Top 10 important predictors of the prediction models for risperidone and aripiprazole*
| Risperidone | ||||||
| Rank | Backward stepwise LR | LASSO | Ridge | Elastic net | Random forest | GBM |
| 1 | Number of outpatient visits, non-MHD | Number of inpatient visits, non-MHD | Number of EM visits, MHD | Number of inpatient visits, MHD | Age | Age |
| 2 | Number of any route visits, non-MHD | Age | Number of inpatient visits, MHD | Age | Number of any route visits, non-MHD | Number of visits, non-MHDs |
| 3 | Number of inpatient visits, MHD | Number of EM visits, non-MHDs | Number of EM visits, non-MHDs | Number of EM visits, non-MHDs | Number of outpatient visits, non-MHD | Number of outpatient visits, non-MHD |
| 4 | Number of EM visits, non-MHDs | Number of EM visits, MHD | Age | Number of EM visits, MHD | Number of outpatient visits, MHD | ASD |
| 5 | Age | Anticholinergic drugs | Anticholinergic drugs | Anticholinergic drugs | Number of any route visits, MHD | Number of any route visits, MHD |
| 6 | Anticholinergic drugs | EBD | ASD | EBD | Number of outpatient visits, non-MHD | Number of outpatient visits, MHD |
| 7 | EBD | ASD | EBD | ASD | First diagnosis of Schizophrenia | Mental retardation |
| 8 | Bipolar disorder | Insurance type: Medical aid | Number of inpatient visits, non-MHD | Insurance type: Medical aid | Sex: female | Insurance type: Medical aid |
| 9 | ASD | Mental retardation | Other congenital or genetic defect | Mental retardation | Insurance type: Medical aid | Seizure |
| 10 | Insurance type: Medical aid | Sex: female | Insurance type: Medical aid | Reaction to severe stress and adjustment disorder | Other MHDs | Antipyretics, analgesics |
| Aripiprazole | ||||||
| Rank | Backward stepwise LR | LASSO | Ridge | Elastic NET | Random forest | GBM |
| 1 | Number of outpatient visits, non-MHD | Age | Number of EM visits, MHD | Age | Age | Age |
| 2 | Number of any route visits, non-MHD | Number of inpatient visits, non-MHD | Number of inpatient visits, non-MHD | Number of inpatient visits, non-MHD | Number of outpatient visits, non-MHD | Number of any route visits, non-MHDs |
| 3 | Number of EM visits, MHD | Number of EM visits, MHD | Number of EM visits, non-MHDs | Number of EM visits, MHD | Number of any route visits, non-MHD | Number of outpatient visits, non-MHD |
| 4 | Number of inpatient visits, non-MHD | Lithium | Age | Lithium | Number of any route visits, MHD | ASD |
| 5 | Age | Number of EM visits, non-MHDs | Lithium | Number of EM visits, non-MHDs | Number of outpatient visits, MHD | Number of any route visits, MHD |
| 6 | Lithium | Bipolar disorder | Anticholinergic drugs | Bipolar disorder | First diagnosis of Schizophrenia | Number of outpatient visits, MHD |
| 7 | Tic disorder | Sex: female | Bipolar disorder | Sex: female | Antipyretics, analgesics | Mental retardation |
| 8 | MAO inhibitor | Number of any route visits, MHD | ASD | Number of any route visits, MHD | Sex: female | Insurance type: Medical aid |
| 9 | Bipolar disorder | Anticholinergic drugs | Diabetes mellitus | Anticholinergic drugs | Antihistamines | Seizure |
| 10 | ED | MAO inhibitor | MAO inhibitor | MAO inhibitor | Other MHDs | Antipyretics, analgesics |
*The relative importance of predictors in SuperLearner was not evaluated because this algorithm ensembles the predicted probabilities of each model used in construction.
ASD, autism spectrum disorder; BP, bipolar disorder; EBD, other behaviour and emotional disorders with onset generally occurring in childhood and adolescence; ED, emotional disorders with onset specific to childhood; EM, emergency room; GBM, gradient boosting machine; LASSO, least absolute shrinkage and selection operator; MAO, monoamine oxidase; MHD, mental health disorder.