Identifying national and sub-national data of maternal cause of death: challenges in epidemiologic searching of bibliographic databases




Poster session 2


Monday 17 September 2018 - 12:30 to 14:00

All authors in correct order:

Skidmore B1, Allen T2, Chou D3, Ganatra B3, Garritty C4, Henschke N5, Moller AB3, Pestridge C5, Say L3, Villanueva G5
1 Independent Information Specialist, Canada
2 Library and Information Networks for Knowledge, WHO, Switzerland
3 Department of Reproductive Health and Research, WHO, Switzerland
4 Ottawa Hospital Research Institute, Canada
5 Cochrane Response, UK
Presenting author and contact person

Presenting author:

Becky Skidmore

Contact person:

Abstract text
Identifying accurate and robust information on causes of maternal death beyond civil registration data requires looking for information in population- and facility-based studies, specialized surveys, registries and elsewhere. Bibliographic databases (BDs) offer one source but current epidemiologic search filters are suboptimal. We explore development of a robust strategy for searching BDs for epidemiologic information on maternal death.

To identify national and sub-national data of maternal causes of death, particularly from low- and middle-income countries (LMICs), published in BDs.

The multidisciplinary research team, including two information specialists, examined previous, similar strategies. We discussed the new strategy extensively to ensure sensitivity for current needs and future updates. We iteratively tested vocabulary pertaining to the study population, death by any cause in pregnant/postpartum women, and associated epidemiologic information. We resolved challenges in distinguishing physical death from its epidemiologic aspects (i.e. mortality). When testing showed missed potentially relevant references using the epidemiologic filter alone, we applied a parallel LMIC filter to identify information from these regions. We searched multiple BDs (MEDLINE, Embase, Popline, Web of Science, Global Index Medicus). We deemed Russian and Chinese BDs to be critical, and engaged teams with requisite language and skills for these searches. We removed animal-only and male-only records from results, where possible, to contain volume.

Our topic required multiple BDs and extensive vocabulary to identify the desired data. We deduplicated ~110,000 records and are screening ~54,000. We will explore the distribution of included records, particularly for LMICs, in the various BDs.

A robust, comprehensive search on worldwide maternal mortality requires extensive testing and understanding of vocabulary pertaining to the population and epidemiologic data. It requires searching multiple BDs, including non-English ones. A LMIC filter is required, demonstrating the need for multi-pronged searching for epidemiologic data.

Patient or healthcare consumer involvement:
Hidden data on causes of maternal death, especially for LMICs, are often provided by women, households, or community (e.g. verbal/social autopsies). We incorporated vocabulary and methods not commonly associated with epidemiologic searches to identify these unique sources of data.

Relevance to patients and consumers: 

This initiative provides women evidence to advocate for informed decisions on how to improve maternal health programs. It highlights underestimated or hidden causes of death, thus promoting healthcare enhancements. This hidden information is often generated by women, households, or community (e.g., verbal/social autopsies), demonstrating uniquely important roles played by community in providing data. Identifying and ranking causes of maternal death lead to new programmes/strategies for better health services provision. This empowers patients to actively engage in cause prevention through education and development of patient-centred outcomes.