Background: Lay counsellor-delivered psychosocial interventions are increasingly used to address workforce shortages in mental health care. While randomised trials commonly report mean improvements, explicit assessment of clinical deterioration is rare. This secondary analysis evaluated whether a lay counsellor intervention for early psychosis was associated with evidence of deterioration compared with usual care. Methods: Patient-level data from a randomised controlled trial in Thailand (n = 255) were analysed. Deterioration was defined as worsening between baseline and the 6-month follow-up across functional, behavioural, and service utilisation domains. Risk differences were estimated using Newcombe confidence intervals, and risk ratios were calculated using standard methods with the Haldane–Anscombe correction applied, where required. Analyses were conducted for the full sample (UC n = 125; LICM n = 130) and stratified by baseline severity (none/borderline: UC n = 103, LICM n = 103; mild-to-severe: UC n = 22, LICM n = 27). Results: In the full sample, deterioration rates were similar across most domains. A statistically significant reduction in deterioration related to disturbing or aggressive behaviour was observed in the LICM arm (risk difference −14.1%; 95% CI −26.8% to −0.6%; risk ratio 0.45; 95% CI 0.26 to 0.79). No statistically significant excess deterioration was observed in other domains. In severity-stratified analyses, no subgroup showed a statistically significant increase in deterioration attributable to the intervention. However, among participants with mild-to-severe baseline illness, although no statistically significant harm signal was detected, the adverse risk differences and risk ratios observed in socially useful activities, self-care, regular outpatient follow-up visits, and medication adherence among participants with greater baseline severity underscore the importance of careful monitoring in higher-risk subgroups. Conclusions: No statistically significant evidence of excess deterioration was observed in either the full sample or subgroup analyses between the intervention and control arms. However, the adverse absolute difference observed in multiple patient-related domains, among participants with mild-to-severe baseline illness, suggests that lay-counsellor interventions may require a stepped-care approach to safely address the mental health needs of patients with higher levels of severity. Psychosocial trials should routinely report deterioration and subgroup analyses alongside mean improvements.
Nachiket Mor (Thu,) studied this question.