Setting: Two University hospitals in Eastern African capital cities where large prospective studies had been carried out on hospitalized patients to determine the cause of their respiratory diseases.
Objective: To identify features that differentiated between tuberculosis (TB) and non-tuberculous respiratory disease (non-TB) in hospitalized patients from Bujumbura, Burundi (n = 111) and Dar es Salaam, Tanzania (n = 71) whose sputum smears were negative on microscopic examination for acid-fast bacilli (AFB).
Design: Review of clinical findings, radiologic abnormalities, and laboratory test results from 182 patients, first by univariate and then by multivariate (stepwise logistic regression) analysis to assess the contribution of each factor to the final diagnosis.
Results: Of the 182 patients with two or more negative AFB smears, 41 had TB and 141 had non-TB. Stepwise regression analysis revealed four easily ascertained symptoms were associated with TB: 1) cough > 21 days; 2) chest pain > 15 days; 3) absence of expectoration; and 4) absence of shortness of breath. Any two of the four diagnosed TB with 85% sensitivity and 67% specificity; any three of the four with 49% sensitivity and 86% specificity. Multivariate analysis showed that adding lymphadenopathy and hematocrit < 30% improved discrimination.
Conclusion: This methodological approach provides a means for diagnosing TB among all AFB smear-negative hospitalized patients. In this setting, simple clinical symptoms alone are helpful. Similar studies are needed to develop a system for out-patient TB suspects.