Pneumocystis carinii pneumonia is often difficult to diagnose in an ambulatory care setting. Previous reports have identified elements of the clinical history, physical examination, and clinical testing that are useful predictors of P carinii pneumonia. We analyzed published data on these predictors and measured them against methodologic standards for clinical prediction rules. Variables with high negative or positive predictive value for P carinii pneumonia, low error rates, or compelling biologic credibility were then selected to develop an untested clinical prediction rule for P carinii pneumonia. We suggest that dyspnea, oral lesions, chest roentgenographic examination, and pulse oximetry may be used to select patients requiring sputum testing and/or bronchoscopy for the diagnosis of P carinii pneumonia. The role of pulse oximetry in the diagnosis of P carinii pneumonia merits further study.
(Arch Intern Med. 1992;152:623-624)