Clinical predictors of Pneumocystis carinii pneumonia, bacterial pneumonia and tuberculosis in HIV-infected patients

Clinicians are frequently faced with the differential diagnosis between Pneumocystis carinii pneumonia (PCP), bacterial pneumonia, and pulmonary tuberculosis in HIV-infected patients. To identify features that could help differentiate these three pneumonia types at presentation by evaluating the clinical characteristics of the three diagnoses among patients at two urban teaching hospitals. Retrospective chart review. Cases were HIV-infected patients with a verified hospital discharge diagnosis of PCP (n = 99), bacterial pneumonia (n = 94), or tuberculosis (n = 36). Admitting notes were reviewed in a standardized manner; univariate and multivariate analyses were used to determine clinical predictors of each diagnosis. Combinations of variables with the highest sensitivity, specificity, and odds ratios (OR) were as follows: for PCP, exertional dyspnea plus interstitial infiltrate (sensitivity 58%, specificity 92%; OR, 16.3); for bacterial pneumonia, lobar infiltrate plus fever ≤ 7 days duration (sensitivity 48%, specificity 94%; OR, 14.6); and for tuberculosis, cough > 7 days plus night sweats (sensitivity 33%, specificity 86%; OR, 3.1). On regression analysis, independent predictors included interstitial infiltrate (OR, 10.2), exertional dyspnea (OR, 4.9), and oral thrush (OR, 2.9) for PCP; rhonchi on examination (OR, 12.4), a chart mention of ‘toxic’ appearance (OR, 9.1), fever ≤ 7 days (OR, 6.6), and lobar infiltrate (OR, 5.8) for bacterial pneumonia; and cavitary infiltrate (OR, 21.1), fever > 7 days (OR, 3.9), and weight loss (OR, 3.6) for tuberculosis. Simple clinical variables, all readily available at the time of hospital admission, can help to differentiate these common pneumonia syndromes in HIV-infected patients. These findings can help to inform clinical decision-making regarding choice of therapy, use of invasive diagnostic procedures, and need for respiratory isolation.