Step-by-Step treatment effect estimates
| Post-treatment | Follow-up | |||||||||
| Model 1 MI | Model 2 RFLB | Model 1 MI | Model 2 RFLB | |||||||
| B MI (SE) | P value | Effect size (95% CI) | B Lower bound (SE) | B Upper bound (SE) | B MI (SE) | P value | Effect size (95% CI) | B Lower bound (SE) | B Upper bound (SE) | |
| Primary | ||||||||||
| PHQ | −4.18 (0.78) | <0.01 | 0.71 (0.45 to 0.97) | −6.73 (1.14) | −1.57 (0.94) | −3.07 (0.90) | <0.01 | 0.52 (0.22 to 0.82) | −5.17 (1.28) | −0.36 (1.15) |
| WHODAS | −3.95 (1.04) | <0.01 | 0.43 (0.21 to 0.65) | −7.79 (1.68) | 0.99 (1.76) | −1.57 (1.22) | 0.20 | 0.17 (−0.09 to 0.43) | −5.03 (2.01) | 2.64 (1.97) |
| Secondary | ||||||||||
| WHO-5 | 2.01 (0.68) | <0.01 | 0.37 (0.12 to 0.62) | −0.47 (1.05) | 4.74 (0.94) | 2.42 (0.76) | <0.01 | 0.44 (0.17 to 0.71) | 0.31 (1.11) | 4.81 (1.04) |
| GAD-7 | −4.14 (0.71) | <0.01 | 0.74 (0.49 to 0.99) | −6.98 (1.03) | −1.50 (1.08) | −2.86 (0.79) | <0.01 | 0.49 (0.22 to 0.76) | −5.30 (1.25) | −0.50 (1.27) |
| PCL-5 | −3.86 (0.94) | <0.01 | 0.53 (0.27 to 0.79) | −7.31 (1.28) | −0.77 (1.32) | −2.70 (1.15) | 0.02 | 0.35 (0.06 to 0.64) | −6.54 (1.74) | 0.31 (1.86) |
| PSYCHLOPS | −2.79 (0.67) | <0.01 | 0.56 (0.29 to 0.83) | −5.34 (0.97) | −0.88 (0.92) | −2.64 (0.81) | <0.01 | 0.47 (0.19 to 0.75) | −5.33 (1.24) | −0.64 (1.24) |
*Treatment effects in model 1 are derived through multiple (150) imputations, assuming missing at random, applying multivariate normal regression, imputation by experimental group and accommodating arbitrary missing outcome value patterns using an iterative Markov Chain Monte Carlo method. Prescore and background controls (gender and a dummy for whether a participant had missing values on 1 or more of these characteristics) are included for improved precision of the regression point estimates reported in model 1. Effect sizes reported in model 1 are Hedges g, combining MI estimation results using Rubin’s rules. Treatment effects in model 2 are derived through RFLBs procedure, assuming missing not at random for the differential attrition between treatment arms.
GAD-7, Generalised Anxiety Disorder 7; MI, multiple imputation; PCL-5, Post-Traumatic Stress Disorder Checklist 5; PHQ, Patient Health Questionnaire; PSYCHLOPS, Psychological Outcomes Profile; RFLBs, Random Forest Lee Bounds; WHODAS, WHO Disability Assessment Schedule-12.