Coercion In Mobile Crisis: A Latent Class Analysis of Six Interventions

Background & Purpose: Police involvement in behavioral health crises (BHCs) can result in harmful, even fatal outcomes, for people in crisis. In response, the 988 Suicide & Crisis Hotline was launched in 2020 to redirect crisis calls from police to mental health professionals, including mobile crisis (MC) teams. While MC teams are framed as a safer alternative to police, they may still deploy involuntary interventions that result in iatrogenic harm. Research on coercion in mental healthcare has predominantly focused on voluntary and involuntary interventions; however, this framing imposes a binary that does not capture the complexity of intervention decision-making. MC workers’ intervention decisions are poorly understood, and their potential for reproducing harm via purportedly alternative interventions remains.

Methods: We conducted latent class analysis on intervention endorsement scales from an experimental 2x2 factorial vignette survey examining the impact of race and prior criminal-legal involvement on MC workers' use of involuntary hospitalization and police involvement in a national sample of MC workers (n = 369). Participants were randomly assigned to one of four crisis vignettes, differing only service user race and prior criminal-legal involvement. To assess decision-making stability across agitation levels, participants rated the appropriateness, likelihood, and supervisor's recommendation of six interventions across low/moderate- and moderate/high-agitation scenarios. Intervention endorsement was binarized (low vs. high) for referral to community-based services (CBS), voluntary crisis stabilization (CS), voluntary CS with involuntary hospitalization if declined (VHIH), involuntary hospitalization (IH), specially trained law enforcement (SLE), and any law enforcement (LE).

Results: A 4-class solution was selected as the most theoretically meaningful solution despite yielding the second-best model-fit indices, as the addition of a 5^th class was not theoretically meaningful (log likelihood = -9715.44; BIC = 19,790.43; AIC = 19,520.87; entropy = 0.84). Voluntary Intervention Users (Class 1, 32.49%) highly endorsed voluntary interventions while avoiding any involuntary interventions. Law Enforcement-Averse Intervention Users (Class 2, 17.92%) highly endorsed CS, VHIH and IH but avoided law enforcement involvement. Contingent Involuntary Intervention Users (Class 3, 32.45%) highly endorsed all involuntary interventions while avoiding voluntary interventions except voluntary CS. Direct Involuntary Intervention Users (Class 4, 17.14%) highly endorsed involuntary hospitalization and involving law enforcement but, unlike Class 3, did not highly endorse provider persuasion to crisis stabilization services.

Conclusions and Implications: Resulting classes indicate theoretically meaningful differences in MC providers' intervention endorsement patterns, confirming that MC providers exercise considerable discretion when responding to BHCs. Findings have direct implications for MC service delivery as agency approaches, team composition, and provider attitudes towards coercion may influence the range of MC interventions deployed with service users. MC training and supervision efforts may benefit from addressing provider attitudes toward involuntary hospitalization and law enforcement-involved responses to decrease unnecessary use of involuntary interventions that can result in iatrogenic harm to service users. Involuntary endorsement patterns in Classes 2 and 3 suggest increased investment in high-quality, community-based crisis stabilization services may reduce unnecessary involuntary hospitalization and law enforcement involvement. Future research should replicate these findings and extend them to other mental health service settings beyond mobile crisis.