Every neighborhood in America has a lead risk score. The science behind it just passed peer review.
On July 27, 2026, the journal GeoHealth (American Geophysical Union, published by Wiley) accepted our paper describing the national lead risk map on this site. The map predicts childhood lead exposure risk for all 83,388 census tracts in the country using only public data, and it was checked against real measured blood lead in eleven states. It is free to use and free to read. This page explains exactly how it works, where it holds up, and where it does not.
Where the paper is right now. Accepted after independent peer review and in production at Wiley. The final article gets a DOI when it publishes online, expected late summer 2026. The full text will be linked here that day. The interactive map named in the paper is already live at detectlead.com/lead-risk-map.
83,388census tracts scored, every one in the country
3,222counties scored, including DC and Puerto Rico
11states where predictions were checked against measured childhood blood lead
54,825,220individual homes with year built on record in our database
283,269federal housing properties tracked
33,772ZIP codes with housing age coverage
How the model works
Two public numbers, published by the Census Bureau for every neighborhood in the country. No secret data, no login, no model you have to take on faith. Anyone can rebuild it.
1
Start with housing age. The Census publishes when the homes in every tract were built. Lead paint was banned in 1978, and the older the home, the more likely lead paint is still in it. The model weights the three eras roughly 8 : 4 : 1, so a neighborhood of pre-1940 homes counts about eight times as heavily as one built in the 1960s and 70s.
2
Add poverty. The Census also publishes the share of each tract living below the poverty line. Old paint that gets maintained is a smaller hazard than old paint that peels. Poverty is the public signal for where repair money runs short.
3
Put both on the same national scale. Each input becomes a z-score against every tract in the country, then they combine: 58% housing age, 42% poverty. Housing age gets the bigger share because it is the direct source of the lead.
4
Rank everything 0 to 100. The combined score becomes a percentile. A tract at 90 is predicted to carry more lead exposure risk than 90 percent of tracts nationwide. Same math for all 83,388 tracts and all 3,222 counties.
5
The output is a screening rank, not a diagnosis. The map does not measure lead in any home and does not test any child. It says where to look first. The paper's recommended next step for a flagged area is physical confirmation: a home risk assessment, an XRF reading, a dust wipe, or a blood lead test.
housing age A = 100 × (0.619·pre1940 + 0.309·built40-59 + 0.075·built60-79)
combined R = 0.58·z(A) + 0.42·z(poverty)
final score P = percentile rank of R, 0 to 100
One thing never goes in: blood lead data. The model is built only from housing age and poverty. That is what makes checking it against real blood tests a genuine test instead of a circle.
Does it actually work?
The predictions were compared with measured childhood blood lead surveillance in eleven states the model never saw. The score is a rank correlation (Spearman's rho): 1.0 would mean the map orders neighborhoods exactly as the blood tests do, 0 would mean no relationship. Values in the 0.4 to 0.6 range are conventionally called moderate. The Michigan and Ohio tract rates come from the same surveillance the source EPA study used, roughly 4.2 million children's test results (Michigan 2006 to 2016, Ohio 2005 to 2018), so those two are consistency checks rather than independent tests.
State
Scale
Areas compared
Correlation (rho)
Test type
New Jersey
counties
21
0.77
independent
Illinois
counties
56
0.68
independent
Wisconsin
census tracts
1,308
0.64
independent
Ohio
census tracts
2,534
0.62
consistency
Massachusetts
census tracts
796
0.61
independent
Missouri
counties
40
0.56
independent
Iowa
counties
56
0.55
independent
Michigan
census tracts
2,156
0.54
consistency
Rhode Island
census tracts
204
0.50
independent
New York
counties
61
0.48
independent
Minnesota
census tracts
1,147
0.43
independent
Full confidence intervals, spatial autocorrelation corrections, and two finer-scale ZIP code checks (Missouri rho 0.60, New York climbing to 0.69 among well-tested ZIPs) are in the paper.
3.2–3.8×how much higher elevated blood lead rates run in the map's top 10% of tracts vs the state average (MI, OH, WI)
~halfof each state's total elevated blood lead burden sits in the map's top 20% of tracts (50 to 55%)
0.86–0.89AUC when the map is used to find the worst fifth of tracts in MI, OH, and WI
What the map cannot do
The paper spends as much space on limits as on results, on purpose. A screening map that overstates itself gets people hurt. Here is the honest version.
It scores areas, not houses. A high risk tract still contains remediated, lead-free homes. A low risk tract can still hold a pre-1940 house with failing paint. Reading a neighborhood score as a verdict on one address is the exact mistake the paper warns against.
Housing age and poverty explain about half of the variation. Cross-validated models on the same public inputs top out around R² of 0.45 to 0.48 at the tract level. The map ranks areas in roughly the right order. It does not capture everything that drives exposure.
The validation states are all older-housing regions. All eleven are Northeast, Midwest, or Great Plains states. Low risk calls for the West and Sun Belt are extrapolation, not a tested result, and most of those states publish too little blood lead data to check against.
The underlying blood lead data is thin. CDC receives about 3 million child blood lead results a year against roughly 22 million children under six, and testing targets kids already judged at risk. The paper corrects for what it can and says so where it cannot.
It scores some communities less accurately than others. In the paper's residual analysis on Michigan and Ohio tracts, the strongest predictor of what the model under-predicts was percent Black population. In plain terms, the map can under-rank the highest risk Black and historically disinvested neighborhoods relative to their real measured burden. The paper is explicit that this is a descriptive pattern in two states, not a proven cause, and that race should never be built into the score itself.
Disclosure, stated plainly. The paper's author founded Fluoro-Spec Inc., the company behind DetectLead, which sells a lead test kit, and also owns Spirochaete Research Labs LLC, which has lead detection interests of its own. Both are disclosed inside the paper itself. The map is free, requires nothing from you, names no product as the required next step, and is built entirely from public data anyone can use to reproduce or challenge it. The disclosed conflict was part of the record when the paper went through review.
The paper scores neighborhoods. The database behind this site goes house by house.
The model in the paper runs on Census tract data. Alongside it, we assembled something the paper only hints at: year built records for individual American homes, pulled from county assessor rolls across 41 states plus DC. That is what powers the address lookup below, and it is the backbone of the next paper.
54,825,220homes with a documented year built, individually searchable
8,683,449built before 1940, the era where about 87 in 100 homes contain lead paint
19,593,539built 1940 to 1979, the declining-but-real lead paint decades
26,548,232built 1980 or later, after the ban did its job
Counts queried live from our database on July 28, 2026. Coverage grows as more counties publish records. Era lead paint rates from the federal housing surveys cited in the paper.
Use it right now
This is the same data the research runs on. Type your ZIP code for your area's picture, then your street address to find your own home's year built. Free, instant, no email. Nothing you type here is stored.
Housing ages from U.S. Census records for 33,772 ZIP codes.
Now go one house deeper
Street number and name is enough. If your county publishes building records, we will show the year your home was built and what that means.
How to cite this work. Ritter, E. C. (2026). Childhood Lead-Exposure Prevention Through Predictive Mapping of At-Risk Housing Nationwide: A Free, Public-Data Screening Map Checked Against Measured Blood Lead in Eleven States. GeoHealth. In press, accepted July 27, 2026. DOI and full text will be posted here on publication. Press and data inquiries: eric@fluorospect.com.
Fluoro-Spec Inc. · DetectLead · eric@fluorospect.com · 631-461-1838
The risk map and lookups on this page are free screening tools built from public data. They do not measure lead in any specific home and do not diagnose any child. For product safety information see safety. To check your area's housing directly, see know your risk.