Skip to main content
. Author manuscript; available in PMC: 2017 Jul 18.
Published in final edited form as: Lancet. 2015 Nov 13;387(10017):462–474. doi: 10.1016/S0140-6736(15)00838-7

Table 3.

Overview of data sources that are included in the maternal mortality model and the systematic and random errors associated with the observations for each source type that are taken into account in the data model.

Data source type Information used to construct maternal mortality estimates Assumptions about systematic errors (reporting issues which result in biases) Assumptions about random errors
 A1. Civil registration vital statistics, ICD 9 (VR-ICD9) PM (including late maternal deaths)

  1. Misclassification of maternal deaths and/or incompleteness

  2. Inclusion of late maternal deaths

 Observations are subject to stochastic errors
 A2. Civil registration vital statistics, ICD 10 (VR-ICD10) PM Misclassification of maternal deaths and/or incompleteness
 B. Specialized studies Maternal deaths are used a model input if a rigorous assessment of the total number of all- cause deaths was also carried out in the study to confirm that all deaths to women of reproductive age were captured. Otherwise, the PM or MMR is used*. No systematic errors
 C. Other data sources reporting on maternal mortality PM or MMR* Underreporting of maternal deaths Observations may be subject to sampling, stochastic and/or additional random error
 D. Other data sources reporting on pregnancy-related mortality, e.g. through sibling histories Pregnancy-related PM or pregnancy- related MMR*

  1. Underreporting of pregnancy-related deaths

  2. Over-reporting of maternal deaths due to the inclusion of pregnancy-related deaths that are not maternal
*

PM takes precedence over the MMR.

Additional definitions used: Stochastic errors refer to differences between observed PMs and expected PMs due to the randomness associated with the event of a maternal death, i.e., when considering the event of a maternal death as the outcome of a random variable with a Bernouilli distribution with the probability of a maternal death given by the expected PM. Sampling error arises in observed PMs that are obtained from samples that are a subset of the population, e.g. in surveys or sample registration systems. In addition to sampling or stochastic errors, observations may be due to additional random errors, which are non-systematic errors that may occur at any point during the data collection process, e.g. due to how a questionnaire was administered or due to data entry errors. Abbreviations: PM = proportion of all-cause deaths that are maternal; MMR = maternal mortality ratio; VR = vital registration.