Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 BMC Pregnancy and Childbirth https://doi.org/10.1186/s12884-023-05707-1 R E S E A R C H Open Access The dose-response association between LEAP 1000 and birthweight – no clear mechanisms: a structural equation modeling approach Sarah Quinones1*, Shao Lin2,3, Lili Tian4, Pauline Mendola1, Jacob Novignon5, Clement Adamba6 and Tia Palermo1 Abstract Background Birthweight is an important indicator of maternal and fetal health globally. The multifactorial origins of birthweight suggest holistic programs that target biological and social risk factors have great potential to improve birthweight. In this study, we examine the dose-response association of exposure to an unconditional cash transfer program before delivery with birthweight and explore the potential mediators of the association. Methods Data for this study come from the Livelihood Empowerment Against Poverty (LEAP) 1000 impact evaluation conducted between 2015 and 2017 among a panel sample of 2,331 pregnant and lactating women living in rural households of Northern Ghana. The LEAP 1000 program provided bi-monthly cash transfers and premium fee waivers to enroll in the National Health Insurance Scheme (NHIS). We used adjusted and unadjusted linear and logistic regression models to estimate the associations of months of LEAP 1000 exposure before delivery with birthweight and low birthweight, respectively. We used covariate-adjusted structural equation models (SEM) to examine mediation of the LEAP 1000 dose-response association with birthweight by household food insecurity and maternal- level (agency, NHIS enrollment, and antenatal care) factors. Results Our study included a sample of 1,439 infants with complete information on birthweight and date of birth. Nine percent of infants (N = 129) were exposed to LEAP 1000 before delivery. A 1-month increase in exposure to LEAP 1000 before delivery was associated with a 9-gram increase in birthweight and 7% reduced odds of low birthweight, on average, in adjusted models. We found no mediation effect by household food insecurity, NHIS enrollment, women’s agency, or antenatal care visits. Conclusions LEAP 1000 cash transfer exposure before delivery was positively associated with birthweight, though we did not find any mediation by household- or maternal-level factors. The results of our mediation analyses may serve to inform program operations and improve targeting and programming to optimize health and well-being among this population. Trial Registration The evaluation is registered in the International Initiative for Impact Evaluation’s (3ie) Registry for International Development Impact Evaluations (RIDIESTUDY- ID-55942496d53af ) and in the Pan African Clinical Trial Registry (PACTR202110669615387). *Correspondence: Sarah Quinones squinone@buffalo.edu Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 2 of 12 Keywords LEAP 1000, Birthweight, Ghana, Empowerment, Antenatal care, Food insecurity, SEM, Cash transfers Background beneficiaries adhere to certain behaviors, such as mater- Infant birthweight is a critical metric of maternal and nal and child healthcare visits or school enrollment and fetal health and a key predictor of child and adult health attendance, to receive payments. A recently published outcomes globally. Low birthweight (LBW; < 2,500  g) study was the first to identify positive impacts of a UCT infants have increased risk of morbidity, mortality, mal- on birthweight and LBW in Africa [12]. nutrition, and chronic disease throughout the life course This study seeks to contribute to this evidence base by compared to healthy weight infants [1–4]. Mothers born examining (1) the association between months of expo- LBW are more likely to give birth to a LBW infant, sug- sure to a UCT program before delivery and birthweight gesting intergenerational persistence of impaired fetal and (2) the pathways through which these associations health and development [5]. Despite global reductions materialize. We hypothesize that a UCT program cou- of LBW, prevalence remains high in African countries pled with health insurance enrollment targeted to preg- with 14% prevalence in sub-Saharan Africa in 2015 [6]. nant women will increase birthweight and decrease LBW Further, average birthweight has trended downward risk among infants through the pathways of household in Africa in the 21st century [7]. Taken together, these food security, antenatal care (ANC), women’s agency, and trends suggest a need for interventions to improve birth- health insurance [13–15]. weight outcomes in this region. LBW is a multifactorial birth outcome which arises Materials and methods from preterm birth (PTB; delivery before 37 completed Livelihood empowerment against poverty (LEAP) 1000 weeks of gestation), intrauterine growth restriction program (IUGR; infant growth did not reach full biological poten- In 2008, the Ministry of Gender, Children and Social tial), or a combination of the two. Prevention of LBW Protection (MoGCSP; Government of Ghana) imple- relies on comprehensive interventions that target risks mented LEAP, its flagship national social protection pro- to the health of the mother and the developing fetus [8, gram. The purpose of LEAP was to reduce poverty in the 9]. The multifactorial origins of LBW risk present many short-term and improve human capital development in opportunities for intervention and risk reduction. Com- the long-term [16]. To achieve these objectives, LEAP prehensive interventions that target the myriad risk fac- provided cash payments to households living in extreme tors of reduced birthweight and increased LBW may poverty with a household member from a vulnerable serve as cost-effective approaches to improved health, demographic group (i.e., orphan or vulnerable child, though evidence on such interventions is lacking. Most elderly person, or a person with a severe disability). Then, birthweight interventions have focused primarily on in 2011, the National Health Insurance Authority (NHIA) nutrition during pregnancy [9]. However, there are sev- and the MoGCSP collaborated to enroll LEAP beneficia- eral other predictors of birthweight, many of which are ries into the National Health Insurance Scheme (NHIS) poverty-related, worth targeting for LBW risk reduction under the NHIA ‘indigent’ exemption, which waives in low-resource populations of Africa. NHIS enrolment and other fees, including card process- Social protection, defined as “the set of policies and ing, premiums, and renewals. As of 2017, LEAP reached programs aimed at preventing or protecting all peo- more than 200,000 households in Ghana, and as of 2022, ple against poverty, vulnerability and social exclusion it now reaches 550,000 households nationally. throughout their lifecycle, with a particular emphasis In 2015, a pilot program within LEAP – LEAP 1000 - towards vulnerable groups,” [10] is a potential cost-effec- expanded program eligibility to pregnant and lactating tive intervention for LBW risk reduction. Specifically, women living in extremely impoverished, rural house- cash transfers (CTs), whereby recipients receive sched- holds in districts of Northern and Upper East Ghana. The uled and predictable amounts of cash based on poverty objective of the LEAP 1000 pilot program was to reduce or other criteria, have been associated with reduced malnutrition and stunting. To achieve this objective, LBW risk in various contexts [11]. However, there is a LEAP 1000 aimed to target children in the first 1,000 days dearth of evidence on the (1) dose-response associa- of their lives (i.e., from conception to age 2 years). Using tions between CT program participation and birthweight a multi-stage targeting approach, communities in 10 dis- and (2) mediators by which CTs, specifically uncondi- tricts of Northern and Upper East Ghana were identified tional CTs (UCTs), improve birthweight. UCTs require using district-level poverty rankings and then households no actions on the part of the recipients to be eligible for in the poorest communities (with priority given to those payments. In contrast, the literature includes studies that not already covered by LEAP) were selected based on evaluate conditional CTs (CCTs) [11], which require that proxy means test (PMT) scores that served as measures Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 3 of 12 of household poverty status. PMTs were administered Measures to households containing women of reproductive age Our dependent variables included infant birthweight (15–49 years) who were eligible if (1) they were pregnant (measured in kilograms; from maternal recall and or (2) they had a child 12 months of age or younger and records on health cards) and LBW (birthweight < 2.5 kg). could present health facility documentation to confirm The independent variable was months of LEAP 1000 their status. treatment received before infant delivery, which was cal- The effectiveness of the LEAP 1000 pilot program was culated based on the difference in months between infant tested in an impact evaluation, led collaboratively by the birth date and LEAP 1000 implementation (September UNICEF Office of Research – Innocenti, the University 2015). All comparison infants and treatment infants born of North Carolina at Chapel Hill, the Institute for Social, before program implementation were classified as having Statistical, and Economic Research (ISSER), and the zero months of exposure before delivery. Navrongo Health Research Centre. The impact evalua- Potential mediators evaluated in this study are shown tion was conducted between 2015 and 2017 in 5 of the in Fig.  1. These mediators were selected based on the 10 initial districts where LEAP 1000 was piloted (Bongo, LEAP 1000 conceptual framework (Supplementary East Mamprusi, Garu-Tempane, Karaga, and Yendi). Fig. 1) and the results of the LEAP 1000 impact evalua- Power calculations run for the original impact evalua- tion showing positive impacts on these variables [13]. tion found that program impacts on the primary out- Potential mediators included current NHIS enrollment (a comes of interest (stunting, wasting, and underweight) current and valid NHIS card observed by the enumera- could be observed with a sample size of 2,500 households tor), self-reported ANC visits with a skilled provider, (1,250 comparison and 1,250 treatment). However, these number of ANC visits during pregnancy, and number of power calculations were not conducted with secondary meals reported per day. Household food insecurity score outcomes, such as LBW, in mind. The impact evaluation was calculated based on the sums of the following indica- sample selection was inspired by a Regression Discon- tors: (1) the household head reported worrying that their tinuity Design (RDD) identification strategy that lever- household didn’t have enough food more than once in the aged a PMT score threshold to select a census of 1,250 past 4 weeks (0: Never; 2: Rarely; 3: Sometimes; 4: Often) comparison households just above the threshold and and (2) the household reported that a household mem- 1,250 treatment households just below the threshold for ber went an entire day and night without food more than interviews to maximize comparability between groups. once in the past 4 weeks (0–4). Additionally, women’s At baseline, 2,497 eligible households (1,235 compari- agency was included as a potential mediator, informed by son and 1,262 treatment) were included. By endline, 6% the literature that suggests CTs improve agency and that of baseline households were lost to follow-up, leading agency is a salient predictor of maternal and child health to panel sample of 2,331 households (1,146 compari- outcomes [17–21]. The definition of women’s agency was son and 1,185 treatment) used for the impact evaluation based on the sum of the following indicators [22]: In the and which serves as the sample for this secondary data past 12 months, how often did you feel that (a) Your life is analysis. determined by your own actions; (b) You have the power to make important decisions that change the course of Data collection your own life; (c) You have the power to make important Household questionnaires were administered to house- decisions that change the wellbeing of your children; (d) hold heads and/or LEAP 1000 eligible women (one per You have the power to make important decisions that household) by trained enumerators at baseline (July – change the wellbeing of your household; (e) You are capa- September 2015) and endline (June to August 2017). ble of protecting your own interests within your house- Topics covered by the household questionnaire included hold; and (f ) You are capable of protecting your own housing conditions and WASH, food security, time use interests outside of your household. Reponses were on a and employment, productive livelihoods, non-farm scale of 1 (never) to 5 (very often/always). Prior to sum- enterprises, reproductive health, and household con- ming these indicators for a total women’s agency score, sumption. Topics covered by the LEAP 1000 beneficiary each indicator was dichotomized as classified as 1 if at questionnaire included birth history, contraception and least sometimes and classified as 0 otherwise for a total fertility preferences, women’s agency, stress and prefer- score range of 0–6. ences, nutrition and feeding knowledge, and intimate partner violence. Lastly, LEAP 1000 beneficiaries were Statistical analysis asked about their children in the questionnaires using Bivariate analyses were conducted to test any differences the following topics: maternal and newborn health, child in household-, maternal-, and infant-level characteris- health, immunizations, child nutrition and feeding, birth tics between infants who did not receive any LEAP 1000 registration and child development, and anthropometry. treatment before delivery (either comparison infants or Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 4 of 12 Fig. 1 Hypothesized pathways between months of LEAP 1000 exposure before infant delivery and birthweight treatment infants born before LEAP 1000 implementa- the household-level in the SEM and regression models. tion) and those who received 1 or more months of LEAP All analyses were conducted using Stata version 16 (gsem 1000 treatment before delivery. We present bivariate command for mediation analysis [nlcom command for analyses using logistic regression for dichotomous out- individual and total effects]) [23]. We describe results as comes and linear regression for continuous outcomes, statistically significant at an alpha less than 5%, though adjusted for PMT score. To test the dose-response rela- in mediation analyses we highlight results at a p-value of tionship between the number of months of LEAP 1000 10% using boldface. exposure before delivery and birthweight, we estimated coefficients and confidence intervals (CI) using crude and Sensitivity analyses adjusted linear regression models. Crude and adjusted We ran generalized SEM with LBW as a dichotomous logistic regression models were used to estimate the dependent variable as a test of our main results for con- associations between months of LEAP 1000 and LBW tinuous birthweight. These models were adjusted for the with odds ratios (OR) and 95% CI. same covariates as in the main analysis, SEs were clus- Model covariates were selected if differences were sta- tered at the household level, and the generalized model tistically significant across categories of months of LEAP was specified with a ‘binomial’ family and a ‘logit’ link 1000 exposure during pregnancy (p < 0.1). Covariates function. These models estimate the change in log odds included total number of children under age 5 in the of LBW with 95% CI in response to changes in the inde- household, parity, household has an improved lighting pendent variable (months of LEAP 1000) and the media- source, and district of residence. Given that sample selec- tors in the model. tion was informed by the threshold based on PMT score distributions, we also adjusted for PMT score. Addition- Results ally, to adjust for potential time trends, month and year of The final analytic sample for this study included 1,439 birth were included as covariates in final models. infants born to women who were part of the LEAP 1000 Mediation analysis was conducted using an adjusted impact evaluation (treatment and comparison groups) generalized Structural Equation Model (SEM). SEM from 2015 to 2017 with complete information on birth- presents the direct and indirect of months of LEAP 1000 weight (~ 50% of the full sample of infants), birth date, on birthweight through ANC, NHIS, women’s agency, and other model mediators and covariates (Fig. 2) [13]. and household food insecurity measured at endline to Among the 1,439 infants in the analytic sample, 129 ensure temporality in these associations, while adjust- (9%) were exposed to at least one month of LEAP 1000 ing for all covariates outlined above. SE were clustered at before delivery. Comparisons of household-, maternal, Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 5 of 12 Fig. 2 CONSORT diagram of LEAP 1000 impact evaluation and study sample selection and infant-level characteristics by number of months of was significantly lower among those exposed before treatment (adjusted for PMT score) are shown in Table 1. delivery (0.60 ± 0.52) compared to those who were not Comparison and treatment infants born before program (3.79 ± 1.99; p < 0.001). We observed no significant differ- implementation were generally comparable to infants ences by category of LEAP 1000 exposure before delivery who received one or more months of LEAP 1000. Infants for any measure of ANC. who received one or more months of LEAP 1000 were The unadjusted and adjusted associations between less likely to be born in the rainy season (57%) than those months of LEAP 1000 treatment before delivery and who did not receive LEAP 1000 before delivery (66%; birthweight are presented in Table  2. In this sample, p = 0.009). A lower proportion of infants treated before birthweight was normally distributed with normally dis- delivery resided in Bongo (19%) than those who did tributed errors when regressed with months of LEAP not receive LEAP 1000 before delivery (26%; p = 0.028), 1000 exposure before delivery. On average, a 1-month though a higher proportion resided in East Mamprusi increase in LEAP 1000 exposure before delivery was (47 vs. 41%, respectively; p = 0.054). Higher maternal par- associated with a 9-gram increase in infant birthweight ity, number of children under the age of 5 years in the in the adjusted model (p = 0.015). Also, increased parity household, and improved lighting sources were observed was marginally associated with increased birthweight among the infants treated by LEAP 1000 before delivery (p = 0.054), while later year of birth was associated with compared to those not treated. Current NHIS enrollment decreased birthweight (p = 0.027), suggesting a negative was observed to be lower among women who received trend in birthweight over time in this sample. at least 1 months of LEAP 1000 treatment (66%) than In Table 3, we present the logistic regression estimates those who did not (72%; p = 0.008). Women’s agency was of the unadjusted and adjusted ORs and 95% CI for the significantly lower among those not exposed to LEAP association between months of LEAP 1000 before deliv- 1000 before delivery (3.02 ± 1.95) compared to those with ery and LBW. A 1-month increase in LEAP 1000 expo- at least 1 month of LEAP 1000 exposure before delivery sure before delivery was associated with 7% reduced odds (4.11 ± 1.79; p < 0.001). Household food insecurity score of LBW in the adjusted model (p = 0.024). Increasing Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 6 of 12 Table 1 Descriptive statistics of sample characteristics by Table  4 presents the independent variable-mediator number of months of LEAP 1000 before delivery among a sample and mediator-dependent variable associations estimated of infants with complete birthweight born to women receiving using adjusted linear regression models for months of the LEAP 1000 cash transfer (N = 1,439) LEAP 1000 exposure before delivery, potential house- Number of months of LEAP 1000 hold- and maternal-level mediators, and birthweight, Mean ± SD or N (%) respectively. Increasing months of LEAP 1000 exposure 0 1+ p-value# before delivery was associated with a significant reduc- Household-level tion in household food insecurity score (β=-0.068; 95% Number of children under 5 in 1.76 ± 0.71 2.09 ± 0.62 < 0.001 CI: [-0.085, -0.05]; p < 0.001). We also observed a mar- household ginal improvement in women’s agency in response to Household head married 1,245 (95) 125 (97) 0.275 increased number of months of LEAP 1000 before deliv- Female household head 133 (10) 8 (6) 0.132 ery (β = 0.023; 95% CI: [-0.004, 0.05]; p = 0.1). No other Age of household head 39.3 ± 13 41.4 ± 12.2 0.457 associations were statistically significant. Household has improved 406 (31) 52 (40) 0.004 Table  5 is organized to show the independent and lighting source mediator variables estimated in the SEM in column (1), Household food insecurity 3.79 ± 1.99 0.60 ± 0.52 < 0.001 the direct effects of each variable on birthweight in col- score (0–8) umn (2), the indirect effects of months of LEAP 1000 Maternal-level through all mediators combined and each mediator indi- Number of meals consumed 2.6 ± 0.6 2.6 ± 0.58 0.851 per day vidually in column (3), and the percent due to mediation Current NHIS enrollment 948 (72) 85 (66) 0.008 (indirect/total effect) in column (4). In the SEM, months Mother’s age 29.2 ± 6.52 30.8 ± 6.21 0.117 of LEAP 1000 exposure before delivery had significant Women’s agency (0–6) 3.02 ± 1.95 4.11 ± 1.79 < 0.001 direct effects on birthweight in the adjusted models Parity (Total number of live 3.67 ± 1.96 4.78 ± 2.17 < 0.001 (β = 0.01; p = 0.006). Mediation through all mediating births) variables accounted for 1% of the effect of months of Sought ANC from skilled 1,218 (93) 120 (93) 0.291 LEAP 1000 on birthweight. Women’s agency did not have provider any significant direct or indirect effects on birthweight, Number of ANC visits 5.89 ± 1.73 5.92 ± 1.66 0.897 though mediation accounted for 2% of the adjusted LEAP Infant-level 1000-birthweight association. Similarly, no statistically Singleton 1,263 (96) 127 (98) 0.164 significant direct or indirect effects were observed for Delivered in a health facility 1,220 (93) 117 (91) 0.325 household food insecurity, NHIS enrollment, or ANC, Infant birthweight (kg) 3.01 ± 0.46 3.07 ± 0.46 0.195 though household food insecurity accounted for 8% of Infant low birth weight (< 2.5 95 (7) 8 (8) 0.501 the LEAP 1000-birthweight association. kg) Mediated effects of LEAP 1000 of LBW by the same set Infant female 655 (50) 56 (43) 0.177 0.009 of mediators are presented in Supplementary Table 1. The Infant born during the rainy 868 (66) 73 (57) season (March - Sept) overall direct effect of each month of LEAP 1000 before District delivery on LBW was statistically significant (-0.084; 95% East Mamprusi 537 (41) 61 (47) 0.054 CI: [-0.151, -0.018]; p = 0.013) and mediation by these Karaga 60 (5) 6 (5) 0.810 factors accounted for 7.7% of the total effect. Current Yendi 104 (8) 10 (8) 0.973 NHIS enrollment had a marginally significantly negative Bongo 346 (26) 24 (19) 0.028 association with LBW (-0.407; 95% CI: [-0.849, 0.036]; Garu-Tempane 263 (20) 28 (22) 0.959 p = 0.072) but did not mediate the association between N 1,310 129 months of LEAP 1000 and birthweight. Similar to the ANC: Antenatal care; NHIS: National Health Insurance Scheme; PMT: Proxy findings in Table  5, no mediating effects were observed means test; SD: Standard deviation for current NHIS, ANC, women’s agency or household # Tests of significance conducted using logistic and linear regression models adjusted for PMT score for dichotomous and continuous variables, respectively food insecurity, though the latter accounted for 7.6% of the association between months of LEAP 1000 and LBW. PMT score was associated with large reductions in odds Discussion of LBW (OR = 0.07; 95% CI: (0.005–0.910); p = 0.042). We found a 9-gram increase in average birthweight and Current enrollment in the NHIS was marginally associ- 7% reduced odds of LBW for each additional month of ated with reduced odds of LBW (p = 0.068) and living in LEAP 1000 exposure before delivery in adjusted lin- Bongo versus East Mamprusi was associated with 50% ear and logistic regression models, respectively. These reduced odds of LBW (p = 0.015). findings were confirmed in the SEM models with a 10-gram increase in birthweight and log odds of LBW Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 7 of 12 Table 2 Unadjusted and adjusted associations between reduced by 8.4 in response to a 1-month increase in number of months of LEAP 1000 treatment before delivery exposure to LEAP 1000 before delivery. We observed and birthweight among a sample of 1,439 Ghanaian infants; no mediating effect of ANC, current NHIS enrollment, 2015–2017 women’s agency, or household food insecurity on these Infant birthweight (kg) β (95% CI) associations. p-value The evidence that CTs improve birthweight and reduce Unadjusteda Adjustedb LBW risk is limited in general, and virtually nonexis- Months of LEAP 1000 Treatment 0.005 0.009 tent in Africa [11]. A previous study by our team was (-0.001–0.011) (0.002–0.017) the first to examine whether a UCT in Africa impacted 0.095 0.015 birthweight and LBW [12]. Saville and colleagues found Household-level that a Participatory Learning and Action women’s group Food insecurity score (0–8) -0.002 with food transfers increased average birthweight by (-0.018–0.014) 78 g compared to a control group in Nepal [24]. Barber 0.763 and Gertler found that Mexico’s Oportunidades CT pro- Improved lighting source 0.031 gram increased average birthweight by 102–127  g and (-0.028–0.089) decreased LBW by 4.4–4.6% points [25, 26]. In Colom- 0.302 bia, Attanasio and colleagues found a 578-gram increase Number of children under 5 in -0.007 household (-0.046–0.032) in the birthweight of urban infants born to women who 0.729 participated in Familias en Accion CT program. And, PMT score 0.006 0.111 Amarante and colleagues found the PANES CT pro- (-0.305–0.316) (-0.209–0.431) gram in Uruguay to increase average birthweight by 31 g 0.971 0.495 and decrease LBW by 1.9–2.5% points [27]. A review by Maternal-level Glassman and colleagues included myriad studies from Parity 0.014 8 countries that examined CT impacts on maternal and (-0.000–0.027) neonatal health and found improved prenatal moni- 0.054 toring, increased births attended by a skilled provider, Current NHIS enrollment 0.037 greater health facility deliveries, mixed results on fertility, (-0.020–0.094) and decreased LBW risk [28]. 0.199 While the evidence on dose-response impacts of CTs Infant-level on health outcomes is limited, there are studies that Month of birth -0.004 (-0.013–0.005) support our approach and findings. In Brazil, a dose- 0.383 response association was observed between the Bolsa Year of birth -0.041 (-0.077 Familia Programme, both in terms of cash amounts - -0.005) and program duration, and reduced maternal mortality, 0.027 which was explained by prenatal care visits and case- Infant born during the rainy season 0.005 fatality during delivery [29]. Relatedly, ANC was a pos- (March - Sept) (-0.050–0.060) ited mediator in our study given that LEAP 1000 was 0.858 associated with an 11.4% point increase in ANC from a District [ref: East Mamprusi] skilled provider during pregnancy [13]. Further, ANC is Karaga -0.029 associated with improved birth outcomes and increased (-0.172–0.113) birthweight [30]. However, ANC was not shown to be a 0.687 mediator in our study’s SEM analyses. These disparate Yendi 0.036 findings may be explained by the differences in Bolsa (-0.057–0.130) Familia and LEAP 1000. Bolsa Familia is a conditional 0.447 Bongo -0.018 CT program that imposes ‘soft conditionalities’ on ben- (-0.082–0.047) eficiaries to attend prenatal care for continued payment 0.597 whereas LEAP 1000 has no such conditions. Nonethe- Garu-Tempane -0.003 less, both studies demonstrate improvements in maternal (-0.072–0.067) health resulting from CTs. 0.940 We found no mediation by current NHIS enrollment or N 1,439 household food insecurity. Pregnant women enrolled in PMT: proxy means test; NHIS: National Health Insurance Scheme. aModel NHIS receive myriad services for free including mater- adjusted for PMT score; bModel adjusted for PMT score, parity, improved lighting source in household, number of children under 5 years old in the nity care [31]. These free healthcare services may then household, district of residence, year of birth, infant born in the rainy season, positively influence health-seeking behavior [32], which current NHIS enrollment, household food insecurity score, and month of birth. Standard errors clustered at household level Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 8 of 12 Table 3 Unadjusted and adjusted associations between number of months of LEAP 1000 treatment before delivery and LBW among a sample of 1,439 Ghanaian infants; 2015–2017. PMT: proxy means test; LBW: Low birthweight; NHIS: National Health Insurance Scheme aModel adjusted for PMT score; bModel adjusted for PMT score, parity, improved lighting source in household, number of children under 5 years old in the household, district of residence, year of birth, infant born in the rainy season, current NHIS enrollment, household food insecurity score, and month of birth. Standard errors clustered at household level Low birthweight OR (95% CI) p-value Unadjusteda Adjustedb Months of LEAP 1000 Treatment 0.960 (0.907–1.017) 0.928 (0.869–0.990) 0.164 0.024 Household-level Food insecurity score (0–8) 0.899 (0.783–1.032) 0.130 Improved lighting source 0.842 (0.536–1.323) 0.456 Number of children under 5 in household 1.023 (0.773–1.352) 0.875 PMT score 0.220 (0.016–2.992) 0.068 (0.005–0.910) 0.256 0.042 Maternal-level Parity 0.911 (0.801–1.037) 0.157 Current NHIS enrollment 0.657 (0.419–1.032) 0.068 Infant-level Month of birth 0.979 (0.906–1.058) 0.593 Year of birth 1.068 (0.817–1.395) 0.632 Infant born during the rainy season (March - Sept) 1.204 (0.752–1.928) 0.439 District [ref: East Mamprusi] Karaga 1.180 (0.425–3.275) 0.750 Yendi 0.679 (0.291–1.581) 0.369 Bongo 0.495 (0.281–0.873) 0.015 Garu-Tempane 0.541 (0.279–1.048) 0.069 N 1,439 we found to be the case in the LEAP 1000 impact evalu- Solidarias Rurales program on skilled attendance at ation [13]. LEAP 1000 increased NHIS enrollment by birth and birth in health facilities, which they posit to 14.1% points and increased ANC utilization by 11.4% be attributed to supply-side service improvements and points overall [13, 14]. The null findings among our enhancements in women’s agency [34], which inspired sample of infants may suggest the need to address addi- our assessment of women’s agency as a potential media- tional supply- and demand-side barriers to NHIS annual tor but also points to care quality as a potentially impor- renewal and health care utilization, such as the quality of tant mediator that we did not explore in our study due to health facilities in the area [33]. Furthermore, all preg- lack of data. nant women in Ghana were entitled to NHIS fee waiv- The absence of a mediating effect by household ers, likely diluting the impacts of LEAP 1000 treatment food insecurity can be explained, in part, by the intra- on NHIS enrollment. Moreover, de Brauw and Peter- household, gendered dynamics that allocate household man found robust impacts of El Salvador’s Comunidades resources to men and boys as opposed to women and Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 9 of 12 Table 4 Adjusted associations between mediators, months of LEAP 1000 before delivery, and birthweight; N = 1,439 Independent variable - mediator Mediator – dependent variable Months of LEAP 1000 before delivery Birthweight (kg) β coefficient (95% CI) p-value β coefficient (95% CI) p- value Current NHIS enrollment 0.0002 (-0.006, 0.007) 0.961 0.033 (-0.023, 0.089) 0.253 Number of ANC visits -0.007 (-0.032, 0.017) 0.548 -0.006 (-0.02, 0.008) 0.42 ANC from a skilled provider -0.001 (-0.005, 0.003) 0.521 0.009 (-0.093, 0.112) 0.860 Number of meals consumed per day -0.003 (-0.011, 0.006) 0.572 0.013 (-0.033, 0.059) 0.590 Food insecurity score (0–8) -0.068 (-0.085, -0.05) < 0.001 -0.001 (-0.016, 0.015) 0.948 Women’s agency score (0–6) 0.023 (-0.004, 0.05) 0.1 0.009 (-0.003, 0.022) 0.154 ANC: Antenatal care; PMT: Proxy means test. All regressions are adjusted for PMT score, month and year of birth, district of residence, total number of children under the age of 5 in the household, parity, improved source of lighting in the household, and season of birth (rainy v. dry). Standard errors were clustered at the household level Table 5 Adjusted mediation effect of maternal and household-level characteristics on the association between months of LEAP 1000 treatment before delivery and birthweight among the sample of 1,439 infants Variable Direct effect Indirect effect Percent due to (1) (2) (3) mediation (4) Months of LEAP 1000 0.01 (0.003, 0.018) 0.0001 (-0.001, 0.001) 0.0001/0.01 = 1% 0.006 0.808 Women’s agency 0.009 (-0.004, 0.021) 0.0002 (-0.0002, 0.001) 0.0002/0.009 = 2.2% 0.188 0.298 Household food insecurity score (0–8) 0.001 (-0.015, 0.017) -0.00004 (-0.001, 0.001) 0.00004/0.0005 = 8% 0.946 0.946 Current NHIS enrollment 0.032 (-0.024, 0.088) 5.15e-06 (-0.0002, 0.0002) 0.000005/0.03 = < 1% 0.263 0.961 ANC from a skilled provider 0.012 (-0.088, 0.113) -0.00002 (-0.0001, 0.0001) 0.00002/0.01 = < 1% 0.811 0.807 ANC: Antenatal care; LEAP 1000: Livelihood Empowerment Against Poverty 1000 program; NHIS: National Health Insurance Scheme; PMT: Proxy means test. Models adjusted for PMT score, parity, improved source of lighting in the household, number of children under 5 years old in the household, district of residence, year of birth, infant born in the rainy season, and month of birth. Standard errors clustered at household level. Boldface results are those with a p-value less than 10% girls [35]. Moreover, household food security may not The LEAP 1000 impact evaluation offers a unique translate to or adequately capture individual nutrition opportunity to conduct research on the dose-response and consumption behaviors. This may be a direct artefact relation between the number of months of program of how household food insecurity was measured in this exposure before pregnancy and birthweight. LEAP 1000 study – we use only two out of nine validated measures is one of only a few CT programs in the African conti- of food insecurity [36] to calculate our score as the nine nent with primary objectives to reduce stunting in chil- items were assessed at endline only. Thus, our measure of dren under 5 years old that explicitly targets pregnant household food insecurity may not adequately measure women, which is imperative to achieve program objec- the true experiences of food insecurity in this sample. tives. Evaluations of other programs that target preg- The absence of a mediating effect by women’s agency nant women – Zambia’s Child Grant Program [39] and conflicts with the literature. Women’s empowerment, Mozambique’s Child Grant [40] – do not assess infant which captures agency, is considered to be a salient fac- birthweight, which misses an important opportunity as tor in the improvement of maternal and child health birthweight is antecedent to stunting [41]. outcomes [19, 21, 37] and has also been shown to be increased by CTs [17, 18]. Improvements in women’s Strengths empowerment translates to improved decision-making We used quasi-experimental, longitudinal data collected power which allows a woman to participate in deci- among pregnant and lactating women in high-poverty, sions related to household resource and food allocation, rural Ghana to examine the dose-response associations healthcare seeking for themselves and their children, and between months of LEAP 1000 exposure and birth- demanding better quality of care from providers. We may weight. This is the first study to examine duration of not have been able to observe a mediating effect because cash transfer exposure in utero and birthweight and also of the way we defined women’s agency, which is notably addresses the dearth of evidence of unconditional cash difficult to define [38]. transfers and birthweight more broadly. Additionally, we Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 10 of 12 use SEM, a statistical tool used to analyze complex rela- collected in the impact evaluation. We also were unable, tionships among variables [42], to explore mediators of due to lack of data and/or lack of statistical power, to these dose-response associations. We conduct sensitivity include other mediators worthy of examination such as analyses to test our assumptions and appraise the valid- maternal nutrition, energy expenditures, and WASH ity of our main findings. The additional mediation analy- indicators. The null findings in the SEM for LBW indi- ses of this study provide useful information for program cate greater need for inputs that could present clinically development and implementation as we highlight what is meaningful improvements in maternal and infant health or isn’t working with program selection and operations beyond what is already allocated by LEAP 1000. that may be improved in future iterations. Conclusions Limitations Our findings suggest that LEAP 1000 exposure before This study has limitations that warrant discussion. In this delivery can increase birthweight and lower the risk of sample of LEAP 1000 eligible women, only 60% reported LBW. An absence of mediation in our study may serve delivering in health facilities and 50% of infants had a to inform future program development and data col- birthweight (either recorded on a health card or recalled lection strategies to better measure potential media- by the mother). Our complete-case sample approach may tors. However, the low percentage of births exposed to result in biased findings owed to selection. The inclusion CTs suggests that increased efforts are needed to target of infants born as part of multiple births and with weight women earlier in pregnancy (and roll out cash payments recorded by maternal recall likely biased our results faster) or prior to conception for CT receipt. Given that upward as multiple births generally have lower birth- the treatment households studied here have now been weights. Relatedly, we have a generally small sample that, exposed to seven years of CTs, and more births will have though is well-powered to detect associations in linear occurred, additional follow-up surveys should be con- regression, may be too small to reach statistical signifi- ducted to understand impacts on birthweight and media- cance in the SEM models that use maximum likelihood tors of impact among a larger sample than feasible in the estimation approaches. Further, the estimation of the current study. effects of mediators on birthweight may still suffer from bias, as these are simultaneously assessed with effects List of abbreviationsANC Antenatal care of treatment on mediators in SEM. Also, issues of resid- CI Confidence interval ual and uncontrolled confounding are likely to bias our CT Cash transfer results as there are certain measures likely to confound CCT Conditional cash transferISSER Institute for Social, Statistical, and Economic Research the mediator-outcome association that we have not con- IUGR Intrauterine growth restriction sidered. Further, it is possible that the household- and LEAP L ivelihood Empowerment Against Poverty maternal-level variables assessed as potential mediators LBW L ow birthweightMoGCSP Ministry of Gender, Children, and Social Protection do operate along the pathway between LEAP 1000 and NHIA N ational Health Insurance Authority birthweight, but we were unable to capture their effects NHIS National Health Insurance Scheme due to how these variables were measured and defined. OR Odds ratioPMT P roxy means test In our study, only 10% of infants were exposed to any PTB Preterm birth number of months of LEAP 1000 before delivery due to SE S tandard errors the time-consuming processes of registration (occur- SEM Structural equation modelingUCT U nconditional cash transfer ring in March 2015) and payment delivery (starting from UNICEF U nited Nations Children’s Fund September 2015). Also, pregnant and lactating women with an infant up to 12 months old were targeted, mean- Supplementary Information ing women who were visibly pregnant (usually around 4 The online version contains supplementary material available at https://doi. months) and those who have already given birth likely org/10.1186/s12884-023-05707-1. comprise most of our targeted sample, which explains Additional file 1: Supplementary Figure 1 the low exposure prevalence in this sample. There was Additional file 2: Supplementary Table 1 also a 6-month delay between targeting and enrolment (March 2015) and the first cash transfer receipt (Septem- ber 2015), therefore many women had already given birth Acknowledgements Not applicable. before receiving cash. All these contributors to small sample size and limited exposure limit the power of our Authors’ contributions analyses. In epidemiology, birthweight is a hotly debated SQ and TP conceptualized the study. TP and CA were involved in data curation. SQ, TP, LT, SL, and PM contributed to formal data analysis. SQ, TP, LT, outcome that is often considered a nebulous outcome and SL developed the study methodology. SQ drafted the original manuscript. in the absence of gestational age [43], which was not Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 11 of 12 All authors contributed to the final manuscript draft, study interpretations, and 8. World Health Organization. Global Nutrition Targets 2025: Low birth weight all approved the final submitted draft. policy brief. 2014. Global Nutrition Targets. 2015;2025. 9. Bhutta ZA, Das JK, Rizvi A, Gaffey MF, Walker N, Horton S, et al. Evidence- Funding based interventions for improvement of maternal and child nutrition: what The authors have no funding to report. can be done and at what cost? Lancet. 2013;382(9890):452–77. 10. Alfers L, Moussié R. The ILO world social protection report 2017–19: an assess- Data Availability ment. Dev Change. 2020;51(2):683–97. The data used in this analysis are publicly available from the University of 11. Leroy JL, Koch B, Roy S, Gilligan D, Ruel M. Social Assistance Programs and North Carolina Population Center (https://data.cpc.unc.edu/projects/13/ Birth Outcomes: a systematic Review and Assessment of Nutrition and Health view# res_ 226). Pathways. J Nutr. 2021. 12. Quinones S, Mendola P, Tian L, Lin S, Novignon J, Angeles G et al. Ghana’s livelihood empowerment against poverty (1000) program seasonally Impacts Declarations Birthweight: a difference-in-differences analysis. Int J Public Health. 2023;68. 13. Ghana, LEAP 1000 Evaluation Team. Ghana LEAP 1000 Programme: Endline Competing interests Evaluation Report. 2018. The authors have no competing interests to disclose. 14. Palermo TM, Valli E, Ángeles-Tagliaferro G, de Milliano M, Adamba C, Spada- fora TR, et al. Impact evaluation of a social protection programme paired with Ethics approval and consent to participate fee waivers on enrolment in Ghana’s National Health Insurance Scheme. BMJ The LEAP 1000 evaluation study was reviewed by the Ethics Committee for open. 2019;9(11):e028726. the Humanities of the University of Ghana. The evaluation is registered in the 15. de Groot R, Yablonski J, Valli E. The impact of cash and health insurance on International Initiative for Impact Evaluation’s (3ie) Registry for International child nutrition during the first 1000 days: evidence from Ghana. Food Policy. Development Impact Evaluations (RIDIESTUDY- ID-55942496d53af ) and 2022;107:102217. in the Pan African Clinical Trial Registry (PACTR202110669615387). The 16. Angeles G, Chakrabarti A, Handa S, Darko Osei R, Osei-Akoto I. Livelihood current analysis uses de-identified data and was exempted from IRB review Empowerment Against Poverty (LEAP) Programme Endline Impact Evalua- at the University at Buffalo. Verbal consent was ascertained by all survey tion Report. 2018. respondents. 17. Ambler K, De Brauw A. The Impacts of Cash Transfers on Women’s Empower- ment. 2017. Consent for publication 18. Bonilla J, Zarzur RC, Handa S, Nowlin C, Peterman A, Ring H, et al. Cash for Not applicable. women’s empowerment? A mixed-methods evaluation of the government of Zambia’s child grant program. World Dev. 2017;95:55–72. Author details 19. Essilfie G, Sebu J, Annim SK. Women’s empowerment and child health out- 1Department of Epidemiology and Environmental Health, School comes in Ghana. Afr Dev Rev. 2020;32(2):200–15. of Public Health and Health Professions, University at Buffalo, State 20. Galiè A, Teufel N, Girard AW, Baltenweck I, Dominguez-Salas P, Price MJ, et al. University of New York, Buffalo, NY 14214, USA Women’s empowerment, food security and nutrition of pastoral communi- 2Department of Environmental Health Sciences, One University Place, ties in Tanzania. Global Food Security. 2019;23:125–34. 212D University at Albany, State University of New York, Rensselaer, 21. Kabir A, Rashid MM, Hossain K, Khan A, Sikder SS, Gidding HF. Women’s NY 12144, USA empowerment is associated with maternal nutrition and low birth weight: 3Department of Epidemiology and Biostatistics, One University Place, evidence from Bangladesh Demographic Health Survey. BMC Womens 212D University at Albany, State University of New York, Rensselaer, Health. 2020;20(1):93. NY 12144, USA 22. Rotter JB. Generalized expectancies for internal versus external control of 4Department of Biostatistics, State University of New York, 717 Kimball reinforcement. Psychol monographs: Gen Appl. 1966;80(1):1. Tower University at Buffalo, Buffalo, NY 14214, USA 23. StataCorp. Stata Statistical Software: release 16. College Station. TX: StataCorp 5Department of Economics, Kwame Nkrumah University of Science and LLC; 2019. Technology, Kumasi, Ghana 24. Saville NM, Shrestha BP, Style S, Harris-Fry H, Beard BJ, Sen A, et al. Impact on 6Institute of Statistical, Social and Economic Research, University of birth weight and child growth of participatory learning and action women’s Ghana-Legon, P.O. Box LG 74, Legon-Accra, Ghana groups with and without transfers of food or cash during pregnancy: find- ings of the low birth weight South Asia cluster-randomised controlled trial Received: 3 January 2023 / Accepted: 16 May 2023 (LBWSAT) in Nepal. PLoS ONE. 2018;13(5):e0194064. 25. Barber SL, Gertler PJ. The impact of Mexico’s conditional cash transfer programme, Oportunidades, on birthweight. Tropical Med Int Health. 2008;13(11):1405–14. 26. Barber SL, Gertler PJ. Empowering women: how Mexico’s conditional cash transfer programme raised prenatal care quality and birth weight. J Dev Eff. References 2010;2(1):51–73. 1. Barker DJ. Fetal origins of coronary heart disease. BMJ. 1995;311(6998):171–4. 27. Amarante V, Manacorda M, Miguel E, Vigorito A. Do cash transfers improve 2. Wilcox AJ, Skjaerven R. Birth weight and perinatal mortality: the effect of birth outcomes? Evidence from Matched Vital Statistics, Program, and Social gestational age. Am J Public Health. 1992;82(3):378–82. Security Data. Am Economic Journal: Economic Policy. 2016;8(2):1–43. 3. Gage TB, Fang F, O’Neill EK, DiRienzo AG. Racial disparities in infant mortality: 28. Glassman A, Duran D, Fleisher L, Singer D, Sturke R, Angeles G, et al. Impact of what has birth weight got to do with it and how large is it? BMC Pregnancy conditional cash transfers on maternal and Newborn Health. J Health Popul Childbirth. 2010;10(1):86. Nutr. 2013;31(4 Suppl 2):48–S66. 4. Negrato CA, Gomes MB. Low birth weight: causes and consequences. Diabe- 29. Rasella D, Alves FJO, Rebouças P, de Jesus GS, Barreto ML, Campello T, et al. tol Metab Syndr. 2013;5(1):1–8. Long-term impact of a conditional cash transfer programme on maternal 5. Liu D, Lin G, Su D, Alexender JM, Sun X, Qu M. Intergenerational associa- mortality: a nationwide analysis of brazilian longitudinal data. BMC Med. tions of adverse birth outcomes: a surveillance report. Prev Med Rep. 2021;19(1):127. 2020;20:101226. 30. Bater J, Lauer JM, Ghosh S, Webb P, Agaba E, Bashaasha B, et al. Predictors of 6. Blencowe H, Krasevec J, de Onis M, Black RE, An X, Stevens GA, et al. low birth weight and preterm birth in rural Uganda: findings from a birth National, regional, and worldwide estimates of low birthweight in 2015, cohort study. PLoS ONE. 2020;15(7):e0235626. with trends from 2000: a systematic analysis. The Lancet global health. 31. NHIS, Benefits Package Accra. Ghana2022 [Available from: https://www.nhis. 2019;7(7):e849–e60. gov.gh/benefits.aspx. 7. Marete I, Ekhaguere O, Bann CM, Bucher SL, Nyongesa P, Patel AB, et al. 32. Mitra S, Palmer M, Pullaro S, Mont D, Groce N. Health Insurance and Chil- Regional trends in birth weight in low- and middle-income countries dren in Low- and middle-income countries: a review. Economic Record. 2013–2018. Reproductive Health. 2020;17(3):176. 2017;93(302):484–500. Quinones et al. BMC Pregnancy and Childbirth (2023) 23:364 Page 12 of 12 33. Otieno P, Angeles G, Quiñones S, van Halsema V, Novignon J, Palermo T, et 39. Seidenfeld D, Handa S, Tembo G, Michelo S, Harland Scott C, Prencipe L. The al. Health services availability and readiness moderate cash transfer impacts impact of an unconditional cash transfer on food security and nutrition: the on health insurance enrolment: evidence from the LEAP 1000 cash transfer Zambia Child Grant Programme. 2014. program in Ghana. BMC Health Serv Res. 2022;22(1):599. 40. UNICEF. 24-Month Impact evaluation of the child grant 0–2 component 34. de Brauw A, Peterman A. Can conditional cash transfers improve maternal in the Nampula Province in Mozambique. Maputo, Mozambique: UNICEF health care? Evidence from El Salvador’s Comunidades Solidarias Rurales Mozambique; 2022. program. Health Econ. 2020;29(6):700–15. 41. Aryastami NK, Shankar A, Kusumawardani N, Besral B, Jahari AB, Achadi E. 35. Haddad LJ, Peña C, Nishida C, Quisumbing AR, Slack A. Food security and Low birth weight was the most dominant predictor associated with stunting nutrition implications of intrahousehold bias: a review of literature. 1996. among children aged 12–23 months in Indonesia. BMC Nutr. 2017;3(1):1–6. 36. Coates J, Swindale A, Bilinsky P, Food, Nutrition Technical Assistance Project 42. Beran TN, Violato C. Structural equation modeling in medical research: a (FANTA). Household Food Insecurity Access Scale (HFIAS) for measurement primer. BMC Res Notes. 2010;3(1):267. of Food Access: Indicator Guide (v. 3). Washington, DC: Food and Nutrition 43. Wilcox A. On the importance—and the unimportance—of birthweight. Int J Technical Assistance Project; 2007. Epidemiol. 2001;30(6):1233–41. 37. Santoso MV, Kerr RB, Hoddinott J, Garigipati P, Olmos S, Young SL. Role of Women’s empowerment in Child Nutrition Outcomes: a systematic review. Adv Nutr. 2019;10(6):1138–51. Publisher’s Note 38. Peterman A, Schwab B, Roy S, Hidrobo M, Gilligan DO. Measuring women’s Springer Nature remains neutral with regard to jurisdictional claims in decisionmaking: Indicator choice and survey design experiments from cash published maps and institutional affiliations. and food transfer evaluations in Ecuador, Uganda and Yemen. World Dev. 2021;141:105387.