Freedom from Within: A Meta-Analytic Review of Mindfulness Meditation-Based Interventions in Prisons

meditation-based interventions


Introduction
Research has frequently indicated the counterproductive nature of incarceration to serve as a rehabilitation space with high recidivism rates highlighting insufficiencies in correctional institutions and the difficulties prisoners face circumventing re-incarceration (Day, 2020;Pratt, 2019).Prison imposes severe deprivations on incarcerated individuals who often already possess psychological disturbances.Two-thirds of those incarcerated report the use of illicit drugs and/or alcohol directly prior to their incarceration (De Andrade et al., 2018;Lyons & Cantrell 2015).Correspondingly, recidivism rates for parolees combined with relapse rates for substance use in incarcerated people are greater than 50%, with illegal drug-seeking behaviour and participation in unlawful activities leading to repeated involvement in the criminal justice system (Hamilton & Belenko, 2019).Amidst the prison environment, correctional policies assume rehabilitation can occur through educational programs, behavioral interventions and socialization improvements.Granados et al. (2023) argue that programs that exclude psychological development, physical wellness, mental balance, and empathy call into question the efficacy of such initiatives, whereby the separation of these qualities falls short of deeply rehabilitating the offender.
Over the past several decades, the increasing popularity of mindfulness has been incorporated into Western practices as a fundamental technique for introspection and a noted therapeutic tool (Spina, 2023).This increased awareness of mindfulness has contributed towards a progression in attitudes regarding institutional reform and the rehabilitation of incarcerated people (Hunnicutt & Rhodes, 2015).Differential procedures have been used in prisons, commonly in the form of transcendental meditation, mindfulness-based interventions (mindfulness-based stress reduction, mindfulness-based cognitive therapy) and Vipassana meditation.Although these techniques vary in delivery, all are aimed towards the individual acquiring life-long inner resources in self-awareness, self-regulation and healthy coping strategies (Davies et al. 2021;Spina, 2023).For the current study, the name Mindfulness Meditation-Based Interventions (MMBIs) was composed to encompass the variety of structured prison meditation programs aimed at generating awareness of moment-to-moment understanding encountered by the mind to reduce reactivity to acute experiences, whether positive, negative, or neutral, in an unfiltered manner, thus increasing well-being (West, 2016).
Although there is a growing body of meta-analytic investigations on meditation practices within clinical populations, previous reviews within prisons are limited.These have usually examined a broad range of undefined dependent outcomes (such as psychological well-being and behavioral functioning) rather than specific components of criminogenic behavior (such as substance use and recidivism) (see Auty et al., 2017;Derlic, 2020).This paper addresses the gap in the literature by conducting a meta-analytic review of MMBI impact of behavioralspecific evidence of substance use and recidivism.

The Ideology of Mindfulness Meditation and Application for Incarcerated Offenders
Mindfulness meditation diverges from other rehabilitative initiatives as an individual learns to understand their mental life by observing the interaction of the body and the mind (Hunnicutt & Rhodes, 2015).The aim is to develop non-judgmental relationships to thoughts as they occur through an objective, detached form of self-observation without reaction (West, 2016).Accordingly, Roos et al. (2020) posit that meditation-based programs may promote awareness of internal and external contextual circumstances and a greater nonreactivity toward negative thoughts stemming from distressing events.Further, Garland et al. (2009) contend that mindfulness reinforces meaning-focused coping by facilitating positive reappraisals.Known as decentering, increases in mindfulness shift cognitive sets, enabling the realization that emotions are separate from facts, resulting in alternative appraisals (Grecucci et al., 2015).Sears and Kraus (2019) found that those who practiced meditation had significant anxiety and negative affect reductions, exhibiting a meaningful driving mechanism in reducing cognitive distortions.Similarly, Jones et al. (2019) further examined the effects of mindfulness meditation on coping flexibility, which continued weeks post-intervention.This 'mindful reappraisal' increased attentional flexibility and broadened awareness, increasing psychological and behavioral functioning that self-perpetuates in an upward spiral over time, thus reducing maladaptive coping strategies (Garland et al., 2009).
Respectively, the evolving role of MMBIs in prison settings has demonstrated promise for incarcerated populations to transform negative thought patterns and decrease the likelihood of impulsive criminal habits such as substance use and cycles of recidivism.Findings from Davies et al. (2021) demonstrate the mechanism of MMBIs as a tool that reinforces the ability of individuals to engage in healthy connections, releasing the built-up self-protective stance constructed in prisons, with Bouw et al. (2019)

Theoretical Framework
The underpinnings this research was approached through the positivist paradigm.In accordance with Khanna (2019), the essence of positivism is one of a scientific method, combining mainly deductive logic and employing predominantly quantitative methods to seek an explanation and prediction of the phenomena observed.Within such a paradigm, knowledge is obtained from a factual and impartial source through an objective lens (Hasan, 2016).The researchers believe that it is appropriate to examine the effectiveness of prison MMBIs using a positivist approach as data will be data gathered and aggregated from pre-existing statistical explorations of prison MMBI research (Quick & Hall, 2015).Hence, observable, quantifiable numerical data will be examined to summarize information regarding the extent this intervention impacts behaviour.

Systematic Review & Meta-Analysis
This research is based on a quantitative methodology, employing a systematic review and metaanalysis to produce an unbiased summary, drawing a robust conclusion of the aggregate evidence regarding prison MMBIs and their effect on behavioral changes within incarcerated individuals (Gough et al., 2017).Additionally, a meta-analysis was performed to observe the strength of the relationship between the two variables, providing a more accurate interpretation of the intervention's effectiveness (Siddaway et al., 2019).The objective was to calculate the standardized effect on various prison MMBIs and whether they impact incarcerated individuals with regard to the specified behavioral covariates, substance use and recidivism.These behavioral changes have been analyzed through official records pertaining to substance and alcohol use, prison rule infractions, re-arrests, re-incarceration, parole outcomes and selfreports.

Defining Inclusion & Exclusion Criteria
Studies were eligible for inclusion if they satisfied the following criteria (4) the study only looked at psychological changes; or (5) anecdotal or preliminary reports.

Search Strategy
The Searches were not limited by study design or sample size.Other potential sources beyond the databases included citations and references from significant peer-reviewed articles from existing research.

Risk of Bias within Individual Studies
The risk of bias within individual studies was assessed using the Cochrane Collaboration tool (Higgins et al., 2011).This consisted of items pertaining to research design, method group allocation, concealment, details of the intervention, MMBI facilitators, explanations of attrition, and an overall summary assessment of the risk of bias within the individual studies.

Approach to Data Synthesis and Analysis
A random-effects model analysis was employed to account for between-study variance, allowing for real differences in the treatment effect (Nagashima et al., 2019).The meta-analysis of the research contained three mandatory phases: (1) measures of effect size with its 95% confidence intervals and summary of the individual studies; (2) measures of heterogeneity; and; (3) integration of the studies statistically (Gough et al., 2017).The analysis aimed to identify trends across the studies and assess whether the strength of the MMBI effect exists in a positive or negative direction.All means, standard deviations and percentages found within the selected studies were then calculated using the statistical programming meta package R to determine effect sizes, confidence intervals, standardized mean differences, standard errors and odds ratios.
Data obtained from the selected studies determined the meta-analysis model.Standardized mean differences were used to build forest plots of continuous data to evaluate differences in MMBIs and low or high frequency of substance use.Odds ratios were used to build forest plots of dichotomous data to assess differences in MMBIs and recidivism.Forest plots visually exhibited the weight of each MMBI study and their confidence intervals with its pooled effect, the overall observed effect of the comparative studies, heterogeneity, and statistical significance (Gough et al., 2017).Furthermore, Borenstein's et al. (2009) I2 statistic was used to test heterogeneity as it measured the proportion of total variability in the pooled effect size influenced by heterogeneity rather than variations due to sampling error.
The risk of bias across studies and potential publication bias was graphically represented using Egger's et al. (1997) funnel plot.Plots presented the variability of the individual studies (standard error) against the mean effect sizes and odds ratios.Finally, Rosenthal's (1979) failsafe N, Begg and Mazumdar's (1994) rank correlation test and Egger's et al. (1997) regression intercept test were used to test funnel plot asymmetry and to quantify publication bias.

Study Selection
The research objective was to identify MMBIs among incarcerated populations with reference to substance use and recidivism.Screening began with identifying inclusive and exclusive studies that were relevant to the title and abstract.Total search results yielded 92 citations.Following duplicate citation removal and abstract screening, 40 articles remained.Full-text articles were retrieved, and the remaining articles were located within reference lists, with 16 being excluded, having not met the inclusion criteria.Of the 32 articles remained, 15 were excluded for reasons summarized in Figure 1.Conclusively, the remaining 17 full-text articles were reviewed and confirmed eligible for inclusion, having met all the outlines criteria of MMBI's as the independent variable.Of these 17 studies, 8 corresponded with substance use and 10 studies with recidivism, with one of these studies (Malouf et al., 2017) examining both covariates.One study was based in Taiwan, one in India and the remaining in the United States.
All studies were written in English and published between 1973 and 2018.

Classification of Dependent Measures
The classification of the dependent measures was determined by the data found within the 17 full-text articles.Each dependent measure examined throughout the selected studies was summarized into 8 different categories to be coded under the two covariates -substance use and recidivism.Substance use contained the variable codes; alcohol use, illicit drug use, locus of control/cravings, and expectancy of use.Recidivism consisted of the variable codes; recidivism/re-incarceration, re-offending/re-arrest, parole revocations/outcomes, and prison rule infractions/criminal propensity.Only the dependent measures in the aforementioned inclusion criteria were examined.Dependent measures were classified under a positivenegative dimension to indicate whether changes led to a reduction or increase in the dependent measures.

Data Extraction and Summary of Study
Table 1 displays the extracted data summarizing the key study characteristics included in this review.To avoid data extraction inconsistencies or the possibility of misinterpreting data leading to error or bias, the co-authors completed data extraction separately to ensure interrater reliability (Higgins et al., 2011).Minimum, medium and maximum correctional facilities can be observed in the review, demonstrating a diverse range of incarcerated populations.
Males were the predominant sex, with a mean age of 33.96.The mean sample population consisted of n=66.22 for MMBIs and n=127.67 for active controls.

Risk of Bias within Individual Studies
Table 2 summarizes the quality measures for each selected study.Only one study employed a blind allocation design (Alexander, 1982), three were RCTs (Lee et al., 2011;Lyons et al., 2018;Malouf et al., 2017), nine were longitudinal trials, and the remaining were quasiexperimental.The facilitator's MMBI training style was not specified by one study (Lyons et al., 2018).Two studies contained a high risk of bias as they contained a relatively small sample size and lacked a long-term follow-up (Khurana & Dhar, 2000;Lee et al., 2011).Khurana and Dhar (2000) was not derived from an independent source and proved statistically insignificant.The risk of bias within individual studies was presented through high attrition levels.In many studies, the attrition rates ranged from 12% to 47%, with two studies (Alexander et al., 2003;Murphy, 2002) failing to report attrition rates.Only one study (Bleick, 1983) reported no attrition and had extensive follow-up periods, while two studies (Khurana & Dhar, 2000;Lee et al., 2011) revealed no attrition as they only measured pre-and post-scores with no extensive follow-up periods.Half of the selected studies contained high attrition rates.

Synthesis of Substance Use and Recidivism Results
The primary outcome measures were defined as the standardized mean score difference in substance use and the odds ratio in recidivism, the MMBI group, following the intervention.A secondary outcome measure was calculated, reflecting the difference in mean gain score in substance use and percentage gain score in recidivism between MMBIs and active controls.This was performed to provide confidence towards the validity of the pre-and post-intervention analysis.Additionally, pooled standard deviations were weighted by sample size to adjust for biases in small sample size studies.The random-effects model was employed to allow for statistical inference from the variability in the data, offering a better understanding and characterization of the results (Nagashima et al., 2019).
Figure 2 shows the effects of MMBIs on substance use.Alcohol use, illicit drug use, locus of control, cravings, and expectancy of use were the most commonly used measures across the selected studies and were coded and collated to form the covariate substance use.It should be noted one study only recorded excessive alcohol use (Bowen et al., 2007).All other studies recorded a variety of illicit substances to measure the frequency of use between MMBIs and active controls (Bowen et al., 2006;Marlatt et al., 2004;Simpson et al., 2007;Witkiwitz et al., 2005), and two studies examined substance use with its relationship to cravings and expectancy (Lee et al., 2011;Lyons et al., 2018).The data was averaged, as opposed to summed, to measure substance use within the incarcerated individuals across the studies for both MMBIs and active controls.For each of the eight papers, means and standard deviations was collected, along with sample sizes for each group.The Forest plot was then constructed based on the average number of substances used for the MMBI group and compared to the active controls.Since the 95% confidence interval does not contain >0, it can be concluded that there is a statistically significant effect of MMBI on substance use.Since 0 lies to the left of the confidence interval, the true difference is negative, meaning that MMBIs used fewer substances than the active controls.The standardized differences in the forest plot show MMBI were all negative, demonstrating MMBI reduced substance use across all studies.Two studies (Bowen et al., 2007;Malouf et al., 2017) revealed that their confidence intervals crossed the line of no effect, demonstrating that these studies were not statistically significant.The pooled results of the meta-analysis were statistically significant, revealing that the overall standardized mean difference was -0.94 with a 95% confidence interval [-1.98 to -0.09], suggesting a small to moderate reduction of substance use.I2 demonstrated definitive confidence intervals across all heterogeneity assessment measures.Heterogeneity for the effect of MMBIs on substance use was I2 = 44%, p <0.03, meaning that the null hypothesis of homogeneity of studies was rejected.This value signifies slight to moderate heterogeneity, suggesting a degree of similarity in the results of the individual studies, and thus statistical tests could be performed.The overall effect p value test is below 0.05 (p <0.05), suggesting a statistically significant positive correlation between MMBIs and decreased substance use.
Figure 3 shows the effects of MMBIs on recidivism.Data was collected from ten papers.Recidivism, re-incarceration, re-offending, re-arrests, parole revocations and outcomes, prison rule infractions, and criminal propensity were collated and used to measure overall recidivism across all sourced studies.One study measured criminal propensity (Khurana & Dahr, 2000) and two studies recorded rule infractions (Ballou, 1973;Perelman et al., 2012).All other studies examined recidivism, with two investigating parole revocations and outcomes (Bleick, 1983;Bleick & Abrams, 1987).Percentage data was obtained from each of the ten papers.The odds ratio was obtained by calculating the probability of recidivists divided by non-recidivists by the sample size (Cooper, 2016).Odds ratios of MMBI groups and active controls were then simply defined and summed, giving a ratio measure from the percentages of recidivism for both MMBI and active control groups.The forest plot was constructed based on the odds ratio of recidivism for MMBIs versus active controls.Since the 95% confidence interval does contain 1, it can be concluded that there was a statistically significant effect of MMBIs on recidivism compared to active controls.All but one study (Perelman et al., 2012) in the forest plot was greater than 1.Hence, the meta-analysis shows that MMBIs do reduce recidivism.All but two studies (Khurana & Dhar, 2000;Perelman et al., 2012) revealed that their confidence interval crossed the line of no effect.Furthermore, the effect sizes of these studies were below 1 or slightly past 1, demonstrating that these studies were statistically insignificant.The overall pooled results of the meta-analysis were statistically significant, revealing that the overall odds ratio was 1.34 with a 95% confidence interval [1.17 to 1.49], suggesting a moderate to substantial reduction in recidivism.Additionally, the heterogeneity test for the effect of MMBIs on recidivism was I2 = 58%, p <0.02, meaning that the null hypothesis of homogeneity of studies was rejected.Overall, this value signifies moderate to substantial heterogeneity.Again, the results of the individual studies are statistically sufficiently similar and do not need to be combined.However, given the breadth of the research question and the pooled effect sizes, caution was had when interpreting high heterogeneity.Lastly, the test of the overall effect p value is below 0.05 (p <0.01), indicating a statistically significant correlation between MMBIs and a moderate to substantial effect in reducing recidivism.

Risk of Bias across Studies
Egger 's et al. (1997) funnel plot was used to present potential publication bias graphically.Publication bias was assessed based on the pre-and post-intervention of the included studies and their effect size calculations.The standard errors of effect sizes were plotted against the standardized mean difference for substance use (Figure 4) and odds ratio for recidivism (Figure 5).Upon visual inspection, both funnel plots resembled an inverted funnel, suggesting negative findings seem to have been published and included in the review.Taken together, results from both funnel plots resemble an inverted funnel.Even so, some possible bias was detected.Rosenthal's (1979) fail-safe N was 108, meaning there would need to be 98 studies with effect sizes of zero to make the p value insignificant and exceed .05.In the recidivism meta-analysis of 10 studies (z value = 3.68; p <0.01, two-tailed), the fail-safe N was 124, meaning there would need to be 124 studies with effect sizes of zero to make the p value insignificant and exceed .01.

Discussion
The evidence from the systematic review and meta-analysis in this research suggested that MMBIs in prison settings had positive benefits.The pooled effect sizes from variables were meaningful, demonstrating both tests of overall effect revealed that MMBIs were statistically significant in reducing substance use (p <0.05) and recidivism (p <0.01).MMBIs statistically reduced the use of several substances compared to the active controls.Six of eight studies (Bowen et al., 2006;Lee et al., 2011;Lyons et al., 2018;Marlatt et al., 2004;Simpson et al., 2007;Witkiwitz et al., 2005) reported statistically significant reductions in substance use.The two remaining studies (Bowen et al., 2007;Malouf et al., 2017) showed no statistically significant difference in the treatment condition.It is unclear to what extent these null findings are related to the effectiveness of MMBIs when considering the methodological shortcomings due to the small sample sizes and high attrition rates observed in these two studies.Nonetheless, although all studies varied in their design quality (retrospective surveys to random assignment experiments), MMBIs demonstrated a small to moderate reduction in substance use for those who participated in MMBIs, contributing towards decreased use for substance-involved offenders.Therefore, the results of this study must be interpreted conservatively as these findings only examined short-term effects of MMBIs, with the most extended follow-up period being 6 months.Due to a paucity of extensive follow-up examinations found in prison MMBIs, there was not enough evidence to suggest that long-term changes in this review are unknown.
Regardless, incarcerated individuals in the current MMBI sample showed differential changes towards reductions in substance use compared to active controls, indicating better abstinence and a reduction in cravings.In line with the literature (see Grant et al., 2017;Li et al., 2017), the overall observed reduction in substance use within the MMBI sample in the context of behavioral changes and substance-related management was identified.A similar pattern was identified within emergent research (see Rosenthal et al., 2021;Tang et al., 2016) on the causes and maintenance of addiction, wherein the core processes that led to habitual substance misuse included increased avoidance-oriented coping, impaired self-control, cravings and emotional dysregulation.Corresponding findings from Garland and Howard (2018) contend MMBIs facilitate and reinforce cognitive control of drug cue-reactivity behaviors.However, it was beyond the scope of this current study to evaluate the direct cognitive effects that may have mediated the effects of MMBIs for substance use without an in-depth qualitative investigation.
Furthermore, recidivism data showed statistical reductions across the dependent measurements.Eight of ten studies (Alexander, 1982;Alexander et al., 2003;Bleick, 1983;Bleick & Abrams, 1987;Malouf et al., 2017;Murphy, 2002;Rainforth et al., 2003) demonstrated a reduction in recidivism, while two studies (Khurana & Dhar, 2000;Perelman et al., 2012)  Nonetheless, corresponding research comparing incarcerated MMBI groups and active controls recognized an increase in psychological well-being, such as optimism (Garland et al., 2009), subjective mood states (Samuelson et al., 2007) and self-compassion (Morley, 2017).Reductions of recidivism in the current findings resonated with existing literature whereby MMBIs were reported to inhibit criminal tendencies through the development of healthier psychological functioning, thus contributing to and reinforcing rehabilitative environments in the prison setting (see Bouw et al., 2019;Ronel et al., 2011).Similar reports found a decrease in negative psychological states, such as obsessive-compulsive behavior depression, anxiety anger, hostility and trauma symptomology (Spina, 2023).However, the current study cannot confirm this as further subgroup analysis examining psychological-specific variables was not conducted.

Limitations
Consequently, the findings were interpreted conservatively as several studies revealed shortcomings regarding methodological rigor, limited matched controls, and confidence intervals containing wider variations, thus limiting the external validity (Borenstein et al., 2009).Concerns regarding lack of randomization, small sample sizes, high risk of bias and limited statistical power were evident.These studies revealed smaller effect sizes with larger confidence intervals.Due to the methodological diversity of the selected studies, slight to substantial levels of heterogeneity was unavoidable, given the variations in MMBIs and study duration.Although studies did not equally use the same measurement methods, according to Higgins et al. (2011), this does not necessarily affect the intervention's overall outcome.While the findings support the effects of MMBIs, the researcher notes that publication bias was still possible.Biasing factors may be attributed to less consistent studies publishing secondary outcomes due to a weakness in methodological rigor or selective outcome reporting (Cooper, 2016).
Moreover, attrition, especially in studies of longitudinal design, was an inherent limitation.
When using an incarcerated sample, reductions in program completion and follow-up assessments are common.The average attrition rate was 26%, consistent with a meta-analysis conducted by Olver et al. (2011) of overall prison intervention attrition (27.1%).Upon further investigation, attrition was commonly due to institutional logistics, prisoner transfer, or release upon intervention completion.Though not all were negative circumstances, the statistical power of these studies was affected.Furthermore, an alternative explanation for the current findings could be attributed to self-reported responses bias.Substance use results may have had this possibility.However, regarding recidivism results, this data was compared with the official department of justice records, lending more confidence in the validity of these findings.

Conclusions and Broader Implications
Although the findings demonstrated MMBIs were sufficient to produce reductions in substance use and recidivism, this research does not propose MMBIs are superior to other prison rehabilitation initiatives, nor should they replace current interventions.Instead, MMBIs should be offered in conjunction with other established treatment modalities, especially for those who suffer from chronic comorbidities (Simpson et al., 2007).Safeguarding well-designed MMBI within prison walls is essential to ensure program effectiveness and safety to avoid disturbances in such highly stressed environments (Davies et al., 2021).As such, programs that successfully address empathy, conscience and self-respect may be beneficial in transforming the prison atmosphere.MMBIs may serve as a synergistic extension to existing treatments, whereas in severe cases, it might be suitable as a multi-modal approach, one that combines MMBIs, psychotherapy, and medication (Granados et al. 2023).Adjunct MMBI approaches may eliminate institutional limitations related to prison resources, staff, time and space.This may increase engagement in criminogenic need-specific programs relating to addiction and postrelease recidivism.

Figure 4 .
Figure 4. Funnel Plot Standard Error by Standardized Mean Difference (Substance Use) reporting an increase in a willingness to participate in prison-offered treatment programs and cooperation with prison authorities.Accordingly, MMBIs reflect a transition from the 'problem-and-treatment only' prison paradigm to a holistic model that promotes prison safety, mental health, and positive re-entry outcomes.Therefore, the central aim of the present study is to understand to what extent MMBIs impact incarcerated populations systematically.Based on the appraised literature and the knowledge gap, the following study represents the first meta-analytic review of prison MMBIs focusing on defined covariates -substance use and recidivism -and their specific contributions, quantifying the effect size of MMBIs.Specifically, the research examines whether prison programs that utilize MMBIs improve behavior compared to pre-test measures and active controls of incarcerated populations.Based on these aims, the following research has been guided by the question: To what extent do mindfulness meditation-based interventions influence behavioral changes in incarcerated populations regarding substance use and recidivism?

Table 1 .
Characteristics of Studies Reporting the Effects of MMBIs on Behavioral Changes -Substance Use and Recidivism

Table 2 .
Quality Measures of Studies Reporting the Effects of MMBI on Behavioural Changes Andrade et al. (2018)significant.Upon further investigation, null findings may be due to an absence of continued practice.Though there were discrepancies in the delivery of the MMBIs, meaningful comparisons between the interventions across studies were made, indicating a moderate to substantial reduction for MMBI participants.The current study revealed MMBIs exhibited a significant change towards reductions in recidivism and mirrored the recidivist statistical trends, demonstrating a total mean average recidivism rate of 60% for active controls compared to 44% for MMBI participants.A meta-analysis conducted by Lipsey (2019) examined the impact of multiple rehabilitation interventions for adult offenders and found the most effective programs represented an approximate 20% reduction.The findings from this current study indicated that MMBIs produced a similar reduction of 10% to 20%.Consistent withCullen (2011), this meta-analytic review provided future validation of previous reports of MMBI participants demonstrating positive changes in risk factors associated with recidivism.However, DeAndrade et al. (2018)argue that MMBIs may have attracted incarcerated residents who possessed motivations for rehabilitation prior to the program, thus predictively reducing recidivism.