Conditional cash transfers to improve use of health facilities by mothers and newborns in conflict affected countries, a prospective population based intervention study from Afghanistan

There have been considerable improvements in maternal and newborn health across Asia [1, 2]. However, inequities in access and use of maternal and newborn health services remain widespread. Barriers include: cost, distance, perceived quality of care and poor awareness of healthcare options [1, 2]. ‘Demand side financing’ has been used for many years to incentivise families to use primary care services [3]. The type of scheme varies from unconditional transfers (e.g. money provided following the birth of a child), vouchers (reimbursable for a range of services) and conditional cash transfer (CCT) schemes (payment for specific services) [4].

Many low and middle income countries have implemented CCT schemes. Two systematic reviews have examined impact on use of maternal and child health services in the last 10 years including studies from 17 countries [5, 6]. The authors of the systematic reviews concluded that CCT can have important short term impacts on antenatal care and facility delivery but that effects are dependent on family ‘exposure’ and awareness of the scheme, access to transport, efficiency of payments and staffing in health facilities [5, 6]. Monitoring of non-incentivised services is also important as coverage of non incentivised services may decrease as more attention is provided to incentivised services. However, to date there appears to be only one small study from India which specifically assessed the effects of CCT on non incentivised maternity services (antenatal care [ANC] and postnatal care [PNC]) [7].

No CCT studies also appear to have been conducted in conflict affected areas, though one study in Afghanistan has assessed feasibility [8]. We recently reported that barriers to MCH use in Afghanistan are similar to other Asian countries but lack of decision making ‘power’ of women in the family and the presence of anti-government elements in small villages are also important determinants [9, 10]. Families in conflict affected countries such as Afghanistan are likely to require additional assistance to enable them to use financial incentives. This includes community level support [11].

Thus, in 2015, the Ministry of Public Health (MoPH) in Afghanistan in partnership with Unicef, developed a CCT intervention to assist rural Afghanistan women to access health facilities for maternal and newborn health services in rural areas.

The primary objective of this study was to assess the effectiveness of the MoPH CCT intervention on use of health facilities for delivery. Secondary objectives were to assess effects on (i) maternal care seeking for non-incentivised services such as ANC and PNC and (ii) community health worker (CHW) motivation to perform home visits. In addition we assessed effects in the poorest and least poor quintiles.

Baseline rates of MCH service use were concerningly low in our study area in rural Afghanistan, only 61% of women delivered in a health facility and 76% received at least one ANC visit. Despite increasing levels of conflict, CCT provided to women aged 16–49 years in our study was associated with increases in maternal and newborn service use and CHW home visiting, though only the change in ANC was statistically significant. There was some evidence of greater effect in the least poor quintile compared to the poorest quintile. Program exposure was low, especially in the poorest quintile.

Overall, there was only a 3% increase in facility delivery associated with the CCT intervention, and this increase was not statistically significant. This contrasts with effects of 10–20% reported in other Asian studies [5, 6]. We feel that low CCT program exposure was an important reason for the modest effect of our intervention. Only 27% of women in intervention areas reported that they had heard about financial incentives or our CCT program. When we performed a ‘per protocol’ analysis (i.e. an analysis restricted to women who knew they would receive money if they delivered in a health facility) [26], the effect size increased though results were still not statistically significant. It is important to understand the reasons for the limited impact of our CCT intervention on institutional delivery (i.e. the incentivized behavior). An indepth analysis of ‘implementation’ ‘process’ data is needed, i.e. to understand the ‘reach’ of our IEC delivery channels including billboards, brochures, CHWs and village leaders. This will be reported in a separate manuscript.

Prior to the onset of the study we conducted detailed formative research and our interviewees (women of reproductive age, male heads of households, village leaders) provided consistent feedback that home visiting from CHWs would be of most use to families. An important part of our planned exposure was home visits from CHWs to raise awareness and to plan transport and assistance for facility delivery and cash transfer. Our CHW data demonstrate the complex nature of working in Afghanistan. Conflict has increased over the 12 months since program implementation [14], and has restricted the movements of the Afghanistan volunteer CHW workforce [28]. In our study, CHW home visits reduced in both intervention and control areas over the intervention period but the decrease was greater in the control areas. This suggests that our CCT intervention may have improved CHW motivation in the intervention areas despite the increasing conflict. Ongoing Afghanistan HMIS surveillance [17] will help to understand further trends. We also feel that our CHWs could be better trained in home visiting as a recent study indicates this is an important way of accessing pregnant and postpartum women in Afghanistan and can significantly improve ANC, facility delivery and PNC [29]. Due to funding constraints we only allocated one day to the CHW training and had to cover many topics from record keeping to IEC messages. Future work must include longer training periods and pre and post training assessments.

We actively monitored impact on ANC and PNC services because we were concerned that their coverage might decrease. However, surprisingly, our intervention was associated with improvements in both ANC and PNC coverage, though only the change in ANC was statistically significant. This is most likely due to the effectiveness of the ANC and PNC messages that were included in our IEC campaigns. Our findings are also consistent with previous studies in Afghanistan reporting the effectiveness of basic ANC and PNC messages communicated by CHWs [29, 30]. Also, in Asian studies, mothers report that they prefer travelling for ANC visits compared to PNC or facility delivery because there is more choice of clinic site and timing [31]. Our formative research indicates that the situation is similar in Afghanistan [9].

Disappointingly the effect of our CCT intervention was consistently lowest in the poorest quintile. The low rate of program exposure in the poorest women was also of concern. Poor uptake of CCT interventions in the most disadvantaged communities is reported in many other studies [3, 31]. The importance of designing CCT programs so they are appropriate to the local context and meet the needs of the poorest families is clearly described [31]. We included remote areas and trained CHW to prioritise the poorest families in their home visiting. However, the CHWs reported difficulties in accessing the poorest families due to lack of trust and the distances they needed to travel. More work is needed in our study area to ensure that our interventions are truly progressive and understand the perspectives of the most disadvantaged families.

We are very aware that quality of health service provision is a driver of demand [32, 33]. At the health facility level our CCT intervention focused on efficiency of payments to the family and CHW. We did implement a HSS package in both the intervention and control health facilities but this focused on training and quality of care and we were not able to increase staffing levels. The mothers who participated in our formative research reported that sometimes no midwives were present to assist with deliveries after hours [9]. In Afghanistan, due to low wages, some doctors and nurses work in private clinics in the afternoons and evenings. However, in our study, these problems were likely to be non-differential across intervention and control groups. Most importantly, the MoPH is working to improve workforce conditions and staffing levels for MCH services.

We encountered other challenges especially conflict reducing access to onsite training of CHWs and community leaders. This delayed the start of the project by six months. Monitoring is a challenge for all programs across Afghanistan due to conflict and mountainous regions [34]. However, in our study we included additional funds for overnight accommodation and for male staff to accompany our largely female data collectors and supervisors. We also included a dedicated independent field worker to monitor the cash transfer.

Our study had some limitations. It was non-randomised. Our study findings can only be generalised to public health facilities because there were no private facilities in our study area. Intervention and control districts were purposively chosen to ensure similar socio economic status of families but there were still baseline differences in socio demographic characteristics in the intervention and control districts. However, ‘difference-in-differences’ analyses were used to account for potentially differing trends in control and intervention areas. We also adjusted for all potential confounding factors though residual confounding was still possible. We were not able to collect cost data from families or perform a cost effectiveness analysis. Outcome data were self-reported and verification was not possible due to project restrictions, however we consider that any recall bias was likely to be non-differential. We also only allowed 12 months for implementation and we are aware that CCT projects can take time to embed. Some women might have only been exposed to the intervention for a limited period before they delivered. However, the CCT message was quite simple so we feel that the duration of intervention was of limited importance. We also consider that exposure rates were underestimated in our study. The intervention was delivered to all family members but the ascertainment of ‘exposure’ data relied solely on maternal self report. Literacy (26.2%) and education rates (15.6%) were low in our study and mothers may not have understood some of the more complex exposure questions. If we had been able to ascertain exposure more accurately we may have been able to show a greater impact ‘per protocol’. 10% of our quintile and 7% of our maternal age data were missing however no change to results were observed when these data were included in the primary and secondary analyses.

Difference-in-differences’ analyses require two main assumptions to be fulfilled: ‘common shocks’ and ‘parallel trends’ [23, 35]. We showed stable distribution of covariates between intervention and control groups from baseline to endline indicating no disparate changes in the intervention compared to the control group. There were also no differential policy shocks or changes between the intervention and control groups during the intervention period. However there has been a major escalation in conflict in both the intervention and control districts over the study period. We feel that that the CCT had an important effect in maintaining ANC, PNC and CHW homevisiting levels in the intervention group and without the CCT program the ANC, PNC and CHW levels would have been the same level as the control group. We consider that facility delivery was preserved in the intervention and control districts as there has been a major focus on facility delivery across Afghanistan from the MoPH. We also tested for and did not reject the null of parallel trends. Though it is important to note that there were some systematic differences between intervention and control groups, and our sample size did not provide enough statistical precision to conclusively reject the null.

Strengths included our prospective community and population based design, large sample size and high response rate (94%). We had an intensive data quality and supervision plan which included scheduled and unscheduled revisits. Overall, our study demonstrated that a CCT program provided to women aged 16–49 years can be implemented in a highly conservative conflict affected population and can include fathers and male heads of households as well as female family members. It is likely that if we can improve our focus on the poorest families, CCT will have important impacts on facility delivery, ANC and PNC.

Almost 400 million women and children are estimated to live in conflict-affected areas globally and these numbers are increasing [36]. Our innovative study demonstrates the important ‘proof of principle’ that CCT for mothers and newborns can be implemented in emergency situations and fragile states. There is an imperative to implement and evaluate novel and ‘outside the box’ interventions for the mothers and children that live in deprived and conflict affected communities across the world. We encourage other governments and researchers to also implement and evaluate CCT in conflict affected areas.