Using flexible methods to determine risk factors for ventilator-associated pneumonia in the Netherlands

Abstract

Seven hospitals participated in the Dutch national surveillance for ventilator-associated pneumonia (VAP) and its risk factors. We analysed time-independent and time-dependent risk factors for VAP using the standard Cox regression and the flexible Weighted Cumulative Effects method (WCE) that evaluates both current and past exposures. The prospective surveillance of intensive care patients aged 16 years and ventilated 48 hours resulted in the inclusion of 940 primary ventilation periods, comprising 7872 ventilation days. The average VAP incidence density was 10.3/1000 ventilation days. Independent risk factors were age (16–40 years at increased risk: HR 2.42 95% confidence interval 1.07–5.50), COPD (HR 0.19 [0.04–0.78]), current sedation score (higher scores at increased risk), current selective oropharyngeal decontamination (HR 0.19 [0.04–0.91]), jet nebulizer (WCE, decreased risk), intravenous antibiotics for selective decontamination of the digestive tract (ivSDD, WCE, decreased risk), and intravenous antibiotics not for SDD (WCE, decreased risk). The protective effect of ivSDD was afforded for 24 days with a delay of 3 days. For some time-dependent variables, the WCE model was preferable over standard Cox proportional hazard regression. The WCE method can furthermore increase insight into the active time frame and possible delay herein of a time-dependent risk factor.