The value of clinical judgment in the diagnosis of nosocomial pneumonia

Abstract

Nosocomial pneumonia presents a diagnostic and therapeutic challenge in the care of critically ill patients. The present study was designed to determine as closely as possible the occurrence of nosocomial pneumonia in surgical intensive care unit (ICU) patients using clinical, radiographic, and bacteriological parameters in a prospective concurrent fashion. This clinical study enrolled all surgical, trauma, and neurosurgical patients admitted to a surgical ICU over a 13-month period. Routine surveillance was used to identify those patients suspected of developing nosocomial pneumonia. Numerous clinical parameters concerning ventilatory support, acute lung injury, organ dysfunction, nutrition, and length of stays were used to identify factors disposing to development of pneumonia. Univariate and multivariate analyses were used for this purpose. Patients thought to have pneumonia were then followed concurrently to determine, as closely as possible, whether pneumonia was present by serial examination of clinical, bacteriologic, and radiographic data. Those "validated" by this process were then compared to those "nonvalidated" to see if any distinction could be made. Of the 352 patients enrolled, 46 (13%) were initially labeled as having developed nosocomial pneumonia when compared to the 306 patients without pneumonia. Univariate analysis demonstrated a greater need for intubation and mechanical ventilation, longer mechanical ventilation, more acute lung injury, longer ICU and hospital stays, poorer nutrition, and higher mortality (17% versus 5%, P < 0.01). Multivariate analysis demonstrated only length of ICU stay and length of intubation/mechanical ventilation as longer in the pneumonia group. On further concurrent review, 23 of 46 patients were validated as having pneumonia while the rest were felt not to have pneumonia. When the two groups were compared, only asymmetric and segmental radiographic infiltrates distinguished validated from nonvalidated pneumonia patients and all other clinical parameters, including mortality and length of stay, were similar. Nosocomial pneumonia was initially suspected in 13% of this ICU population. Numerous clinical parameters clearly distinguished these pneumonia patients from others and they suffered a substantially higher mortality. However, within this pneumonia group, only half of the patients could be validated as truly having pneumonia using available clinical parameters. Nevertheless, those validated were indistinguishable in their clinical behavior from those who were not. This calls into question the need for elaborate and sometimes expensive investigations for diagnosis of nosocomial pneumonia.