Protecting Nutrition of Pregnant and Lactating Women and Children in Acute Food Crises A comparative analysis of the effectiveness of food vouchers and mixed transfers Shannon Doocy, PhD ● Associate Professor Johns Hopkins School of Public Health

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

Rationale Cash transfers are significant change to humanitarian assistance. Major commitments to scale up have been made by donors and humanitarian agencies. Cash-based approaches can be more efficient than in-kind assistance and more supportive of local economies, human agency and dignity. There is evidence from non-crises settings of the positive impact of cash- based approaches on dietary diversity and use of health services; but the link between these and improved nutrition outcomes has not been adequately researched in emergencies. Most data is collected at a household level, thus there is an evidence gap for commonly targeted vulnerable groups (PLWs, CU5s) There is little evidence for individual nutrition outcomes (ex: SAM/MAM) – with the exception of recent REFANI findings, most evidence from emergency settings is limited to food security

Objectives and Research Questions Objective 1: Increase the understanding on how different transfer modalities affect nutrition and food security outcomes and health seeking behaviors among pregnant and lactating women (PLWs) and children in their households. Objective 2: Characterize contextual factors, notably migration and role of men, and challenges of cash transfer programing in the Somalia context with the aim of developing actionable recommendations to inform future food and nutrition programming in Somalia Primary Research Questions: Is provision of unconditional household transfers in addition to vouchers more effective than vouchers alone for preventing acute malnutrition and supporting health behaviors among PLWs? What is the added value of unconditional cash transfers with respect to household food security?

PLWs and CU5s in Somalia PLWs and CU5 are considered high priority vulnerable groups in emergencies due to the increased nutritional demands of pregnancy and lactation, the rapid growth and development in early childhood and the high levels morbidity and mortality often observed in these population sub- groups. In the ongoing Somalia crisis, PLW may be especially vulnerable as men are forced to migrate farther with herds, leaving women without access to milk from herds, or the ability to trade livestock for cash or food staples. In Somalia, PLW and CU5 represent a sizeable population affected by the current crisis: the Somalia Nutrition Cluster has estimated that some 800,000 PLWs and 1.4 million CU5s need humanitarian assistance. Assessments and surveys conducted during the April to June 2017 period report critical global acute malnutrition (GAM) rates, with GAM levels more than 15% in all affected areas and >30% in some of the most severely affected areas.

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

Study Design Mixed methods approach: both quantitative and qualitative data were collected. Qualitative findings will be used to enrich understanding of quantitative results and their implications, which is important for application to programming in the current crisis. A 3 comparison group quasi-experimental design was required because humanitarian programs are ongoing World Vision SFP beneficiary households with PLWs eligible Malnourished PLWs excluded and referred for SAM treatment Similar non-beneficiary PLWs will be recruited from nearby List-based sampling from beneficiary databases Data collected for PLWs and all CU5s in household Conducted in Wajid, Somalia—drought/conflict affected ‘Reasonable’ security High numbers of IDPs Wajid District

Comparison Groups Food Voucher Group - consisting of PLWs from households receiving FFP paper vouchers (approximately US$81/month); a “top up” of approximately $35 was provided to study participants to ensure equal transfer amounts between the voucher and mixed transfer groups (total value of US$106). Mixed Transfer Group - consisting of PLWs from households receiving mixed transfers of WFP in-kind food, WFP e-vouchers and UNICEF unconditional cash (combined value of approximately US$106/month). Non-Assistance Group - recruited from outside voucher program coverage areas and were PLWs from households that meet voucher eligibility criteria AND that were not receiving monthly household transfers from other organizations. Non-assistance PLWs may have receive regular individual transfers via blanket SFP programs [as did some voucher beneficiaries]. Planned intervention period of 6 months: Nov 2017 – Apr 2018 Actual intervention period of 4 months: Nov 2017 – Feb 2018 due to issues with transfer continuity

Summary of Interventions & Participants Food Vouchers Mixed Transfers Total transfer value US$96-130/household/month (transfer value varied monthly, however, both groups received the same amount each month) Modalities Paper food voucher In-kind food ($US 32-45) Food e-voucher ($US 32-45) Unrestricted cash (US $30-50) Voucher Commodities whole grains, flours, pasta, legumes/pulses, vegetable oil whole grains, flours, pasta, legumes/pulses, vegetable oil, fruits, vegetables, milk, eggs, meat, sugar, salt, spices Total Beneficiary Households (HH) 1650 3000 HH in study communities* 474 700 HH in study communities with PLWs* 190 280 Study Participants Total FFP Paper Vouchers WFP/UNICEF Mixed Transfers Non-Assistance Group PLWs enrolled at baseline 514 166 288 60 PLWs at endline (% of enrolled) 490 (95.3%) 162 (97.6%) 269 (93.4%) 59 (98.3%) *projected based on program data

Food Security & Nutrition Outcome Measures Household food security Household Hunger Scale (HHS) Meals consumed on preceding day Pregnant and Lactating Women Dietary diversity (Food groups & Minimum Dietary Diversity for Women – MDDW) Meal frequency (meals on preceding day) MUAC (mean MUAC and MUAC<21.0cm as cut-off for acute malnutrition) Children Under Five Diet Dietary diversity (food groups consumed on preceding day) Frequency of consumption (meals/snacks/BF on preceding day) Minimum Acceptable Diet (MAD) as summary measure MUAC (mean MUAC and MUAC<12.5cm as cut-off for acute malnutrition)

Statistical Analysis Exploratory and descriptive statistics used for comparison of all groups at baseline and endline (chi-square for proportions, t-tests for means) Linear and logistic regression models used to calculate baseline/endline change for the various outcomes for each comparison group Due to small sample size and contamination resulting from food assistance received during the study period, the non-assistance group was excluded from statistical hypothesis tests and adjusted analyses Difference-in-difference approach used to compare magnitude of change in key outcomes between food voucher and mixed transfer groups To account for non-randomized design, adjusted analyses were conducted: Inverse probability of treatment weighting (IPTW) based on household-specific propensity scores Propensity scores used to generate household-specific stabilized weights to adjust for baseline imbalances between treatment groups Household-specific stabilized weights were used in linear and logistic models to analyze adjusted change from baseline to endline, as well as differences in change between voucher and mixed transfer beneficiaries

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

Household Food Security Household food security at baseline was better in the mixed transfer group when assessed by both household hunger score and meal frequency. Household food security decreased over the intervention period. At endline, food security was similar between the intervention groups with a mean HHS score of 1.8; approximately 80% of households had moderate hunger The decline in HHS score was significantly greater among the mixed transfer group compared to the vouchers, however, when change in HHS category was compared, there were no significant differences All households consumed >1 meal per day at baseline. By endline, 8.0% (CI: 4.3-13.3) of food voucher and 2.2% (CI: 0.8-4.8) of mixed transfer households were consuming one or fewer meals per day (p=0.005).

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

PLW Meal Frequency Women consumed an average of 2.5 meals per day at baseline There was a slight decrease in meal frequency (-0.2 meals/day) among the food voucher group and a slight increase (0.1 meals/day) among the mixed transfer group; these changes were not statistically significant. In adjusted models, women receiving mixed transfers consumed an average of 0.3 (CI: 0.1-0.5) meals per day more than those receiving vouchers at endline. This difference was statistically significant. Food Vouchers Mixed Transfers p-value Meal frequency at baseline 2.5 .834 Meal frequency at endline 2.4 2.6 .009 Adjusted change from baseline -0.2 (-0.3-0.0) 0.1 (0.0-0.2) .001 Adjusted difference btwn groups 0.3 (CI: 0.1-0.5)

PLW Dietary Diversity PLW Dietary diversity was significantly better in the mixed transfer group at both baseline and endline. Improvements in dietary diversity between baseline and endline were not statistically significant for either group, nor was the difference in change between groups. Average dietary diversity increased by 0.5 and 0.7 food groups in the voucher and mixed transfer groups, respectively, during the intervention period. Food Vouchers Mixed Transfers p-value Food groups consumed at baseline 4.8 5.3 .008 Food groups consumed at endline 5.2 6.0 <0.001 Adjusted change from baseline 0.5(0.0-0.9) 0.7 (0.5-1.0) .118 Adjusted difference btwn groups 0.3 (CI: -0.3-0.8) .324

PLW Nutrition - MUAC Mean MUAC was significantly better in the mixed transfer group at both baseline and endline. Mean MUAC increased by 0.9cm and 1.3cm in the voucher and mixed transfer groups, respectively, during the intervention period. These increases were statistically significant. Mean MUAC increased by an average of 0.4cm more in the mixed transfer group as compared to the food voucher group, however, this difference was not statistically significant. Food Vouchers Mixed Transfers p-value Mean MUAC at baseline 24.4 25.2 <0.001 Mean MUAC at endline 25.4 26.5 Adjusted change from baseline 0.9 (0.6-1.3) 1.3 (1.1-1.5) 0.086 Adjusted difference btwn groups 0.4 (CI: -0.1-0.8)

PLW Nutrition - Acute Malnutrition Prevalence Acute malnutrition prevalence at baseline was 0% among both groups due to eligibility criteria (using national cutoff of 21.0cm). At the end of the intervention period, 3.1% (CI: 1.0-7.1%) of women in the food voucher group were acutely malnourished compared to 0.0% (CI: 0.0- 1.4%) of women in the mixed transfer group (unadjusted). In adjusted models, the difference in change between food voucher and mixed transfer groups was 2.9% (CI: -0.4-6.1%), which was of marginal significance (p=0.086) Despite the lack of statistical significance between groups, mixed transfers were fully successful in preventing acute malnutrition among PLWs

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

Children’s Food Frequency Children consumed an average of 2.1 meals per day at baseline, with approximately one-third of children achieving minimum meal frequency; these findings were similar across groups. There were no significant changes in mean meal frequency or the proportion of children achieving minimum meal frequency Within each group over time Magnitude of change between groups Food Vouchers Mixed Transfers p-value Meal frequency at baseline 1.9 2.2 .252 Meal frequency at endline 2.0 2.3 .335 Adjusted change from baseline 0.3 (CI:-0.4-0.61 0.1 (-0.2-1.0) .825 Adjusted difference btwn groups -0.3 (CI: 00.8-1.0)

Children’s Dietary Diversity Children’s dietary diversity was significantly better in the mixed transfer group at baseline; at endline dietary diversity was similar between groups. The decline in dietary diversity in the mixed transfer group and difference in change over time in mean dietary diversity between groups were not significant. At baseline, more children achieved minimum diversity in the mixed transfer group (57% vs. 31%) whereas they were similar at endline (31%). Compared to the food voucher group, the mixed transfer group saw a 9.4% (CI: -8.7 - -0.4) decline in children achieving target dietary diversity over the intervention period; this difference was significant (p=0.031). Food Vouchers Mixed Transfers p-value Food groups consumed at baseline 2.5 3.3 .003 Food groups consumed at endline 2.6 .964 Adjusted change from baseline 0.1 (-0.8-0.2) -0.9 (-1.3- -0.3) .145 Adjusted difference btwn groups -1.0 (-1.3 – 0.2)

Child Nutrition - MUAC Children’s mean MUAC was significantly better in the mixed transfer group at both baseline and endline. Mean MUAC increased by 0.5cm in the voucher group and 0.1cm in the mixed transfer group, however, these changes were not statistically significant. Mean MUAC increased by an average of 0.4cm more in the food voucher group as compared to the mixed transfer group, however, this difference was not statistically significant. Food Vouchers Mixed Transfers p-value Mean MUAC at baseline 13.8 14.7 <0.001 Mean MUAC at endline 14.2 14.6 .009 Adjusted change from baseline 0.5 (0.0-0.7) 0.1 (-0.1-0.4) 0.199 Adjusted difference btwn groups -0.4 (-0.7-0.1)

Child Nutrition - Acute Malnutrition Prevalence Child acute malnutrition at baseline was significantly greater in the food voucher group (11.4%) than the mixed transfer group (6.1%) (p=0.029). At the end of the intervention, 9.3% (CI: 7.4-16.5%) of CU5s in the food voucher group were acutely malnourished compared to 3.1% (CI: 1.4-5.8%) of CU5s in the mixed transfer group (unadjusted) (p=0.003). In adjusted models, magnitude of change was not significant for either intervention; the difference in change between food voucher and mixed transfer groups was 5.5%, which was not statistically significant. Both interventions appeared to be successful at preventing CU5 acute malnutrition. In the non-intervention group, mean MUAC decreased significantly (-1.2cm) and acute malnutrition prevalence increased significantly from 1.5% to 32.5% (unadjusted)

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

Limitations Limited sample size – security and beneficiary numbers made it difficult to recruit large samples of beneficiary households with PLWs. The non-assistance group was difficult to identify due to numerous other ongoing interventions and was contaminated over the life of the intervention period Interventions changed – in terms of transfer amounts and in the mixed transfer group composition (proportion of in-kind assistance varied monthly) Intervention period was shorter than anticipated – a 6 month intervention period was planned, however, lack of continuity in funding meant that endline data collection had to be conducted earlier than anticipated, as one intervention group was going to have a gap in assistance Security precluded travel to study areas – remote supervision was successful but access was a challenge for the study team

1 2 3 4 5 6 7 PRESENTATION OVERVIEW RATIONALE AND OBJECTIVES INTERVENTIONS AND METHODS 3 HOUSEHOLD FOOD SECURITY OUTCOMES 4 PLW DIET & NUTRITION OUTCOMES 5 CHILD DIET & NUTRITION OUTCOMES 6 LIMITATIONS 7 SUMMARY OF KEY FINDINGS

Summary of Findings Household food security decreased over the study period in all groups according to both Household Hunger Score and meal frequency. Among PLWs, meal frequency did not differ between endline and baseline for either group, but dietary diversity improved significantly for both groups. PLW MUAC increased significantly in both groups, but magnitude of change was greater in the mixed transfer group; mixed transfers were fully successful in preventing acute malnutrition whereas 3.1% of voucher PLWs developed acute malnutrition. Differences in magnitude of change between the groups for nutrition outcomes were marginally significant. With respect to children's diet, meal frequency remained constant in both groups and transfers did not improve dietary diversity. Child nutrition improved in both intervention groups in terms of mean MUAC and acute malnutrition prevalence; however, improvements were not statistically significant.

Conclusions Despite decreases in household food security over the study period, PLW dietary quality and mean MUAC improved. No PLWs in the mixed transfer group had become malnourished by the end of the study period. Transfers were not as protective for children’s diet quality. However, child nutrition improved in both intervention groups in terms of mean MUAC (significant) and acute malnutrition prevalence (not significant). This compares to a large decline in nutrition status in the non-assistance group. Results show promise but do not indicate a clear benefit for mixed transfers as compared to food vouchers – likely a result of study limitations. More research or program evaluations needed to deepen understanding – in particular with larger sample sizes and longer intervention periods.