What Consumers Say about Recovery-Oriented Services: Topic Modeling Analysis of Google Map Reviews of U.S. Mental Health Facilities

Background and Purpose: Recovery-oriented care places the voices and lived experiences of mental health service users at its center. Yet when facilities are evaluated for how well they embody recovery orientation, those voices are largely absent. Existing consumer experience measurement tools, such as the Consumer Assessment of Healthcare Providers and Systems (CAHPS) survey, constrain what consumers can express and reflect institutional priorities as much as consumer ones. More fundamentally, the facility types most central to the recovery-oriented service system, including community mental health centers, residential programs, and outpatient specialty providers, lack representation of consumer perspectives on service access and utilization. This study asks whether naturally occurring Google Maps reviews, written voluntarily by consumers themselves, can serve as an ecologically valid source of consumer voice for assessing recovery-oriented care across U.S. adult mental health facilities, and how consumer experience themes vary by facility type and recovery-oriented service characteristics.

Methods: Consumer reviews were collected in March 2026 via the Google Maps Places API for 2,115 adult mental health facilities from the 2024 National Substance Use and Mental Health Services Survey (N-SUMHSS), achieving a 95.3% facility match rate. Structural Topic Modeling was applied to a corpus of 1,614 facility-level documents derived from 7,250 individual reviews after iterative exclusion of non-mental health content. Facility type (inpatient, residential, community, outpatient) and three binary recovery-oriented service provision indicators (Illness Management and Recovery [IMR], peer support, and consumer-run services) were examined as correlates of topic prevalence. The final number of topics was selected based on semantic coherence, exclusivity, held-out likelihood, and lower bound.

Results: Eight substantively interpretable topics emerged: two positive (Positive Staff and Supportive Care; Recovery Journey and Therapeutic Alliance) and six negative (Appointment and Medication Access Barriers; Community Mental Health and Resource Navigation; Institutional Abuse and Systemic Failure; Residential Group Home Environment; Substance Use Treatment and Child Welfare System Involvement; Inpatient Psychiatric Conditions and Safety). Appointment and Medication Access Barriers was the most prevalent topic (mean proportion = 0.239) and was most strongly associated with lower ratings. IMR was associated with lower prevalence of Residential Group Home Environment concerns, but counterintuitively with higher prevalence of both Institutional Abuse and Inpatient Psychiatric Conditions narratives. Consumer-directed services were associated with lower prevalence of Institutional Abuse narratives. Peer support showed no significant independent association with topic prevalence.

Conclusions and Implications: This study demonstrates that consumer-generated narratives constitute an ecologically valid source of consumer voice for mental health facilities systematically excluded from existing measurement infrastructure. The prominence of appointment and medication access barriers suggests that structural access issues represent the most salient dimension of consumer experience across mental health settings, warranting system-level quality improvement attention. The finding that consumer-directed services, but not IMR, were associated with lower institutional abuse narratives suggests that placing consumers in operational roles may be more consequential for organizational accountability than clinician-delivered programming alone. Findings should be interpreted in light of limitations including self-selection bias, restriction to the five most recent reviews per facility, and service indicators reflecting availability rather than implementation fidelity.