Prognostic indicators in the World Health Organization’s algorithm for seriously ill HIV-infected inpatients with suspected tuberculosis

Journal Article


Journal Title

AIDS research and therapy

Journal ISSN
Volume Title

BioMed Central


University of Cape Town

Background: Criteria for the 2007 WHO algorithm for diagnosing tuberculosis among HIV-infected seriously ill patients are the presence of one or more danger signs (respiratory rate > 30/min, heart rate > 120/min, temperature > 39 °C, and being unable to walk unaided) and cough ≥ 14 days. Determining predictors of poor outcomes among HIV-infected inpatients presenting with WHO danger signs could result in improved treatment and diagnostic algorithms. Methods: We conducted a prospective cohort study of inpatients presenting with any duration of cough and WHO danger signs to two regional hospitals in Cape Town, South Africa. The primary outcome was all-cause mortality up to 56 days post-discharge, and the secondary outcome a composite of any one of: hospital admission for > 7 days, died in hospital, transfer to a tertiary level or tuberculosis hospital. We frst assessed the WHO danger signs as predictors of poor outcomes, then assessed the added value of other variables selected a priori for their ability to predict mortality in common respiratory opportunistic infections (CD4 count, body mass index (BMI), being on antiretroviral therapy (ART), hypotension, and confusion) by comparing the receiver operating characteristic (ROC) area under the curve (AUC) of the two multivariate models. Results: 484 participants were enrolled, median age 36, 66% women, 53% had tuberculosis confrmed on culture. The 56-day mortality was 13.2%. Inability to walk unaided, low BMI, low CD4 count, and being on ART were independently associated with poor outcomes. The multivariate model of the WHO danger signs showed a ROC AUC of 0.649 (95% CI 0.582–0.717) for predicting 56-day mortality, which improved to ROC AUC of 0.740 (95% CI 0.681–0.800; p = 0.004 for comparison between the two ROC AUCs) with the multivariate model including the a priori selected variables. Findings were similar in sub-analyses of participants with culture-positive tuberculosis and with cough duration ≥ 14 days. Conclusion: The study design prevented a rigorous evaluation of the prognostic value of the WHO danger signs. Our prognostic model could result in improved algorithms, but needs to be validated.