Breastfeeding Trends Following the US Infant Formula Shortage

Breastfeeding, a nearly universally available and cost-effective practice, enhances maternal and child health.^1–13 In the United States, despite 95% of women being capable of producing breastmilk,¹⁴ 1 in 6 women does not initiate breastfeeding.¹⁵ Although the Healthy People 2020 (HP 2020) goals for breastfeeding initiation (81.9%) were met in 2014, this achievement did not extend to low-income households, some racial groups, and rural areas.^16,–18 There are structural and policy factors that contribute to these disparities, including lack of adequate community or workplace lactation supports^19,20 and limited access to supportive maternity care practices.^18,21 

In late 2021, pandemic-related supply chain issues led to problems with infant formula access, and shortages were reported.²² The situation worsened in February 2022 as Abbott Nutrition, the largest infant formula manufacturer supplying 40% of the nation’s formula, issued a voluntary infant formula recall and shut down their Michigan facility following a federal investigation into foodborne illness.²³ This closure exacerbated prior shortages, increasing the national out-of-stock rate for powdered infant formula to 23% by May, ranging from 13% to 32% across states.^24,25 Although the variety and volume of available formula improved in the second half of 2022, it remained below the precrisis levels.²⁵ In fall 2022, 34.7% of parents with infants reported experiencing difficulties obtaining formula. The situation improved but did not resolve by summer 2023 when 20% of parents with infants still faced challenges obtaining formula.²⁶ 

Formula shortages and recalls likely enhanced the appeal of breastfeeding due to its greater perceived safety.^27,–29 To the best of our knowledge, only 2 studies analyzed the breastfeeding behavioral response to the formula crisis, yet their focus was confined to low-income communities and did not assess nationwide heterogeneity or average impacts.^27,30 One study observed a statistically significant increase in breastfeeding initiation in rural and medically underserved regions of southern Illinois.²⁷ The other study examined trends among prenatal Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) participants and a control group, also finding a statistically significant increase.³⁰ Our study evaluates US breastfeeding initiation rates during the 2022 infant formula crisis accounting for potential heterogeneity in the breastfeeding response across subpopulations. The study has the following 2 objectives: (1) to measure the overall change in breastfeeding initiation during the 2022 infant formula crisis, and (2) to identify heterogeneous effects across socioeconomic and demographic subpopulations and their impact on preexisting disparities in breastfeeding rates.

This study was based on national birth certificate data from 2016 to 2022 from the National Vital Statistics System, a comprehensive census of all live births in the United States collected via the US Standard Certificate of Live Birth.^31,32 Our analytic sample included infants alive when a birth certificate was completed, not transferred to another facility within 24 hours of birth, with mothers not admitted to the intensive care unit, and with complete covariate data (n = 20,996,516) (Supplemental Information, Supplemental Figure 1A). Excluded from the analysis were the following 3 states: Michigan for not adopting the standard breastfeeding wording until 2021, Utah for data quality concerns stemming from the state legislation change in 2022, and California for the absence of data on breastfeeding.³³ Births in 47 states and the District of Columbia (DC) represented 79.7% of US live births, contributing to nearly national estimates.

The outcome was breastfeeding initiation, determined from a birth certificate’s question about the infant’s breastfeeding status at discharge. Data were extracted by the clerk completing the birth certificate from medical records.^34,35 A set of variables, known as important predictors of breastfeeding, included mothers’ self-reported participation in the WIC, education (less than high school, high school, some college, college or more), and race and ethnicity (ie, Hispanic or non-Hispanic [NH] ethnicity, and American Indian or Alaskan Native, Black or African American, Asian, Native Hawaiian or Pacific Islander, white, or more than 1 race). The payment source for delivery (private insurance, Medicaid, other) was extracted from medical records. County identifiers were based on the residence location and reflected the county population size (100 000 or more inhabitants; less than 100 000 inhabitants).^34,35 

The overall change in breastfeeding initiation was estimated using a Bayesian structural time series (BSTS) model, which required data aggregation by month of birth to construct a time series for analysis. In addition, a linear probability model (LPM) was estimated based on the infant-level data without data aggregation. The 2-model approach was used to ensure result consistency across specifications, whereas the LPM model allowed testing for heterogeneity across subpopulations.

The “intervention” period of the formula crisis was defined as February 2022 (the month of the Abbott’s plant closure and formula recall) to December 2022, whereas January 2016 through January 2022 served as a control period. A longer duration than in prior studies^27,30 was used to account for formula availability and perceptions of formula feeding safety that may have had lasting effects beyond the crisis peak in summer 2022.^25,–29 

The BSTS model offers a convenient alternative to the interrupted time series design, requiring fewer explicit assumptions.³⁶ BSTS can manage temporal dependence effects in time series data, such as trends over time, seasonality, or other autocorrelation effects. This is crucial to avoid biased estimates and forecast a counterfactual trend in a synthetic control scenario devoid of any intervention (Supplemental Information, section 1). The BSTS method has limitations, as it does not allow for the inclusion of interaction terms to test for differences across subpopulations. To address this limitation, we used an LPM approach that does not have similar constraints and can include interaction terms.

The LPM computed the average change in breastfeeding initiation during the crisis using an indicator representing the intervention period (1 for the period after February 2022 and 0 otherwise). This approach is less computationally demanding than other generalized linear models and generates outcomes comparable with large sample sizes.³⁷ A linear trend was included to adjust for potential trend confounders. Seasonal effects were accounted for by incorporating month-fixed effects. County-fixed effects accounted for relevant differences across counties that were stable over the study period, such as the number of maternity facilities or cultural norms. Clustered standard errors at the county level were used to account for the geographic variation in formula shortages (Supplemental Figure 2).

Heterogeneous effects across subpopulations were assessed by incorporating interactions between the intervention period indicator and key covariates, including race and ethnicity, payment source for delivery, mother’s education, WIC participation, and county population (Supplemental Figure 2).

Several states implemented paid family leave policies during our study period (2016–2022), which are known to increase breastfeeding rates.^38,–40 To mitigate potential bias attributable to these policies, we replicated all analyses excluding these states. An additional sensitivity analysis was included based on an earlier cutoff²² and defining November 2021 through December 2022 as the alternative intervention period when shortages (albeit smaller) were still reported²⁵ (Supplemental Figure 3).

The BSTS model was estimated using the CausalImpact package in R [v. 4.2.1]^41,42; other analyses were conducted in Stata 18.⁴³ The University of Connecticut Institutional Review Board (IRB) deemed the study exempt from IRB review (protocol #23–274–910).

Table 1 summarizes maternal characteristics from 2016 to 2022, which showed increasing levels of maternal education, a rise in the share of Hispanic mothers, and a decline in WIC participation. The precrisis period was characterized by stable trends, with the national unadjusted breastfeeding initiation rate averaging 83.98%. Following the onset of the formula crisis in February 2022, breastfeeding rates began to rise, reaching a peak increase of 3.20 percentage points (pp) in June 2022 (87.18% vs 83.98%). Although there was a gradual decline after June, breastfeeding initiation remained above the precrisis levels, with an average rate of 86.34% in July to December 2022 (Figure 1).

Unadjusted trends across subpopulations aligned with the overall breastfeeding patterns until the start of the infant formula crisis in February 2022 but showed substantial heterogeneity after that. This variability was shaped by the precrisis breastfeeding rates. Population groups nearing a saturation point, with rates already exceeding the HP 2020 target of 81.9%, showed smaller increases in breastfeeding initiation, whereas groups further from the saturation point experienced higher growth (Figure 2).

The net (adjusted) increase in the national breastfeeding initiation rate during the infant formula crisis was 1.96 pp (95% credible interval, 1.68–2.23 pp) according to the BSTS model (Figure 3). The robustness of these results was confirmed in sensitivity analyses (Supplemental Figure 3).

In the LPM specification, the net increase in the national breastfeeding initiation rate was 1.88 pp (95% confidence interval [CI], 1.53–2.23), demonstrating consistency with the BSTS results. The increases across subpopulations varied considerably, with the largest gains exhibited in populations with historically lower breastfeeding rates. Our reference group in the LPM model consists of infants born to NH white mothers with a college education, living in counties with more than 100,000 residents, covered by private insurance, and not participating in WIC. This group, chosen as the reference due to its largest size, did not experience any statistically significant increase in breastfeeding rates (0.13 pp [CI, −0.18 to 0.44 pp; P = .41]). Compared with this group, mothers with less than a high school diploma and those with a high school diploma had additional increases in breastfeeding initiation of 2.48 pp (95% CI, 2.07–2.88 pp; P < .001) and 2.18 pp (95% CI, 1.88–2.48 pp; P < .001), respectively. Medicaid coverage was associated with an additional increase of 0.66 pp (95% CI, 0.37–0.95 pp; P < .001), whereas WIC participants saw an additional gain of 1.33 pp (95% CI, 0.99–1.67 pp; P < .001). NH Black mothers increased breastfeeding initiation by an additional 0.69 pp (95% CI, 0.19–1.18 pp; P = .007), whereas Hispanic mothers experienced an additional decrease of 0.98 pp (95% CI, −1.42 to −0.54 pp; P < .001). Finally, mothers living in counties with fewer than 100 000 residents had significantly higher gains in breastfeeding rates (0.79 pp, [CI, 0.33 to 1.24 pp; P < .001]) compared with our reference group (Figure 4). The robustness of these results was confirmed in sensitivity analyses (Supplemental Figure 3).

Less advantaged mothers, specifically those with less than a high school diploma, covered by Medicaid, participating in WIC, and residing in counties with fewer than 100 000 inhabitants, experienced the highest increase in breastfeeding initiation. This increase was most pronounced among infants born to NH Black mothers, who had an average 6.06 pp increase in breastfeeding initiation (95% CI, 5.26 to 6.87 pp; P < .001), more than 4 times the national increase in breastfeeding rates during this period, whereas infants born to Hispanic mothers showed positive but smaller gains (4.40 pp [95% CI, 3.69 to 5.11 pp; P < .001]). In contrast, more advantaged mothers, specifically college educated, covered by private insurance, not WIC participants, and residing in counties with 100 000 or more inhabitants, did not experience a statistically significant increase in breastfeeding initiation, except for infants born to NH Black mothers (0.81 pp [95% CI, 0.19 to 1.44 pp; P = .011]). In rare cases, there was a slight decrease in breastfeeding initiation, as observed in infants born to Hispanic mothers among the advantaged group (−0.85 pp [95% CI, −1.45 to −0.24 pp; P = .006]).

National breastfeeding initiation increased during the infant formula crisis of 2022, reflecting shortages of infant formula and fears about formula availability and the safety of formula feeding. This trend began to diminish after June 2022 with the Abbott’s Michigan plant reopening and the enactment of the Access to Baby Formula Act of 2022, which suspended tariffs on imported infant formula.^44,45 Breastfeeding rates remained elevated above the historical levels throughout 2022.

The observed 1.88 to 1.95 pp average increase in breastfeeding initiation over a 10-month period is on par with achievements from past policy efforts to promote breastfeeding. For example, mandated coverage of lactation support services under the Affordable Care Act increased the breastfeeding initiation rate by 2.5 pp.⁴⁶ Paid maternity leave implementation had mixed impacts on breastfeeding initiation yet increased breastfeeding duration and exclusivity.^38,–40 A national study using shelter-in-place measures during the pandemic as a proxy for paid leave observed increased breastfeeding duration but did not find statistically significant changes in breastfeeding initiation.⁴¹ The 2009 WIC food package revision narrowed the gap in breastfeeding initiation between WIC-eligible participants and nonparticipants by an average of 4 pp over 2009 to 2014.⁴⁷ 

As the shortages of infant formula impacted mothers’ control over infant feeding choices, there were nonuniform effects across subpopulations that changed the long-term trend in sociodemographic disparities. The biggest improvements in breastfeeding initiation during the 2022 infant formula crisis were observed among populations with greater reliance on formula feeding, including mothers with lower education, receiving WIC assistance, residing in less populated counties, covered by Medicaid, and who are NH Black.⁴⁸ The potential for maintaining these beneficial gains beyond the crisis period should be further investigated.

The increase in breastfeeding initiation during the crisis seems to be closely tied to the coping mechanisms used to find infant formula during the shortages, with common strategies involving searching for formula across multiple retail outlets or relying on others for purchases.^26,49 Less advantaged households are hit hardest due to limited time and resources for acquiring formula.⁴⁸ In our analysis, this is evident in the results by education, Medicaid, and WIC participation. Women with lower levels of education often work in service-oriented industries like lodging and retail⁵⁰ that provide less schedule flexibility and funds to actively search for formula. Furthermore, WIC participants initially encountered challenges purchasing imported formula with WIC benefits, and waivers providing additional formula options were not uniformly implemented across states.^51,52 WIC benefits covered only in-store purchases, not online, and reliance on local stock was further complicated by state-specific bans on using WIC benefits across state lines.⁵³ In less populated counties, additional difficulties may arise due to limited access to grocery stores and transportation barriers, which make coping with shortages more challenging and amplify their impact.⁵³ The higher increase observed among NH Black mothers may be related to the greater prevalence of formula feeding in their communities. Other factors, such as mistrust of institutions, could have exacerbated concerns about contaminated formula.⁵⁴ Conversely, the lower increase among Hispanic mothers may be due to their lower prevalence of formula feeding. Additionally, qualitative studies suggest they managed the formula shortage through family support networks.⁵⁵ Strong family ties, which help mitigate poverty’s effects,⁵⁶ may also have helped them better cope with the shortage. Shortages of infant formula in 2022 were linked to notable reductions in preexisting disparities in breastfeeding rates and extend the observed increase in breastfeeding initiation among WIC participants and nonparticipants covered by Medicaid from 2009 to 2017.⁵⁷ 

It remains unknown if and how long the observed reduction in breastfeeding disparities will persist. Similar incidents involving recalls of infant formula in other high-income countries have shown that declines in infant formula consumption can persist for years after the initial event.⁵⁸ Consumer responses to foodborne illness indicate that people often avoid not just the recalled product but the entire product category.⁵⁹ Households with vulnerable members, who are at higher health risks including infants, tend to be more sensitive to food safety news and respond more drastically than less risk-averse consumer segments.⁶⁰ 

The crisis underscores how the widespread marketing of infant formula as a solution to infant growth and health challenges may systematically undermine breastfeeding.⁶¹ Better regulation of infant formula marketing, in line with World Health Organization and United Nations International Children’s Emergency Fund recommendations,⁶¹ offers an opportunity to address longstanding disparities in breastfeeding rates. Policies like the Providing Urgent Maternal Protections for Nursing Mothers Act (“PUMP Act”), which extends the right to pump at work and allows legal action against noncompliant employers, could significantly support breastfeeding mothers.⁶² This is particularly crucial for low-income workers in service industries, where workplace support for breastfeeding is often limited.²⁰ Expanding lactation support through virtual breastfeeding consultations can help underserved mothers, especially in rural areas, who have expressed a need for telelactation services.⁶³ Additionally, broader implementation of mandatory paid family leave programs in the United States could promote equity in paid leave use, potentially reducing disparities in breastfeeding initiation and duration.^38,–40 

This study has several key strengths. It used national birth certificate data (2016–2022), which allowed for the assessment of breastfeeding initiation among nearly all newborn infants from 47 states and Washington DC (79.7% of US live births). Validation from other studies underscores the reliability of breastfeeding information recorded on birth certificates.⁶⁴ The national breastfeeding initiation trends observed in our data closely align with findings from the National Immunization Survey-Child (NIS-Child) (Supplemental Figure 4).¹⁵ Our data enable a detailed examination of different subpopulations with a high degree of accuracy. The BSTS model provides a robust, causally interpretable inference by predicting counterfactual trends, accommodating preintervention extended timeframes to capture long-term dynamics, preserving seasonal patterns, and preventing overfitting with spike-and-slab priors.

Limitations of this study encompass incomplete data from 3 states and the challenge of discerning confounding events that coincided with the infant formula crisis. The BSTS model allows for a causal interpretation of average impacts but does not allow for testing of heterogenous effects across subpopulations. Further, data on infants with food allergies, intestinal failure, kidney disease, and metabolic disorders were unavailable, impeding our capacity to assess the impacts on a group facing heightened formula shortages.⁶⁵ This study focused on breastfeeding initiation only. Understanding how formula shortages affect breastfeeding duration and exclusivity is important for promoting maternal and child health, designing effective policies, and addressing equity in infant feeding practices. This highlights an important area for future research.

During the 2022 infant formula crisis, there was a statistically significant increase in breastfeeding initiation rates nationwide with larger increases among populations with historically greater reliance on formula feeding, including mothers with lower education levels, with infants receiving WIC assistance, residing in less populated counties, covered by Medicaid, and with Black mothers. The disproportionate changes for population groups with lower breastfeeding rates contributed to a reduction in preexisting disparities in breastfeeding initiation, suggesting an underlying potential for improvement. Targeted interventions, including breastfeeding-friendly workplace policies, enhanced lactation support, expanded access to paid family leave, and improved regulation of infant formula marketing, could help reduce breastfeeding disparities in the future.

ACA

Affordable Care Act

BSTS

Bayesian structural time series

CI

confidence interval

Cr

Credible Interval

DC

District of Columbia

FDA

US Food and Drug Administration

LPM

Linear Probability Model

NH

Non-Hispanic

NVSS

National Vital Statistics System

pp

percentage points

WIC

Special Supplemental Nutrition Program for Women, Infants, and Children