Table 2.
Relative predictive value of top 25 variables in the optimal Classification and Regression Tree solution (see Supplementary Table 6 for a complete list).
| Input variable | Relative importance | |
|---|---|---|
| 1 | Socioeconomic status (SEFI-2) | 1.000 (reference) |
| 2 | Lives in long-term care | 0.883 |
| 3 | Segal frailty score | 0.879 |
| 4 | Distance from home to closest high-intensity | 0.860 |
| 5 | McIsaac frailty score | 0.810 |
| 6 | Age | 0.756 |
| 7 | Outpatient clinic visits: 7–12 months prior to start date | 0.587 |
| 8 | Outpatient clinic visits: 13–24 months prior to start date | 0.580 |
| 9 | Outpatient laboratory test counts: 13–24 months prior to start date | 0.577 |
| 10 | ATC4 prescription counts: 7–12 months prior to start date | 0.558 |
| 11 | ATC4 prescription counts: 4–6 months prior to start date | 0.549 |
| 12 | ATC4 prescription counts: 13–24 months prior to start date | 0.535 |
| 13 | ATC4 prescription counts: 0–3 months prior to start date | 0.504 |
| 14 | Hospital-days: 13–24 months prior to start date | 0.458 |
| 15 | Outpatient laboratory test counts: 7–12 months prior to start date | 0.449 |
| 16 | Outpatient clinic visits: 4–6 months prior to start date | 0.444 |
| 17 | Outpatient clinic visits: 0–3 months prior to start date | 0.421 |
| 18 | Outpatient laboratory test counts: 4–6 months prior to start date | 0.416 |
| 19 | Hospital-days: 0–3 months prior to start date | 0.401 |
| 20 | Hospital-days: 7–12 months prior to start date | 0.383 |
| 21 | Hospital-days: 4–6 months prior to start date | 0.301 |
| 22 | Metastatic cancer | 0.285 |
| 23 | Outpatient laboratory test counts: 4–6 months prior to start date | 0.274 |
| 24 | Statistical area classification | 0.200 |
| 25 | Open homecare file | 0.186 |
ATC4, fourth level of the Anatomic Therapeutic Chemical Classification system; SEFI, socioeconomic factor index.