Spatial autocorrelation and regression approach for delineating maternal mortality and its associated factors in Karnataka, India
Kodagehalli Suresh Sahana, Basavegowda Madhu, Mysore Chikkapapanna Manjunatha, Mallaiah Chaithra, Manivasagan Mounika Sree, Berkolly Manjunath Suraj, Anethyar Yashashwini, Basavaraju SnehalathaObjectives
The main objective of this study is to assess the temporal distribution of taluk (sub-district)-level maternal mortality trends within Karnataka and identify the hotspots and medical and non-medical factors that were significantly contributing to maternal deaths.
Design
Spatial patterns and determinants of maternal mortality were investigated using a retrospective ecological study design.
Data source
Maternal mortality data were collected from the Directorate of Health and Family Welfare Services, Government of Karnataka, while political boundaries of state, districts and taluks were downloaded from the Karnataka Geographic Information System Portal.
Methodology
Taluk-wise (sub-district level) maternal mortality ratio was mapped using GeoPandas software. Global and local Moran’s I along with spatial regression was performed using GeoDa software in evaluating the spatial dependence and identifying significant predictors of maternal mortality.
Results
Maternal mortality varied geographically, according to thematic maps, and local indicators of spatial autocorrelation map showed high-high clustering (nine taluks (5.1%)) with positive Moran’s I (0.114). Descriptive analysis of time of death revealed that the majority of maternal deaths (39.82%) occurred within 48 h postpartum, followed by 2–7 (20.95%) and 8–30 (16.23%) days. The spatial error model showed negative associations for antenatal care at the sub-centre, SDHs, deliveries conducted by doctors and private hospital (β=−0.047 to −0.184, p<0.05), and a positive association was found for below poverty line, stillbirths, parity 1 and deliveries at medical colleges (β=0.112–0.758, p<0.05) with a λ value of 0.286. Primary postpartum haemorrhage, sepsis, hypertensive disorder of pregnancy, cardiorespiratory disorders and other direct and indirect causes were identified as major contributors to maternal mortality in the spatial lag model (β=0.273 to 1.926, p<0.05, ρ=−0.077).
Conclusion
Spatial analysis revealed geographic hotspots, temporal risk windows and socio-economic and medical determinants of maternal mortality in Karnataka. These findings provide actionable evidence for spatially targeted, temporally focused and socio-clinically comprehensive maternal health interventions.