Joelle Abramowitz, Ph.D.
University of Michigan
Background: Over 1977-2001, 15 states mandated insurance coverage for infertility treatments in some form. Extensive work has found effects of the mandates on births for women at older childbearing ages, the utilization of treatment, and multiple births, but less work has explored how the mandates affected maternal health outcomes.
Objectives: The paper considers the extent to which mandated insurance coverage of infertility treatment in the United States affected maternal mortality.
Research Design: Retrospective analyses use 1981-1998 Vital Statistics data on births and maternal mortality to examine variation by state and over time in mandated health insurance coverage of infertility treatment.
Results: Results suggest that over 1981-1998, White women ages 35-49 living in states that mandated insurance coverage of infertility treatment had 20 percent lower maternal mortality rates relative to women living in states that did not adopt mandates.
Conclusions: The results of this analysis are important for quantifying effects of existing policies mandating health insurance coverage of infertility treatment as well as considering potential effects of future policies. Future work examining the extent to which different factors contribute to this paper’s findings is warranted.
Impaired fecundity, which includes problems getting pregnant and problems carrying a baby to term, affects many women: over 2006–2010, 10.9 percent of women 15–44 years of age and 21.2 percent of currently married, childless women reported having impaired fecundity. Infertility treatments can be used to help women and couples with fertility problems conceive a child and range from counseling and fertility testing to surgical procedures, such as in vitro fertilization (IVF). The first successful IVF procedure in the United States was performed in 1981, and since that time, such procedures have become much more effective and widely used. Of women 15–44 years of age surveyed over 2006–2010, 11.9 percent had ever received any infertility services, and 1.2 percent had ever undergone artificial insemination procedures.1
While the use of infertility treatments has increased over time, it has been limited by price: procedures are expensive and may require multiple attempts to be successful., These procedures are generally not covered by insurance unless insurers are required by a state mandate to provide insurance plans that cover them. Over 1977-2001, 15 states mandated insurance coverage for infertility treatments in some form. In general, the mandates require coverage of infertility treatment at the same rates as basic health care services or pregnancy-related services.
Extensive work using difference-in-differences approaches, which compare outcomes in states that adopted mandates to those in states that did not adopt mandates before and after the adoption of the mandates, has found effects of the mandates on births for women at older childbearing ages, the utilization of treatment,3,,, and multiple births.6,, These findings suggest that the mandates were effective at increasing access to infertility treatment.
Less work has explored how the mandates affected maternal health outcomes. This paper will examine how mandated insurance coverage of infertility treatment affected maternal mortality. The mandates have the potential to affect maternal mortality as a result of a change in the number of women giving birth as well as a change in the type of procedures chosen. If the mandates lead to more women giving birth, the total number of maternal deaths is likely to increase. The maternal mortality rate, the number of deaths as a share of births, may increase or decrease if the additional women giving birth are more or less at risk, respectively, than the average woman giving birth. The additional women giving birth may be at greater risk if births using infertility treatment or to women facing infertility are riskier. However, the additional women giving birth may be at less risk if they are more likely to have health insurance coverage or higher socio-economic status and if these are associated with better general health or better medical care. If the mandates result in women choosing different types of procedures, we may see increases or decreases in the maternal mortality rate. One such example of a switch to a less risky procedure would be to choose to implant fewer embryos in a given cycle, with the expectation that should the procedure be unsuccessful, the cost of an additional cycle would be covered by insurance. To this point, comprehensive mandates were found to have resulted in significantly less aggressive treatment.7
2.1 Classifying the Mandates
Table 1 lists states that mandated insurance coverage for infertility treatment in some form through 2001. There could be some concern that states that adopted mandates were fundamentally different from those that did not. Anecdotal evidence suggests that states’ adoption of the mandates often comes about due to idiosyncratic factors and was part of a greater trend in the adoption of insurance mandates in the United States between the 1970s and 1990s, and lobbying in favor of the infertility insurance mandates has generally been coordinated at the national, rather than state, level.
To consider how mandated insurance coverage of infertility treatment affected maternal mortality, the analysis uses National Vital Statistics Natality Data and Multiple Cause-of-Death Mortality Data. Population estimates come from the U.S. Census Bureau. The analysis is restricted to 1981-1998. 1981 was the first year included in the analysis to exclude a period of declining maternal mortality through the 1970s, which leveled off in the early 1980s. Data after 1998 were excluded due to changes in classifying maternal death in 1999 and to recording maternal mortality on the death certificate in 2003. Due to these considerations, the analysis excludes the mandates of New Jersey and Louisiana, adopted in 2001, and includes them as control states.
The analysis focuses on women ages 35-49 since women can face infertility at any age, but the incidence of infertility increases with age, and the majority of women using assisted reproductive technology are age 30 or older.4 In addition, age has been shown to be a risk factor for increased maternal mortality. This study specifically examines older women of childbearing age because they are the most likely to use infertility treatment and the most at risk of maternal mortality, and it follows that they also have the potential for the most effect of such procedures on maternal mortality and the potential for the most effect of the mandates to mediate this relationship. Results of analyses examining outcomes for women ages 25-30 and 30-35 did not show any statistically significant effects of the mandates on maternal mortality for these age groups.
To estimate the effects of the mandates on maternal mortality, it is necessary to identify the effects of the mandates separately from other factors that might affect maternal mortality occurring over the same period. Since the mandates apply only to insurers in states that adopted mandates, this analysis uses a difference-in-differences approach to compare outcomes for women living in states that adopted mandates (treatment states) to outcomes for women living in states that did not adopt mandates (control states), before and after the adoption of the mandates. This approach assumes that women living in control states face similar trends in maternal mortality as the women living in treatment states and thus will account for time-varying factors that would have resulted in different rates of maternal mortality after the adoption of the mandates for the states that adopted them.
To control for other factors that might affect the relationship between the mandates and maternal mortality, the paper follows the approach of Schmidt (5) and estimates a difference-in-differences ordinary least squares regression model. The main outcome of interest is the maternal mortality rate, defined as the number of deaths per 100,000 births; additional analyses examine effects on birth rates (births per million women) and maternal death rates (maternal deaths per million women). Outcomes are examined for age group-state-year cells. The effect of the mandates is estimated by controlling for whether the state had a mandate in place for at least two years, which represents the associated difference in the outcome of interest for women living in states with mandates at least two years after their adoption relative to those living in states without mandates, or those living in states that would eventually adopt mandates, prior to their adoption. Consistent with the literature, a two-year lag is used to allow for time for pursuing infertility treatment and becoming pregnant, though results are robust to estimating the effect of the mandates in the year of their adoption.5 Additional controls include state dummy variables to account for differences in time-invariant state characteristics, year dummy variables to account for differences in national-level characteristics over time, dummy variables for five-year age groups to account for differences in baseline characteristics associated with age, and the state-year-level unemployment rate, female labor force participation rate, log median weekly earnings, and log 10th percentile weekly earnings to account for differences in economic conditions. In all regressions, the data are weighted to be population-representative of the female population for each state and year. To control for serial correlation among the outcomes and the mandates, the standard errors of the estimates are estimated using robust methods with clustering at the state level, which corrects for such correlation.
Specifications are run separately for Whites and Blacks; other races are not examined in this analysis due to their rare incidence of maternal mortality associated with small population sizes. Since the mandates vary in whether they require that all plans offered by insurers cover infertility treatment, known as a “mandate to cover,” or only that insurers offer some plans that cover infertility treatment, known as a “mandate to offer,” as well as specifically requiring that IVF be covered or be excluded from coverage, additional specifications consider effects by mandate characteristics. Further specifications also examine effects by cause of maternal death.
Figure 1 shows the number of deaths per 100,000 births by whether the state would adopt a mandate as of 1998 for women ages 35-49 over 1981–1998. The figure suggests that while maternal mortality rates were initially higher in mandate-adopting states, there was a sharp decrease in maternal mortality in mandate-adopting states following 1985, the period of mandate adoption, that was not apparent in states that did not adopt mandates.
Looking more closely at changes in maternal mortality associated with the mandates, Table 2 shows maternal mortality rates across mandate-adopting states and states that did not adopt mandates, before and after their adoption. Since most states did not adopt mandates until 1985 or later and most mandates considered in this analysis were adopted by 1990, Table 2 compares mean maternal mortality rates in states that adopted mandates and those that did not, before (pre-1985) and after (post-1990) their adoption. The 1985-1990 period is excluded since it covers the period of mandate adoption. Consistent with Figure 1, Table 2 shows that for both Whites and Blacks, maternal mortality rates were initially higher in states that would adopt mandates. Table 2 further shows that in the period from before to after mandate adoption, for both Whites and Blacks, maternal mortality rates fell in all states. However, for both Whites and Blacks, maternal mortality rates fell more in states that adopted mandates as compared to states that did not. This “difference-in-differences” estimate suggests 2.3 fewer deaths for Whites and 4.3 fewer deaths for Blacks associated with the mandates.
While Table 2 compares mean maternal mortality rates by mandate status over broad time periods, the difference-in-differences ordinary least squares regression model results provide estimates that account for year variation in the adoption of the mandates and control for other factors that might affect the relationship between the mandates and maternal mortality. These results, presented in Table 3, show the regression estimates for the effect of the mandates, the equivalent of the difference-in-differences estimates in Table 2, but accounting for year variation in the adoption of the mandates and controlling for other factors. The results suggest that, controlling for other factors, the mandates were associated with statistically significantly lower maternal mortality rates for Whites, but not for Blacks, for which the effect is smaller in magnitude and less precisely estimated as the standard errors of the estimate are substantially larger due to smaller sample sizes and more variation in outcomes. For Whites, the mandates were associated with 3.4 fewer deaths per 100,000 births (shown in Table 3), a 20 percent decrease from the mandate-state mean of 16.9 deaths (shown in Table 2) before mandate adoption (3.4 deaths/16.9 deaths). Since the effects of the mandates do not appear to be significant for Blacks, consistent with the literature analyzing the effects of the mandates on use of fertility treatment and likelihood of a first birth,3, results of further specifications will only be presented for Whites.
To understand whether different characteristics of the mandates are associated with different effects on maternal mortality, the paper estimates the main specification including only states that did not adopt mandates and states adopting mandates meeting each specified characteristic. Results of these analyses for White women, presented next in Table 3, are suggestive that more comprehensive mandates, and in particular, those covering IVF, are associated with lower maternal mortality.
Shedding light on the channels through which the mandates might have led to lower maternal mortality rates, Table 4 shows results of the regression estimates for the effect of the mandates by ICD-9 classifications of cause of maternal death for White women. These results suggest meaningful and significant effects of the mandates only for direct obstetric causes (e.g., eclampsia and pre-eclampsia, hemorrhage). The results do not show meaningful effects of the mandates on rates of death as a result of pregnancy with an abortive outcome or indirect obstetric causes (e.g., conditions aggravated by the physiologic effects of pregnancy, such as cardiac diseases or influenza).
Considering drivers of the lower maternal mortality rates for White women, further results in Table 4 show the mandates associated with a greater number of births per population, but do not show the mandates associated with a meaningfully different number of maternal deaths per population. Since the results suggest the mandates were associated with approximately 3.4 fewer deaths per 100,000 births to White women ages 35-49 per year and in 1998, there were approximately 418,218 births to White women ages 35-49, the findings suggest approximately 14 fewer maternal deaths per year occurred than would be expected in the absence of the mandates (3.4 deaths/100,000 births X 418,218 births).
To better understand the evolution of the effects over time, Figure 2 shows effects of the mandates by the number of years since the mandate was adopted, with the year prior to adoption as the reference year. The black line denotes coefficient estimates; gray lines denote 95 percent confidence intervals. The figure shows no statistically significant effect of the mandates in the years prior to their adoption, but does show significant effects in the year of and immediately after adoption and then increasing in magnitude four or more years after adopting a mandate. This result is not surprising as there may be a lag in the time that a mandate is adopted, a woman seeks treatment, and ultimately becomes pregnant and gives birth as well as time for awareness of the mandates to spread.
This paper found that White women ages 35-49 living in states that mandated insurance coverage of infertility treatment had lower maternal mortality rates, related to an increase in the birth rate, but no proportional increase in the maternal death rate, relative to women living in states that did not adopt mandate as compared to prior to mandate adoption. Results may be driven by the characteristics of the additional women choosing to use infertility treatment or by women choosing alternative treatments, e.g. elective single embryo transfer, as a result of insurance coverage of infertility treatment. Additionally, the results may be driven by greater pregnancy monitoring of women using infertility treatment.
Results suggest the mandates were associated with statistically significantly different maternal mortality rates for Whites, but not for Blacks. We may see differential effects of the mandates on maternal mortality by race since baseline maternal mortality rates differ by race and also because access to employer-sponsored insurance coverage differs by race. Along these lines, work on the mandates and racial disparities found that infertility was more common for non-Hispanic Black women and non-Hispanic other race women than for non-Hispanic White women, but found no evidence that the racial disparities are ameliorated by the mandates, suggesting differential effects of the mandates on the use of infertility treatment by race.14
The results of this analysis are important for quantifying effects of existing policies mandating health insurance coverage of infertility treatment as well as considering potential effects of future policies. In Australian and Dutch contexts, research has found women using IVF to be healthier and have lower overall mortality rates than the general population, but to also have higher maternal mortality rates than the general population. The finding that the U.S. mandates are associated with lower maternal mortality could reflect the health insurance landscape in the United States. Since the mandates necessarily function through private health insurance channels, effects could reflect the characteristics of women covered by private insurance or could reflect how insurance rules can induce changes in behavior such as opting for riskier or less risky procedures.7 Results suggest the need for more careful consideration of the more far-reaching effects of health insurance policies before their implementation as well as study of the implementation of new policies.
There are several limitations to this analysis. This analysis does not identify the extent to which alternate phenomena may contribute to the paper’s findings. Future work examining the extent to which different factors contribute to this paper’s findings is warranted. In addition, it is important to note that while this paper’s analysis focuses on the effects of mandates implemented over 1985-1991, effects of these and future mandates may change over time with technological advances and changes in treatment practices.
 Centers for Disease Control and Prevention, National Center for Health Statistics. Key statistics from the National Survey of Family Growth. 2014. Available at: http://www.cdc.gov/nchs/nsfg/key_statistics/i.htm. Accessed October 21, 2019.
 Bitler M, Schmidt L. Utilization of infertility treatments: The effects of insurance mandates. Demogr 2012;49:125–49. doi: 10.1007/s13524-011-0078-4.
 Centers for Disease Control and Prevention, American Society for Reproductive Medicine, Society for Assisted Reproductive Technology. 2010 assisted reproductive technology national summary report. Atlanta, GA: U.S. Department of Health and Human Services. 2012. Available at: http://www.cdc.gov/art/ART2010/PDFs/ART_2010_National_Summary_Report.pdf. Accessed October 21, 2019.
 Schmidt L. Effects of infertility insurance mandates on fertility. J Health Econ 2007;26:431–46. doi: 10.1016/j.jhealeco.2006.10.012.
 Bundorf MK, Henne M, Baker L. Mandated health insurance benefits and the utilization and outcomes of infertility treatments [Internet]. (Working Paper No. 12820), National Bureau of Economic Research, Inc. 2007. Available at: http://www.nber.org/papers/w12820.pdf. Accessed October 21, 2019.
 Hamilton BH, McManus B. The effects of insurance mandates on choices and outcomes in infertility treatment markets. Health Econ 2012;21:994–1016. doi: 10.1002/hec.1776.
 Bitler M. Effects of increased access to infertility treatment on infant and child health: Evidence from health insurance mandates. NBER/IZA/University of California-Irvine: Unpublished Manuscript. 2008.
 Buckles KS. Infertility insurance mandates and multiple births. Health Econ 2013;22:775–89. doi: 10.1002/hec.2850.
 Lahey JN. The efficiency of a group‐specific mandated benefit revisited: The effect of infertility mandates. J Policy Analysis Management 2012;31:63-92. doi: 10.1002/pam.20616.
 Leridon H. Can assisted reproduction technology compensate for the natural decline in fertility with age? A model assessment. Hum Reproduction 2004;19:1548–53. doi: 10.1093/humrep/deh304.
 Centers for Disease Control and Prevention, National Center for Health Statistics, National Vital Statistics System. 2007 Births and Deaths. Available at: https://mchb.hrsa.gov/whusa10/hstat/mh/pages/237mm.html. Accessed December 20, 2019.
 Bertrand M, Duflo E, Mullainathan S. How much should we trust differences-in-differences estimates? Q J Econ 2004;119:249-75.
 Bitler M, Schmidt L. Health disparities and infertility: Impacts of state-level insurance mandates. Fertility Steril 2006;85:858–865. doi:10.1016/j.fertnstert.2005.11.038.
 Berchick E R, Barnett J C, Upton R D. Health Insurance Coverage in the United States: 2018. Current Population Reports, P60-267(RV). Washington, DC: U.S. Government Printing Office. 2019.
 Venn A, Hemminki E, Watson L, Bruinsma F, Healy D. Mortality in a cohort of IVF patients. Human Reproduction. 2001;16:2691–6. doi: https://doi.org/10.1093/humrep/16.12.2691.
 Braat DDM, Schutte JM, Bernardus RE, Mooij TM, van Leeuwen FE. Maternal death related to IVF in the Netherlands 1984–2008. Hum Reproduction 2010;25:1782–6. doi: 10.1093/humrep/deq080.