Table 2.
Multivariable model of risk factors for hematoma (full cohort, sorted by hazard ratio)
| N (cat) | Mean (cont) | Hazard ratio | Lower 95% CI | Upper 95% CI | P value | |
|---|---|---|---|---|---|---|
| Anticoagulant use | 2968 | NA | 2.44 | 1.69 | 3.51 | <.001 |
| History of valve surgery | 603 | NA | 2.11 | 1.42 | 3.15 | <.001 |
| Existing pocket reopened | 5641 | NA | 1.92 | 1.06 | 3.47 | .032 |
| Antiplatelet use | 4109 | NA | 1.66 | 1.14 | 2.42 | .008 |
| Male | 4884 | NA | 1.63 | 1.06 | 2.50 | .027 |
| History of coronary artery disease | 2862 | NA | 1.47 | 1.04 | 2.10 | .031 |
| Lead revised | 2529 | NA | 1.45 | 0.94 | 2.23 | .090 |
| History of nonischemic cardiomyopathy | 2066 | NA | 1.42 | 0.99 | 2.05 | .058 |
| Procedure time (hours increase) | NA | 0.92 | 1.21 | 0.98 | 1.50 | .082 |
| # Previous cardiac device procedures | NA | 1.33 | 1.14 | 0.99 | 1.32 | .064 |
| BMI (unit decrease) | NA | 29.16 | 1.06 | 1.03 | 1.09 | <.001 |
To evaluate risk factors for hematoma, the Akaike information criterion (AIC) was used. Since a lower AIC indicates a better goodness-of-fit/complexity tradeoff, a global model of baseline and procedural characteristics was built using stepwise assessment of Cox proportional hazard regression to identify the model minimizing AIC. Since AIC minimization does not depend on P values, there is no minimal P value restriction. However, P values are provided to determine relative confidence among variables remaining in the model.
BMI = body mass index; cat = categorical variable; CI = confidence interval; cont = continuous variable; NA = not applicable.