Table 2.
Subscriber characteristics and prevalence of the dependent variable in the training and test sets for the Germany predictive model compared with the previously published results of the US predictive model in the test set. P values are reported for differences between German test and training sets, and between US and German test sets.
| Characteristics | Germany predictive model (this study) | US predictive model (from [9]) | ||||
| Training set | Test set | P value (test vs training Germany) | Test set | P value (US vs Germany test) | ||
| General | ||||||
| Prediction dates | Jan 1, 2012, to Jan 1, 2018 | Jan 1, 2012, to Jan 1, 2018 | — | Feb 1, 2014 | — | |
| # of unique PERSa users, n (%) | 4644 (80) | 116 (20) | — | 109,966 (20) | — | |
| # of prediction windows, n (%) | 96,273 (79.5) | 24,847 (20.5) | — | 109,966 (20) | — | |
| Age in years, mean (SD) | 84.0 (8.2) | 83.6 (8.3) | .001 | 81.1 (11.4) | <.001 | |
| Female gender, n (%) | 2997 (64.5) | 743 (64.0) | .37 | 88,433 (80.4) | <.001 | |
| Years on PERS service, n (%) | ||||||
| 0-2 | 2598 (55.9) | 630 (54.3) | .32 | 48,922 (44.4) | <.001 | |
| 2-4 | 1079 (23.2) | 264 (22.7) | .75 | 26,193 (23.8) | .41 | |
| 4 or more | 967 (20.8) | 267 (23.0) | .11 | 34,851 (31.7) | <.001 | |
| Number of PERS self-reported medical conditions, n (%) | ||||||
| None | 265 (5.7) | 73 (6.3) | .49 | 24,910 (22.6) | <.001 | |
| 1-2 | 1325 (28.5) | 329 (28.3) | .93 | 26,515 (24.1) | <.001 | |
| 3-4 | 1397 (30.1) | 370 (31.9) | .25 | 28,561 (26.0) | <.001 | |
| 5 or more | 1657 (35.7) | 389 (33.5) | .18 | 29,980 (27.3) | <.001 | |
| 30-day emergency hospital transport (% of prediction windows) | 1506 (1.6) | 350 (1.4) | .08 | 2455 (2.2) | <.001 | |
aPERS: personal emergency response system.