Table 5 Recommended protocols for collection of high-quality vital events data for mortality estimation in population-based birth cohort studies

1

Identify a high proportion or representative sample of pregnancies (or live births) in a geographic area to avoid selection bias associated with place of delivery

2

Enroll pregnant women, rather than live births at the time of delivery, to capture more live births and early deaths and reduce under-reporting of stillbirths and misclassification of neonatal deaths

3

Consider the impact of an open or closed cohort on mortality estimates given patterns of out-migration and in-migration behaviors in the study population

4

Minimize inclusion/exclusion criteria restrictions (e.g., exclusion of multiple births) on the study population for which vital event data is collected to avoid selection bias and reduce impact on generalizability associated with special populations

5

Attempt to capture vital information on pregnancy outcomes as quickly as possible (i.e., on the day of birth) after the occurrence of the birth outcome to avoid missing deaths (even if immediate follow-up is not required for the study’s primary aim)

6

Understand local reasons for misclassification of stillbirths and neonatal deaths and utilize staff training and study protocols to reduce this bias

7

Train study staff to avoid common epidemiologic biases and data collection errors that affect mortality estimates, such as reporting biases (e.g., recall bias or bias due to stigma of reporting a death) or date heaping

8

Reduce missing birth outcomes and infant vital status data by closely tracking participants through frequent visits, using digital technologies if possible, to reduce selection bias associated with loss to follow-up

9

Utilize post hoc analytical techniques to explore for and report on selection and reporting biases, such as date heaping graphs or comparison of participants fully followed vs. participants lost to follow-up