RESUMO
Background: In recent decades, the world has witnessed unprecedented progress in child survival. However, our knowledge of what is killing nearly 6 million children annually in low- and middle-income countries remains poor, partly because of the inadequacy and reduced precision of the methods currently utilized in these settings to investigate causes of death (CoDs). The study objective was to validate the use of a minimally invasive autopsy (MIA) approach as an adequate and more acceptable substitute for the complete diagnostic autopsy (CDA) for pediatric CoD investigation in a poor setting. Methods and findings: In this observational study, the validity of the MIA approach in determining the CoD was assessed in 54 post-neonatal pediatric deaths (age range: ≥1 mo to 15 y) in a referral hospital of Mozambique by comparing the results of the MIA with those of the CDA. Concordance in the category of disease obtained by the two methods was evaluated by the Kappa statistic, and the sensitivity, specificity, and positive and negative predictive values of the MIA diagnoses were calculated. A CoD was identified in all cases in the CDA and in 52/54 (96%) of the cases in the MIA, with infections and malignant tumors accounting for the majority of diagnoses. The MIA categorization of disease showed a substantial concordance with the CDA categorization (Kappa = 0.70, 95% CI 0.49-0.92), and sensitivity, specificity, and overall accuracy were high. The ICD-10 diagnoses were coincident in up to 75% (36/48) of the cases. The MIA allowed the identification of the specific pathogen deemed responsible for the death in two-thirds (21/32; 66%) of all deaths of infectious origin. Discrepancies between the MIA and the CDA in individual diagnoses could be minimized with the addition of some basic clinical information such as those ascertainable through a verbal autopsy or clinical record. The main limitation of the analysis is that both the MIA and the CDA include some degree of expert subjective interpretation.