Daidone 2014.
| Study characteristics | ||
| Methods |
Study design: cRCT Study grouping: parallel group How were missing data handled? investigated attrition at 24‐month follow‐up by testing for similarities at baseline between 1. treatment and control groups for all non‐missing HHs (differential attrition) and 2. all HHs at baseline and the remaining HHs at the 24‐month follow‐up (overall attrition). Testing these groups on baseline characteristics can assess whether the benefits of randomisation are preserved at follow‐up. There was no significant differential attrition at 24‐month follow‐up, meaning that benefits of randomisation were preserved. There were small differences between study population at baseline and those that remained at 24‐month follow‐up; the remaining HHs were less likely to have experienced a shock, especially flooding or drought at baseline, and they consumed a higher proportion of maize over cassava. The differences from overall attrition were primarily driven by the lower response rate in Kaputa district. Randomisation ratio: about 1:1 Recruitment method: 90/300 CWACs in the 3 districts were randomly selected and ranked through a lottery to be considered in the programme. In second phase, CWAC members and Ministry staff identified all eligible HHs with ≥ 1 child under the age of 3 years living in these 90 randomly selected communities. This resulted in > 100 eligible HHs in each of the CWACs. Sample size justification and outcome used: power analysis completed to ensure study size was able to detect meaningful effects. Sampling method: randomised phase‐in method that included several levels of random selection. First, 90/300 CWACs in the 3 districts were randomly selected and ranked through a lottery to be considered in the programme. In second phase, CWAC members and Ministry staff identified all eligible HHs with ≥ 1 child under the age of 3 years living in these 90 randomly selected communities. This resulted in > 100 eligible HHs in each of the CWACs. After implementing a power analysis to ensure the study was able to detect meaningful effects, 28 HHs were randomly selected for inclusion in the evaluation from each of the 90 communities. This yielded a final study sample of > 2500 HHs. Baseline data collection carried out before CWACs were randomly assigned to treatment and control. Importantly, neither the HHs nor the enumerators knew who would benefit first and who would benefit later. Randomisation was concluded with the flip of a coin and was carried out in public with local officials, Ministry staff and community members. Study aim or objective: CGP has 6 specific objectives: 1. supplement and not replace HH income; 2. increase the number of children enrolled in and attending school; 3. reduce the rate of mortality and morbidity among children aged < 5 years; 4. reduce stunting and wasting among children aged < 5 years; 5. increase the number of HHs owning assets such as livestock and 6. increase the number of HHs that have a second meal a day. Study period: 3‐year RCT; began in December 2010 and ended in 2013. Evaluation occurred at 24‐month follow‐up. Unit of allocation or exposure: HHs |
|
| Participants |
Baseline characteristics Intervention or exposure
Control
Overall
Inclusion criteria: any HHs with a child aged < 5 years in 3 districts (Kalabo, Kaputa and Shangombo) that had not participated in a previous cash transfer programme. Exclusion criteria: NR Pretreatment: treatment HHs were slightly larger than the control group. HHs in Kaputa were bigger compared to the other 2 districts. Attrition per relevant group: NR. Overall attrition rate 8.8% (Diadone 2014) and 9% (Seidenfeld 2013). Attrition rate for treatment 8.1%; attrition rate for control 0.9% (Seidenfeld 2013). Description of subgroups measured and reported: subgroup analyses by districts (Kalabo, Kaputa and Shangombo). Total number completed and analysed per relevant group: total 2519 HHs. Calculated based on response rate; intervention 1158 (91.9%); control 1141 (90.6%) Total number enrolled per relevant group: treatment 1260 HHs (7254 individuals); control 1259 HHs (7091 individuals); total 2519 HHs (14,345 individuals) Total number randomised per relevant group: NR |
|
| Interventions |
Intervention characteristics Intervention or exposure
Control: no intervention |
|
| Outcomes | Proportion of HH expenditure on food: HH monthly expenditure on food and expenditure on individual food groups Food security: proportion people eating > 1 meal/day; proportion not severely food insecure; proportion who ate meat/fish ≥ 5 times in last month HHFIAS Dietary diversity: HDDS (0–12) Anthropometry: WAZ; HAZ; WHZ Cognitive function and development: ECD Index Morbidity: proportion of children aged 0–60 months with ARI; proportion of children aged 0–60 months with diarrhoea |
|
| Identification |
Sponsorship source: From Protection to Production (PtoP) programme, jointly with the UNICEF, is exploring the linkages and strengthening co‐ordination between social protection, agriculture and rural development. The PtoP is funded principally by the UK DfID, the FAO of the UN and the EU. The programme is also part of the Transfer Project, a larger effort together with UNICEF, Save the Children and the University of North Carolina, to support the implementation of impact evaluations of cash transfer programmes in sub‐Saharan Africa. Country: Zambia Setting: communities in the 3 poorest districts of Zambia: Shangombo, Kalabo and Kaputa Comments: both Daidone 2014 and Seidenfeld 2013 used for data extraction. Seidenfeld 2013 was the official programme impact report and more useful in the population and intervention extraction, while Diandone was more useful in the methods extraction. Author's names: Silvio Daidone, David Seidenfeld, Sudhanshu Handa, Benjamin Davis Email: dseidenfeld@air.org; shanda@email.unc.edu; benjamin.davis@fao.org Declarations of interest: NR Study or programme name and acronym: Zambian Child Grant Programme (CGP) Type of record: research report |
|
| Notes | ||
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Random sequence generation (Selection bias) | Low risk | The CGP impact evaluation was designed as an RCT using a randomised phase‐in method (Duflo et al 2008) that included several levels of random selection. First, 90/300 CWACs in the 3 districts were randomly selected and ranked through a lottery to be considered in the programme. In second phase CWAC members and Ministry staff identified all eligible HHs with ≥ 1 child aged < 3 years living in these 90 randomly selected communities. |
| Allocation concealment (Selection bias) | Low risk | Allocation at CWAC level at start of study. Importantly, neither HHs nor enumerators knew who would benefit first and who would benefit later. |
| Baseline characteristics similar (Selection bias) | Low risk | Randomisation was successful, as mean characteristics were balanced across groups (Table 2). |
| Baseline outcome measurements similar (Selection bias) | Low risk | At baseline, majority of indicators were not statistically different at the conventional 5% significance level, with 10 exceptions out of 71 (Table 2). 4 indicators had standardised differences > 10, but they were all < 15. |
| Blinding of participants and personnel (Performance bias) | Low risk | Blinding not done but unlikely to influence behaviour of personnel and participants. |
| Blinding of outcome assessment (Detection bias) | High risk | Neither HHs nor enumerators knew who would benefit first by receiving the case grant (treatment) and who would benefit later by receiving the case grant after the RCT (control) at baseline. However, blinding was not possible. Some outcomes were subjective and could have been influenced by knowledge of intervention allocation. |
| Protection against contamination (Performance bias) | Unclear risk | Possible that increases in treatment HHs agricultural productivity could have had a spillover effect on controls as they resided in the same community. |
| Incomplete outcome data (Attrition bias) | Low risk | Seidenfeld et al (2013) investigated in detail both differential and overall attrition. Differential attrition relates to baseline characteristics between treatment and control HHs that remain at follow‐up. Overall attrition looked at similarities at baseline between the full sample of HHs and the non‐attriters. They found no significant differential attrition after 24 months, meaning that the benefits of randomisation were preserved. The differences in overall attrition were primarily driven by the lower response rate in Kaputa district. |
| Selective outcome reporting (Reporting bias) | Unclear risk | Authors did not refer to a protocol. |
| Other bias | Low risk | None identified. |