Peninsular overview
Estimated prevalence
Districts with 95% credible intervals
Model sequence
Same data, covariates added in sequenceReading this page
- Compare districts within a state and use the estimates to set local baselines.
- A wide interval means few sampled blocks nearby; treat that estimate as provisional.
- Covariates are model inputs. Their coefficients are not read as associations.
- The surface is smoothed over sampled blocks; do not read it at neighbourhood scale.
About these estimates
The tool presents the estimated prevalence of diabetes among adults aged 18 years and above across Peninsular Malaysia, with 95% credible intervals, from four Bayesian geostatistical models fitted to NHMS 2023 data at enumeration-block centroids. Diabetes follows the NHMS 2023 definition: known diabetes plus raised blood glucose among adults not known to have diabetes.
Model 0 contains the intercept and a spatial field. Models 1 to 3 add built-environment covariates in sequence: green space and air quality, then urbanicity and socioeconomic deprivation, then the food environment. District summaries are population-weighted means of grid-cell estimates.
Undiagnosed diabetes and the diagnosis gap
Undiagnosed diabetes is raised blood glucose among adults not known to have diabetes. The diagnosis gap is the percentage of adults with diabetes who are undiagnosed (undiagnosed divided by total). Both are shown with the same models; state values follow the same NHMS 2023 table.
Prototype data
State means are set to the NHMS 2023 age-standardised state prevalence (Technical Report, Table 4.1.4). District values, the grid surface, the credible intervals and all covariate values are synthetic and serve only to test the interface. The display grid is 4 km; the thesis analysis uses the WorldPop 1-km grid.
Limits
- Variation below the spacing of sampled blocks is not identified.
- Estimates far from a sampled block depend mainly on the covariates and the spatial field.
- The model does not use the survey weights; estimates may differ from design-based state figures.
- Covariate coefficients do not establish associational or causal effects.
Covariates and sources
| Domain | Measure | Source |
|---|---|---|
| Green space | NDVI, 2023 | Sentinel-2, Google Earth Engine |
| Air quality | Annual mean PM2.5, 2022–2023 | ACAG surface V6.GL.03 |
| Urbanicity | Urbanisation rate by district | DOSM Census 2020 |
| Deprivation | Median household income by district, 2022 | DOSM HIES |
| Food environment | Unhealthy outlets per 10,000; unhealthy-to-healthy ratio | OpenStreetMap |
| Population | Gridded population, 1 km | WorldPop 2020 |