326. Confidential review of maternal deaths in the Community Level Interventions for Pre-eclampsia (CLIP) Mozambique Trial

Abstract

Introduction Computer based algorithms, such as the WHO’s verbal autopsy tool (InterVA), are often used to attribute cause of death (COD) in low-resource settings despite their variable agreement with physician review. After a large scale community-level intervention trial for pre-eclampsia in Mozambique, maternal deaths were assessed in order to identify COD and barriers to care. Objective To compare COD for women who died in the Community Level Interventions in Pre-eclampsia (CLIP) Mozambique Trial (NCT01911494), 2015–2017, by physician review vs. the InterVA output. Methods Two physicians independently reviewed maternal mortality data from the trial (i.e., verbal autopsy [2012 WHO Instrument], baseline data and outcomes, and in the intervention arm, data from a mobile health application that included blood pressure readings), and assigned COD (by International Classification of Disease-Maternal Mortality). Disagreement was resolved by a third reviewer. Assigned COD was compared between physician review and InterVA output (set for high prevalence of malaria and HIV) using Cohen’s Kappa. Results Of the twenty-one (0.14%) maternal deaths reported, most were postpartum (14/21, 66.7%), and due to indirect causes (12, 57.1%). Malaria during or prior to pregnancy (14/21, 66.7%) and HIV/AIDs (11/21 reported, 52.3%) were found to significantly affect pregnancy. The three most common causes of death were 1) non-obstetric complications, 2) obstetric hemorrhage and 3) pregnancy-induced hypertension. Agreement between physicians and InterVA was fair, K = 0.40 [0.10–0.69]). Compared with InterVA, physicians were more likely to assign COD due to non-obstetric complications, specifically infectious diseases (9/21, 42.9%), including malaria and HIV/AIDS. Discussion Addressing infectious diseases, such as malaria and HIV/AIDs, obstetric hemorrhage and pregnancy-induced hypertension as most important causes of death is important to reduce maternal mortality in Mozambique. We found a fair level of agreement between COD assigned by InterVA and physician review at an individual case level, particularly for indirect causes of death.