What do administrative diabetes indicators measure? Treatment‐side and screening‐side evidence from nationwide NDB Open Data in Japan

Abstract Aims Regional administrative diabetes indicators are increasingly used for population surveillance and planning, yet they are generated through the healthcare system rather than measuring disease directly. We examined what two complementary indicator families—treatment‐side prescription quantities and screening‐side HbA1c—measure across 47 Japanese prefectures. Methods Cross‐sectional ecological study using the 10th NDB Open Data release. Indicator A was oral antidiabetic agent dispensing volume per population (drug classification 396), related to ageing, income, education and population density by ordinary least squares with VIF assessment and AIC‐based comparison. Indicator B comprised the HbA1c high‐value rate, mean HbA1c and the D005 laboratory‐testing rate, related to three claim‐rate outcomes by ordinary least squares with HC3 errors (five sensitivity specifications). Results For Indicator A, ageing was the only significant univariable predictor ( β = 9961; R 2 = 0.324); the lowest‐AIC model retained ageing and income ( R 2 = 0.394; VIF = 2.18; income β = 56.6, p = 0.030). For Indicator B, the HbA1c high‐value rate was associated with chronic dialysis management ( β = 195.1/%pt., p = 0.006, R 2 = 0.271); mean HbA1c showed a comparable association; the primary association was stable across all five specifications. Conclusions Neither indicator family provided a direct estimate of disease burden. Prescription quantities reflected demographic and income structure, whereas HbA1c indicators showed a selective ecological association with downstream service use. Both are better interpreted as ecological surveillance indicators than direct measures of disease burden or prevalence.