Skip to main content

Table 1 Model inputs and outputs

From: Explainable AI enables clinical trial patient selection to retrospectively improve treatment effects in schizophrenia

Model input features

PAI model output

BRL model output

Demographics

- Age

- Sex

Symptom scales

- 30 Baseline individual item PANSS scores

- Baseline total PANSS score

- Baseline daytime drowsiness and quality of sleep scores from the Sleep Visual Analog Scale (VAS)

- Baseline Personal and Social Performance Scale score

- Baseline total Schizophrenia Quality of Life Scale score

Others

- Randomized arm (treatment or placebo; PAI only)

- Interactions of randomized arm with the other features (PAI only)

- Predicted actual and counterfactual week 6 PANSS scores

- Numerical PAI score (predicted outcome on treatment—predicted outcome on placebo)

- Labels: treatment-indicated, rest-indicated created from thresholded PAI score

- Labels: treatment-indicated and rest-indicated

  1. Several sequential outputs are generated during PAI modeling and are listed in order of generation