Author Reply

This is the authors' abstract. We don't add key points for this paper.

Dr. Deneux-Tharaux [1] criticises our use of death certificate data for illustrating the merits of classification systems based on multiple causes of maternal death [2]. As acknowledged in our study and elsewhere [2], death certificates underestimate maternal death rates and are less accurate with regard to cause-of-death information (compared with expert medical reviews). However, we were dismayed by Dr. Deneux-Tharaux's blanket (and somewhat solipsistic) dismissal of a less-than-ideal system for maternal death ascertainment that remains extant in many high-income countries. The objective of our study was to examine differences between two cause-of-death classification schemes, and we did this using a convenient dataset. We are not aware of evidence suggesting that cause-of-death inaccuracies in death certificates differ in magnitude for different leading obstetric causes of death, which would be necessary for invalidating the results of our study. Our study was motivated by a conundrum presented by the UK Confidential Enquiry into Maternal Deaths and the Dutch Audit Committee for Maternal Mortality and Morbidity [2]. If a woman with preeclampsia dies of an ensuing haemorrhage, how does one reconcile the underlying cause of death assigned in the Netherlands (viz. preeclampsia) with the underlying cause of death assigned in the United Kingdom (viz. haemorrhage)? These august committees concluded their discussion with a plea for suggestions to help resolve the impasse. This discussion may benefit further from reference to two issues related to the attribution of effects based on a singular underlying cause versus multiple causes. The World Health Organization (WHO), which equates severe maternal morbidity (SMM) with ‘near miss’, recommends that the ‘same classification of underlying causes [be] used for both maternal deaths and near misses’ [3]. Nevertheless, most SMM studies in the literature ignore this long-standing WHO injunction [4]. SMM rates and SMM component rates are estimated with women being assigned one or more severe maternal illnesses, rather than a singular underlying illness; studies show an exponential increase in case fatality among women with 1, 2, 3 or more severe maternal illnesses [4]. The second issue relates to the performance of cause-of-death classification systems, given temporal changes in causes of death. Increases in advanced maternal age and obesity have resulted in a rise in maternal multi-morbidity, and in the proportion of chronic disease-associated maternal deaths [5]. If a pregnant woman with pre-existing hypertension, diabetes or other chronic diseases develops a fatal case of severe preeclampsia, embolism or stroke, the underlying cause of death assigned would likely be one of the latter severe pregnancy complications. However, a focus on the singular underlying cause of death would obscure the rising rate of chronic disease-associated maternal deaths, irrespective of whether the cause-of-death information was obtained from death certificates or expert review. On the other hand, a multiple cause-of-death analysis would show the rise in chronic disease-associated maternal deaths [5]. In conclusion, we reiterate that cause-of-death assignment based on multiple causes of death is consistent with contemporary multi-factorial models of causation and will improve the effectiveness of clinical and public health programs aimed at maternal death prevention. All authors contributed to the conception, drafting and revision of this letter. All authors approved the final version for submission. The authors declare no conflicts of interest. There is no data presented in this Letter to the Editor (response).