Ventilator-associated pneumonia and mortality

Abstract

Ventilator-associated Pneumonia (VAP) is one of the most common nosocomial infections among patients admitted to the intensive care unit. It is generally believed that VAP increases the mortality of patients, however exact determination of the attributable mortality of VAP is challenging. This thesis demonstrates that the attributable mortality of VAP can be investigated in several ways: 1. The use of observational studies of patients at risk for VAP, of whom some will develop VAP (and some will not) and subsequent comparison of crude mortality rates of those with and without VAP. This is obviously flawed by the many variables - aside from developing VAP or not - that differ between both patient groups that also influence mortality. 2. The use of the same observational studies, but now with additional adjustments for differences that are not equally distributed between those developing and not developing VAP. Yet, in many studies these methods overestimate the attributable mortality of VAP because of – inevitable – incomplete adjustment for differences between groups – both unmeasured variables at baseline, as well as changes in clinical condition during ICU stay. 3. One could - theoretically but not ethically - randomize patients to receive VAP or not and count deaths in each group. 4. Using the data of randomized controlled trials on VAP prevention. This provides the benefit of randomization, which excludes the effects of measured and unmeasured differences between groups. 5. Using novel statistical methods, such as competing risk survival analyses including VAP as a time dependent exposure. From the studies described in this thesis it can be concluded that the estimated attributable mortality of VAP is around 10% (9% in chapter 3 and 13% in chapter 4), which is much lower than estimates provided in previous studies. Subgroup analyses revealed an absence of attributable mortality in trauma patients, medical patients and patients with a low or high severity of illness at admission. Highest attributable mortality rates are found in surgical patients and patients with an intermediate score of severity of illness at admission. The lower estimates of attributable mortality and the differences of mortality in subgroups of patients are of critical importance for the design and interpretation of VAP prevention trials. Many studies were hugely underpowered to demonstrate improvements in patient outcome. The new knowledge on attributable mortality underpins the need of large studies (>1000 patients per study group) to demonstrate whether VAP prevention improves patient outcome. Furthermore, it is rational to focus on the subgroup of patients with the highest mortality rates, considering the large differences in mortality between subgroups. Further methodological improvements are needed in determining the attributable mortality of VAP, especially concerning methods to limit the influence of confounding. The use of multistate models might improve the determination of the attributable mortality of VAP. As these methods enable various intermediate events, and thus could offer more flexibility in the analyses and better capture the (chronological) complexity of factors influencing mortality