Methods/Statistics
Obtaining a Probability Sample of a Pregnancy Cohort of Births Michael Elliott* Michael Elliott Jean Kerver Alexa Drew Kaitlyn Watson Breanna Kornatowski Gwendolyn Norman Glenn Copeland Eva Leissou Terri Ridenour Shonda Kruger-Ndiaye Tengfei Ma Douglas Ruden Charles Barone Daniel Keating Robert Sokol Christine Johnson Nigel
The Michigan Archive for Research on Child Health (MARCH) study produced a probability sample of Michigan births between 2017 and 2023, with data collection beginning at first prenatal visit and continuing up to age 4. Birth certificate data were used to create a sampling frame of hospitals and associated obstetric clinics, from which a probability-proportional-size sample of 10 hospitals was drawn. Approximately 100 pregnancies were then recruited in clinics serving each sampled hospital, yielding an approximately equal probability sample of 1,021 births. This sample was supplemented with 109 births from a certainty selection of a Flint, MI hospital, for a total sample of 1,130. The resulting response rate was extremely high, with 100% of sampled hospitals and 65% of sampled clinics participating. Comparing the resulting sample with all 2017-2023 Michigan births showed close correspondence with respect to birth outcomes (birthweight, gestational age, Apgar scores, gestational diabetes) and mothers’ demographics (age, race, education, marital status), with underrepresentation of Hispanic ethnicity and overrepresentation of reported smoking. Given the recent failures of two major prospective birth cohorts (the US National Childrens’ Study and the UK Life Study), our work shows a way forward for obtaining probability samples for pre- and post-natal studies of births