Social support, behavioural activation and problem management carried the benefit of task-shared therapy; relaxation detracted

Design

Bayesian individual participant data component network meta-analysis of randomised trials

Population

10,612 adults with depression or anxiety in 30 trials of non-specialist-delivered interventions

Primary outcome

Symptom reduction at study endpoint

Effect

Social support −9.48 (−13.29 to −6.60); problem management −4.08 (−5.37 to −2.83); relaxation +7.97 (3.35 to 12.11)

This Bayesian component network meta-analysis used individual participant data from 30 randomised trials and 10,612 adults with depression or anxiety, all testing psychosocial interventions delivered by non-specialists. Each intervention was broken down into its components to estimate what each added.

Three components were associated with the most benefit: strengthening social support (incremental mean difference −9.48, 95% credible interval −13.29 to −6.60), behavioural activation (−4.15, −7.48 to −0.05) and problem management (−4.08, −5.37 to −2.83). Two were associated with worse outcomes when added: relaxation (7.97, 3.35 to 12.11) and, less clearly, cognitive reframing (5.17, 0.78 to 9.15). The effect of components varied with baseline severity and sociodemographic factors.

This is the most direct evidence yet on what lay counsellors should spend their limited sessions on. Many brief programmes still include relaxation exercises by default. India's task-shared programmes, where behavioural activation and problem-solving are already central, sit well with these findings; a programme padded with relaxation may be diluting itself. The evidence is associational within trials, so it points to redesign and testing rather than to removing components overnight.

• Build brief lay-delivered programmes around behavioural activation and problem management
• Make strengthening social support an explicit session goal, not an afterthought
• Review whether relaxation exercises are taking session time from more useful components
• Supervise counsellors on these core skills rather than on a broad menu
• Use the authors' web tool to estimate likely benefit for individual patient profiles

Why it matters

Session time in task-shared care is scarce, and some commonly included components may be reducing its effect.

Don't overread it

Component effects are estimated indirectly across trials, so they suggest what to test, not proof that dropping relaxation improves outcomes.

The statistics, in plain English

An incremental mean difference estimates how much adding one component changes the symptom score; negative is better here. A 95% credible interval is the Bayesian range within which the true value probably lies. The behavioural activation interval ends at −0.05, very close to zero, so that benefit is less certain than for social support or problem management. Component estimates come from comparing different packages across trials, which is indirect.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.