Reliability of the validated clinical diagnosis of pneumonia on validated outcomes after intracranial hemorrhage

Purpose: Reducing the incidence of hospital-acquired pneumonia (PNU) is important but depends on accurate assessment. We sought to determine the interrater reliability of diagnosis of PNU and its impact on resource utilization and functional outcomes in a high-risk population. Materials and Methods: Patients admitted in 2007 with intracranial hemorrhage were prospectively identified. Pneumonia was prospectively diagnosed by Centers for Disease Control criteria by a neurointensivist and infection control. An independent retrospective determination was made by a fellow, an infectious disease attending physician, and a pulmonologist after review of the electronic medical records and radiographs. Interrater reliability was analyzed with kappa statistics. One and 3-month outcomes were measured with the modified Rankin scale. Results: Of 103 patients, the incidence of PNU ranged from 5% to 25%. Interrater reliability was poor (median kappa = 0.30 [0.19-0.42]; P < .001). Any ascertainment of PNU was associated with longer intensive care unit length of stay, more fever and ventilator dependence, and worse functional outcomes. Conclusions: Pneumonia had poor interrater reliability despite highly trained reviewers and validated criteria. Although the clinical assessment of PNU is difficult, it was associated with greater resource use and worse outcomes. Diagnosis of clinical PNU may be suboptimal for measuring quality of intensive care. (C) 2012 Elsevier Inc. All rights reserved.