Multivariable mixed-effects binary Poisson regression for risk ratio of ≥ 2 reoperations versus 1 reoperation*
| Covariate | Risk ratio ± SE (95% CI) | p value |
|---|---|---|
| Dislocation | 4.3 ± 0.9 (2.9–6.3) | < 0.001 |
| Infection | 3.1 ± 0.7 (2.0–4.9) | < 0.001 |
| Alcohol abuse | 1.8 ± 0.3 (1.3–2.4) | < 0.001 |
| Dementia† | 1.1 ± 0.2 (0.81–1.6) | 0.48 |
CI = confidence interval; SE = standard error.
↵* n = 229 reoperations in 154 hips (154 first reoperations in 154 hips; ≥ 2 reoperations, 75 reoperations in 46 hips); 1 additional hip had 1 reoperation that was excluded because of missing risk factor data. Results shown only for significant covariates except dementia.
↵† For the 229 reoperations, 101 reoperations were for hips in patients with no dementia and no dislocation; 46 reoperations with dementia and dislocation; 32 reoperations with dementia and no dislocation; and 50 reoperations with dislocation and no dementia (χ2 test, p < 0.001).
Therefore, the predictor variables (independent variables) dementia and dislocation were colinear (associated with each other) and could not independently predict the risk of the dependent variable (≥ 2 reoperations) in the regression model.