Ventilator-Associated Pneumonia: Translating Risk Assessment into Prevention

abstract & commentary

Source: Cook DJ, et al. Incidence of and risk factors for ventilator- associated pneumonia in critically ill patients. Ann Intern Med 1998; 129:433-440.

Cook and colleagues prospectively followed 1014 patients who underwent mechanical ventilation for at least 48 hours in a multicenter study of nosocomial pneumonia in intensive care units in Canada and France. They used four different definitions of pneumonia, including a blinded adjudication panel, modified CDC criteria, a Clinical Pulmonary Infection score, and the results of bronchoalveolar lavage or protected specimen brush cultures. A large number of independent variables were recorded; these included demographics, underlying diagnoses, several illness severity measures (e.g., Apache II score, Glasgow coma score), feeding method, and drug exposures. Variables associated with pneumonia on univariate analysis were entered into a multivariate analysis, as well as a Cox hazard model, in order to assess the influence of independent variables over time.

Of the 1014 patients, 17.5% developed pneumonia; the overall incidence was 14.8/1000 patient days. The daily risk of pneumonia peaked on day 5 (3.3% per day) but decreased rapidly thereafter, to a risk of 1.3% per day by day 15. The majority of independent predictors of pneumonia were related to underlying illness, including burns, trauma, central nervous system disease, respiratory disease, and cardiac disease. Witnessed aspiration (risk ratio 2.3, CI 1.1-4.7) and receipt of paralytic agents (RR 1.6, CI 1.03-2.4). Receipt of antibiotics had a protective effect (RR 0.37, CI 0.27-0.51), which however, declined with time in the ICU; a significant protective effect was no longer demonstrable by day 15. The risk factors for pneumonia were the same no matter which of the four diagnostic criteria were used.

Comment by Robert Muder, MD

This study of ventilator-associated pneumonia is notable for its size, its prospective, multicenter design, the complexity of the data base, and the rigor of the statistical analysis. Its results are likely to be generalizable to critical care units in other large tertiary care centers. This study was conducted concurrently with a randomized trial of sucralfate vs. ranitidine for prophyaxis of gastrointestinal bleeding (which, incidently, found no difference in pneumonia between the treatment groups).

Although this study sheds considerable light on the complex interrelationships between risk factors for nosocomial pneumonia in the ICU, it is less clear how this data can be used to decrease the risk of pneumonia. The greatest risk is conferred by the patient’s underlying condition—this cannot, of course, be modified. The potentially modifiable risk factors are aspiration, exposure to paralytic agents, and exposure to antibiotics.

It is well known that aspiration increases the risk of pneumonia in virtually every setting; measures to decrease the risk of aspiration are clearly warranted. However, it is not clear how this should be accomplished. The study found no relationship between nasoenteral alimentation and pneumonia, for example. Indeed, the existing literature on this relationship is conflicting. The relationship between pneumonia and paralytic agents is a bit puzzling; paralytic agents predispose to pneumonia by inhibiting cough reflex and preventing effective swallowing. However, it is not clear how this would adversely affect a patient who had already been intubated, in whom effective coughing was prevented by the endotracheal tube. Cook et al did not specify which paralytic agents were used, but some agents could have adverse effects on gastrointestinal motility, facilitating reflux of gastric contents into the upper aerodigestive tract.

Cook et al found a protective effect of antibiotic exposure, and this was most pronounced during the first ten days of ICU stay. The most likely explanation for the waning of the protective effect over time is that the risk of pneumonia itself decreased rapidly after day 5. Patients received a wide variety of antibiotics, and many received multiple agents. Cephalosporins were given most often (63% of patients). Some comparative trials (using a variety of regimens) have shown that prophylactic antibiotics reduce the incidence of pneumonia in ICU patients,1,2 while others have not.3 Not surprisingly, use of prophylactic antibiotics may increase the risk of colonization by multiresistant organisms. Thus, the precise role of antibiotic prophylaxis in this setting is uncertain. There is a great need for well-designed clinical trials to assess the efficacy of antibiotic prophylaxis in the ICU and to determine the "best" regimen for prevention of pneumonia. Any study should also carefully assess the effect of prophylaxis on carriage of and infection with multiresistant organisms in both pulmonary and extra-pulmonary sites.


1. Sirvent JM, et al. Am J Respir Crit Care Med 1997; 155:1729-1734.

2. Quinio B, et al. Chest 1996;109:765-772.

3. Mandelli M, et al. Crit Care Med 1989;17:501-505.