by Christine Piette Durrance, Department of Public Policy, University of North Carolina at Chapel Hill and Melanie Guldi, Department of Economics, University of Central Florida
American Journal of Health Economics 1(3): 1-29, 2015
The United States experiences unusually high rates of premature birth and infant mortality relative to countries of similar economic development. In the United States in 2009, just over 12 percent of all babies were born prematurely (before 37 weeks). Of these, 70 percent were late preterm births (34–36 weeks) and about 16 percent were very preterm (before 32 weeks) (Martin et al. 2011). Moreover, Thirty-six percent of the infant deaths in the United States in 2007 were “preterm” related (Matthews and MacDorman 2011). Understanding the underlying causes and potential remedies is important since negative shocks to early child health influence both short- and long- run outcomes (Almond and Currie 2010). Interventions prior to birth, rather than early in childhood, may reduce the probability or the severity of the health shock and may be relatively cost effective. In our recent article in the American Journal of Health Economics, “Maternal Bed Rest and Infant Health”, we examine one such intervention, bed rest during pregnancy, which is frequently prescribed to reduce the likelihood of preterm birth.
Medical professionals considering prescribing antepartum bed rest or restricted activity for a patient face a trade-off. Bed rest may improve infant health and consequently decrease infant medical care costs. At the same time, it has an adverse effect on maternal health and potentially imposes other (pecuniary or nonpecuniary) costs on the mother both during and after pregnancy. Prior evidence suggests that bed rest is recommended in 18.2 percent of pregnancies (Goldenberg et al 1994). Using more recent data from the Pregnancy Risk Assessment Monitoring System (PRAMS), we report that 19.1 percent of women were recommended bed rest (of two days or more) during their pregnancy and 29.3 percent of women experiencing at least one medical issue during their pregnancies were given a recommendation.
Recent medical literature on the efficacy of bed rest is mixed, but tilts towards a negative or zero relationship between bed rest and infant health, suggesting that this intervention does not improve initial infant health status. This is not surprising given the difficulty of navigating a crucial threat to identification: conditions leading to bed rest recommendations are correlated with infant health. We overcome this hurdle by using a data set with detailed information on mothers and their children, the Pregnancy Risk Assessment Monitoring System (PRAMS), along with standard econometric methods. We begin by replicating the findings of much of the medical literature and show that a naïve regression will produce estimates indicating a negative relationship between bed rest and infant health. When examining the sample of mothers who are likely to be recommended bed rest, but some of whom are not, the estimates reveal a more nuanced relationship.
Among women who experienced a medical issue during pregnancy, we compare outcomes of infants born to mothers who are observationally similar yet differ in whether or not they are recommended bed rest. The credibility of our estimates hinges on the ability to control for the types of observable characteristics that physicians would rely upon when making a recommendation for bed rest or not. We argue that the most important observables to control for in this setting are the set of problems experienced by the mother during pregnancy, since they are often mentioned in the medical literature as key drivers of a bed rest recommendation. We show that a large number of other individual-level covariates that are often included in studies of infant health (including social and demographic variables) are less substantial predictors of a bed rest recommendation than pregnancy problems and that the bias from omitting the other individual-level variables is fairly small relative to the substantial bias that occurs when pregnancy problems are omitted from the analysis.
We use several approaches to estimate the effect of bed rest on birthweight and gestation. Our main results utilize OLS methods, controlling for maternal pregnancy problems. Our estimates for the typical infant health margins indicate a positive relationship between bed rest and the likelihood of low birth weight (<2500g) and prematurity (<37 weeks), which is consistent with prior medical literature that suggests bed rest is not helpful for and may actually harm infant health. Yet, when we examine more extreme margins of infant health—the margins likely to result in the most costly infant outcomes—our results show that bed rest decreases the incidence of very low birth weight (<1500g) by at least -15.4 percent and very premature outcomes (<33 weeks) by at least -7.7 percent. We also find a reduction for infant death. Finally, we also employ propensity score nearest neighbor matching, inverse probability weighting, and entropy balance techniques and show that our main findings are robust to model choice. In sum, our results suggest bed rest shifts initial health status upward along the potential birth weight distribution. Given these results, our back-of-the-envelope calculations suggest that when accounting for cost savings over both the short- and long- run, bed rest may be a more cost-effective treatment than post-birth interventions. Taken together, our findings suggest that in studies of infant health, special attention should be paid to the margin of infant health examined since standard low birth weight or prematurity margins may be misleading.
Almond, Douglas, and Janet Currie. 2010. “Human Capital Development before Age Five.” NBERWorking Paper No. 15827. http://www.nber.org/papers/w15827.
Goldenberg, Robert L., Suzanne P. Cliver, Janet Bronstein, Gary R. Cutter, William W. Andrews, and Stephen T. Mennemeyer. 1994. “Bed Rest in Pregnancy.” Obstetrics and Gynecology 84 (1): 131–36.
Martin, Joyce A., Brady E. Hamilton, Stephanie J.Ventura, Michelle J. K. Osterman, Sharon Kirmeyer, T. J.Matthews, and Elizabeth C. Wilson. 2011. “Births: Final Data for 2009.” National Vital Statistics Reports 60 (1): 1–72. http://www.cdc.gov/nchs/data/nvsr/nvsr60/nvsr60 01.pdf
Matthews, T. J., and Marian F. MacDorman. 2011. “Infant Mortality Statistics from the 2007 Period Linked Birth/Infant Death Data Set.” National Vital Statistics Reports 59 (6).