Table 1.

Prediction of post-treatment depression by linear regression model including only pre-treatment assessment of outcome (benchmark), additional variance explained beyond benchmark model by ensemble model (model gain), and total variance explained.

	Prediction R2	95% CI
HRSD
Benchmark	0.17	0.07–0.26
Random forest	0.23	0.14–0.31
Elastic net	0.24	0.14–0.33
Random forest/elastic net ensemble	0.25	0.16–0.33
Gain for ensemble model	+0.08	+0.008 to +0.15
Disability
Benchmark	0.20	0.10–0.31
Random forest	0.24	0.13–0.34
Elastic net	0.24	0.15–0.33
Random forest/elastic net ensemble	0.25	0.16–0.35
Gain for ensemble model	+0.05	−0.003 to +0.10
IDAS-Well Being
Benchmark	0.18	0.08–0.27
Random forest	0.26	0.19–0.34
Elastic net	0.29	0.19–0.40
Random forest/elastic net ensemble	0.29	0.21–0.38
Gain for ensemble model	+0.12	+0.05 to +0.19

95% CIs for prediction R² were based on the standard error formula applied to the 10 × 10 cross-validation estimates; 95% CIs for gain (the increase in predicted R² over benchmark) were estimated by bootstrap.