Displacements in reported date of birth and differences in age at death in surveys of child mortality: a record linkage study in Guinea-Bissau

Child mortality data in low- and middle-income countries are often derived from survey data prone to imprecision due to recall. Such imprecisions can affect the accuracy of date of birth (DOB) and age at death (AAD). Bandim Health Project runs two Health and Demographic Surveillance Systems (HDSSs) collecting prospective information on pregnancies, births, and deaths in Guinea-Bissau. In a survey, HDSS resident women were interviewed on their births and survival status of their children born within 5 years prior to the interview. We matched child records in the survey data to their corresponding records in the HDSS data based on sex, name, birth order, twinning status, and sibling names. We compared the risk and magnitude of displacement in DOB and AAD in regression models to identify risk factors for displacement. A total of 7679 out of 9960 survey-reported live births (78%) were matched to HDSS-reported births. Displacement in month or year of birth was observed for 9%, with a median displacement of 30 days backwards. Child death was associated with larger risk of displacement in DOB (adjusted risk ratio: 5.16 (95% confidence interval: 4.44–5.99)) and greater magnitude of displacement. A longer recall period was also associated with a greater risk of displacement. Among 339 children who had died, 68% had a discrepancy in AAD with 34% of these exceeding 1 week. Discrepancies were common and more likely for longer recall periods and children who had died. Hence, the interpretation of temporal mortality patterns from survey data may be compromised.