Stroke Death Rates

Limitations

Stroke death rates here are modeled estimates for adults 35 and older: CDC age-standardizes every rate and spatially smooths county rates, borrowing deaths from neighboring counties (the Georgia row is not smoothed). Each year is a 3-year window named by its last year (2024 = 2022-2024 deaths), and consecutive windows share two of their three years, so they are not an annual series; there is no 2020-2022 window (no 2022). Rates are deaths per 100,000 adults 35 and older a year, directly age-standardized to the 2000 U.S. standard population, from death certificates (NVSS; ICD-10 I60-I69); county rates are smoothed with a Local Empirical Bayes method. Every window's dataset description states the same method (3-year average, age-standardized, county rates spatially smoothed), and its metadata the same ICD codes; CDC's statistical methods page names one age standard and one smoother and mentions no change between releases. The one change found in the rows between windows is the race groups (below). Each window's dataset description gives its data source as 'National Vital Statistics System', but the 2022-2024 one gives 'National Vital Statistics System and U.S. Census Bureau'. On data.cdc.gov the rows of the five windows 2015-2017 through 2019-2021 were last updated in January-February 2025, after those datasets were first posted (whether values changed then cannot be told). Years 2020, 2021, 2023 and 2024 (windows 2018-2020 through 2022-2024) include COVID-19 pandemic years (2020, 2021 or 2022); only 2017, 2018 and 2019 are pre-pandemic. Race groups: hispanic is Hispanic of any race and every other group is non-Hispanic, as CDC's atlas shows them (the window files do not say so). Through 2018-2020 (year 2020) the files have four race groups besides Hispanic, with Asian and Pacific Islander combined (asian_pacific_islander); from 2019-2021 (year 2021) six, with asian, pacific_islander and multiracial separate. Race-specific rates are not comparable across that change (the all-race rates are), and never add or compare asian_pacific_islander with asian. That the change is a switch from bridged-race to single-race groups is our inference from the group lists; CDC's window datasets do not state it. Georgia's multiracial rate is the lowest of any group's (2022-2024: 17.9 against 84.1 for all races), probably because death certificates record more than one race less often than population estimates do (our reading). The 39 published county multiracial rates (in 6 counties) range from 0.0 to 124.4 (women in Cobb County, 2021-2023), and five are above the all-race rate of the same county, sex and year. Don't compare multiracial rates with other groups. CDC withholds a rate ('Insufficient Data') where too few deaths or people over the window make it unreliable: NULL here, never filled in or derived. The all-race, both-sexes rate is published for Georgia and every county in every year. Counties with a both-sexes rate, by group (fewest-most over the years): white 159, Black 152-154, Hispanic 36-47, Asian 25-27, Asian and Pacific Islander 24-25, multiracial 4-6. Native Hawaiian/Pacific Islander and American Indian/Alaska Native rates are never published, Georgia included. CDC's rules for a smoothed county rate and an unsmoothed rate differ (see the rate's null_meaning), so a county rate can be published where Georgia's for the same group is withheld: multiracial residents (2021), multiracial men (2021, 2023 and 2024) and multiracial women (2021, 2023 and 2024). Small groups can get extreme smoothed estimates, kept as CDC publishes them: seven rates are exactly 0 (Hispanic men in Chatham County, 2017 and 2018, Houston County, 2020 and 2024, and Colquitt County, 2024; Hispanic residents of Glynn County, 2017; multiracial residents of Chatham County, 2024), and 14 more are above 0 but below 5 per 100,000, all for small Hispanic, Asian and Pacific Islander, or multiracial groups. A 0 is a model estimate for a group with no or almost no stroke deaths over the three years (our reading of the method), not a measured absence of risk; the highest is 406.8, Black men in Turner County, 2021-2023 (year 2023). No deaths, population or confidence interval is published, so county rates cannot be summed, averaged or pooled into a regional rate, and the Georgia row (computed from all Georgia deaths, unsmoothed) is not an average of its counties. A smoothed county rate is pulled toward its neighbors', most for small counties and groups. Male and female rates are each age-standardized, so the both-sexes rate is not their average. Not comparable head to head with OASIS's deaths by cause (health/oasis/deaths_by_cause_nchs_rankable, cerebrovascular diseases): those are single-year, all-age rates, unsmoothed and with counts; these are smoothed 3-year rates for adults 35 and older. CDC's separate 2000-2019 annual county series (a Bayesian spatiotemporal model, age bands 35-64 and 65+, a 2010 standard population) is a different product and not in this table. tract_geoid is always NULL.