alzhub.org
Get help now — free helplines

You can call any of these for free. You do not need a diagnosis, an appointment, or a crisis to call — questions and plain exhaustion are reasons enough.

  • Alzheimer's Association 24/7 Helpline 800-272-3900 24 hours a day, 365 days a year Free, confidential support and information for people with dementia, caregivers, families and the public. Master's-level clinicians. Also Spanish.
  • 988 Suicide & Crisis Lifeline 988 24/7 Call or text 988, or chat at 988lifeline.org. For anyone in emotional distress or a mental-health crisis, including caregivers.
  • Alzheimer's Foundation of America (AFA) Helpline 866-232-8484 9:00 am - 9:00 pm ET, seven days a week Licensed social workers providing support, info and referrals. Also text message and web chat. Text: 646-586-5283

In a medical emergency, call 911. More numbers, and what each line does, on the help page.

Home › Research › Dementia by county

Dementia, county by county

Until this week, this site had no county-level dementia outcome data at all. Now it has diagnosed prevalence, incidence and mortality for 3207 US counties, drawn from the NORC Dementia DataHub's own 2020 Medicare claims file. This page is built to be read carefully rather than skimmed: the honest version of a county map has more caveats than colors.

Read this before the map forms an opinion in your head

Every rate on this page counts diagnosed cases from Medicare claims — not disease itself. Two counties can have identical numbers of people actually living with dementia and still show very different rates here, if one diagnoses more of them than the other. A county with a low measured rate may have less disease, or it may simply have less access to a diagnosis — more primary care, more neurologists, more people who can get to an appointment. This file cannot tell those two apart, and neither can this page.

These are also 2020 figures for people covered by Medicare specifically (mostly age 65+, plus some under-65 beneficiaries with disabilities or ESRD) — not the whole population of any county, and not a more recent year.

12.33%

Median county diagnosed prevalence

Middle half of counties: 10.88%–13.86% Source: NORC Dementia DataHub, 2020
4.34%

Median county diagnosed incidence

New diagnoses in 2020 alone, range 1.71%–10.75% Source: NORC Dementia DataHub, 2020
18.23%

Median mortality among the diagnosed

Share of already-diagnosed cases who died in 2020 — not a population death rate Source: NORC Dementia DataHub, 2020

How prevalence is spread across counties

A single national number would hide everything interesting here. Diagnosed prevalence runs from about 4.1% to 35.55% across 3207 counties — but the extremes on both ends are mostly places with a few thousand people, where a handful of diagnoses either way swings the rate a lot. The middle half of counties sits in a much narrower band, 10.88% to 13.86%.

How diagnosed prevalence is spread across counties

Median 12.3%, middle half of counties between 10.9% and 13.9%.

Distribution of diagnosed dementia prevalence across US counties, 2020Histogram of 3181 county prevalence rates, median 12.3%, interquartile range 10.9% to 13.9%, full range 4.1% to 35.5%.020040060080010000%5%10%15%20%25%30%35%median 12.3%Q1 10.9%Q3 13.9%diagnosed prevalence rate (% of Medicare beneficiaries)number of countiesn = 3181 counties with a reported rate, 2-point bins, 202026 counties are excluded here because the source suppressed their count (fewer than 11people) -- blank, not zero, and not counted toward this median.Lowest: Oliver County, ND (4.1%, FIPS 38065) -- highest: Wayne County, IA (35.5%, FIPS19185).

This chart is wider than your screen — drag it sideways, or open the table underneath.

n = 3181 of 3,207 counties (the other 26 have a suppressed, not zero, prevalence count and are excluded from this median and every other average on this page that uses this column). The five highest and five lowest counties are almost all places with populations under 10,000 -- Kenedy County, TX (highest bracket) has 343 residents. A rate built from a small number of Medicare beneficiaries swings further on a handful of diagnoses either way, so the extremes of this distribution should be read as noisy, not as the most/least at-risk places in the country.
The ten highest and lowest counties, as a table
RankCountyFIPSPrevalence rate
LowestOliver County, ND380654.1%
LowestSioux County, ND380855.5%
LowestFerry County, WA530195.7%
LowestDaggett County, UT490095.8%
LowestMeagher County, MT300595.8%
HighestWayne County, IA1918535.5%
HighestKenedy County, TX4826132.1%
HighestStarr County, TX4842726.6%
HighestWillacy County, TX4848925.4%
HighestAlcorn County, MS2800324%

Suppression, made visible

26 counties have a suppressed — not zero — prevalence count, 66 have a suppressed incidence count, and 127 have a suppressed mortality-among-diagnosed count, because the source found fewer than 11 people in that cell. Every median and mean on this page is computed only over the counties that were not suppressed for that specific measure; dropping the suppressed counties silently, instead of counting and naming them, would bias every one of those averages in the same direction.

Suppression, made visible

A blank cell in this data means the source found fewer than 11 people in a county -- not zero. Mortality is suppressed far more often than prevalence, because dying in a given year is a rarer event than living with a diagnosis.

How many counties are suppressed, by measureDiagnosed prevalence: 3181 reported, 26 suppressed; Diagnosed incidence: 3141 reported, 66 suppressed; Mortality among the diagnosed: 3080 reported, 127 suppressed.0%20%40%60%80%100%Diagnosed prevalence26 suppressedDiagnosed incidence66 suppressedMortality among the diagnosed127 suppressedreported (a real count)suppressed (blank, not zero)n = 3,207 counties per measure. Suppression follows CMS's small-cell rule: any cellrepresenting fewer than 11 people is blanked at the source, before this project eversees it.Dropping these counties instead of reporting them would silently understate everyaverage on this page in the same direction, toward whatever is true of smallercounties.

This chart is wider than your screen — drag it sideways, or open the table underneath.

Source: NORC Dementia DataHub, 2020 Public Use File. Every mean, median and rate on this page is computed only over the reported (non-suppressed) counties for that specific measure, and this figure is where that exclusion count lives for all three.
Every number in this chart, as a table
MeasureReportedSuppressedShare suppressedMedian rate (reported only)
Diagnosed prevalence3181260.8%12.3%
Diagnosed incidence3141662.1%4.3%
Mortality among the diagnosed30801274%18.2%

Black population share and diagnosed prevalence

This is the one cross-sectional pattern in the new data strong enough to call a finding: counties with a higher Black population share tend to show higher diagnosed dementia prevalence, and the pattern survives controlling for income. Measured counts on both sides, but a county-level correlation describes places, not people — it cannot show that any individual's race changes their odds of developing dementia. One plausible explanation this data cannot rule out is differential access to a diagnosis rather than differential disease: a hypothesis, not a conclusion the numbers below can settle.

Black population share and diagnosed prevalence

The clearest cross-sectional pattern in this data -- and the one that most needs the caveats stated next to it, not below it.

Diagnosed dementia prevalence by county Black population share, grouped into five bins0-0.5% Black, mean prevalence 11.2%, n=619; 0.5-1.3% Black, mean prevalence 11.8%, n=619; 1.3-3.7% Black, mean prevalence 12.2%, n=619; 3.7-13.5% Black, mean prevalence 13%, n=619; 13.5-85.9% Black, mean prevalence 13.9%, n=620. Raw r=0.30, partial r (controlling income)=0.27. County-level correlation; not evidence of individual risk.0%2%4%6%8%10%12%14%0%–0.5% Black11.2% (n=619)0.5%–1.3% Black11.8% (n=619)1.3%–3.7% Black12.2% (n=619)3.7%–13.5% Black13% (n=619)13.5%–85.9% Black13.9% (n=620)Counties grouped into five equal-sized bins by Black population share (lowest tohighest); each bar is that group's mean diagnosed prevalence rate.Raw correlation across all 3096 counties: r = 0.30. Controlling for medianhousehold income, the partial correlation is r = 0.27 (p = 1.3e-51) -- itsurvives, roughly unchanged, once income differences between counties are heldconstant.This is a county-level (ecological) correlation: a property of PLACES, not ofPEOPLE. It cannot show that any person's race changes their odds of developingdementia.One hypothesis this data cannot rule out: places with more Black residents may havemore diagnosis of dementia that already exists elsewhere too, not more of the diseaseitself -- differential access to a diagnosis, not differential risk.

This chart is wider than your screen — drag it sideways, or open the table underneath.

n = 3096 counties with a reported prevalence rate, a Census race share and a median household income (of 3,207 total; the rest are missing one of those three, mostly suppressed prevalence counts or 8 Connecticut county-equivalents the Census Bureau retired mid-decade). Exposure: US Census Bureau ACS 5-year estimates, 2020-2024. Outcome: NORC Dementia DataHub, 2020. Ecological correlation only: this describes how county averages move together, and cannot be read as a claim about any individual’s risk.
Every number in this chart, as a table
Black population share (bin)CountiesMean diagnosed prevalence
0%–0.5%61911.2%
0.5%–1.3%61911.8%
1.3%–3.7%61912.2%
3.7%–13.5%61913%
13.5%–85.9%62013.9%

For context, five other county- and state-level checks run against this same new data (exposures: air quality, social ties, living alone, primary-care access, poverty) came back weak or negligible — this Black-population-share pattern is the one that stood out, not the typical result of running this kind of check. See correlation studies for all of them, including how each was tested for whether it survives controlling for likely confounders.

Incidence and mortality: how they relate to prevalence

Diagnosed incidence — new diagnoses within 2020 alone — runs from 1.71% to 10.75% across counties, a narrower spread than prevalence because it only counts one year's new cases rather than everyone currently living with a diagnosis. The mortality measure here is different from a normal death rate: it is the share of people already diagnosed who died that year — a case-fatality rate, not a count of dementia deaths per resident. A county can show a low number here and still have many dementia deaths, if it also has many diagnosed cases.

Higher-prevalence counties tend to show lower case-fatality

How the mortality-among-diagnosed measure moves against diagnosed prevalence, grouped into five bins.

Mortality among diagnosed cases, by county diagnosed-prevalence bin5.7-10.7% prevalence, mean case-fatality 19.1%, n=616; 10.7-11.9% prevalence, mean case-fatality 19%, n=616; 11.9-12.9% prevalence, mean case-fatality 18.4%, n=616; 12.9-14.3% prevalence, mean case-fatality 18.2%, n=616; 14.3-35.5% prevalence, mean case-fatality 16.8%, n=616. Raw r=-0.26.0%5%10%15%20%5.7%–10.7% prevalence19.1% (n=616)10.7%–11.9% prevalence19% (n=616)11.9%–12.9% prevalence18.4% (n=616)12.9%–14.3% prevalence18.2% (n=616)14.3%–35.5% prevalence16.8% (n=616)Counties grouped into five bins by diagnosed prevalence; each bar is that group’s meanmortality-among-the-diagnosed rate. Raw r = -0.26 across 3080 counties.This is a share of already-diagnosed cases who died that year, not a populationdeath rate -- a county can have a low number here and still have many dementiadeaths, if it has many diagnosed cases.Read with caution: the diagnosed-case count is the denominator of one rate and thenumerator of the other, which can itself pull two ratios toward moving oppositely -- aknown statistical trap with paired ratios, not necessarily a fact about dementia care.

This chart is wider than your screen — drag it sideways, or open the table underneath.

n = 3080 of 3,207 counties with both rates reported (not suppressed). The death share is deaths divided by diagnosed cases, and prevalence is diagnosed cases divided by Medicare beneficiaries. The diagnosed-case count sits in both, as the bottom of one ratio and the top of the other, so part of this pattern could be a mechanical artifact of the two ratios sharing a term rather than a fact about care or survival. Shown descriptively, not as a validated finding.
Every number in this chart, as a table
Prevalence binCountiesMean case-fatality (of diagnosed)
5.7%–10.7%61619.1%
10.7%–11.9%61619%
11.9%–12.9%61618.4%
12.9%–14.3%61618.2%
14.3%–35.5%61616.8%

What this page cannot tell you

It cannot tell you whether a specific county has more dementia, or just diagnoses more of it. It cannot tell you anything about any one person's risk, from any chart on this page — every figure here describes averages across thousands of people in a place, not an individual. And it cannot separate "this county's rate is genuinely rising or falling" from "this county's diagnosis practices changed," because this is a single year, 2020, not a trend.

What it can tell you: this data exists now, where for years it did not, and every number on this page traces back to a real county, a real Medicare claims record, and a real, stated suppression rule — never to a guess filling in for one.

Where this county data comes from

Records shown
of the 3207 in the data behind this page
Distinct sources
2
Years covered
2020
Basis, all 3207
3207 measured · 0 modelled · 0 not stated
Data collected
2026-09-27
Source files
county_dementia_outcomes.csv (~350 KB, 3207 rows), panel_county.csv (~1.1 MB, 3143 rows)

Most-cited sources on this page

What could make these numbers misleading

These are DIAGNOSED dementia cases identified from Medicare claims, not a model of true disease prevalence. Diagnosis rates for the same underlying symptoms vary by place, so a county with a low measured rate may under-diagnose rather than have less disease -- and this file cannot tell those two apart. Every row is the source's own published county grand total (all stratifiers blank), never a sum this project built across strata. Blank cells are the source's own suppression of any count under 11 people, and are excluded from every average with the exclusion count stated, never treated as zero.

Every figure on this page is read straight out of a file in our data folder; nothing is typed in by hand. You can download the raw data and check any row yourself. This is a summary of published research, not medical advice.