Returns to childbirth technologies: Evidence from preterm births

Medical expenditures increased tremendously over the last few decades throughout the entire developed world. The growth of medical expenditures could be slowed down if the use of inefficient technologies is reduced, since there is substantial heterogeneity in treatment effects across patients, with some technologies highly effective for some populations while having little benefits for others (Chandra and Skinner, 2012). One such strategy is the expanded use of lower-cost physician extenders instead of higher-cost physicians (Fuchs, 1998).

In this paper, we examine the effects of increased use of physician extenders in the context of childbirth technologies. In particular, we investigate the impact of obstetrician/gynecologist (OB/GYN) supervision of birth, as opposed to midwife supervision, on the health (7-day and 28-day mortality and Apgar score) of low-risk newborns.

Simple estimations of the returns to OB/GYN supervision are likely to be biased as women expecting a complicated birth may give birth under the supervision of an OB/GYN. Our empirical approach eliminates this bias by exploiting a policy rule in the Netherlands that provides an exogenous source of variation in the type of medical professional supervising a low-risk birth. In the Dutch obstetric care system, women without known medical risk factors (low-risk women) have their pregnancies and births supervised by a midwife (who is prohibited by law from performing any medical intervention), with no OB/GYN present. However, if the onset of labor is premature (before 37 completed gestational weeks), the woman is generally referred to an obstetrician who will supervise the birth. The discontinuity in the probability of being treated by an OB/GYN at gestational week 37 generated by this “week-37 rule” motivates our identification strategy – a regression discontinuity design.

Using data on the universe of births in the Netherlands between 2000–2008, we show that the week-37 rule generates an almost 60 percentage point increase in the probability that a spontaneous low-risk birth is supervised by an OB/GYN as the 37-week threshold is crossed from the right. We confirm that the variation in OB/GYN supervision generated by the week-37 rule is as good as random in two ways. First, we show that there are no heaps in the frequency of births around the cutoff. Second, we show that the distribution of a wide range of covariates is smooth around the week-37 cutoff. We proceed to estimate the causal effect of obstetrician supervision on infant health outcomes by using the variation induced by the week-37 rule in an instrumental variable (IV) framework. Despite the substantial variation in OB/GYN supervision, our results indicate that average newborn health outcomes are remarkably similar across the week-37 cutoff. Our IV estimates of the OB/GYN effect are consistently insignificant and of the wrong sign, indicating no health benefits from obstetrician supervision. Our results point to potential cost savings from increased use of midwifery care for low-risk births that are close to the week-37 cutoff.