital	.018	.008	<.035
Human Capital	.030	.008	<.000
Cultural Capital	.020	.007	<.006

Discussion

This study had two aims, to hypothesize overall recovery capital growth trajectories (Aim 1] and to make preliminary estimates of per time-unit growth for Total Recovery Capital (TRC) and each specific RC domain [i.e., Social Capital (SC), Physical Capital (PC), Human Capital (HC), and Cultural Capital (CC); (Aim 2)]. Several key points can be drawn from the findings, including, importantly, that this preliminary analysis indicates that recovery capital growth does happen. In bivariate and in controlled models, length of time in recovery was a statistically significant predictor of TRC scores and each subscale score.

Growth trajectory plots suggest a pattern of general increase over time. For TRC, SC, and PC, growth was observed in the short term, with each showing growth in the initial 12 months of recovery. It appears, however, that growth in HC and CC was delayed, showing a flat early trajectory, and then later growth over a longer period of one to five or more years. The reasons for this are undoubtedly contextual and case-dependent, but may be explained in part by people’s ability to establish supportive social networks and garner social support in early recovery (Stevens et al., 2015) as well as utilize social services to bolster access to housing, financial assistance, and other sources of physical capital (Connolly & Granfield, 2017; Weisner et al., 2003). In contrast, changes in attitudes, spirituality, health, coping skills, or other personal characteristics (HC) or in cultural milieu and surrounding environment (CC) appear to be more time-intensive efforts, delaying HC and CC development to the longer term.

Modeling results suggested that on average, we could predict a 1.08 unit per annum increase in TRC based on length of time in recovery alone. Based on MIRC baseline estimates from Bowen et al. (2023), at this rate of growth, we would expect the recovery capital of the average person to increase approximately one standard deviation over eight years – a substantial improvement. We do not have longer term data available to analyze the sustainability of growth, but if that average growth rate were to persist over a 10 or greater year period, people in recovery could move from approximately the 10^th percentile to the median or from the median to the 90^th percentile. SC, PC, HC, and CC models predict per annum growth in the range of .36-.48 units whereby on average, a person in recovery may experience a one standard deviation increase in these domains over approximately 10 years, on the assumption of consistent linear growth in that period.

With respect to practice implications for social workers and other recovery professionals, these results and considerations imply several key take aways. Firstly, recovery capital growth is achievable and people in recovery may be encouraged as such, showing empirical data to support optimism and motivate growth. Secondly, on average, we observed in the time frame of the study that recovery growth may accrue at approximately 1 point per year when assessing with the MIRC. This can amount to quite substantial growth over a long term recovery period, for example potentially moving from low recovery capital to moderate recovery capital (e.g., 10^th–50t^h percentile). It is important to note however, that some data from this study and from prior studies (e.g., Kelly et al. 2018) suggest that growth rates may be higher in the initial stages of recovery and then attenuate to lower levels of growth over time. This means that we may not be able to project growth linearly over long periods and that overall long term gains in RC could be somewhat less than linear projections. Summarily, there is at present, empirical reason from this and other studies, to expect that on average, overall RC growth occurs immediately and is sustained at some level, albeit perhaps more limitedly over time, and that in long term recovery situations, individuals may progress from low to moderate or greater RC.

Additionally, this study suggests that social workers and other recovery professionals should not necessarily be alarmed by delays they may observe in early-stage growth of the HC and CC domains, given that in preliminary data, these delays appear to be standard. However, they may choose to focus on these factors, depending on their perceived clinical relevance, especially given that barriers to growth in these domains appear present at the outset. Recovery professionals may also choose to focus on social and physical capital, given that early evidence suggests all recovery capital domains are of approximately equivalent importance (Bowen et al., 2023) and people in recovery may benefit from appreciable short term recovery successes.

Moving forward, the global need for well-trained professionals in the recovery field appears likely to remain (Gust et al., 2021). The results of this paper provide a preliminary estimate of recovery capital growth, useful for both treatment progress assessment and research purposes. The MIRC can thus be used to monitor progress over the course of treatment and can be built into treatment planning. Practitioners can use the MIRC in both initial assessment and in review or treatment plans and recovery evaluation, with preliminary baseline averages scores already established and growth guidelines estimated here to contextualize progress in recovery. The MIRC may also be used to identify treatment targets or to identify strengths to buttress. This information and the MIRC generally might also be employed by social workers and other recovery professionals outside of format treatment settings. Program administrators, peer mentors, case managers, and other professionals or paraprofessionals may benefit for utilizing the MIRC in non-clinical settings. Further, the substantial number of individuals in ‘natural recovery’ outside of any formal recovery setting at all, may engage in self-monitoring and evaluation (Gass et al., 2021) and buttress their efforts by using the MIRC guided by these data. In any application, we would strongly caution users about the possibility of harming one’s recovery by falling into negative cognitive feedback loops based on comparative self-judgement (Mackey, 2008; McGaffin et al., 2013).

A key limitation to this study stems from the cross-sectional and categorical nature of the data. It is possible that differences in TRC and any recovery capital domain are not attributable to growth, but rather to another phenomenon such as a selection effect whereby persons with greater baseline TRC at the onset of recovery last longer in recovery, giving the appearance that growth has occurred when in fact we only observe greater baseline recovery capital reflected in greater time in recovery. Likewise, the categorical nature of the data limits the number of timepoints available for analysis. Another caveat to consider is that while the sample included people who identified as having problems with a wide variety of substances, the inclusion criterion was based on alcohol, so these finding may or may not be applicable to people whose histories of problematic substance use do not include alcohol. Finally, it is always the case that a measure will not be fully inclusive or relevant parties or factors. The set of control variables employed in the analysis is not fully exhaustive of all relevant factors, nor is the specific intersection of those variables accounted for. Additionally, the sample used to construct the measure, while greatly expanded relative to previous attempts, comprised only English-speaking, Americans.

Despite this, these preliminary estimates are useful as a working hypothesis. Future research should include longitudinal studies to rule out the potential for fallacious inferences about growth and more closely identify patterns and factors in early- and middle-stage recovery capital growth. Further research on how recovery capital is related to recovery outcomes, such as moderation and abstinence as well as quality of life, would also be beneficial. Finally, researchers should investigate recovery capital growth in individuals and among specific demographic groups to detect factors associated with non-standard growth, as well as further explore the application of recovery capital in clinical settings.

Conclusion

In this study, we endeavored to identify preliminary evidence that recovery capital growth occurs, and if so, at what rate. We find evidence that TRC growth may be appreciable within one year and could be substantial across several years or more. We also find that growth trajectories (linear or curvilinear) for TRC, SC, and PC appear to show quick early-stage growth. Contrastingly, HC and CC may not show much growth until a year or more of time in recovery is accrued. Social workers and other recovery professionals may wish to encourage people in recovery of growth potential, monitor progress through use of the MIRC in treatment planning, and normalize perceptions of the pace of growth in recovery or build specific recovery targets based on these observations. Finally, ongoing research should assess individual growth trends longitudinally, such that the ecological fallacy can be disconfirmed, and more nuanced analyses of specific racial, ethnic, and other demographic groups should be undertaken.

Highlights:

• Recovery capital increases over time.

• Total recovery capital, social capital, and physical capital begin to grow immediately.

• Human capital and cultural capital grow after approximately the first year.

• Recovery capital may grow by one standard deviation or more over time.

Data Availability:

Data used in this study are now in the public domain, accessible through the NIH/NIAAA data archive [https://nda.nih.gov/niaaa/].