Unconditional cash transfers (UCTs; provided without obligation) for reducing poverty and vulnerabilities (e.g. orphanhood, old age or HIV infection) are a type of social protection intervention that addresses a key social determinant of health (income) in low‐ and middle‐income countries (LMICs). The relative effectiveness of UCTs compared with conditional cash transfers (CCTs; provided so long as the recipient engages in prescribed behaviours such as using a health service or attending school) is unknown.

To assess the effects of UCTs for improving health services use and health outcomes in vulnerable children and adults in LMICs. Secondary objectives are to assess the effects of UCTs on social determinants of health and healthcare expenditure and to compare to effects of UCTs versus CCTs.

We searched 17 electronic academic databases, including the Cochrane Public Health Group Specialised Register, the Cochrane Database of Systematic Reviews (the Cochrane Library 2017, Issue 5), MEDLINE and Embase, in May 2017. We also searched six electronic grey literature databases and websites of key organisations, handsearched key journals and included records, and sought expert advice.

We included both parallel group and cluster‐randomised controlled trials (RCTs), quasi‐RCTs, cohort and controlled before‐and‐after (CBAs) studies, and interrupted time series studies of UCT interventions in children (0 to 17 years) and adults (18 years or older) in LMICs. Comparison groups received either no UCT or a smaller UCT. Our primary outcomes were any health services use or health outcome.

Two reviewers independently screened potentially relevant records for inclusion criteria, extracted data and assessed the risk of bias. We tried to obtain missing data from study authors if feasible. For cluster‐RCTs, we generally calculated risk ratios for dichotomous outcomes from crude frequency measures in approximately correct analyses. Meta‐analyses applied the inverse variance or Mantel‐Haenszel method with random effects. We assessed the quality of evidence using the GRADE approach.

We included 21 studies (16 cluster‐RCTs, 4 CBAs and 1 cohort study) involving 1,092,877 participants (36,068 children and 1,056,809 adults) and 31,865 households in Africa, the Americas and South‐East Asia in our meta‐analyses and narrative synthesis. The 17 types of UCTs we identified, including one basic universal income intervention, were pilot or established government programmes or research experiments. The cash value was equivalent to 1.3% to 53.9% of the annualised gross domestic product per capita. All studies compared a UCT with no UCT, and three studies also compared a UCT with a CCT. Most studies carried an overall high risk of bias (i.e. often selection and/or performance bias). Most studies were funded by national governments and/or international organisations.

Throughout the review, we use the words 'probably' to indicate moderate‐quality evidence, 'may/maybe' for low‐quality evidence, and 'uncertain' for very low‐quality evidence. UCTs may not have impacted the likelihood of having used any health service in the previous 1 to 12 months, when participants were followed up between 12 and 24 months into the intervention (risk ratio (RR) 1.04, 95% confidence interval (CI) 1.00 to 1.09, P = 0.07, 5 cluster‐RCTs, N = 4972, I² = 2%, low‐quality evidence). At one to two years, UCTs probably led to a clinically meaningful, very large reduction in the likelihood of having had any illness in the previous two weeks to three months (odds ratio (OR) 0.73, 95% CI 0.57 to 0.93, 5 cluster‐RCTs, N = 8446, I² = 57%, moderate‐quality evidence). Evidence from five cluster‐RCTs on food security was too inconsistent to be combined in a meta‐analysis, but it suggested that at 13 to 24 months' follow‐up, UCTs could increase the likelihood of having been food secure over the previous month (low‐quality evidence). UCTs may have increased participants' level of dietary diversity over the previous week, when assessed with the Household Dietary Diversity Score and followed up 24 months into the intervention (mean difference (MD) 0.59 food categories, 95% CI 0.18 to 1.01, 4 cluster‐RCTs, N = 9347, I² = 79%, low‐quality evidence). Despite several studies providing relevant evidence, the effects of UCTs on the likelihood of being moderately stunted and on the level of depression remain uncertain. No evidence was available on the effect of a UCT on the likelihood of having died. UCTs probably led to a clinically meaningful, moderate increase in the likelihood of currently attending school, when assessed at 12 to 24 months into the intervention (RR 1.06, 95% CI 1.03 to 1.09, 6 cluster‐RCTs, N = 4800, I² = 0%, moderate‐quality evidence). The evidence was uncertain for whether UCTs impacted livestock ownership, extreme poverty, participation in child labour, adult employment or parenting quality. Evidence from six cluster‐RCTs on healthcare expenditure was too inconsistent to be combined in a meta‐analysis, but it suggested that UCTs may have increased the amount of money spent on health care at 7 to 24 months into the intervention (low‐quality evidence). The effects of UCTs on health equity (or unfair and remedial health inequalities) were very uncertain. We did not identify any harms from UCTs. Three cluster‐RCTs compared UCTs versus CCTs with regard to the likelihood of having used any health services, the likelihood of having had any illness or the level of dietary diversity, but evidence was limited to one study per outcome and was very uncertain for all three.

This body of evidence suggests that unconditional cash transfers (UCTs) may not impact a summary measure of health service use in children and adults in LMICs. However, UCTs probably or may improve some health outcomes (i.e. the likelihood of having had any illness, the likelihood of having been food secure, and the level of dietary diversity), one social determinant of health (i.e. the likelihood of attending school), and healthcare expenditure. The evidence on the relative effectiveness of UCTs and CCTs remains very uncertain.