Gangopadhyay 2015.

Study characteristics
Methods	Study design: RCT Study grouping: parallel group How were missing data handled? Midline survey results indicated that only 4 HHs that received the cash transfer (4%) did not want to continue, so dropped out of the analysis. Missing data were excluded from the analyses. Randomisation ratio: 1:1:1 – bank account and UCT (100 HHs); bank account and no UCT (100 HHs); no bank account or UCT (100 HHs) Recruitment method: study authors collaborated with the Self‐Employment Women's Association to explain the experiment to a 12‐block community of Raghubir Nagar. Awareness campaign ran for 2 weeks (first week of August 2010 to 13 August 2010). It initially targeted groups of 15–20 people, but, because ration shop owners tried to influence people to avoid participation in experiment, group sizes were reduced to 5 or 6 people at a time, to make their participation less noticeable. Sample size justification and outcome used: NR Sampling method: random selection of 350 HHs that had agreed to participate, of which 50 dropped out. These 300 HHs were selected for treatment group and control groups 1 and 2. Random selection of 150 HHs that did not want to participate were selected for control group 3. Study aim or objective: to compare effects of replacing welfare transfers in‐kind with a UCT on food security (measured as food consumption) to determine the impact of the cash transfer and the bank account. Study period: January 2011 to December 2011 Unit of allocation or exposure: HHs
Participants	Baseline characteristics Intervention: bank account + UCT Age: NR Place of residence: poor community in Delhi Sex: NR Ethnicity and language: NR Occupation, %: HHs self‐employed: 44; regular salary employed: 34; casual labour: 15; other: 7 Education: education level of HH head, %: primary: 78; secondary: 19; above secondary: 3 SES: poor HHs BPL Social capital: NR Nutritional status: data reported were unclear; per capita Kcal consumption, mean: 47,480 (SD 23,004) Morbidities: NR Concomitant or previous care: previous care: PDS BPL card – in‐kind food transfer Control 1: bank account and no cash transfer Age: NR Place of residence: poor community in Delhi Sex: NR Ethnicity and language: NR Occupation, %: HHs self‐employed: 36; regular salary employed: 39; casual labour: 22; other: 3 Education: education level of HH head, %: primary: 62; secondary: 36; above secondary: 2 SES: poor HHs BPL Social capital: NR Nutritional status: per capita Kcal consumption, mean: 43,954 (SD 13,446). Morbidities: NR Concomitant or previous care: previous care: PDS BPL card – in‐kind food transfer Control 2: no bank account + cash transfer Age: NR Place of residence: poor community in Delhi Sex: NR Ethnicity and language: NR Occupation, %: HHs self‐employed: 43; regular salary employed: 30; casual labour: 23; other: 4 Education: education level of HH head, %: primary: 68; secondary: 30; above secondary: 2 SES: poor HHs BPL Social capital: NR Nutritional status: per capita Kcal consumption, mean: 47,398 (SD 21,908) Morbidities: NR Concomitant or previous care: previous care: PDS BPL card – in‐kind food transfers Control 3: random selection of 150 HHs that did not want to participate. Not included in analyses. Overall Age: NR Place of residence: poor community in Delhi Sex: NR Ethnicity and language: NR Occupation: NR Education: NR SES: poor HHs BPL Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: previous care: PDS BPL card – in‐kind food transfers Inclusion criteria: community in Delhi that received Government's BPL card (part of PDS programme; in‐kind transfer programme) Exclusion criteria: participants who were not BPL cardholders and lived outside of Raghubir Nagar Pretreatment: groups of self‐selected HHs were similar, such that randomisation apparently resulted in balanced groups (according to Table 3). Per capita expenditure on non‐food items was significantly different between the control groups 1 and 2; which was linked to differences in personal care expenditures between men and women. Attrition per relevant group: attrition in treatment group: 6 (6%); control group 1: 3 (3%); control group 2: 9 (9%); control group 3: 14 (9.3%) Description of subgroups measured and reported: N/A Total number completed and analysed per relevant group: total number completed and analysed in treatment group 94; control group 1: 97; control group 2: 91; control group 3: 136 Total number enrolled per relevant group: total number enrolled to participate 300 HHs. 100 HHs per group (intervention group, control group 1 and control group 2) Total number randomised per relevant group: 100 HHs per group (intervention group, control group 1 and control group 2)
Interventions	Intervention characteristics Intervention: bank account and UCT Food access intervention category: increase buying power Intervention type: UCT Description: bank accounts opened in name of women in the HH and UCTs started. Government stamps on ration cards indicated that they could not use their rations for 1 year; instead, they received a monthly cash transfer of INR 1000 (about USD 18), with no conditions on how to spend it, Deposited every month, from January 2011 and ending in December 2011. Duration of intervention period: 12 months: January–December 2011 Frequency: monthly Number of study contacts: 3: baseline, midline (July 2011) and endline (December 2011) Providers: researchers Delivery: bank accounts created for cash transfers. UCT included an exit option for all recipients after 6 months, which was important because the UCT replaced a public programme to which HHs already had access. Therefore, given option to go back to the PDS (4% did, but 96% wanted to continue UCTs and not in‐kind transfers). According to author analyses, it appeared that money transferred was used to buy food, as they showed that after initiating UCTs there were no changes to non‐food items or alcohol. Co‐interventions: none Resource requirements: NR Economic indicators: monthly cash transfer of INR 1000 Control 1: bank account and no cash transfer Food access intervention category: increase buying power Intervention type: bank account opened Description: bank accounts opened in the name of the women in HHs but no cash transfers done. Duration of intervention period: 12 months: January 2011 to December 2011 Frequency: once Number of study contacts: 3: baseline, midline (July 2011) and endline (December 2011) Providers: researchers Delivery: other than creation of bank accounts nothing else reported Co‐interventions: NR Resource requirements: NR Economic indicators: NR Control 2: no bank account or cash transfer Control 3: 150 HHs that did not want to participate, therefore no bank account or cash transfer. Not included in analyses.
Outcomes	Per capita expenditure on non‐cereal food items (pulses, milk, eggs, fish and meat, fruits and vegetables) Per capita calories consumed from cereals
Identification	Sponsorship source: NR Country: India Setting: poor communities in Raghubir Nagar (West Delhi) Author's name: Robert Lensink Email: b.w.lensink@rug.nl Declarations of interest: yes; no potential conflict of interests Study or programme name and acronym: N/A Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Unclear risk	Authors mentioned this was an RCT but did not describe how the random sequence was generated.
Allocation concealment (Selection bias)	High risk	Unit of allocation were HHs and HHs self‐selected into intervention group.
Baseline characteristics similar (Selection bias)	Unclear risk	Although authors stated that characteristics of HHs were similar, percentages were reported with no CIs.
Baseline outcome measurements similar (Selection bias)	Low risk	No important differences in outcomes were present at baseline according to author's analyses.
Blinding of participants and personnel (Performance bias)	Low risk	No blinding done, but it was unlikely to have influenced participant's and personnel behaviour during trial.
Blinding of outcome assessment (Detection bias)	High risk	No blinding was done and outcomes were based on self‐reports, which may have been influenced due to knowledge of treatment allocation.
Protection against contamination (Performance bias)	Low risk	Intervention was delivered as planned in intervention and control groups, and no‐one in control group received the intervention, aside from those that chose to be excluded from the study after it started (4 HHs).
Incomplete outcome data (Attrition bias)	Low risk	Comment: there was little attrition in all groups. Authors tested whether attrition was random through (quote) "… estimating a logit model that explained attrition using a vector of baseline variables …" and concluded that "… there is little reason to anticipate that our statistical results might be compromised by non‐random attrition biases."
Selective outcome reporting (Reporting bias)	Unclear risk	No protocol available. Outcomes were not explicitly stated in the Methods section of the paper
Other bias	Unclear risk	Misclassification bias: unlikely. Measurement bias: high risk. Measurement of outcomes was through a questionnaire; it was unclear which specific tools were used to ascertain dietary intake and HH expenditure and whether these had been validated.