BACKGROUND: The National Children's Study (NCS) was established as a national probability sample of births to prospectively study children's health starting from in utero to age 21. The primary sampling unit was 105 study locations (typically a county). The secondary sampling unit was the geographic unit (segment), but this was subsequently perceived to be an inefficient strategy. METHODS AND RESULTS: This paper proposes that second-stage sampling using prenatal care providers is an efficient and cost-effective method for deriving a national probability sample of births in the US. It offers a rationale for provider-based sampling and discusses a number of strategies for assembling a sampling frame of providers. Also presented are special challenges to provider-based sampling pregnancies, including optimising key sample parameters, retaining geographic diversity, determining the types of providers to include in the sample frame, recruiting women who do not receive prenatal care, and using community engagement to enrol women. There will also be substantial operational challenges to sampling provider groups. CONCLUSION: We argue that probability sampling is mandatory to capture the full variation in exposure and outcomes expected in a national cohort study, to provide valid and generalisable risk estimates, and to accurately estimate policy (such as screening) benefits from associations reported in the NCS.
BACKGROUND: The National Children's Study (NCS) was established as a national probability sample of births to prospectively study children's health starting from in utero to age 21. The primary sampling unit was 105 study locations (typically a county). The secondary sampling unit was the geographic unit (segment), but this was subsequently perceived to be an inefficient strategy. METHODS AND RESULTS: This paper proposes that second-stage sampling using prenatal care providers is an efficient and cost-effective method for deriving a national probability sample of births in the US. It offers a rationale for provider-based sampling and discusses a number of strategies for assembling a sampling frame of providers. Also presented are special challenges to provider-based sampling pregnancies, including optimising key sample parameters, retaining geographic diversity, determining the types of providers to include in the sample frame, recruiting women who do not receive prenatal care, and using community engagement to enrol women. There will also be substantial operational challenges to sampling provider groups. CONCLUSION: We argue that probability sampling is mandatory to capture the full variation in exposure and outcomes expected in a national cohort study, to provide valid and generalisable risk estimates, and to accurately estimate policy (such as screening) benefits from associations reported in the NCS.
Authors: Jean M Kerver; Michael R Elliott; Gwendolyn S Norman; Robert J Sokol; Daniel P Keating; Glenn E Copeland; Christine C Johnson; Kendall K Cislo; Kirsten H Alcser; Shonda R Kruger-Ndiaye; Beth-Ellen Pennell; Shobha Mehta; Christine L M Joseph; Nigel Paneth Journal: Paediatr Perinat Epidemiol Date: 2013-05 Impact factor: 3.980
Authors: Joseph B Stanford; Ruth Brenner; David Fetterer; Leslie Palmer; Kenneth C Schoendorf Journal: BMC Med Res Methodol Date: 2015-09-24 Impact factor: 4.615