Participation in Conditional Cash Transfer Program During Pregnancy and Birth Weight–Related Outcomes

This cohort study investigates the association of participation in the Bolsa Família cash transfer program with birth weight–related outcomes.


Introduction
2][3][4][5] Low birth weight (LBW) is a marker of risk among newborns, with short-and long-term consequences, and is therefore a concern, particularly in low-and middle-income countries. 6In Latin America and the Caribbean, 8.7% of live births are considered LBW 6 and 12.5% are considered small for gestational age (SGA). 7The prevalence of LBW in Brazil is approximately 8.7%, 8 and this has not significantly decreased in the last 15 years. 6SGA births 9 correspond to 7.8% of births in the 100 Million Cohort. 101][12][13][14][15][16][17][18][19] Conditional cash transfer (CCT) programs have emerged in Latin America beginning in the 1990s as a strategy for social protection and poverty reduction. 20,21Complementary to unconditional cash transfer programs (UCTs), which provide only monetary transfers, CCTs incorporate the fulfillment of conditionalities (typically, adherence to a health and education agenda) as a requirement for continued receipt. 20,21[24][25] CCTs have been associated with lower child 26 and maternal mortality, 27 improvements in child nutrition and health, 28,29 preventive behavior, and an increase in the use of health services. 23spite this potential to stimulate positive health-related behaviors, a recent literature review indicated that due to the CCT health conditionality component characteristic, there was a lack of understanding about whether cash transfers are more effective in specific subgroups of the population than others. 30e Bolsa Família Program (BFP) is one of the world's pioneering CCTs.It has more than 13 million beneficiary families per year. 31Although Brazil was one of the pioneers in implementing CCTs in Latin America and there has been some evaluation of the association of this program with child health, [22][23][24]26 there is still a lack of evidence to support an association of the BFP with birth weight indicators.
Our objective was to estimate the effectiveness of PBF, focusing on its potential association with a decreased likelihood of LBW and SGA, as well as improved birth weight (in grams).It is recognized that the association of BFP with birth weight indicators may vary by population subgroup.

Ethical Considerations
The Research Ethics Committee of the Institute of Collective Health, Federal University of Bahia approved the protocol for this cohort study and waived informed consent because this study uses electronic data without any personally identifiable information.The Reporting of Studies Conducted Using Observational Routinely Collected Health Data (RECORD) statement has been followed.

Study Population
The eligible study population consisted of children from live births in the Centro de Integracao de Dados e Conhecimentos Para Saude (CIDACS) Birth Cohort 32 33,34 (birth weight <500 g or born before 22 gestational weeks) and ( 2) multiple births and newborns with congenital anomalies (given that these conditions are associated with adverse birth weight indicators 35 ).

Exposure
Live births were classified as being exposed to BFP if the mother started receiving BFP at any time during the cohort period, considering an exposure window of at least the estimated period of a complete pregnancy (9 months), without interruptions.Mothers who discontinued receipt were not considered in the analysis.Newborns of mothers who did not receive BFP at any time before delivery were classified as unexposed.BFP eligibility criteria are CadÚnico registered family per capita income and family composition (such as the presence of children, adolescents, and pregnant individuals).
6 Ideally, cash payments are directed toward women, contingent on the fulfillment of specific program requirements (conditionalities). 35,36These criteria encompass the necessity for consistent school attendance and use of health care services throughout childhood (including maintaining an up-to-date vaccination schedule), during pregnancy (prenatal consultations), and in the postpartum period. 37Further details on eligibility criteria and program characteristics are described in eAppendix 2 in Supplement 1.

Statistical Analysis
We estimated the association of BFP with birth-related indicators using propensity score (PS)-based methods.Analyses were described in detail in the research protocol. 40PS estimation was performed using complete data (main analysis).The descriptive analysis based on missing data is available in eTables 1-4 and eAppendix 4 in Supplement 1.Additionally, we performed the PS estimation incorporating a missing data category (category 7 = missing data) (eTable 5 and eAppendix 5 in Supplement 1).Our analysis involved a PS estimation through a logistic model to estimate the probability of receiving BFP based on confounding variables observed (eAppendix 5, eFigure 4, and eTables 6-7 in Supplement 1) and year of cohort entry.BFP beneficiary and nonbeneficiary individual weights were estimated from the PS through kernel-based matching. 41A 2-sided P value < .05indicated statistical significance.
To estimate the association of BFP with LBW and SGA, we used logistic models weighted and adjusted for the following risk factors (categorical variables): gestational age, sex of live-born infant, maternal age at birth, and type of delivery.Adjusted risk ratio was calculated using the δ method to verify discrepancies in the odds ratio (OR) (eTables 8-9 and eAppendix 6 in Supplement 1).A linear model was used to estimate the association of BFP with birth weight (continuous, by 1-g increase in weight) weighted and adjusted by gestational age in weeks, maternal age at birth in years, sex of the live-born infant, and type of delivery (eTable 6 in Supplement 1).We used inverse probability of treatment weighting (IPTW) 42 as an alternative approach (eTable 10 and eAppendix 7 in Supplement 1) to estimate the association between BFP participation and birth weight indicators.
Analyses were performed using Stata statistical software version 16 (StataCorp).Data were analyzed from January 3 to April 24, 2023.
We aimed to explore BFP association with birth indicators across subgroups according to selfreported maternal race (Asian, Black, Indigenous, Parda, and White), educational level (Ն8, 4-7, and Յ3 years), and attendance at prenatal appointments (<7 and Ն7 appointments).All PSs were estimated separately for each population subgroup, with the variable defining the subgroup excluded from the calculation.Similarly, kernel-weighted logistic and linear models were calculated overall and separately within each population subgroup (eTable 11 and eAppendix 8 in Supplement 1).
,44 Additionally, we evaluated unadjusted associations of BFP, maternal education, and self-reported race with attendance of prenatal appointments (eTable 14 and eAppendix 9 in Supplement 1).Furthermore, we examined the association of BFP with birth outcomes among multiparous mothers, while also accounting for characteristics of their prior pregnancies through use of weighted and adjusted models (eTable 12 in Supplement 1).Additionally, we investigated the association between BFP and birth outcomes based on PS quintile (eTable 13 in Supplement 1).

Results
Of

Discussion
In this cohort study, we found that BFP participation was associated with reduced chances of LBW and an increase in birth weight in grams.6][47] However, we also explored the association of BFP with outcomes in pregnant individuals from different social and ethnic subgroups, showing greater changes in outcomes among the highest-risk groups.In addition, use of information on previous childbirths enabled adjustment for birth intervals, previous LBW, and previous prematurity. 19,48Although CCTs have an association with an increased interval between births, 49 the association with beneficiary fertility among mothers is controversial. 50The first pregnancy and grand multiparity are risk factors for LBW and SGA. 10,19e magnitude of outcomes associated with other CCTs and UCTs has varied by program characteristic.A study of the effectiveness of the Oportunidades program, a CCT implemented in Mexico, demonstrated a 127-g increase in mean weight at birth among beneficiary children and a 4.6% lower prevalence of LBW in this group. 45A randomized study conducted in rural villages in Togo, West Africa, found that receiving a UCT reduced the chance of having a baby with LBW (adjusted OR, 0.29; CI 95%, 0.10-0.82). 46In Colombia, a study on the Familias en Acción program showed a 578-g increase in birth weight in urban treatment locations. 47Increased birth weight in the US Food Stamp Program (currently known as the Supplemental Nutrition Assistance Program) provides further evidence that prenatal nutritional intake may play a role in child birth outcomes. 51In the US, Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) services have also been associated with reduced LBW and increased birth weight in grams, especially among subgroups of Black women and those with late prenatal care or no prenatal consultation. 52sh transfer strategies are also implemented in high-income countries. 53,54In the US, poverty relief during the prenatal period (an income tax credit) was associated with an increase in birth weight of 15.7g (12.5 g when adjusted for smoking). 53In retrospective cohort studies, cash transfers during the prenatal period provided to women with lower incomes who were residents in a municipality of Canada (through the Healthy Baby Prenatal Benefit UCT) were associated with a 26% 55 and 29% 56 lower risk of LBW.
To the best of our knowledge, this is the first study to evaluate the association of CCT with weight standards at birth by gestational age in a high-risk Brazilian population.In a study in Canada, the only study found that assessed the association of a CT (specifically, a UCT) with SGA, an association was found with a decrease in SGA births (adjusted risk ratio, 0.90; 95% CI, 0.81-0.99). 55e difference between CT designs may explain the variability reported in estimates (heterogeneity of findings).We may consider 2 hypotheses for the mechanisms behind the a In logistic regression results, the analysis was kernel weighted and adjusted for gestational age, sex of the live-born child, mother's age at birth, and type of delivery.
b LBW was defined as birth weight less than 2500 g, and not LBW was defined as birth weight 2500 g to less than 4000 g.
c SGA was defined as weight for gestational age at birth less than the 10th percentile of weight for gestational age according to sex, and not SGA (ie, appropriate for gestational age) was defined as weight for gestational age at birth in the 10th to 90th percentile.
d In linear regression results, the analysis was kernel weighted and adjusted for gestational age, sex of the live-born child, mother's age at birth, and type of delivery.increased social capital, and female decision-making power. 57,58[61] Despite the significant increase in attending prenatal appointments in Brazil between 2000 and 2015, inequality remains pronounced, particularly among Black and Indigenous women and those with a lower level of education. 62These groups include individuals experiencing more deprivation with greater difficulty in accessing this service.Our investigation found associations between BFP and increased birth weight and decreased odds of LBW within specific subgroups.These subgroups included mothers who attended fewer prenatal appointments; those who were Black, of mixed race, and Indigenous; and those with lower levels of education.Furthermore, our study found that BFP beneficiaries had lower odds of SGA only in subgroups of Indigenous mothers and those with lower education levels.These results suggest that beneficiaries and nonbeneficiaries may be more homogeneous in relation to characteristics not observed in subgroups of mothers at higher risk.The group with a lower number of prenatal consultations likely consisted of individuals who faced greater challenges in accessing this service, particularly those with lower socioeconomic status.Education was the only variable that was not well-balanced between groups.Therefore, disparities persisted, with a higher percentage of mothers in the group with fewer consultations having low levels of education.These findings are consistent with those of a recent review on CCTs and child health in low-and middle-income countries showing that these programs exhibited considerable heterogeneity among subgroups by socioeconomic status indicator. 30W, as a result of poverty, can contribute to worse health status over time and consequently maintain inequality from generation to generation. 53The difficulty of reducing birth weight-related outcomes indicates the need to intensify policies with this focus. 6Thus, there is a need to strengthen social, redistributive, and health policies that act on the negative consequences of inequalities, seeking to minimize their effects on health, striving for food and nutritional security, prenatal care, and assistance during labor. 12

Strengths and Limitations
This study used PS-based approaches to evaluate the association of BFP with maternal-child health results in a population of low-income and extremely low-income Brazilian families.The study followed a previously defined and published research protocol, 40 providing data analysis transparency and greater result comparability.Several strengths can be highlighted in this study.The population-level database encompasses a wide range of socioeconomic variables at family and personal levels and a variety of risk factors, which are rarely available in administrative data.A robust analytical approach using kernel-based PS weighting and IPTW was used to account for observed confounding factors in the study.Beneficiary and nonbeneficiary groups were well-balanced for covariate distributions.
Several limitations should also be considered.Receiving BFP is not a random attribution but the result of a self-selection process by families.A BFP selection bias was reported in another study, 26 which dealt with the issue in a similar way to our study, by following a kernel matching approach to select a set of nonbeneficiary BFP observations within the CIDACS 100 Million Brazilian Cohort.This method enabled us to balance groups by observable characteristics.The external validity of the study was affected by the population choice given that we considered only 1 child per mother.BFP is a binary variable in our study, and this proposal did not investigate nuances related to the value received and poverty levels.Another limitation of this study is the bias related to unmeasured confounding.Important unmeasured factors should be considered, particularly family income, which could not be included in this study.Moreover, we were unable to investigate the distribution of some

JAMA Network Open | Health Policy Conditional
from 2012 to 2015 among mothers aged Cash Transfer Program During Pregnancy and Birth Weight-Related Outcomes years who were registered on CadÚnico (the Brazilian national social program register) at any time from 2004 to 2015 (Figure).Births before the mother entered the cohort, births before the study period, and individuals with inconsistencies in variables (eg, mother's age) and missing data on outcomes were considered ineligible for the study (eFigure 1 and eAppendix 1 in Supplement 1).Our selection was limited to births that occurred between 2012 and 2015 due to a change to birth certificates in 2011.The live birth certificate includes crucial variables for our study, such as JAMA Network Open.2023;6(11):e2344691. doi:10.1001/jamanetworkopen.2023.44691(Reprinted) November 28, 2023 2/16 Downloaded from jamanetwork.comby guest on 12/24/2023 10 to 49

Table 1 .
Birth Weight Indicators of Live-Born Children as weight for gestational age at birth less than the 10th percentile of weight for gestational age according to sex, and not SGA (ie, appropriate for gestational age) was defined as weight for gestational age at birth in the 10th to 90th percentile.populations, the International Fetal and Newborn Growth Consortium for the 21st Century Consortium considers only infants born between 24 and 42 gestational weeks.Live births that were not within this range were considered missing.
a Appointments categorized by the median.b LBW was defined as birth weight less than 2500 g and not LBW as birth weight 2500 g to less than 4000 g. c SGA was defined d To calculate SGA

Table 2 .
Variables With Complete Data Used for PS

Table 2 .
Variables With Complete Data Used for PS (continued) JAMA Network Open.2023;6(11):e2344691. doi:10.1001/jamanetworkopen.2023.44691(Reprinted) November 28, 2023 7/16 Downloaded from jamanetwork.comby guest on 12/24/2023 translates from Portuguese as "brown," is used to denote individuals whose racial background is predominantly Black and those with multiracial or multiethnic ancestry, including European, African, and who attended fewer than 7 antenatal care appointments; were Black, Indigenous, or Parda; and less educated (Յ3 years of formal education).An association between BFP participation and decreased odds of SGA was found among Indigenous mothers and those with less education.

Table 3 .
Association of Bolsa Família Participation With Birth Weight Indicators

Table 4 .
57,47,56ion of Bolsa Família Participation With Birth Weight Indicators by SubgroupAnalytical steps (propensity score estimation, kernel matching, and weighted regression) were conducted separately within each level of education, self-reported maternal race, and number of appointments.The analysis was kernel weighted and adjusted for gestational age, sex of the live-born child, mother's age at birth, and type of delivery.LBW and SGA analyses used logistic regression, and the birth weight analysis used linear regression.Parda, which translates from Portuguese as "brown," is used to denote individuals whose racial background is predominantly Black and those with multiracial or multiethnic ancestry, including European, African, and Downloaded from jamanetwork.combyguest on 12/24/2023 association of CCTs with birth weight outcomes.First, CTs enable the family to diversify the food purchased (consuming more vegetables, fruit, and meat, which are sources of minerals and vitamins), which is associated with family food security indicators,46,47,56psychosocial health,57 c d established biological risk factors associated with LBW and SGA, including chronic diseases, JAMA Network Open | Health Policy Conditional Cash Transfer Program During Pregnancy and Birth Weight-Related Outcomes Receiver Operating Characteristic Curve of 100 Million Brazilian Cohort and Live Birth Information System (2001-2015) Linkage: Approach 1 eFigure 3. Receiver Operating Characteristic Curve of 100 Million Brazilian Cohort and Live Birth Information System (2001-2015) Linkage: Approach 2 eAppendix 4. Missing Data eTable 1. Missing Data for Propensity Score Variables for Total Population eTable 2. Distribution of Missing Data eTable 3. Description of Study Population for Entire Period (2004-2015) in Accordance With Missing Data Pattern eTable 4. Crude Odds Ratio of Propensity Score Variables With Missing Data Category eTable 5. Adjusted and Weighted Coefficients Considering Propensity Score Variables With Missing Data eAppendix 5. Propensity Score eTable 6. Variables Used in Study eFigure 4. Common Support Area of Exposed Over Unexposed Group eTable 7. Propensity Score Description in Accordance With Confounding Covariate eAppendix 6. Adjusted Risk Ratio With δ Method eTable 8. Adjusted Risk Ratio With δ Method of Bolsa Família Beneficiaries on Birth Weight Indicators eTable 9. Adjusted Risk Ratio With δ Method of Bolsa Família Beneficiaries on Birth Weight Indicators in Accordance With Subgroup Analysis eAppendix 7. Analysis of Robustness for Propensity Score-Based Methods eTable 10.Coefficients of Adjusted and Weighted Logistic and Linear Regressions of Bolsa Família Beneficiaries on Birth Weight Indicators eAppendix 8. Subgroup Analysis eTable 11.Variables Used in Subgroup Analysis eTable 12. Bolsa Família and Birth Weight Indicators of the Population of Second Live Births eTable 13.Adjusted and Weighted Coefficients of Bolsa Família Beneficiaries on Birth Weight Indicators Considering Propensity Score Quintiles eAppendix 9. Bivariate Analyses eTable 14.Crude Analysis for the Association of Bolsa Família, Level of Education, and Race With Appointments eReferences.