| English | 中文 |
Health-impact estimates require population, baseline outcomes, exposure and a risk function referring to compatible populations. Acquiring each file does not establish compatibility. Use WorldPop for spatial population estimates and GBD for baseline health estimates.
Obtain exposure-response coefficients or curves separately from their published source using the health-impact method and skill. A baseline mortality export is not a risk curve. Match pollutant, exposure statistic/window, units, endpoint and eligible population before linking an upstream prediction; a PM2.5 response cannot be applied to an ozone prediction by renaming its column.
Specify geography/boundary vintage, year/reference date, age boundaries, sex and cause definitions. WorldPop and GBD may use different total benchmarks and age structures; justify a chosen reference rather than assuming interchangeable denominators. Prefer compatible products from one release across the period and treat revisions separately from observed change.
| Quantity | Alignment needed |
|---|---|
| Population counts | Count/density units, reference date, grid/boundaries, age/sex coverage |
| Baseline deaths | Cause hierarchy, count versus rate, age/sex and location definitions |
| Baseline rate | Correct denominator and unit scale; age-standardized rates cannot reconstruct actual counts |
| Concentration | Exposure period, spatial support and population distribution |
| Risk function | Applicable causes/ages, concentration units, exposure window and counterfactual |
For an age-specific baseline rate r per 100,000 and compatible population P, expected baseline deaths are P × r / 100000. This conversion requires a crude or matching age-specific rate; it is not valid for an age-standardized rate applied to actual population. Aggregate deaths and population first when constructing a combined rate, rather than averaging subgroup rates.
Use nonoverlapping ages, sexes and regions; never add totals to their components. Preserve missing combinations and distinguish absent information from zero. Check subgroup totals, population conservation across spatial processing and consistency of concentration/population grids.
Retain baseline uncertainty and exposure/risk uncertainty separately. Derived totals and differences require a justified dependence structure or compatible joint draws; adding uncertainty bounds does not create a valid combined interval. Document population downscaling, age disaggregation and gap-filling as additional assumptions.
Acquisition steps are in the WorldPop skill and GBD skill. Apply the health-impact method after alignment.