BACKGROUND: Clinico-pathological discrepancies are more frequent in settings in which limited diagnostic techniques are available, but there is little information on their actual impact. AIM: We assessed the accuracy of the clinical diagnoses in a tertiary referral hospital in sub-Saharan Africa by comparison with post-mortem findings. We also identified potential risk factors for misdiagnoses. METHODS: One hundred and twelve complete autopsy procedures were performed at the Maputo Central Hospital (Mozambique), from November 2013 to March 2015. We reviewed the clinical records. Major clinico-pathological discrepancies were assessed using a modified version of the Goldman and Battle classification. RESULTS: Major diagnostic discrepancies were detected in 65/112 cases (58%) and were particularly frequent in infection-related deaths (56/80 [70%] major discrepancies). The sensitivity of the clinical diagnosis for toxoplasmosis was 0% (95% CI: 0-37), 18% (95% CI: 2-52) for invasive fungal infections, 25% (95% CI: 5-57) for bacterial sepsis, 34% (95% CI: 16-57), for tuberculosis, and 46% (95% CI: 19-75) for bacterial pneumonia. Major discrepancies were more frequent in HIV-positive than in HIV-negative patients (48/73 [66%] vs. 17/39 [44%]; p = 0.0236). CONCLUSIONS: Major clinico-pathological discrepancies are still frequent in resource constrained settings. Increasing the level of suspicion for infectious diseases and expanding the availability of diagnostic tests could significantly improve the recognition of common life-threatening infections, and thereby reduce the mortality associated with these diseases. The high frequency of clinico-pathological discrepancies questions the validity of mortality reports based on clinical data or verbal autopsy.