Table 3.
Summary of fit results for study sample stratified by case-mix group
| Case-mix group | Model | N clients | MAPE (h) | CPM (%) | ± sd* (%) | RMSE (h) | R-squared (%) | ± sd* (%) |
|---|---|---|---|---|---|---|---|---|
| ST-H | INT | 694 | 1.0 | 0 | 0 | 2.0 | 0 | 0 |
| DEMO | 1.0 | − 0.6 | 0.3 | 1.7 | − 1.7 | 0.4 | ||
| DEMO + NANDA-I | 1.0 | − 0.5 | 0.4 | 1.7 | − 1.9 | 1.1 | ||
| ST-F | INT | 754 | 1.2 | 0 | 0 | 2.3 | 0 | 0 |
| DEMO | 1.2 | − 1.3 | 0.3 | 2.0 | − 1.7 | 0.6 | ||
| DEMO + NANDA-I | 1.1 | 9.4 | 0.6 | 1.9 | 13.8 | 2.1 | ||
| LT-SOM | INT | 2359 | 2.1 | 0 | 0 | 3.9 | 0 | 0 |
| DEMO | 2.1 | 0.5 | 0.1 | 3.7 | 0.1 | 0.1 | ||
| DEMO + NANDA-I | 1.8 | 15.4 | 0.3 | 3.4 | 17.5 | 1.1 | ||
| LT-PG | INT | 389 | 2.0 | 0 | 0 | 3.6 | 0 | 0 |
| DEMO | 2.0 | − 0.8 | 0.6 | 3.0 | − 3.6 | 1.6 | ||
| DEMO + NANDA-I | 1.8 | 6.6 | 0.7 | 2.9 | 8.4 | 1.4 |
MAPE mean absolute prediction error, CPM Cumming’s prediction measure, RMSE root mean squared error, h hours, sd standard deviation
*Standard deviation is generated using 30 repeated tenfold cross-validation