Estimating uncertainty in geospatial modelling at multiple spatial resolutions: the pattern of delivery via caesarean section in Tanzania

Abstract

Visualising maternal and newborn health (MNH) outcomes at fine spatial resolutions is crucial to ensuring the most vulnerable women and children are not left behind in improving health. Disaggregated data on life-saving MNH interventions remain difficult to obtain, however, necessitating the use of Bayesian geostatistical models to map outcomes at small geographical areas. While these methods have improved model parameter estimates and precision among spatially correlated health outcomes and allowed for the quantification of uncertainty, few studies have examined the trade-off between higher spatial resolution modelling and how associated uncertainty propagates. Here, we explored the trade-off between model outcomes and associated uncertainty at increasing spatial resolutions by quantifying the posterior distribution of delivery via caesarean section (c-section) in Tanzania. Overall, in modelling delivery via c-section at multiple spatial resolutions, we demonstrated poverty to be negatively correlated across spatial resolutions, suggesting important disparities in obtaining life-saving obstetric surgery persist across sociodemographic factors. Lastly, we found that while uncertainty increased with higher spatial resolution input, model precision was best approximated at the highest spatial resolution, suggesting an important policy trade-off between identifying concealed spatial heterogeneities in health indicators.

Geospatial estimation of reproductive, maternal, newborn and child health indicators: a systematic review of methodological aspects of studies based on household surveys

“…Only five of the 82 studies presented estimates at both grid cell and administrative levels. There is much discussion regarding the ideal level of aggregation, as it depends on multiple factors including the outcome, the objective of the analysis, how decentralized decision-making is within the country and the trade-off between precision and resolution [ 41 , 44 ].…”

“…A total of 55 out of the 82 studies opted for gridded-estimates (Table 1). Apart from three studies [40][41][42], the approximate cell size (grid) for all reported resolutions ranged from 1 × 1 km to 10 × 10 km, with 5 x 5 km being the most common one. Smoothed maps were presented by 12 studies without specifying the originally estimated resolution.…”

“…Within reproductive health, we identified four studies focusing on contraception [45,[115][116][117] and two studies on undesired adolescence pregnancies [118,119]. From pregnancy to child birth, four studies focused on antenatal care, skilled birth attendance, c-section and postnatal care [40,41,120,121]. Diarrhea [32,122,123] and respiratory infections [33] were the focus of a total of seven studies, as they are still among the leading causes of death for children in the poorest countries.…”

A simulated ‘sandbox’ for exploring the modifiable areal unit problem in aggregation and disaggregation

“…Decision-making criteria regarding the spatial scale and zonation of areal units has a fundamental impact on the nature of geographic spatial analysis 1 – 7 . While this phenomenon has been acknowledged in the geographic literature for decades 1 – 6 , being cognizant of the implications in how underlying data is constructed matters for any field, with particularly useful examples noted for demographic 8 , health 9 , 10 , urban 11 , and ecological 12 applications. Considering that no rule set or agreed upon standards currently exist for areal aggregation in spatial analysis 1 , 2 , 4 , 5 , it is critical to determine the underlying rationale for a given spatial resolution in geographical analysis as the sensitivity of associated outcomes is tied directly to the decision-making criteria of model development and the underlying characteristics of the data 1 , 2 , 5 .…”

A review of geospatial methods for population estimation and their use in constructing reproductive, maternal, newborn, child and adolescent health service indicators

“…Uncertainty intervals in estimates tend to be much larger at small scales and reduce as the data are aggregated. Recent work applying Bayesian statistical models to capture uncertainty provides quantitative measures of uncertainty in population estimates at grid square, district, province and national levels, highlighting how uncertainty intervals shrink as data is aggregated, and the need to account for this in applications [7,24,37].…”