Panel 3.

Counting maternal deaths in the COVID-19 era: the importance of health and demographic surveillance sites

This paper has highlighted an important gap in the empirical data on maternal mortality in the COVID-19 era; a lack of population-based data capturing births and maternal deaths that occur outside of health facilities. Civil and vital registration systems (CVRS) are very useful for providing national-level information on all births and deaths within a country, but recent estimates suggest that only 73% of countries have a comprehensive system capturing at least 90% of births and 68% a system that captures at least 90% of deaths [46]. Unsurprisingly, it is largely countries with relatively low levels of maternal mortality that have data from CVRS available. In the absence of CVRS, many countries have instead relied on data from health and demographic surveillance sites (HDSS) which collect detailed longitudinal data on geographically defined populations, including on births, death and migrations in the study population. HDSS have been widely used to quantify levels and causes of maternal death [47–49], although not without challenges. The frequency of data collection and the reliance on information reported for all household members by a household head has been found to impact on the quality of data on pregnancies and their outcomes [50,51]. Moreover, many of the HDSS cover relatively small geographical areas, leaving such individual HDSS unable to document changes in rare outcomes – such as maternal deaths – rigorously. The establishment of networks of HDSS, of which Professor Peter Byass was key player, have facilitated both comparative and pooled analyses of mortality in pregnant and postpartum women across Asia and Africa, overcoming some of the challenges posed by the small sample sizes [52]. By drawing on data from regular population-based HIV surveys across six HDSS, we were able to document that women living with HIV were nearly eight times more likely to die during the pregnancy and postpartum period compared with their uninfected counterparts [53]. Given their geographical reach, population-based coverage and wealth of historical data on levels and causes of mortality as well as on births, HDSS are uniquely placed to help us to understand the impact of COVID-19 pandemic on levels and causes of maternal mortality in some of the highest burden countries. There are already examples of HDSS which have adapted their data collection procedures in response to COVID-19 by, for example, conducting data collection over the phone rather than through household visits, and embedding screening for COVID-19 in their surveys [54]. Professor Byass repeatedly highlighted the importance of empirical data from HDSS [55,56], and our paper is a timely reminder of the value of these data.