Asfaw 2014.
Study characteristics | ||
Methods |
Study design: cRCT Study grouping: parallel group How were missing data handled: data from HHs that were LTFU were excluded from the analysis (Ward 2010). For the HH spending study, a further 45 HHs were excluded due to very large values for total adjusted expenditure as well as missing values (The Kenya CT‐OVC team 2012). Randomisation ratio: 1:1 Recruitment method: NR Sample size justification and outcome used: NR Sampling method: within 7 districts, 4 locations (clusters) were selected randomly after excluding those with particularly low poverty rates or an inadequate capacity to supply the relevant health and education services, or large existing orphans and vulnerable children support programmes. List of eligible HHs was compiled in the intervention locations according to standard programme operation guidelines. In control districts, programme targeting was 'simulated' in order to identify a sample of HHs that were comparable to eligible HHs in intervention areas. Study aim or objective: 1. to determine if the Kenya CT‐OVC led to an increase in investment in agricultural and non‐agricultural productive assets and activities; increased food consumption obtained from own production; resulted in a shift in adult labour towards own agricultural and non‐agricultural activities and away from casual labour; resulted in heterogeneous impact by gender; and reduced the time children spend at work. 2. to investigate whether the CT‐OVC had changed the preferences of HHs in terms of their consumption behaviour. Study period: 4 years (March–August 2007 to May–July 2011) Unit of allocation or exposure: location within eligible district |
|
Participants |
Baseline characteristics Overall (n = 1783)
Intervention or exposure group (n = 1265)
Control group (n = 518)
Inclusion criteria: ultra‐poor HHs and contain an OVC (defined as a HH resident aged 0–17 years with ≥ 1 deceased parent, or who was chronically ill, or whose main carer is chronically ill) in selected locations within 4 districts (Nyanza, Nairobi, Garissa, Kwale) Exclusion criteria: NR Baseline imbalance: in Asfaw et al 2014: intervention HHs had significantly older HH heads, more female‐ and elderly‐headed HHs, smaller HH size, lower education of HH head and spouse, fewer number of young and middle‐aged HH members and ill HH heads, and more elderly HH members. Intervention HHs were also less likely to use unprotected water sources, have various HH assets, but more HHs in which agriculture was the main source of income and less in which salaried employment was the main source of income. Intervention communities had more access to a road to the village and less distance to the local market, but a lower share of HHs which could make telephone calls. Intervention individuals were significantly older and more likely to be disabled, unemployed or in casual work; and had lower levels of education. For the HH spending paper (Kenya CT‐OVC 2012): intervention HHs expended significantly more of their monthly budget on tubers. Attrition per relevant group: at 24 months (Asfaw 2014): intervention group: 231/1542 (15%); control group: 184/571 (24.4%). Attrition by district (intervention clusters vs control clusters) Garissa (11.6% vs 31.6%); Homabay (14% vs 21.1%); Kisumu (15% vs 25.3%); Kwale (13.2% vs 16.1%); Migori (16.3% vs 18.7%); Nairobi 16.9% vs 47.6%); Suba (10.9% vs 20.8%) (Ward 2010). Study authors reported that the loss of HHs was partly due to postelection violence. At 48 months (Asfaw 2014): intervention group: 262/1542 (16.9%); control group: 224/755 (29.7%). The HH spending paper reported only total attrition, i.e. 16.9% (387/2294). Further post‐hoc exclusion of 45 HHs due to large values for total adjusted expenditure and missing values leads to a total 'attrition' of 18.8%. Description of subgroups measured and reported: HH size 5 vs ≥ 5; female‐headed vs male‐headed HHs Total number completed and analysed per relevant group: in the HH spending paper (Kenya CT‐OVC 2012) 1907 HHs completed baseline and 2‐year (2009) follow‐up, but only 1862 HHs were analysed due to post‐hoc exclusions. In the food consumption paper (Asfaw et al 2014), 1280 intervention and 531 control HHs completed baseline and 4‐year (2011) follow‐up, but only 1265 intervention and 518 control HHs were included. Total number enrolled per relevant group: intervention group: 1542 HHs; control group: 755 HHs Total number randomised per relevant group: 4 locations within each of 7 districts randomised to either the intervention or control group. Intervention group: 14 locations (clusters); control group: 14 locations (clusters) |
|
Interventions |
Intervention: UCT
Control: no intervention |
|
Outcomes | Proportion of HH expenditure on food: total monthly consumption expenditure (per AE); proportion of HH monthly consumption expenditure on food; total monthly food expenditure (cereals, tubers, meat/fish, dairy, fruit/vegetables, other food, food eaten out) Dietary diversity: DDS; proportion of HHs that consumed individual food groups (e.g. cereal, fruit, etc.) in the preceding 7 days Anthropometry: HAZ; WAZ; WHZ; stunting; underweight; wasting Morbidity: number of children with reported symptoms of upper respiratory illness |
|
Identification |
Sponsorship source: UK DfID; US National Institute of Mental Health; Eunice Kennedy Shriver National Institute of Child Health and Development Country: Kenya Setting: ultra‐poor HHs in rural areas with high prevalence of HIV/AIDS Authors' names: Solomon Asfaw; Tia Palermo; Patrick Ward Email: solomon.asfaw@fao.org; tiampalermo@gmail.com; patrick.ward@opml.co.uk Declarations of interest: no Study or programme name and acronym: Kenya CT‐OVC (Cash Transfer Programme for Orphans and Vulnerable Children) Type of record: journal articles, operational and impact evaluation report Trial registration: none reported Protocol availability: no |
|
Notes | Population: AE: children aged < 15 years were considered as 0.75 of an adult; children aged ≥ 15 years were considered a full adult. | |
Risk of bias | ||
Bias | Authors' judgement | Support for judgement |
Random sequence generation (Selection bias) | Unclear risk | Authors mentioned random selection of intervention and control locations within 7 districts, but did not report how the random sequence was generated. |
Allocation concealment (Selection bias) | Unclear risk | No information provided on how the locations within districts were randomised; or how this randomisation sequence was protected. |
Baseline characteristics similar (Selection bias) | Low risk | Baseline differences reported between groups. The heads of treatment HHs were older (P ≤ 0.01), more likely to be male (P < 0.05) and to have less education (P < 0.01) than those in control HHs. Intervention HHs also had significantly fewer assets (99% CI). The proportion of control HHs that had agriculture as the main source of income was significantly lower (90% CI). The 24‐month analyses adjusted for baseline HH demographic composition (Kenya CT‐OVC team 2012), whereas the analysis after 4 years used estimated propensity scores to account for these baseline differences (Asfaw 2014). |
Baseline outcome measurements similar (Selection bias) | Low risk | No baseline non‐equivalence was detected for the HH spending (Kenya CT‐OVC 2012) paper; with the exception of proportion AE monthly expenditure for tubers which was significantly higher in the intervention group (P = 0.005). Outcomes such as HH expenditure on food and proportion of children with underweight or stunting, or both (< –2SD z‐scores) were similar between the groups. There was no baseline non‐equivalence for the HH spending paper; with the exception of proportion AE monthly expenditure for tubers, which was significantly higher in the intervention group (P = 0.005). All outcomes for the food consumption paper were NR for the baseline survey. |
Blinding of participants and personnel (Performance bias) | Low risk | Given the way in which the intervention was rolled out, it is not possible for participants to be blinded. However, it is unlikely that lack of blinding influenced behaviour of participants and personnel beyond that expected by the intervention. |
Blinding of outcome assessment (Detection bias) | High risk | There was no blinding. As outcomes were predominantly self‐reported it is likely that the lack of participant blinding would have affected the measurement of outcomes. |
Protection against contamination (Performance bias) | High risk | Quote: "Taylor et al. (2012) simulated the local economy impact and revealed a minimal inflationary impact and real production value added multipliers of Ksh1.58 [KES] per shilling transferred, which suggests that the programme may have led to spillovers." |
Incomplete outcome data (Attrition bias) | High risk | High attrition and no ITT analysis performed. Differential attrition between the 2 groups. |
Selective outcome reporting (Reporting bias) | Unclear risk | Study protocol N/A. |
Other bias | High risk | Misclassification bias: high risk due to self‐report of CCT receipt by HHs. Measurement bias: unlikely. Incorrect analysis: unlikely. Recruitment bias: high risk. Clusters were assigned before recruitment of HHs; which may have lead to a bias in participation, especially for control HHs. Other bias: the introduction of punitive conditionalities in some intervention clusters, but not others, may have lead to bias in attrition or bias in outcome measurement. |