Vilsaint · Frontiers in public health 2025 · Systematic review of RCTs and quasi-experimental designs · n=5 studies (11 reports; sample sizes 150-470)

Recovery housing for substance use disorder: a systematic review.

Cited 20 times in the scientific literature.

Level 2 - randomized trial

Systematic review including both randomized trials (n=3) and quasi-experimental designs (n=2)

PubMed 40115346 · doi:10.3389/fpubh.2025.1506412 · record verified 2026-08-26

What was done

A systematic review searched PubMed, EMBASE, CINAHL, PsycINFO, and CENTRAL through February 2024 for studies comparing recovery housing alone to non-recovery housing conditions for substance use disorder. Eligible designs included randomized controlled trials (RCTs) and quasi-experimental designs (QEDs). The review identified 5 eligible studies (3 RCTs, 2 QEDs) across 11 reports, with sample sizes ranging from 150 to 470 participants and follow-up durations from 6 to 24 months.

What was found

The abstract reports directional findings without quantitative effect estimates. Recovery housing performed better than continuing care as usual or no intervention on abstinence, income, employment, criminal charges, and to a lesser extent incarceration. Compared to other residential interventions (such as therapeutic communities), recovery housing showed higher alcohol abstinence, fewer substance use days, increased income, higher employment rates, and higher cost-effectiveness. However, in samples comprising 90% or more formerly incarcerated women, benefits were limited: reduced substance use was the only benefit relative to other residential interventions, and outcomes were inconsistent compared to continuing care as usual.

Why it matters

This review synthesizes the controlled trial and quasi-experimental evidence supporting recovery housing as an effective, cost-saving intervention for substance use disorder, while highlighting that standard models may fail to produce consistent benefits for formerly incarcerated women.

Limits

The total evidence base is small (5 studies). The abstract reports no numerical effect sizes, odds ratios, or confidence intervals. Two of the five included studies were quasi-experimental, introducing potential selection bias.

Cited by