Objectives: To identify predictors of Pneumocystis jiroveci pneumonia (PCP) or pulmonary tuberculosis (TB) in acid-fast bacillus smear-negative HIV-infected patients and to develop clinical prediction rules.
Design: A cohort study conducted in consecutive hospitalized Asian patients.
Methods: Multivariate analyses were performed on the Cambodian sample to determine clinical, radiological, and biological predictors of PCP or TB at hospital admission. The Vietnamese sample was kept for independent validation.
Results: In Cambodia, the gold standard technique for TB and PCP were fulfilled in 172 (27 cases) and 160 (84 cases) patients, respectively. For TB, independent predictors included the following: headache [odds ratio (OR) 3.0; 95% confidence interval (CI) 1.04 to 8.6], localized radiological opacity (OR 5.8; 95% CI 1.9-17.9), and mediastinal adenopathy (OR 10.1; 95% CI 3.5 to 29.0); and for PCP: resting oxygen saturation <90% (OR 3.3; 95% CI 1.3 to 8.5 for resting arterial oxygen saturation >or=80%; and OR 9.1; 95% CI 1.8 to 44.5 for resting arterial oxygen saturation <80%), trimethoprim-sulphamethoxazole prophylaxis (OR 0.1; 95% CI 0.04 to 0.6), and diffuse radiological shadowing (OR 7.0; 95% CI 2.7 to 18.6). PCP risk predicted by a score based on these 3 factors ranged from 3% to 92% (Cambodia). When tested on Vietnamese patients (n = 69, 38 with PCP), the score maintained correct predictive ability (c-index = 0.72) but with poor calibration.
Conclusions: The PCP score could provide a useful clinical tool to identify PCP among acid-fast bacillus smear-negative pneumonia and start specific therapy.