Answers are grounded in this paper and cite their sources.
Background Routine administrative records are increasingly mapped for health surveillance, but boundary changes, population composition and place-of-care attribution complicate interpretation. We examined the geography of recorded public diabetes care in Paraguay and the contribution of demographic standardisation, spatial dependence statistics, Bayesian disease mapping and cluster detection. Methods Nationwide ecological analysis of annual counts of people attended for diabetes (ICD-10 E10–E14) in outpatient services of the Ministry of Public Health, 2023–2025. The 263 current districts were aggregated to 246 polygons of the 2012 open cartography. Expected counts were indirectly standardised for sex and age with district projections and the 2022 census; crude normalisation was a sensitivity analysis. We applied empirical Bayes Moran statistics, Kulldorff's scan statistic and BYM2 models, including a referral-hospital adjustment. Results We analysed 510,221 records (26.7 per 1000 population annually); from 15 years of age, rates in women were 1.6–1.9 times those in men. Standardisation changed expected counts by factors of 0.68–1.94 and reduced the empirical Bayes Moran's I from 0.39 to 0.11 (p = 0.005); one local Moran cluster remained after false discovery rate control. The BYM2 latent standard deviation was 0.47 (95% credible interval 0.42–0.55) and the spatial mixing parameter 0.56 (0.21–0.86). Fifty-seven areas, mainly in the south-centre of the Eastern Region and often departmental capitals, had a posterior probability ≥ 0.80 of a standardised ratio above 1.25, and 64, notably in the Chaco, along the Brazilian border and around Asunción, of a ratio below 0.80. Areas with a referral hospital had higher recorded use (relative ratio 1.42; 1.21–1.64), but adjustment changed the pattern little. Annual rankings were stable (Spearman 0.77–0.88). Every candidate scan cluster reached the minimum Monte Carlo p-value. Conclusions Recorded public diabetes care in Paraguay is geographically heterogeneous, but much of the regional structure of crude rates reflects population sex and age composition. Because events are attributed to the place of care, standardised patterns describe service activity, not prevalence, coverage or unmet need. Demographic standardisation should precede spatial inference on routine data; with large counts, Bayesian models add uncertainty-based classification more than smoothing, and scan p-values discriminate poorly.
No takes yet. Share an insight, caveat, or question.
Desvars et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: