Effect of the Affordable Care Act on Breastfeeding Outcomes

Tami Gurley-Calvez is with the University of Kansas Medical Center, Kansas City. Lindsey Bullinger is with the School of Public and Environmental Affairs, Indiana University, Bloomington. Kandice A. Kapinos is with the RAND Corporation, Arlington, VA.

Correspondence should be sent to Kandice Kapinos, RAND Corporation, 1200 S Hayes St, Arlington, VA 22202 (e-mail: gro.dnar@sonipakk). Reprints can be ordered at http://www.ajph.org by clicking the “Reprints” link.

T. Gurley-Calvez contributed to the study design, carried out the initial analyses, drafted the initial article, and approved the final article as submitted. L. Bullinger contributed to the study design, reviewed and revised the article, and approved the final article as submitted. K. A. Kapinos conceptualized and designed the study, assisted with analyses, reviewed and revised the article, and approved the final article.

Abstract

Objectives. To assess how the 2012 Affordable Care Act (ACA) policy change, which required most private health insurance plans to cover lactation-support services and breastfeeding equipment (without cost-sharing), affected breastfeeding outcomes.

Methods. We used a regression-adjusted difference-in-differences approach with cross-sectional observational data from the US National Immunization Survey from 2008 to 2014 to estimate the effect of the ACA policy change on breastfeeding outcomes, including initiation, duration, and age at first formula feeding. The sample included children aged 19 to 23 months covered by private health insurance or Medicaid.

Results. The ACA policy change was associated with an increase in breastfeeding duration by 10% (0.57 months; P = .007) and duration of exclusive breastfeeding by 21% (0.74 months; P = .001) among the eligible population. Results indicate no significant effects on breastfeeding initiation and age at first formula feeding.

Conclusions. Reducing barriers to receiving support services and breastfeeding equipment shows promise as part of a broader effort to encourage breastfeeding, particularly the duration of breastfeeding and the amount of time before formula supplementation.

A large body of literature links breastfeeding to positive near-term child and maternal health outcomes. 1,2 In particular, breastfeeding has been associated with a reduced risk of acute otitis media, gastrointestinal infections, respiratory tract diseases, childhood obesity, and type 2 diabetes for normal-term infants. 2,3 For mothers, breastfeeding is associated with a lower incidence of type-2 diabetes, hyperlipidemia, hypertension, and cardiovascular disease, and lower risk of breast and ovarian cancers. 4,5 Evidence from randomized controlled trials of breastfeeding-support services suggests positive effects of breastfeeding on infant cognitive ability and health outcomes. 6

Although the percentage of mothers reporting any breastfeeding has increased over time from 74% in 2008 to 80% in 2014, only about half of mothers who initiate breastfeeding are still breastfeeding at 6 months with 22% breastfeeding exclusively through 6 months. Only about 30% breastfeed the full year recommended by the American Academy of Pediatrics. 1,7 Moreover, evidence suggests that mothers would like to breastfeed longer than they are able. 8

There have been several recent calls and policy efforts for more public health promotion of breastfeeding, including the 2011 US Surgeon General’s Call to Action to support breastfeeding, 9 and provisions in the Affordable Care Act (ACA) of 2010. Specifically, the ACA required large employers to provide reasonable break time and a private place for expressing breastmilk, and mandated insurance coverage of lactation-support services and equipment without cost-sharing for new health insurance policies beginning on or after August 1, 2012 (section 2713). The ACA’s coverage requirement applies to private, nongrandfathered insurance plans. Of note, this change largely did not apply to mothers insured through the Medicaid program.

Use of lactation-support services can increase a woman’s commitment to breastfeeding long term. 10–12 Breast pumps are critical for mothers who need to be separated from their infants for work or school and can also stimulate milk production to address low milk supply issues. 13–15 Breastfeeding promotion interventions, including the provision of lactation-support services, peer counselors, and breast pumps, have shown promising evidence of increasing breastfeeding rates. 10,11,13–16 These breastfeeding promotion interventions may mitigate challenges a new mother faces and increase her confidence in her ability to breastfeed.

Early empirical evidence of the ACA policy changes suggests there have been effects on breastfeeding behaviors. Workplace requirements increased the likelihood of exclusive breastfeeding at 6 months for mothers who gave birth in 2011 and 2012, but only 40% of the study population reported access to time and space accommodations. 17 Mandated private health insurance coverage of lactation-support services and breast pumps was associated with increased breastfeeding at the time of hospital discharge. 18 Little is known, however, about the effects of the ACA policy changes on sustained, or longer-term, breastfeeding and infant feeding practices. The ACA lactation-support coverage provision might affect the duration of breastfeeding if lactation consultants can assist with early difficulties, or if breast pumps make work and school transitions easier, allowing women to successfully follow through with intentions to breastfeed. 8,19,20 This study fills this gap in the literature by examining the effect of the ACA mandate, including coverage of lactation consultant visits and breast pumps, on longer-term breastfeeding behaviors.

METHODS

We used 2008 to 2014 data from the US National Immunization Survey (NIS), a nationally representative survey of mothers with children aged 19 to 35 months sponsored by the National Center for Immunizations and Respiratory Diseases (NCIRD) and conducted jointly by NCIRD and the Centers for Disease Control and Prevention (CDC). 21 Although the survey primarily addresses childhood immunizations, the questionnaire contains detailed questions on breastfeeding behavior. Response rates for the NIS household interview are around 82%. 21

We restricted the sample to children aged 19 to 23 months at the time of the survey (Figure A, available as a supplement to the online version of this article at http://www.ajph.org). In the public-use version of the NIS, child age is available in 3 categories: 19 to 23 months, 24 to 29 months, and 30 to 35 months. We excluded the 2 older age groups, as none of these children would have been born after the ACA mandate. Most children in the group aged 19 to 23 months in the 2014 NIS would have been born after the ACA mandate. Importantly, these children were born before the January 1, 2014, implementation of many other ACA provisions that might bias our results. To estimate the policy effect, we restricted the sample to children who were covered by private health insurance (treatment group) or Medicaid (best-available comparison group). Our final analytic sample size was 38 842 children.

Measures

Following earlier studies, we defined breastfeeding initiation as equal to 1 if the child was ever breastfed. In addition, we analyzed a broader range of longer-term outcome measures including breastfeeding duration (in months), time spent exclusively breastfeeding without supplementation (in months), an indicator variable equal to 1 if the child was ever formula fed, and the age at which the child received his or her first formula (in months). We note that the exclusive breastfeeding rate excludes time when the child was fed formula, but not other foods.

We adjusted for several maternal and household characteristics known to influence breastfeeding behaviors as well as state fixed effects to capture geographic differences in breastfeeding promotion campaigns as well as in breastfeeding attitudes. Control variables included child sex, race, and ethnicity; mother’s age, education, and marital status; and household participation in the Special Supplemental Nutrition Program for Women, Infants, and Children (ever) and number of children. Maternal employment information was not available.

Statistical Analyses

We used a quasi-experimental statistical design that allowed us to compare changes over time across “treatment” and “comparison” groups to isolate the impact of the policy from other impacts on outcomes that are attributable to secular changes. 22 Our treatment group included children covered by private health insurance, as the preventive services coverage mandate only applied to new private health insurance plans. Uninsured women and children or those covered by Medicaid were not directly impacted by this ACA provision and represent potential comparison groups for the analysis. The exception would be mothers newly eligible for Medicaid in states that expanded Medicaid early with eligibility thresholds near or greater than thresholds in the ACA provision (California, Connecticut, District of Columbia, Minnesota, New Jersey, and Washington). We address this possibility in our sensitivity analysis.

In this framework, we used the ACA policy change as a natural experiment or exogenous shock to test the effects of access to no-cost lactation-support services and breast pumps on child feeding outcomes. The statistical validity of this approach relies on the assumption that trends in breastfeeding behaviors were similar (parallel) for both the treatment and comparison group before the ACA policy changes. We used the control group (Medicaid mothers and children) to set the baseline of what would be expected for the treatment group (privately insured mothers and children) if there were no policy change. We examined trends in breastfeeding measures over time for privately insured children, those covered by Medicaid, and those without health insurance. We also formally tested for statistically significant differences in the prepolicy trends by comparing the slopes across the different groups. As discussed in the Results section, trends in breastfeeding outcomes among the Medicaid group were not statistically different from trends in the privately insured group, implying that the group of Medicaid mothers was the best-available comparison group for our estimation.

We used a regression-adjusted difference-in-differences approach, comparing adjusted breastfeeding rates of children covered by private health insurance to those covered by Medicaid. We estimated models with binary outcomes (ever breastfed and ever formula fed) with probit models. For the duration outcomes (months breastfeeding and months exclusively breastfeeding), we only observed duration for those who reported “yes” to ever breastfeeding. We accounted for this selection by estimating a 2-part model that first predicted the probability of breastfeeding and then estimated duration conditional on any breastfeeding. 23 We estimated age at first formula for those ever receiving formula similarly. In all results, we report marginal effects holding all covariates at their means. We considered 2-sided P values of less than .05 to indicate statistical significance. All estimations included state fixed effects, standard errors clustered at the state level, and have been weighted with NIS sampling weights. We used Stata version 13 to conduct the analyses (StataCorp LP, College Station, TX).

Most children covered by Medicaid would not have been affected by the ACA mandate, but newly eligible Medicaid enrollees enrolled as a result of the state Medicaid expansions would have been covered by the preventive services mandate. Therefore, children in our comparison group residing in early Medicaid expansion states whose mothers were newly enrolled at the time of their birth would be “treated,” which would dilute the true policy effect. In other words, we might observe increases in breastfeeding for those children covered by Medicaid as well, resulting in a difference-in-differences effect smaller than the true effect. To address this issue, as a sensitivity check, we re-estimated all models restricting our sample to non–Medicaid expansion states. 24

RESULTS

Figure 1 presents trends in mean rates of breastfeeding duration (similar figures for ever breastfed, ever formula fed, and age at first formula by child’s insurance status are in Figure A). Although mothers with privately insured children tend to breastfeed longer than other mothers, the trends in breastfeeding duration of children insured by Medicaid were mostly similar to those of privately insured children before August 2012, particularly from 2010 to 2013. In most cases, the pre-2012 slopes were not statistically significantly different for Medicaid and privately insured children. However, breastfeeding behaviors among the uninsured appeared to be noisier, suggesting that they would be a less suitable control group. The fact that mothers covered by Medicaid tend to be less likely to breastfeed and to breastfeed for shorter durations did not bias our estimation because we were examining changes in breastfeeding trends—not levels—so any time-invariant differences in Medicaid compared with privately insured women were controlled for by our model.

Mean Breastfeeding Duration in Months, by Insurance Status: National Immunization Survey, United States, 2008–2014

Note. HI = health insurance. The whiskers indicate 95% confidence intervals.

Summary statistics for outcome and control variables are presented in Table 1 for the sample of privately insured and Medicaid mothers. Privately insured mothers were more likely to report ever breastfeeding (84%) than were Medicaid mothers (69%). Mothers with private insurance breastfed about 2 months longer than did those covered by Medicaid and breastfed exclusively more than 1 month longer on average. Those privately insured also used formula less and began formula feeding later.

TABLE 1—

Sample Descriptive Statistics by Health Insurance Coverage: National Immunization Survey, United States, 2008–2014

Note. WIC = Special Supplemental Nutrition Program for Women, Infants, and Children. Entries represent percentages except for breastfeeding months, exclusive breastfeeding months, and age at first formula, which are presented as mean number of months. All differences between privately insured and Medicaid statistically significant at the 1% level except female (child). Entries based on weighted statistics.

With the exception of the child’s sex, means for the control variables were statistically different at the less-than-1% level across all control variables. Children in the private insurance category were less likely to be Hispanic; to live in a household that ever received Special Supplemental Nutrition Program for Women, Infants, and Children benefits; to have siblings; and were more likely to live in the south and to be identified as White. Privately insured mothers were older, more likely to have a college education, and more likely to be married. These differences should not bias our results, however, unless there was a change in the characteristics of mothers who obtained private insurance that was correlated with the timing of the policy change in August 2012. Although several provisions of the ACA may change the composition of the insured, 25 these other provisions went into effect at different times from the policy change we evaluated.

We present regression-adjusted means in the pre- and postperiods for both children covered by private insurance and those covered by Medicaid in Table 2 , first differences, and the difference-in-differences estimates for all outcomes. These were all computed by using the margins command in Stata (as predicted probabilities or means, holding all other covariates at their means).

TABLE 2—

Regression-Adjusted Means, First Differences, and Difference-in-Difference Estimates of Breastfeeding by Insurance Status: National Immunization Survey, United States, 2008–2014

Note. ACA = Affordable Care Act; CI = confidence interval. Models with ever breastfeed and ever formula feed were estimated as probits; breastfeeding and exclusive breastfeeding duration conditional on ever breastfeeding, and age at first formula conditional on ever using formula were estimated as two-part models adjusting for the mass of infants breastfed zero months or first fed formula at birth. Additional individual-level control variables included child sex, race, and ethnicity; household participation in the Special Supplemental Nutrition Program for Women, Infants, and Children (ever) and number of children; mother’s age, education, and marital status; and state fixed effects. Sample sizes differ because of missing values in the outcome variable, but results were consistent using the smallest sample. All estimates were weighted with National Immunization Survey sample weights.

Overall, we found that breastfeeding initiation increased for both groups with a slightly larger increase for children covered by Medicaid relative to private insurance, but with no overall change in breastfeeding initiation. Similarly, the percentage of children who were ever fed formula decreased significantly for both groups of children 14 to 17 percentage points (P < .001). Although the decline was greater for children covered by Medicaid, this difference was not statistically significant.

Our results suggest statistically significant changes in breastfeeding duration for privately insured children. We found that privately insured children were breastfed about 0.83 additional months in the postperiod (P = .001), versus a nonsignificant increase of 0.26 months for Medicaid children. These changes net an additional 0.57 months (P = .007) in breastfeeding duration following the mandate, an almost 10% increase. Children covered by private insurance and Medicaid were breastfed exclusively for an additional 1.44 months (P < .001) and 0.70 months (P < .001), respectively, yielding an overall increase in exclusive breastfeeding duration of 0.74 months (P = .001) for those privately insured, a roughly 21% increase. The results for age at first formula mirrored those for ever-fed formula. Both private insurance and Medicaid groups saw little change in age at first formula feeding.

If mothers covered by Medicaid in states that expanded Medicaid early were affected by the ACA’s lactation-support services coverage mandate, then our results might be understated as some of the control group would have had access to lactation support covered by insurance. In Table 3 , we present the results after we excluded all children in states that expanded Medicaid. Overall, the results were similar to those reported in Table 2 . The magnitudes were larger—an impact of 0.80 additional months in breastfeeding duration (P = .07)—but less precisely estimated. Privately insured children were breastfed 1.15 additional months exclusively (P = .005). These results suggest that our main results likely underestimated the true effect.

TABLE 3—

Regression-Adjusted Means, First Differences, and Difference-in-Difference Estimates for Non–Medicaid Expansion States of Breastfeeding by Insurance Status: National Immunization Survey, United States, 2008–2014

Note. ACA = Affordable Care Act; DD = difference in differences. Models with ever breastfeed and ever formula feed were estimated as probits; breastfeeding and exclusive breastfeeding duration conditional on ever breastfeeding, and age at first formula conditional on ever using formula were estimated as 2-part models adjusting for the mass of infants breastfed zero months or first fed formula at birth. Additional individual-level control variables included child sex, race, and ethnicity; household participation in the Special Supplemental Nutrition Program for Women, Infants, and Children (ever) and number of children; mother’s age, education, and marital status; and state fixed effects. Sample sizes differ because of missing values in the outcome variable, but results were consistent using the smallest sample. All estimates were weighted with NIS sample weights. Robust standard errors are in parentheses.

DISCUSSION

In our investigation of the effect of ACA-mandated coverage of lactation-support services, we found an increase in both the total number of months spent breastfeeding and months exclusively breastfeeding (without supplementation of formula) for children with private health insurance relative to children covered by Medicaid. Our finding of an increase of 0.57 and 0.74 months in breastfeeding and exclusive breastfeeding months, respectively, is consistent with a meta-analysis that documented that breastfeeding-support services increased breastfeeding duration by 8 to 11 percentage points. 26

These estimates are likely a lower bound of the policy effect as limitations in our data and measures create a downward bias. The children aged 19 to 23 months in the 2014 NIS had birthdates between February 2012 and July 2013, implying that some children in the 2014 survey would have been born before the mandate (Figure B, available as a supplement to the online version of this article at http://www.ajph.org). In addition, health insurance coverage was measured at the time of the survey. Although it is likely highly correlated with coverage at the time of the child’s birth, it is possible that measurement error is systematically different for privately insured children compared with those with public coverage. If, for example, disproportionately more children were covered by Medicaid at the time of the survey but were actually covered by private insurance at the time of the birth, our estimates would be biased downward. Finally, we cannot identify mothers who actually received ACA-related services and breast pumps in our data. Some mothers would have had coverage of these services before the ACA and others enrolled in grandfathered plans would have remained without coverage. In this respect, our analysis is akin to an intent-to-treat approach and our estimates should be considered a lower bound of the true policy effect. 27

We found little evidence of a differential increase in breastfeeding initiation or formula feeding for children covered by private insurance relative to those covered by Medicaid. In our earlier study, which used the National Vital Statistics System data (birth records), we found an increase in breastfeeding initiation as of hospital discharge based on hospital administration records. This distinction is important. Our measure of breastfeeding initiation with the NIS data was based on the mother’s self-report of whether the child was ever breastfed as of age 19 to 23 months. Adjusted breastfeeding initiation rates with the National Vital Statistics System data ranged from 58% to 70%, 18 whereas similarly adjusted rates with the NIS data ranged from 75% to 81%. Thus, it is not surprising that we did not observe much change in NIS breastfeeding initiation rates because they were already relatively high and more accurately reflect actual breastfeeding initiation relative to rates based on breastfeeding as of hospital discharge.

Taken together, our results for breastfeeding initiation, breastfeeding duration, and formula feeding suggest that the policy did not change the composition of mothers who breastfeed, but increased the amount of time spent breastfeeding to the point at which an infant can transition to solid foods. However, caution should be taken in interpreting the formula-feeding results, as there is more noise in this outcome measure. The percentage of children who were ever fed formula decreased significantly for both children with private insurance and those covered by Medicaid over time. This is consistent with CDC efforts to increase the number of “Baby-Friendly”–designated hospitals—the designation requires hospitals to “give infants no food or drink other than breast milk, unless medically indicated (step 6)”—through the 2011–2015 Best Fed Beginnings program. 28 The Maternity Practices in Infant Nutrition and Care Survey also suggests an increase in the number of hospitals that reported “limiting supplements” for newborns and a decrease in the percentage of hospitals distributing infant formula at discharge. 29,30 Thus, detecting an effect of the ACA mandate on both breastfeeding and formula initiation is likely limited by other recent efforts, especially if those efforts disproportionately affected our comparison group.

Limitations

Our analysis is subject to standard concerns of secondary survey data analyses. These include recall bias for breastfeeding outcomes that might have occurred weeks or months in the past and respondents’ perceptions of the social desirability of breastfeeding. Although these issues could have affected our measures, it is unlikely that recall bias affected our treatment and control groups differentially. If recall bias affected our treatment and control groups similarly, our difference-in-differences framework mitigates this concern.

In addition, the survey did not include other pieces of information including birth outcomes that are often associated with breastfeeding (e.g., premature birth or Cesarean delivery). Although these birth outcomes may be systematically different across our treatment and control groups (i.e., higher rates of Cesarean delivery or premature birth among Medicaid births), the difference-in-differences approach removes these group-specific unobserved differences as long as they do not change over time. Finally, because we limited our analysis to mothers with children aged 19 to 23 months to capture those who would be affected by the policy, we may not have observed the full breastfeeding duration of older toddlers.

Public Health Implications

Our results are consistent with the critical role of lactation consultants in the weeks after discharge for new mothers at risk for stopping breastfeeding because of pain, latch problems, and insufficient milk supply. 31 Access to a breast pump may also be important for improved rates of sustained breastfeeding with 74% of women of childbearing age in the labor force. 32

Although many attempts to improve overall breastfeeding outcomes occur at more direct levels of intervention—such as a hospital—recent state and federal policies are proving to be complementary initiatives. Though cost of and access to lactation-support services and equipment address only a few of the challenges to breastfeeding for new mothers, they appear to make a difference in moving the dial in achieving sustainable breastfeeding goals.

ACKNOWLEDGMENTS

The authors received funding from Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau (Title V, Social Security Act) grant R40MC28305 to complete this work.

HUMAN PARTICIPANT PROTECTION

The RAND Corporation institutional review board deemed this study as exempt as it relied on secondary de-identified data only.