Development and testing of a composite index to monitor the continuum of maternal health service delivery at provincial and district level in South Africa

Introduction The continuum of care is a recommended framework for comprehensive health service delivery for maternal health, and it integrates health system and social determinants of health. There is a current lack of knowledge on a measurement approach to monitor performance on the framework. In this study we aim to develop and test a composite index for assessing the maternal health continuum in a province in South Africa with the possibility of nationwide use. Materials and methods The composite index was computed as a geometric mean of four dimensions of adequacy of the continuum of care. Data was sourced from the district health information system, household surveys and the census. The index formula was tested for robustness when alternative inputs for indicators and standardization methods were used. The index was used to assess performance in service delivery in the North West province of South Africa, as well as its four districts over a five-year period (2013–2017). The index was validated by assessing associations with maternal health and other outcomes. And factor analysis was used to assess the statistical dimensions of the index. Results The provincial level index score increased from 62.3 in 2013 to 74 in 2017, showing general improvement in service delivery over time. The district level scores also improved over time, and our analysis identified areas for performance improvement. These include social determinants of health in some districts, and access and linkages to care in others. The provincial index was correlated with institutional maternal mortality rates (rs = -0.90, 90% CI = (-1.00, -0.25)) and the Human Development Index (r = 0.97, 95% CI = (0.63, 0.99). It was robust to alternative approaches including z-score standardization of indicators. Factor analysis showed three groupings of indicators for the health system and social determinants of health. Conclusions This study demonstrated the development and testing of a composite index to monitor and assess service delivery on the continuum of care for maternal health. The index was shown to be robust and valid, and identified potential areas for service improvement. A contextualised version can be tested in other settings within and outside of South Africa.

The authors have declared that no competing interests exist. NO   Introduction: The continuum of care is a recommended framework for comprehensive health 21 service delivery for maternal health, and it integrates health system and social determinants of 22 health. There is a current lack of knowledge on a measurement approach to monitor 23 performance on the framework. In this study we aim to develop and test a composite index 24 for assessing the maternal health continuum in South Africa.

68
The framework outlines interventions from pre-pregnancy to childhood; the maternal health 69 interventions encompass reproductive health, antenatal, delivery and postnatal care. The 70 framework for SA also outlines "intersectoral factors", which represent social determinants of 71 health such as housing, water and sanitation, and education (Fig 1). The implementation of 72 the framework is expected to improve health outcomes by improving coverage and 73 comprehensiveness of services, mitigating duplication of resources, and improving 74 integration of health services(11,13,14).

75
A crucial barrier to the implementation of the continuum of care framework in SA and many 76 LMICs is the lack of a comprehensive monitoring tool for service delivery (15,16). The   determinants of health (16) (Fig 2). The adequacy construct complements the COC 120 framework by adding elements of quality of and linkages to care, and proposing that all four 121 dimensions be monitored, not just access and or utilization. Indicators of service delivery 122 across all dimensions should therefore be sought from local data sources, with consideration 123 for their relevance, feasibility and validity(21,22). Health service indicator data may be missing due to lack of services and under-performing 145 systems for data collection and reporting. These are systematic issues that are considered to affect availability of data for indicators completely at random. As such, we conducted single 147 value imputation using the indicator value observed from the adjacent year (36). In the Results 148 section we discuss the impact of the remaining data gaps on the index findings. Single value 149 imputation was also applied to indicators from community survey and census to allow 150 calculation of index at district level.

151
Multivariate analysis 152 We used exploratory factor analysis to assess dimensionality of the data, in order to compare The indicator score is calculated on a scale between 0 -100; the ideal score is the 161 maximum attainable score, which is a 100; the target is the ideal performance of the 162 indicator; and performance is the observed value of the indicator during a given time period.

Weighting and aggregation 171
The comprehensive continuum of care for maternal health index (C3MH index) was computed 172 as a geometric mean of equally weighted sub-indices reflecting the four adequacy 173 dimensions. We chose equal weighting since this was estimated the most reasonable 174 approach, and evidence on the relative importance of each sub-index is lacking (30,31,39).

197
Performance at provincial level 198 In the North West province, we combined 12 indicators of access and utilization to care, two 199 quality of care, two linkages of care, and 9 social determinants of health indicators to 200 measure the C3MH index (Table 1).    performance to demonstrate effect on overall scores, using 2016 as a reference year (Fig 4).  There was no significant difference in district ranks between index scores calculated with 237 linearly scaled indicators (our method) and z-score standardization (rs=0.83, 95% CI =(0.49-238 0.95)) ( Table 2). There was also no significant difference between index scores when 239 geometric and arithmetic aggregation techniques were used (rs =0.95, 95% CI=(0.78-0.99)).

240
The median absolute difference in index rankings at district level when linear and z-score 241 standardization were compared was 2 ranks with an interquartile range (IQR) of 0-3. There 242 was no difference in rankings (IQR= 0-1) observed at district level when indices computed 243 with arithmetic and geometric means were compared.

246
All the index values across alternative indicator selections were highly correlated (Table 3).
247   (Table 4). A two-factor 266 model accounted for 0.72 cumulative proportion of variance of the data but did not reveal any 267 informative conceptual groupingsall factor loadings were relatively high for one factor.

268
The variables for linkages of care and one variable for quality of care were also not included 269 in the results of the model due to missing data.
270    This was a case study of five subnational geographical areas over a five-year period. We  This study shows the feasibility to monitor and assess service delivery for the continuum of 323 care for maternal health using indicators from different sectors with a composite index. The 324 index allows monitoring of performance over time and across geographical areas. From our 325 analyses, we concluded the index to be robust and valid, with potential to guide policy and 326 planning to improve maternal health outcomes and service delivery from a multisectoral 327 perspective. More comprehensive monitoring of social determinants at district level and 328 health information system strengthening can further improve and extend the use of this index.

329
The index is amenable for testing with data from different South African and international