Logistic regression models of potential determinants of severe OSA defined as Apnoea–hypopnoea index ≥30 events·h−1
| Crude models | Adjusted model# | |||||||
| OR | 95% CI | R2¶ | C statistic+ | OR | 95% CI | R2¶ | C statistic+ | |
| Age years | 1.02 | 1.006–1.03 | 0.007 | 0.55 | 1.03 | 1.01–1.04 | ||
| Body mass index kg·m−2 | 1.11 | 1.08–1.15 | 0.040 | 0.63 | 1.11 | 1.08–1.15 | ||
| Triglycerides cg·dL−1 | 1.03 | 1.007–1.04 | 0.007 | 0.55 | 1.03 | 1.01–1.05 | ||
| Peak troponin deciles§ | 1.07 | 1.02–1.11 | 0.007 | 0.56 | 1.07 | 1.02–1.12 | ||
| Killip class >I | 2.31 | 1.39–3.88 | 0.010 | 0.53 | 2.06 | 1.21–3.52 | ||
| Full model | 0.072 | 0.67 | ||||||
Positive predictive value over imputed data for the adjusted model=57.15%. Negative predictive value over imputed data for the adjusted model=66.04%. #: all covariates are together in a single model; ¶: R2 using Fisher's z over imputed data; +: C statistic over imputed data; §: test-specific deciles of peak troponin (computed separately for each testing method to account for differences in sensitivity among centres).