Effects of the revised WIC food package on women’s and children’s health: a quasi-experimental study

Background: The Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) was revised in 2009 to be more congruent with national dietary guidelines. There is limited research examining effects of the revision on women’s and children’s health. The objective of this study was to evaluate whether the revised WIC food package was associated with various indicators of physical and mental health for women and children. Methods: We used 1998–2017 waves of the National Health Interview Survey (N = 81,771 women and 27,780 children) to estimate effects of the revised WIC food package on indicators of health for both women (self-reported health and body mass index) and children (anemia, mental health, and parent-reported health). We used difference-in-differences analysis, a quasi-experimental technique that assessed pre-post differences in outcomes among WIC-recipients while “differencing out” the secular underlying trends among a control group of non-recipients. Results: For all outcomes evaluated for women and children, we were unable to rule out the null hypothesis that there was no effect of receiving the revised WIC food package. These findings were confirmed across several secondary analyses conducted to assess heterogeneity of effects and robustness of results. Conclusion: While we did not find effects of the revised WIC food package on downstream health indicators, studies using similarly robust methods in other datasets have found shorter-term effects on more proximal outcomes related to diet and nutrition. Effects of the modest WIC revisions may be less impactful on longer-term indicators of health, and future studies should examine the larger COVID-19-era expansion. Supplementary Information The online version contains supplementary material available at 10.1186/s12884-022-05116-w.


Difference-in-differences analysis
The effect of the revised food package on women's and children's health outcomes was estimated using difference-in-differences (DID) analysis. The equation for this model for each outcome Y was specified as follows, for each individual i (woman or child) in time t in state s: = 0 + 1 × + 2 + 3 + 4 + 5 + 6 + 1 is the coefficient of interest in the DID model and represents the effect of the revised WIC food package on the outcome . represents a vector of women's or children's covariates, State represents state fixed effects, and represents year fixed effects.
We used linear models for both continuous and binary outcomes. Linear models are preferred for DID analyses due to differences in the interpretation of interaction terms in non-linear models (1,2). For binary outcomes, this means that the regression represents a linear probability model, and the DID effect is therefore interpreted as a percentage-point change in risk.
Several key assumptions underly DID analyses. First, DID estimation relies on the assumption that the outcome trends in the period following the revision would have been the same for WIC recipients and non-recipients had the food package not been revised. Although this counterfactual condition cannot be observed, several analyses were conducted to evaluate the validity of this assumption. We first qualitatively inspected whether the trends in outcomes among recipients and non-recipients were parallel during the period prior to the revision (Supplemental Figure 2). We additionally tested the parallel trends assumption quantitatively, by restricting the data to the period before the WIC food package revision was implemented and including an interaction term between a binary variable for WIC receipt and a continuous variable for time (Supplemental Table 1). A null effect in this analysis would reassure us of no statistically significant difference in the pre-revision parallel trends.
Second, DID estimation assumes that any observed effects are not due to differential changes in the underlying composition of the treatment versus the control groups. To test this assumption, we evaluated whether characteristics in the women and children samples differed between the pre-and post-revision periods. This was done by modeling each covariate as the dependent variable in separate regression models, in which the primary predictor is the interaction term between an indicator for WIC receipt and an indicator for whether the interview was conducted after the revision was implemented (Supplemental Table 2).

Supplemental Figure 2. Graphical evaluation of the parallel trends assumption for women's and children's health outcomes
Note: The study sample was drawn from the National Health Interview Survey for years 1998-2017. The implementation of the revised WIC food package was staggered across states during 2009, ranging from January to November. The dotted vertical line on each graph indicates when the revised WIC food package was revised for each individual's state of residence. Abbreviations: WIC, Special Supplemental Nutrition Program for Women, Infants, and Children * For children's health outcomes, we removed observations prior to 100 months before the WIC revision because trends in earlier periods were not parallel, as shown in this Figure. Supplemental The study sample was drawn from the National Health Interview Survey for years 1998-2017. We regressed each outcome on an indicator for WIC receipt, a variable representing the difference between interview date and date of implementation (computed based on state and year for each individual), and an interaction between the two. The above estimates represent the coefficient on the interaction term. A non-statistically significant coefficient provides evidence of similarity in the trends between recipients and non-recipients before the revision was implemented. Abbreviations: WIC, Special Supplemental Nutrition Program for Women, Infants, and Children Note: The study sample was drawn from the National Health Interview Survey for years 1998-2017. The above coefficients represent the interaction between a binary variable for being interviewed in the time period after the WIC revision was implemented and an indicator for whether or not the participant was a WIC recipient. These models examine whether there are differences in demographic characteristics between recipient/non-recipients and eligible/noneligible individuals before and after the implementation of the food package revision. Null results indicate that there are no differences in sample composition for a given covariate. Abbreviations: WIC, Special Supplemental Nutrition Program for Women, Infants, and Children