Counted Health is a burden-mapping tool built on public federal data. It is not a diagnosis and gives no medical advice.

How it works

How Counted Health works

Counted Health takes health information the country already collects, brings it together for each person, and maps where the burden of invisible illness really falls. Here is how it does that — and, in one place at the end, exactly what the evidence shows and where we are careful about it.

What we look at

Everything comes from a large, trusted U.S. government health survey (NHANES). For each person it gathers far more than one doctor usually sees at once. We use three kinds of information about the same person:

Normally these sit in separate places, so no one sees them together. We bring them into a single burden scorefor each person — the higher the score, the more these signals are off at once.

How we build the map

We line everyone up by their burden score, lowest to highest, and split the line into five equal groups. Group 1 is the lowest burden; group 5 is the highest. These are just groups along a scale — not diagnoses, and not something a computer decided on its own.We fixed the way we sort people up front, so it can’t be tuned to produce a flattering answer.

The evidence — and where we’re careful

Two findings stand out. First, chronic illness climbs steadily across the groups: in the lowest-burden group, about 1 in 3 people live with a diagnosed chronic condition; in the highest, it is closer to 3 in 4— more than double. That gap holds even when we compare people of the same age, sex, and body weight, so it is not simply that older people are sicker.

Second — and this is the strongest test — we matched these same people to official records of who later died, records the burden score never used. Over the next 6 years, people in the highest-burden group died at about 5 times the rate of those in the lowest, even after accounting for age, sex, and weight. When a map can tell you that, it is measuring something real.

The burden map predicts survival

We linked the same respondents to CDC death records — a federal dataset the score never used. Over 6 years (285 deaths in 3,915 adults), all-cause mortality climbs across every burden region. This is the check that matters most: the outcome is a hard endpoint, so the gradient cannot be an artifact of self-report.

Highest vs lowest region
5.0×
adjusted hazard of death (95% CI 2.72–9.23, p < 0.001)
Objective signals only
1.53×
per SD from labs + activity alone (p < 0.001)
Region 11.5%Region 23.4%Region 34.3%Region 49.3%Region 517.8%
View as table
All-cause mortality by burden region — NHANES 2013-2014 x NCHS Public-Use Linked Mortality (through Dec 2019). Cox proportional hazards, adjusted for age, sex, and BMI.
RegionnDeathsCrude mortalityMean age
Region 1782121.5%42
Region 2783273.4%47
Region 3783344.3%49
Region 4784739.3%54
Region 578313917.8%57
Cox proportional-hazards models, adjusted for age, sex, and BMI.
ComparisonHazard ratio95% CIp
Region 5 vs 15.012.72–9.23< 0.001
Region 4 vs 12.821.51–5.27= 0.001
Per region (trend)1.551.39–1.73< 0.001
Severity per SD1.591.44–1.76< 0.001
Objective-only per SD1.531.40–1.67< 0.001

Where we’re careful

The precise figures, for the record
Headline numbers
Chronic-condition share by group, lowest to highest: 33% → 72% (survey-weighted). Highest-vs-lowest group, adjusted for age, sex, race, and BMI: odds ratio 2.45 (95% CI 1.92–3.12). All-cause mortality, highest-vs-lowest group, adjusted for age, sex, and BMI: hazard ratio 5.0 (95% CI 2.79.2). Built on NHANES 2013–2014 (n ≈ 3,900) linked to the NCHS public-use mortality file. Both are U.S. federal open data.

That is the whole of the science on this site, in plain terms. The full statistical write-up — every dataset, variable, model, adjustment, confidence interval, and sensitivity check, plus how to reproduce it — is laid out on our methods & reproducibility page. Counted Health itself is built to be used, not read like a study. Where the data comes from →

← The full mapMethods & reproducibility →Where the data comes from →