The uncertainty about the real burden of causes of death (CoD) is increasingly recognized by the international health community as a critical limitation for prioritizing effective public health measures. The minimally invasive autopsy (MIA) has shown to be a satisfactory substitute of the complete diagnostic autopsy (CDA), the gold standard for CoD determination in low- and middle-income countries. However, more studies are needed to confirm its adequate performance in settings with different epidemiology. In this observational study, the CoD obtained with the MIA were compared with the clinical diagnosis and the results of the CDA in 61 deaths that occurred in an infectious diseases referral hospital in Manaus, Brazilian Amazon. Concordance between the categories of diseases obtained by the three methods was evaluated by the Kappa statistic. Additionally, we evaluated discrepancies between clinical and complete diagnostic autopsy diagnoses. The MIA showed a substantial concordance with the CDA (Kappa = 0.777, 95% CI 0.608–0.946), and a perfect or almost perfect coincidence in specific diagnosis (ICD-10 code) between MIA and CDA was observed in 85% of the cases. In contrast, the clinical diagnosis showed a fair concordance with the CDA (Kappa = 0.311, 95% CI 0.071–0.552). Major clinico-pathological discrepancies were identified in 49% of cases. In conclusion, the MIA showed a substantial performance for CoD identification. Clinico-pathological discrepancies remain high and justify the need for post-mortem studies, even in referral hospitals. The MIA is a robust substitute of the CDA for CoD surveillance and quality improvement of clinical practice in low- and middle-income settings.