Table 6.
Performance of the goodness-of-fit of each algorithm reported as the Hosmer-Lemeshow calibration statistic for Framingham Stroke Risk Score (FSRS), Stroke Riskometer™ and QStroke against observed stroke events at 5-years for the Russian and 5-years and 10-years for the Rotterdam and combined (ARCOS, Russia and Rotterdam data set)
| Data | Risk score | Subset | Hosmer-Lemeshow Test | |
|---|---|---|---|---|
| 5-year risk | 10-year risk | |||
| RUSSIA | FSRS | Females | χ2 = 58·12, P ≤ 0·0001 | |
| Males | χ2 = 133·65, P ≤ 0·0001 | |||
| Stroke Riskometer™ | Females | χ2 = 321·92, P ≤ 0·0001 | ||
| Males | χ2 = 36·84, P ≤ 0·0001 | |||
| QStroke | Females | χ2 = 3·33, P = 0·912 | ||
| Males | χ2 = 318·81, P ≤ 0·0001 | |||
| Rotterdam | FSRS | Females | χ2 = 69·95, P ≤ 0·0001 | χ2 = 222·02, P ≤ 0·0001 |
| Males | χ2 = 100·58, P ≤ 0·0001 | χ2 = 356·01, P ≤ 0·0001 | ||
| Stroke Riskometer™ | Females | χ2 = 298·95, P ≤ 0·0001 | χ2 = 588·20, P ≤ 0·0001 | |
| Males | χ2 = 2247·03, P ≤ 0·0001 | χ2 = 20 297·53, P ≤ 0·0001 | ||
| QStroke | Females | χ2 = 21·68, P = 0·006 | χ2 = 70·10, P ≤ 0·0001 | |
| Males | χ2 = 796·93, P ≤ 0·0001 | χ2 = 949·04, P ≤ 0·0001 | ||
| Combined | FSRS | Females | χ2 = 196·70, P ≤ 0·0001 | χ2 = 304·91, P ≤ 0·0001 |
| Males | χ2 = 153·78, P ≤ 0·0001 | χ2 = 726·04, P ≤ 0·0001 | ||
| Stroke Riskometer™ | Females | χ2 = 547·29, P ≤ 0·0001 | χ2 = 1 811·14, P ≤ 0·0001 | |
| Males | χ2 = 1699·96, P ≤ 0·0001 | χ2 = 11 552·55, P ≤ 0·0001 | ||
| QStroke | Females | χ2 = 1441·52, P ≤ 0·0001 | χ2 = 270·42, P ≤ 0·0001 | |
| Males | χ2 = 1587·38, P ≤ 0·0001 | χ2 = 1 822·10, P ≤ 0·0001 | ||