Bottleneck analysis at district level to illustrate gaps within the district health system in Uganda

Background: Poor quality of care and access to effective and affordable interventions have been attributed to constraints and bottlenecks within and outside the health system. However, there is limited understanding of health system barriers to utilization and delivery of appropriate, highimpact, and cost-effective interventions at the point of service delivery in districts and sub-districts in low-income countries. In this study we illustrate the use of the bottleneck analysis approach, which could be used to identify bottlenecks in service delivery within the district health system. Methods: A modified Tanahashi model with six determinants for effective coverage was used to determine bottlenecks in service provision for maternal and newborn care. The following interventions provided during antenatal care were used as tracer interventions: use of iron and folic acid, intermittent presumptive treatment for malaria, HIV counseling and testing, and syphilis testing. Data from cross-sectional household and health facility surveys in Mayuge and Namayingo districts in Uganda were used in this study. Results: Effective coverage and human resource gaps were identified as the biggest bottlenecks in both districts, with coverage ranging from 0% to 66% for effective coverage and from 46% to 58% for availability of health facility staff. Our findings revealed a similar pattern in bottlenecks in both districts for particular interventions although the districts are functionally independent. Conclusion: The modified Tanahashi model is an analysis tool that can be used to identify bottlenecks to effective coverage within the district health system, for instance, the effective coverage for maternal and newborn care interventions. However, the analysis is highly dependent on the availability of data to populate all six determinants and could benefit from further validation analysis for the causes of bottlenecks identified.

The Tanahashi model used in this study as described earlier was modified for its use in the MBB tool, which was developed to enable LIC at the national level to plan for marginal allocations to health services, cost and budget for these allocations, and access their potential effect on health coverage [19]. One modification to the Tanahashi model still focuses on determinates of effective coverage which is defined as coverage of sufficient quality to reach a defined health impact [14,16,17] and not merely geographic access [18]. Each determinant is analogous to a Tanahashi stage leading towards effective coverage. Another modification was divided, however, the determinant ‘availability’ into the availability of human resources and availability of commodities. This was thought to reflect the types of data that are available and still allow for a stepwise approach to identification of bottlenecks to achieving effective coverage [15]. The modified Tanahashi model therefore has six determinants for effective coverage. The first three determinants – accessibility, availability of human resources, and availability of essential health commodities – are supply-side determinants of the health system, while initial utilization, continuous utilization, and effective coverage focus on the demand side, as illustrated in Figure 1. Supply-side determinants are defined as those factors that influence the health care production function. Demand-side determinants are those that operate at the community, household, and individual levels and are influenced by demand [24]. The six coverage determinants are explained in Table 1. Similar to the original model, the six determinants reflect six distinct aspects of service provision that can be used for a stepwise assessment. Examining the largest differences between each determinant indicates the larger losses of health system effectiveness, thus pointing to those areas of service provision that need to be prioritized. This loss of effectiveness is referred to as a ‘bottleneck’ within the health system [15]. Definition of coverage determinants used in the Tanahashi model. Modified Tanahashi model applied to analyze bottlenecks. The three supply-side determinants do not have the same denominator and the presence or absence of one of them does not necessarily affect the presence or absence of the other supply-side determinants, although they tend to be positively correlated, while the demand-side determinants assume a linear relationship. We selected a set of interventions to function as proxies for analyzing health system bottlenecks. After reviewing approximately 190 maternal and child health interventions, Kerber et al. [25] proposed three main channels for their delivery: community-based preventive and health promotion services, outreach services, and clinical and curative services. Due to the specific nature of the delivery channel, it tends to experience similar bottlenecks across interventions. Therefore ‘tracer’ interventions can be selected to serve as proxies to determine bottlenecks for interventions delivered in a similar way. For example, indicators that assess access, availability, initial and continuous utilization, and effective coverage for treatment of pneumonia can be used as proxies to determine bottlenecks for other curative and clinical services, for example, treatment of diarrhea and malaria [26]. The following criteria were used to select interventions: (1) their relevance and appropriateness to maternal and newborn care in Uganda; (2) being internationally recommended for maternal and newborn care; (3) having evidence of measurable impact on outcomes; and (4) availability of data for all six coverage determinants. For this study we chose: (1) use of iron and folic acid supplements to prevent anemia during pregnancy; (2) intermittent presumptive treatment for malaria; (3) HIV counseling and testing, and (4) syphilis testing during antenatal care (ANC) [27–30]. Interventions provided during ANC visits were specifically selected because data were available on these interventions and because good-quality ANC has been documented to improve health outcomes for maternal and newborn care in LIC [31,32]. In Uganda, ANC is provided through clinical services in public health facilities at no cost to the pregnant woman. This was a descriptive cross-sectional study using household and facility census surveys. The study was conducted in the two rural districts of Mayuge and Namayingo in the Eastern region of Uganda. The district population is approximately 474,000 and 216,000, respectively [33]. In these districts maternal and newborn care is predominantly provided by public health facilities at different levels. In Uganda, health services are provided through health facilities (including Health Centres [HC] IIs, IIIs, and IVs and hospitals) and Village Health Teams (comprised of community volunteers) [34,35]. Although a lot of progress has been made, maternal and newborn deaths are still unacceptably high in this area with the estimated maternal mortality ratio at 438 per 100,000 live births and an estimated neonatal mortality rate of 23 per 1000 live births [36]. It is, therefore, essential to identify bottlenecks to effective service delivery for maternal and newborn care in order to improve the quality of care in those districts. Data for this study were collected through household surveys and repeat health facility censuses in Mayuge and Namayingo districts from November 2011–April 2014 [37,38]. This was during the Expanded Quality Management Using Information Power project (EQUIP), which focused on improving the quality of care for maternal and newborn care. Sample size for the project was calculated to estimate coverage of key maternal and newborn interventions with 80% power at the district level [38]. The household surveys used continuous cluster sampling of 10 household clusters using the probability of selection being proportional to the population size. Each cluster had 30 randomly selected households [38]. For the purposes of this study, data from survey interviews with 6513 women who were pregnant 12 months prior to data collection were included (see Table 2). Data sources used for the study. The health facility census was repeated every four months in all government-owned health facilities in the districts. Data from 50 facilities (30 in Mayuge and 20 in Namayingo) were used in this study. These included 20 HC IIs, 7 HC IIIs, and 3 HC IVs (see Table 2). Facility readiness was assessed by interviewing health facility managers and by use of a checklist to determine the routine care and services provided. Data were analyzed using STATA 13 and construction of the bottleneck analysis graphs was done in Excel 2010. Coverage for each determinant was calculated as a proportion of the target population/supply for which a particular determinant was met. See Table 3 for details on the analysis of the coverage levels for each of the determinants and assumptions made. Coverage measures used for each of the modified Tanahashi model determinants and assumptions made.

N/A

• henriksson, dorcus kiwanuka
• fredriksson, mio
• waiswa, peter
• selling, katarina ekholm
• peterson, stefan