Background
Maternal mortality remains a major public-health problem in sub-Saharan Africa (SSA) and in many resource-limited settings. Globally, SSA accounted for 179, 000 (62%) of all estimated 289,000 maternal deaths in 2013. In that year, the maternal mortality ratio (MMR) in low-income countries was 14 times higher than in high-income countries, with SSA having the highest MMR at 510, compared to the global average of 210 per 100,000 live births. The estimated lifetime risk of maternal mortality in high-income countries was 1 in 3400 in comparison to 1 in 52 in low-income countries [
1‐
5].
Maternal mortality has enormous negative effects on child survival, family dynamics, and household economies, as well as national development [
6‐
8]. Consequently, several global, regional, and national initiatives have been employed to reduce maternal mortality and its negative effects, with varying degrees of success. These initiatives include strengthening health systems, implementing safe motherhood strategies, and developing maternal and newborn health networks, which are part of the United Nations Millennium Development Goals (MDGs) and the UN Secretary General’s Strategy, ‘Every woman every child’, among other programmes [
9‐
11]. Reviews, progress reports, and evaluation studies on such initiatives consistently show better outcomes in middle and high-income countries than in low-income countries [
12‐
14], especially SSA. The key reasons often cited for these disparities include weak health systems, low government health expenditures that often translate to catastrophic health expenditures for households, inequities, lack of policies that support the delivery of evidence-based interventions, and lack of access to skilled birth attendants [
15,
16].
Uganda is one of the several SSA countries that did not achieve the fifth MDG target by September 2015 [
15,
17]. The MMR and institutional delivery rates stood at 435 per 100,000 live births and 52.7%, respectively [
18,
19]. In addition to the reasons cited above, many barriers to accessing healthcare exist in Uganda. These include poor geographical access due to distance and transport issues, lack of decision-making power among women, shortage of professional health workers, poor attitudes of some health workers, and preference for traditional birth attendants [
2,
4,
5].
Oyam District has one of the highest MMRs (500/100,000 live births) in Uganda [
20]. The district is located in a rural post-conflict region in the northern part of the country and had a population of 388, 011 at the time of this study. Over 50% of the population live below the poverty line (US$1.25), more than 70% of the health facilities are health centre IIs, and less than 40% of the population live within 5 km of a health facility. Eighty-nine percent of the pregnant women receive antenatal care (ANC) at least once, but only 48% make four ANC visits (previous recommendation), and 42% use institutional delivery services. The overall rate of caesarean sections is 2.1%, well below the minimum 5% recommended by the WHO [
20‐
24]. This situation has prompted several non-governmental organisations (NGOs) and implementing partners to support healthcare delivery services in the district. If this move is sustained, it could promote institutional deliveries and improve birth outcomes, in line with evidence from literature [
25].
The presence of skilled attendants at birth is considered the single most important factor in preventing maternal deaths [
26‐
29], particularly in resource-limited settings. To that end, several innovative approaches that target both demand and supply-side barriers are being implemented. These include incentive-based interventions using strategies, such as conditional cash transfers, clean birth kits and transport voucher schemes. Although most of the studies of such interventions have been conducted in Asia [
30‐
32] and Latin America [
33] rather than Africa [
34‐
36], the results have been positive [
34,
37] in all the settings.
Few studies have assessed the effects of demand -side incentives, such as a transport voucher scheme, on the proportion of institutional deliveries in Uganda. Nonetheless, there are promising results from one of the preliminary studies [
34]. The Ministry of Health (MoH) launched the
Maama kit initiative in Uganda in 2003 to promote clean and safe deliveries [
38]. Under that initiative,
Maama kits were distributed to pregnant women in mostly rural districts during antenatal visits or community outreach visits [
38,
39]. We hypothesised that encouraging pregnant women to deliver at health facilities and providing them with a kit that reduces the cost of newborn care could increase the demand for institutional deliveries. We call this a “baby kit”, contrary to the
Maama Kit.
We evaluated the effects of providing transport vouchers and baby kits on changes in the number of institutional deliveries, four ANC visits, and postnatal care (PNC) visit. We also measured the proportion of women ‘bypassing’ maternal health services inside their residential sub-counties, in favour of services outside, with respect to four ANC visits, institutional deliveries, and PNC services. As a ‘side objective’, we hypothesised that given the inadequate number and disproportionate distribution of health facilities in the district, this study could help, to some extent, document the extent of ‘bypassing’ in the study’s sub-counties before and during the interventions. We also examined the financial costs of the two interventions in the promotion of institutional deliveries, to scale up to other sub-counties in Oyam District.
Discussion
We found that the baby-kit and the transport-voucher schemes markedly increased the service coverage of institutional deliveries at the intervention facilities over the relatively short study period. The DID analysis indicated that 30.0% of the deliveries at Ngai HC III during the intervention period were attributable to the provision of the baby kit. Ngai HC III is the only HC in Ngai sub-county. This element, combined with the HC’s proximity for the women living in the Ngai sub-county and with the baby- kit intervention, could indeed build up to a motivational effect for mothers to deliver at the facility.
The transport vouchers had a greater effect (42.9%) on the coverage of institutional deliveries when we compared the two interventions. This is a surprising finding, given that the transport vouchers were implemented at HC IIs, which have less technical capacity to conduct deliveries and that are considered to offer health services of poorer quality. Nonetheless, we believe that a combination of health system strengthening efforts with the introduction of transport vouchers may have led to positive synergies in terms of facilitating institutional deliveries.
Similarly, the service coverage of four ANC visits and at least one PNC visit increased during the interventions, with the transport vouchers demonstrating greater effect. Essentially, these outcomes are repetitive health-seeking activities that require improved access to promote utilisation. Our findings, therefore, suggest improved access to, and use of these services over time. The trends in ANC, delivery, and PNC services use by the mothers were pronounced between the transport-voucher facilities compared to their control facility than between the baby-kit and its control facility (figs.
1 and
2). Indeed, ample published literature demonstrates that the vouchers improved service utilisation and health outcomes among the target populations [
54]. The demand-side financing intervention using vouchers has been associated with increased use of ANC, institutional deliveries, PNC, and reduced inequities in institutional deliveries in Pakistan [
55,
56]. A study in Kenya found that some of the women who purchased vouchers meant to cover direct healthcare costs (service vouchers), did not use them because of high transport costs to the health facilities [
57].
We also found that the level of ‘bypassing’ increased considerably during the study period, and it was transport-voucher-related, mainly. Although several factors may explain this finding, the unequal distribution of health facilities and the low quality of services provided by many of the facilities in the district, remain key in promoting “bypassing.” A previous study conducted in the district, and others performed elsewhere demonstrate that distance, perceived quality of care, and affordability are principal determinants of utilisation of delivery services, among others [
17,
24,
58]. Hence, it is highly likely that the transport vouchers removed major barriers, particularly, distance and transport costs and encouraged pregnant women to use the facilities of their choice for maternal health services. Indeed, the DID analysis showed that the transport vouchers promoted ‘bypassing’ which in turn moderately improved service coverage in the transport voucher areas.
As illustrated above, the baby kits and factors such as proximity and a relatively higher quality of services may have attracted women from Ngai sub-county and some nearby villages to deliver at Ngai HC III, where the baby kits were implemented. Logically, the above scenario could encourage ‘bypassing’ from neighbouring communities to some extent. Notwithstanding, for women living in distant communities, the baby kits and the perceived quality of services did not appear to have had any effects on the barriers posed by long distances and the costs of transport to the centre, when compared with the effects of the transport vouchers.
Approximately 37% of the increase in pregnancy and labour-related referrals to a higher level of care was attributable to the interventions. The referrals were predominantly to Anyeke HC IV and Aber Hospital. As shown earlier, the overall facility utilisation increased with the interventions; thus, the increase in referrals might reflect the increase in the proportion of women with complications or at risk of developing complications. Considering that over 70% of the health facilities in the district are HC IIs without adequate capacities to manage complications related to pregnancy and labour, the increase in referrals is consistent with expectation. A recent study reported that although Anyeke HC IV plays a vital role in managing obstetric emergencies, only Aber Hospital has the capacity to perform all comprehensive and basic EmONC functions within the district’s health system [
23]. This calls for an effective referral system that ensures a reliable and timely transfer of patients from lower level health facilities to referral centres. We believe that by strengthening the referral system through effective and free ambulance services, and free caesarean sections at Anyeke HC IV and Aber Hospital, the incentive schemes may have indirectly promoted women’s access to and use of skilled attendants during deliveries in the district [
9,
27].
Of note, our interventions also resulted in unintended but largely positive effects that were predominantly associated with the transport-voucher system. From our experience, the transport-voucher system was a new concept in the district, as such it triggered excitement, and encouraged participation with the potential for some people to make small economic gains. As an illustration, some VHTs and family members or friends who owned bicycles and or motorbikes, used their knowledge of the context to sensitise households and community members about the intervention while advertising themselves as transporters.
Studies report that male involvement in maternal and newborn health services is low in Uganda [
59‐
61]. Nevertheless, as news of the transport-voucher intervention spread through communities, some husbands and partners began to transport their spouses to the health facilities to redeem the transport vouchers. Based on the accounts of midwives and nurses responsible for the study facilities, many of the transporters were either husbands or partners of the pregnant women, who used the facilities during the intervention period. While at the health facilities, the health workers engaged most men in discussions about maternal, newborn, and reproductive health issues. Consequently, the transport vouchers seemed to have generated interest in some segments of the communities in the district, and indirectly promoted male involvement in maternal and newborn health services, to some extent. Pariyo et al. [
62] reported some of these unintended effects. That study also elaborated on the potential risks and the need to enforce traffic regulations and safety precautions when using local transport mechanisms (e.g. bicycles and motorbikes) to promote maternal and newborn health services in rural settings.
Financial costs of the interventions
The cost analysis revealed that it cost US$15.9 and US$30.6 for every additional institutional delivery attributable to the transport-voucher and baby-kit schemes, respectively. Our results imply that providing transport vouchers is less costly than baby kits in facilitating institutional deliveries. The baby kit probably has less scope for usage (reducing costs related to newborn care while encouraging facility deliveries) than the transport-voucher scheme, which facilitates the use of the institutional delivery system by addressing major barriers (distance and transport costs), to accessing facility-based services. Since health services are predominantly facility-based, and majority of the district’s population live beyond 5 km of a health facility, the transport-voucher scheme translated into having a wider impact on the study’s outcomes.
The two incentive schemes were implemented simultaneously to improve access to and use of maternal and newborn healthcare services in the district. The resultant effects were notable and indicated which intervention to prioritise in the event of budgetary constraints.
As of yet, there are few publications on the cost-analysis of transport vouchers and hardly any research literature on baby kits, which makes this study a significant contribution to the literature. Moreover, regardless of the lack of comparable literature, in a setting where more than half of the population lives below the poverty level [
20], our interventions may not be affordable or sustainable without substantially more public-health funding from the government and development or implementing partners.
Study limitations
We acknowledge several limitations to our study that we think might be related to the setting and methodological challenges posed by implementing a study within a larger programme. The control and intervention groups were not randomly selected, which could lead to possible biases. As described in the Methods, the intervention facilities were purposively selected based on their poor performance to improve the maternal and newborn health services in the affected sub-counties. Nonetheless, their controls were comparable. Furthermore, the use of aggregated health-facility data did not allow the use of statistical analyses to adjust for confounders and to account for uncertainty in our estimates by calculating 95% confidence intervals. We are, however, confident that the DID analysis enabled us to demonstrate the effects of the interventions on the study outcomes, given the limitations.
We think that some deliveries that could be attributed to the interventions, particularly the transport vouchers, may have occurred at health facilities outside our study catchment areas. As such, those deliveries were not included in the analysis according to our study design. This means that we might not have captured the full effects of the interventions regarding institutional deliveries.
Likewise, the cost analysis did not capture the effects of the interventions on other services, such as changes in the uptake of ANC and PNC services. Consequently, the analysis probably underestimated the full effects of each intervention.
The district-wide measures to strengthen the health system were congruent with the need for a stronger health system to implement the interventions and achieve the study outcomes [
63,
64]. Notwithstanding, some of the measures might have resulted in synergies that are desirable; yet, they might also confound some of the effects of the interventions, especially the effects on referrals and ‘bypassing’.
The overlap of the catchment areas of the health facilities or the boundaries of the study’s sub-counties could have led to over or underestimation of the calculated percentage of ‘bypassing’. Although we made efforts to minimise this phenomenon, the quest for better quality of care and affordable services might have been a distraction from some of those efforts. Considering these methodological challenges, we urge caution in interpreting our findings. Additionally, we think that a qualitative study on these interventions could complement our findings.
We also recognise that this study was conducted in a rural post-conflict district in northern Uganda, where the context might not permit generalisation of our findings to other settings. Despite the limitations, we firmly believe that our findings remain valid as well as relevant.
Acknowledgements
We thank all the women and families that participated in this study and all the study’s staff members. We are grateful to the Oyam District Local Government and Health Authorities for supporting this study in various ways. Simon Amandi (Assistant District Health Officer and focal person for maternal health and nursing), John Bosco Orech (District Health Educator) and other members of the district health management team played key roles in sensitising and mobilising the communities in the implementation of the study. Sarah Awor (District Biostatistician) provided data from the district health office.
Peter Ongom (Project Social Worker) worked with VHTs to sensitise and mobilise the communities during the implementation of the study. He also participated in data collection from the district health office and health facilities.
We appreciate the inputs provided by Joke Bilcke (University of Antwerp, Belgium), Dr. Gideon Ndawula, Dr. Giuseppe Baracca, Dr. Bruno Turri, and other collaborators. Additionally, Mr. Tito Dal Lago (Administrator), Mr. Elvis Otim (Accountant), CUAMM office staff members, project drivers and several people contributed to this study in one way or another.
William Massavon benefited from TDR 2015 Training Grant, co-sponsored by UNICEF, UNDP, the World Bank and WHO. We are grateful.