HealthDeedLegal
Methodology
Last updated 13 October 2026
The HealthDeed Score is an environmental screening index of the evidence available for an address. It helps identify questions to investigate; it does not establish that a home is safe or predict illness.
What qualifies for scoring
Long-term air observations, facility proximity, matched supplier concerns and an explicit mapped flood zone can contribute. Tract health outcomes, housing-age distributions, county water compliance and county radon potential are shown separately as context. These do not establish a finding inside this home. Missing evidence never receives an invented neutral score.
| Input | Source | Scope and treatment |
|---|---|---|
| Location | Census Geocoder / OpenStreetMap | Street-address match required; may be interpolated. ZIP-only locations cannot receive a property score. |
| Facility proximity | EPA TRI / NPL records | Modeled screening from EPA site status, release quantities, toxicity weights and distances. Not measured exposure. |
| Long-term air | EPA AQS annual summaries | Multi-year observations at an eligible regional monitor within 50 km. Today’s AirNow AQI is context only. |
| Flood zone | FEMA NFHL | Published zone at the geocoded point. No polygon means unknown, not Zone X. |
| Drinking water | EPA SDWIS / ECHO | Unambiguous supplier matching and open health-based violations can contribute. A clean record does not establish safe tap water. |
| Health and housing | CDC PLACES / Census ACS | Tract context only; does not establish exposure or the age of this building. |
| Radon | EPA radon zones / county records | Area potential only. A home test is required. |
| Terrain | USGS 3DEP | Elevation context, not a validated flood model. No scoring adjustment. |
How the number is calculated
Eligible indicators are normalized to 0–100, with higher values indicating fewer concerns in those records, and averaged within their category using evidence weights. The composite is the weighted average of categories that have eligible data. Assigned weights are:
- Water Quality: 20%
- Air Quality: 20%
- Soil & Contamination: 18%
- Natural Hazards: 16%
- Community Health Context: 16%
- Radon: 10%
These are modeling choices, not clinically validated weights. With the current sources, Air, Soil, Natural Hazards and matched Water concerns can score. Community Health and Radon remain contextual. We do not manufacture six scores to make a report look complete. No interaction multiplier, confidence discount or cosmetic stretch is applied.
What the scales mean
Annual PM2.5 uses screening anchors of 5–9 µg/m³; long-term ozone uses 60–70 ppb. Regional monitor values do not establish indoor conditions. EPA facility records use site status, toxic releases, toxicity weights and distance, rather than registrations alone. These are screening models, not measured chemical doses. Flood zones use an internal ordinal scale; their score is not an annual flood probability. These transforms still require independent validation.
There is no facility adjustment added to AirNow and no inferred Zone X from nearby map polygons. Corridor models cannot replace a failed facility lookup in scoring. Terrain samples do not establish rainfall-driven flooding at a building.
Coverage and concern labels
Coverage means the assigned weight of categories with scored data, not the percentage of hazards tested. One indicator can support a category while many hazards remain unknown. Labels describe concern in the scored evidence: Low at 80–100, Low to moderate at 65–79, Moderate at 55–64, Elevated at 36–54 and High below 36. Below half of the assigned category weight, the label is Limited assessment. These cut points are reporting policy, not safety thresholds.
We do not display a statistical confidence interval because one has not been validated. Confidence labels describe evidence resolution, not a probability that a property is safe. Modeled evidence keeps a category at Moderate even if another input is measured.
Missing data and comparisons
The score describes only available eligible inputs. Losing a source can change the average in either direction without a real change at the property. Reports with different evidence sets or engine versions should not be ranked against each other. Category highlights require matching evidence scope; observation dates can still differ.
Monitoring does not announce a score improvement or decline across a change of engine version, evidence scope or coordinates. A qualifying notification describes a change in the screening records, not a proven change inside the home. Current air conditions can move independently of the building.
Reproducibility and limitations
Engine version: ghfre-2026-10-20+evidence-v8+air-blend+water-tiers. Inputs and retrieval dates are stored with each report. Historical scores retain their original methodology and are not rewritten. Retrieval time does not establish when a source observation was made. The same inputs and engine version produce the same result.
Parcel and building attributes, confirmed water suppliers, facility release quantities and site status, long-term exposure, and independent ground-truth validation remain incomplete. Straight-line proximity does not establish a pathway or exposure. A favorable average can coexist with a serious individual finding; read the findings and source gaps first.