Evaluation of the World Health Organization algorithm for the diagnosis of HIV-associated sputum smear-negative tuberculosis

BACKGROUND: Outcomes from the World Health Organization's (WHO's) recommendations for the diagnosis of smear-negative tuberculosis (SNTB) in high human immunodeficiency virus prevalence settings are unknown. METHODS: We retrospectively applied the WHO algorithm for SNTB without danger signs to a prospectively enrolled cohort of ambulatory adult SNTB suspects in KwaZulu-Natal, South Africa. Participants fulfilling specified criteria for SNTB started empiric anti-tuberculosis treatment; the rest of the cohort was observed. All were followed for 8 weeks. Confirmed TB was defined as positive culture or granulomata plus acid-fast bacilli on histology. RESULTS: In total, 221 participants retrospectively fulfilled the WHO ambulatory SNTB algorithm entry criteria. The diagnostic performance of the WHO algorithm was: positive predictive value 0.34 (95%CI 0.26-0.43), negative predictive value 0.86 (95%CI 0.76-0.92), positive likelihood ratio 1.43 (95%CI 1.34-1.48), negative likelihood ratio 0.46 (95%CI 0.38-0.56) and diagnostic odds 3.1 (95%CI 1.52-6.34). Losses to follow-up (n = 4), hospitalisations (n = 6) and deaths (n = 5) did not differ significantly in those who were and were not diagnosed with SNTB. CONCLUSIONS: The WHO ambulatory SNTB algorithm had a reasonably high negative predictive value but low positive predictive value. Mortality over an 8-week period was low in participants who met the entry criteria for the WHO algorithm.