Effects of demand-side incentives in improving the utilisation of delivery services in Oyam District in northern Uganda: A quasi-experimental study

Background: We evaluated the effects and financial costs of two interventions with respect to utilisation of institutional deliveries and other maternal health services in Oyam District in Uganda. Methods: We conducted a quasi-experimental study involving intervention and comparable/control sub-counties in Oyam District for 12 months (January-December 2014). Participants were women receiving antenatal care, delivery and postnatal care services. We evaluated two interventions: the provision of (1) transport vouchers to women receiving antenatal care and delivering at two health centres (level II) in Acaba sub-county, and (2) baby kits to women who delivered at Ngai Health Centre (level III) in Ngai sub-county. The study outcomes included service coverage of institutional deliveries, four antenatal care visits, postnatal care, and the percentage of women ‘bypassing’ maternal health services inside their resident sub-counties. We calculated the effect of each intervention on study outcomes using the difference in differences analysis. We calculated the cost per institutional delivery and the cost per unit increment in institutional deliveries for each intervention. Results: Overall, transport vouchers had greater effects on all four outcomes, whereas baby kits mainly influenced institutional deliveries. The absolute increase in institutional deliveries attributable to vouchers was 42.9%; the equivalent for baby kits was 30.0%. Additionally, transport vouchers increased the coverage of four antenatal care visits and postnatal care service coverage by 60.0% and 49.2%, respectively. ‘Bypassing’ was mainly related to transport vouchers and ranged from 7.2% for postnatal care to 11.9% for deliveries. The financial cost of institutional delivery was US$9.4 per transport voucher provided, and US$10.5 per baby kit. The incremental cost per unit increment in institutional deliveries in the transport-voucher system was US$15.9; the equivalent for the baby kit was US$30.6. Conclusion: The transport voucher scheme effectively increased utilisation of maternal health services whereas the baby-kit scheme was only effective in increasing institutional deliveries. The transport vouchers were less costly than the baby kits in the promotion of institutional deliveries. Such incentives can be sustainable if the Ministry of Health integrates them in the health system. This quasi-experimental study evaluated the effects of the two interventions separately: the provision of transport vouchers and baby kits on the utilisation of maternal health services. The study population consisted of women attending antenatal, delivery and postnatal services at health facilities in the study’s sub-counties. Doctors with Africa CUAMM, an Italian NGO, and hereafter, referred to as CUAMM, implemented the interventions in Oyam District in northern Uganda, from January to December 2014. CUAMM is the main international NGO supporting maternal and newborn healthcare delivery services in Oyam District. This organisation and its operations have been described in previous publications [23, 24, 40, 41]. The papers provide maps of the district and details of the administrative divisions, such as the health sub-districts (HSD), sub-counties, parishes, and villages. Moreover, these papers describe the district’s health system, including the distribution of health facilities and their functional relationships within the district’s health system. Notably, Aber Hospital, a private not-for-profit (PNFP) facility is the only hospital in the district, the single facility capable of providing all emergency obstetric and neonatal care (EmONC) services, and it serves as referral hospital for the district [23]. Table 1 summarises the basic components and services provided at the various levels of the district’s health system. Basic components and services at the various levels of the Oyam District Health System Sources: Health Sector Strategic Plan 2000/01–2004/05, Ministry of Health, Uganda; Health Sector Strategic Plan II 2005/06–2009/10, Ministry of Health, Uganda; Institute for Health Metrics and Evaluation (IHME). Health Service Provision in Uganda: Assessing Facility Capacity, Costs of Care, and Patient Perspectives. Seattle, WA: IHME, 2014 HC, Health Centre. aA HC I is the interface between the formal health system and the community. There are no physical structures, except for a network of Village Health Teams (VHT). The VHTs are non-professional health workers trained to perform several functions. They include health education, disease prevention and health promotion through community sensitisation and mobilisation for various public-health interventions. They are usually residents of the communities and serve as ‘bridges’ between the health facilities and their communities. bSome HC IIs were upgraded and mandated to provide basic maternity services According to the Uganda National Health Policy [42], there should be a HC III in every sub-county and a HC II in every parish. However, Oyam District has only 6 HC IIIs for the 12 sub-counties, 22 HC IIs for the 63 parishes, a HC IV at Anyeke, and 1 PNFP hospital, for a total of 30 health facilities (Oyam District Health Report 2015, unpublished). Although it is desirable to provide maternity services, mainly at the HC III level and above in the Ugandan health system [43–45], an inadequate number of HC IIIs and their geographical inaccessibility make this provision untenable in Oyam. This situation compelled the district’s local government to upgrade and mandate some HC IIs to provide basic maternity services, in line with existing central government provisions [46, 47]. Furthermore, HCs are distributed unevenly, such that some sub-counties and parishes have two or more health facilities, whereas others have none. This disproportionate distribution of HCs in the sub-counties, coupled with the perceptions of poor quality of care leads to ‘bypassing’ the HCs, a phenomenon whereby residents of poorly served sub-counties seek health services from health facilities in neighbouring sub-counties [48, 49]. Following the presidential election campaign in 2001, Uganda officially abolished user fees in all public-health facilities [50], but that did not lead to zero costs. Patients still pay for transport and sometimes also for medications and commodities at the health facilities because of frequent shortages. In 2012, CUAMM launched a 5-year programme to improve access to and use of maternal and neonatal health services in the entire district. The programme adopted a mix of demand and supply-side strategies aimed at strengthening the district’s health system and improving the availability, quality, and use of maternal and newborn healthcare services, particularly, skilled attendants at birth. The programme was integrated into the district’s health system, where services are usually free of charge. The key components of that programme were: The two incentive schemes were implemented within the framework of the programme described above. In selecting the intervention and control sub-counties, we considered institutional delivery service coverage, ‘bypassing’, the logistical feasibility of implementing the interventions, and the similarity of the health facilities. As the main aim of the interventions was to increase institutional deliveries, the sub-counties with the lowest service coverages were purposively selected. To minimise ‘bypassing’, the control and intervention sub-counties were chosen in such a way that they were separated by a buffer zone. The intervention sub-counties were located in the same HSD to make it logistically feasible to implement the study, given the limited resources available. We also ensured that the control and intervention facilities were fairly similar regarding the levels of care in the health system, infrastructure, services provided, and staffing (Table ​(Table11). Based on the above factors, Alao HC II and Atipe HC II in Acaba sub-county were selected for the transport-voucher intervention. Amwa HC II (Myene sub-county) was selected as a control facility for this intervention. Ngai HC III (Ngai sub-county) was selected for the baby-kit intervention, with Agulurude HC III (Loro sub-county) serving as its control. Each of Acaba and Myene sub-counties has two HC IIs but no HC III. Ngai sub-county has only one HC III, while Loro sub-county has one HC III and two HC IIs. The transport-voucher intervention was implemented in Acaba sub-county. The vouchers were given to pregnant women while attending ANC at Alao HC II and Atipe HC II during the intervention period. Some pregnant women in the study’s catchment area learned about the intervention at the time of delivery. To prevent them from feeling being ‘left out’, these women received vouchers to use for PNC services. The voucher allowed women to use any locally available means of transport (motorbike or bicycle) to travel to the health centres for any pregnancy or labour-related condition or emergency, including delivery. The voucher scheme was intended to address geographical inaccessibility. After transport being provided to women, the driver redeemed the voucher at a fixed amount of 10,000 Ugandan Shillings (4 US dollars), at the health facility. To our knowledge, at the time of this study, only Anyeke HC IV in the Oyam District was receiving intermittent supplies of the Ugandan MoH Maama kits. That health centre was not included in the study. There are notable differences and minor similarities between the MoH Maama kit and the Baby kit that we provided, regarding objectives, contents, and distribution strategies. As shown in the background section, the MoH Maama kit initiative was meant primarily to promote clean and safe deliveries. The Maama kit contained a plastic sheet, sterile gloves, razor blades, a cord ligature, a tube of Tetracycline ointment, cotton, sanitary pads and a piece of soap. The kit was offered to pregnant women during antenatal and community outreach visits, which makes the distribution strategy more extensive, compared to our approach. On the other hand, the baby kits were intended to encourage the use of health facility delivery services by reducing costs related to newborn care. Each baby kit consisted of a plastic basin, a bar of soap, a polythene bag, 1/2 kg of sugar, and a piece of cotton cloth for wrapping the baby. The baby-kit intervention was implemented at Ngai HC III, the only health facility in Ngai sub-county. All pregnant women who delivered at that health centre received baby kits before being discharged. The project staff visited the intervention facilities monthly to monitor progress, collect the receipts for the vouchers and baby kits, replenish stocks and ensure accountability. Local radio messages and several stakeholder meetings were used to sensitise and mobilise the target communities for the interventions. In addition, the VHTs, community resource persons, and health facilities worked together to raise awareness about the availability of the incentives and how women could access them. The study outcomes included the service coverage of institutional deliveries, four ANC visits, a minimum of one PNC visit, and the proportion of women ‘bypassing’ local health facilities. Institutional delivery was defined as childbirth in a health facility. ANC was defined as the receipt of pregnancy care from skilled providers, while PNC was defined as care at a health facility after childbirth regardless of the place of delivery. Data on these outcomes were extracted from the ANC, delivery, and PNC registers using a standard form designed specifically for this study’s data collection procedures. The form captured monthly summary data on the numbers of ANC visits, institutional deliveries, PNC visits, and referrals all disaggregated by health facility catchment areas. Baseline and endline data were collected throughout 2013 and 2014, respectively. Data on the number of outpatient visits at each participating health facility during these reference periods were also collected. We calculated costs from the perspective of CUAMM, as the funder of and partner with the district in implementing the interventions. Thus, we did not consider the costs incurred by the users. Data on costs were obtained from the project’s accounting and administrative records for the period January to December 2014. The costs included the values of direct inputs for both intervention arms. This consisted of the cost of each of the items in a baby kit, multiplied by the number of kits distributed, and the total cost of the transport vouchers distributed. Additionally, the costs included labour costs, training costs, sensitisation costs, and administrative support services, notably, the cost of transport for distributing the items and the printing services needed for transport vouchers and receipt books for baby kits. We calculated labour costs by multiplying the number of days in a week spent by drivers and social workers involved in handling vouchers and baby kits by their respective daily net salaries and then annualised the resulting amount. We conducted one joint project start-up training for two staffs from each intervention health facility. Consequently, we allocated training costs in proportion to the number of staff trained for each intervention. We also shared administrative support services costs between the two intervention arms based on the ratio of the institutional deliveries in 2014. Research costs were not included. The costs were handled in Ugandan Shillings (UGX), but for this analysis, we expressed them in US dollars (1 US$ = 2598 UGX as of June 2014) [51]. Details about the cost items are shown in Additional file 1. We calculated the service coverage of the four ANC visits, institutional deliveries and PNC as the number of pregnant women who utilised the facility for each of these services in a 12-month period, divided by the annual number of deliveries expected in the area. We performed these calculations separately for baseline and endline data. We estimated the expected number of deliveries by multiplying the crude birth rate for each year by the health facility’s catchment population for the respective year based on official government data. Based on World Bank statistics, we used a crude birth rate (per 1000 population) of 43.5 for 2013 and 43.00 for 2014 [52]. We included women from the study area who had been transferred to other catchment areas for further care in the calculation. For each outcome, we calculated the ‘bypass’ percentage as the proportion of all users from outside the catchment areas of the study’s health facilities. We did not consider women from the study’s catchment areas who might have utilised services in health centres that were not participating in the study. We performed difference in differences (DID) analyses to estimate the effect of the interventions on each of the outcomes [53]. We calculated the DID estimate of the intervention using the equation: (YI2-YI1)-(Yc2-Yc1), whereby YI2 is the endline value of a given outcome indicator in the intervention group, YI1 is the baseline value of this indicator in the intervention group, Yc2 is the endline value of the indicator in the control group, and Yc1 is the baseline value of the indicator in the control group. The DID estimate is thus, the difference in changes over time in the outcomes between the intervention and control groups. Except for the baby-kit and transport-voucher schemes, the health system strengthening programme mentioned above was implemented equally in the control and intervention areas of study. Therefore, any differences in the outcomes between the control and intervention groups can be attributed to the interventions. The analysis separately compared the two interventions: the transport-voucher system and provision of baby-kit as described in the Methods, to no intervention (the respective control arms). We used institutional deliveries as the main outcome of interest for the cost analysis, and performed a descriptive analysis of the costs for each intervention. Thereafter, we divided the number of institutional deliveries in 2014 in each intervention area by the cost of each intervention to obtain the cost per delivery. We also calculated the incremental cost of each intervention per delivery. To obtain the incremental number of deliveries, we used the formula for the DID analysis presented above, and based the calculations on absolute numbers instead of percentages. We then calculated the incremental cost per delivery by dividing the incremental number of institutional deliveries in each intervention area by the incremental cost of each intervention. All analyses were performed using Microsoft Excel 2010.