Miller 2011.

Study characteristics
Methods	Study design: cRCT Study grouping: parallel group How were missing data handled? Only participants for which there were data for all 3 data collection rounds, were analysed Randomisation ratio: 1:1 Recruitment method: research team attempted to interview all HH heads in the sampling frame by setting up appointments, meeting respondents at their homes or fields, and returning to HHs ≥ 3 times. Sample size justification and outcome used: sample size for this study was determined considering the wide range of expected impacts at the child and HH level. Used the software package R (Version 2.11.1) to perform posthoc power calculations and confirm there was sufficient power to detect statistically significant results at or exceeding the standard 0.80 level for each analysis. Sampling method: 8 village groups were randomly assigned to the intervention or control group. The sampling frame consisted of all cash transfer targeted HHs in these village groups (about 100 villages per village group). Study aim or objective: to examine the impact of the SCTS on food security and diversity at the HH level. Study period: March 2007 to April 2008 Unit of allocation or exposure: village groups
Participants	Baseline characteristics Intervention or exposure Age: HH head, years: 61 Place of residence: NR Sex: female‐headed HH, %: 63 Ethnicity and language: NR Occupation: NR Education: received by HH head, %: no schooling: 44; some primary schooling: 45; some secondary schooling: 1 SES: HH size, n: 4.7; HHs with ≥ 1 chronically ill member, %: 42; HHs with ≥ 1 disabled member, %: 30; HHs with ≥ 1 orphans, %: 74; elderly‐only HHs, %: 12; had a death in the HH in past 5 years, %: 35. Type of housing, %: grass 4; mud 71; mud or burnt brick 26; no toilet, %: 65; pit latrine with no ventilation, %: 35 Social capital: free food maize distribution, %: 1; food for work programme or inputs for work: 0; supplementary inputs for malnourished children: 1; agricultural inputs or other: 1 Nutritional status: FDS: 5; expenditure on food: MWK 129 per week, MWK 24 per capita per week; total expenditure on food, %: 56; number of meals the day before: 1.46 Morbidities: members with HIV, %: 2; members with disability, %: 21; chronically ill adults, %: 35 Concomitant or previous care: NR Control Age: HH head, years: 63 Place of residence: NR Sex: female headed HHs, %: 66 Ethnicity and language: NR Occupation: NR Education: HH head, %: no schooling: 65; some primary schooling: 35; some secondary schooling: 0 SES: HH size, n: 3.5; HHs with ≥ 1 chronically ill member, %: 35; HHs with ≥ 1 disabled member, %: 28; HHs with ≥ 1 orphans, %: 68; elderly‐only HHs, %: 22; had a death in the HH in the past 5 years, %: 26. Type of housing, %: grass 2; mud 41; mud or burnt brick 56; no toilet, %: 64; pit latrine with no ventilation, %: 35 Social capital: free food maize distribution, %: 1; food for work programme or inputs for work: 0; supplementary inputs for malnourished children: 1; agricultural inputs or other: 1 Nutritional status: FDS: 5; expenditure on food: MKW 122 per week, MKW 18 per capita per week; total expenditure on food, %: 52; number of meals the day before: 1.49 Morbidities: members with HIV, %: 3; members with disability, %: 22; chronically ill adults, %: 31 Concomitant or previous care: NR Overall: NR Inclusion criteria: ultra poor HHs in selected villages in the Mchinji district (within the lowest economic quintile, having no assets, or consuming only 1 meal and labour constrained) (dependency ratio > 3, or undefined) Exclusion criteria: child‐headed HHs Pretreatment: intervention group had a greater number of HHs with orphans, whereas the control group had a greater number of elderly headed HHs. According to the study authors these differences were due to differences observed during the targeting process. A greater number of HH heads had no schooling in the intervention group, compared to the control group. HH size was higher in the control group. Attrition per relevant group: intervention group: 42/408 (10.3%); control group: 25/411 (6%); most common reason for LTFU was death (7 deaths in intervention group and 16 deaths in control group). Description of subgroups measured and reported: NR Total number completed and analysed per relevant group: number of HHs: intervention group: 366; control group: 386 Total number enrolled per relevant group: number of HHs: intervention group: 385; control group: 411. 17 intervention HHs were found not to be eligible during recruitment. Total number randomised per relevant group: number of village groups: intervention group: 4; control group: 4
Interventions	Intervention characteristics Intervention or exposure Food access intervention category: increase buying power Intervention type: UCTs Description: HHs received a cash payment (on average MKW 2000 (USD 14) per month), depending on HH size and the number of school‐aged children. Top‐up payments included payments of MKW 200 for children at primary school and MKW 400 for those at secondary school. Duration of intervention period: 12 months Frequency: monthly Number of study contacts: 3 (baseline, 6 and 12 months after first cash payment) Providers: government of Malawi Delivery: NR Co‐interventions: NR Resource requirements: NR Economic indicators: total programme expenditure in Mchinji district (USD 121,000 per month) by February 2009 Control: no intervention for duration of the study; however, after the study (May 2008), the control group also received the intervention. Co‐interventions: single food bucket valued at USD 8.80 containing oil, sugar, tea, salt, soap and beans before the midline follow‐up.
Outcomes	Proportion of HH expenditure on food; proportion of total weekly expenditure on food; weekly HH food expenditure; per capita total weekly expenditures Food security: proportion of HHs reporting not consuming enough food per day/eating ≥ 2 meals per day/reporting hunger after meals/reporting insufficient food for > 8 days/month Dietary diversity: food diversity composite score based on number of food groups consumed in the past week.
Identification	Sponsorship source: USAID and UNICEF Country: Malawi Setting: HHs that were ultra‐poor and labour constrained; Mchinji district Author's name: Candace M Miller Email: candace@bu.edu Declarations of interest: yes; none reported. Study or programme name and acronym: Malawi Social Cash Transfer Scheme (SCTS) Type of record: journal article, policy research working paper
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Unclear risk	Method to generate random sequence NR.
Allocation concealment (Selection bias)	Unclear risk	Allocation concealment NR.
Baseline characteristics similar (Selection bias)	Low risk	Some baseline characteristics differed between the groups: intervention group had fewer HH heads with no schooling (P < 0.0001), larger HH size (P < 0.0001), fewer elderly‐only HHs (P < 0.001). The study authors used the DID method to account for these differences in their analysis.
Baseline outcome measurements similar (Selection bias)	Low risk	Quote: "… the outcome variables of interest were statistically similar between study households at baseline, so that these groups generally experienced the same level of food insecurity" at baseline.
Blinding of participants and personnel (Performance bias)	Low risk	Participants were not blinded, but this was unlikely to have influenced the participant behaviour.
Blinding of outcome assessment (Detection bias)	High risk	Unclear whether the data collectors were blinded towards the group allocations. Outcomes were self‐reported, which could have been influenced by lack of blinding.
Protection against contamination (Performance bias)	Low risk	HHs were monitored by the district to prevent contamination.
Incomplete outcome data (Attrition bias)	Low risk	No clusters were lost. In terms of HHs: 42/408 (10.3%) from the intervention group and 25/411 (6.1%) from the control group were not part of the final analysis.
Selective outcome reporting (Reporting bias)	Unclear risk	Protocol citation or trial registration number NR. All expected outcomes were reported on in the methods section, and reported on in the results section.
Other bias	Unclear risk	Misclassification bias: unlikely. Measurement bias: low risk. Repeated 7‐day dietary recall administered by trained research assistants: (quote) "We trained the team of research assistants (RAs) over 1 week prior to each round of data collection." "Study supervisors checked surveys on a daily basis for consistency and completeness." Incorrect analysis: low risk. Quote: "These villages are homogeneous however; and as was expected, simple OLS regression and regression models that account for village level clustering yielded nearly identical effect sizes and standard errors". Seasonality bias: low risk. Recruitment bias: low risk. The study authors stated that baseline differences between the 1 groups most likely resulted from differences in the prioritisation of either elderly‐only HHs or HHs with orphans in the intervention group during targeting process. However, analyses adjusted for these differences.