International Journal of Population Data Science (Dec 2020)

Estimating Maternal Mortality Rates During The 1918 Flu Using Birth to Death Linkage

  • Peter Christen,
  • Eilidh Garrett,
  • Beata Nowok,
  • Alice Reid,
  • Lee Williamson,
  • Chris Dibben

DOI
https://doi.org/10.23889/ijpds.v5i5.1534
Journal volume & issue
Vol. 5, no. 5

Abstract

Read online

Introduction The Digitising Scotland project (https://digitisingscotland.ac.uk/) has transcribed all Scottish birth, death, and marriage certificates from 1855 to 1974. The linkage of these data will provide formidable challenges for linkage experts and a multitude of opportunities for health and social science researchers. Objectives and approach We linked birth between November 1916 and December 1923 to death of women aged 15 to 49 who died between January 1917 and December 1923. We only linked a death up-to 42 days after a birth. We compared parent names with those of the deceased and her spouse, as well as their address, using string matching functions. Given the lack of ground truth data, we conducted a sensitivity analysis using different similarity thresholds to classify birth linked to death as matches (if their similarity was at least a given threshold), or otherwise as non-matches. We then calculated the maternal mortality rate (MMR) on a monthly basis for January 2017 to July 2018 (the pre-flu period), and October 1918 to March 1919 (the flu period) as the number of matched birth to death certificates divided by the average monthly number of birth in the month of death and the previous month. Results Puerperal risk of death during this period was much higher than for the comparable female age group during the different phases of the 1918 flu. During the October 1918 wave, puerperal women were at greater risk than women of the same age, while the risk dropped below the expected in the summer of 1919. We carried out a sensitivity analysis by linking with different similarity thresholds and found our findings were robust to these decisions. Conclusions We have shown how a large and complex data collection can be successfully linked, resulting in new opportunities for various studies in the health and social sciences.