{"main_topic":"health","sub_topic":"cdc_miov","topic":"homicide_deaths","key":"health/cdc_miov/homicide_deaths","version":"1.0.0","year_min":2019,"year_max":2024,"year_gaps":[],"update_frequency":"annual","detail_levels":["counties"],"default_detail":"counties","path_template":"health/homicide_deaths/year={year}/{detail}.parquet","grain":["year","county_fips","tract_geoid","method"],"title":"Homicide Deaths","summary":"Homicide and firearm homicide death rates per 100,000 and deaths, by Georgia county, 2019-2024 (CDC MIOV).","description":"Homicide deaths among the residents of each of Georgia's 159 counties, 2019-2024, all methods and by firearm: CDC's crude death rate per 100,000 residents (the observed rate, or an estimate from CDC's Bayesian model, with its 95% credible interval, where the county had 1-9 deaths) and the number of deaths, from CDC's Mapping Injury, Overdose, and Violence (MIOV) dashboard (National Vital Statistics System death records). Counts of 1-9 are withheld by CDC.","explorer":true,"search_aliases":{"en":["murders","killings","gun homicides","firearm homicides","violent deaths","people killed","homicide mortality","death by homicide","dying from violence","assault deaths"],"es":["asesinatos","homicidios con armas","muertes violentas","muertes por homicidio","mortalidad por homicidio","defunciones por homicidio","muerte por violencia","violencia armada","personas asesinadas"]},"source":"Centers for Disease Control and Prevention, National Center for Injury Prevention and Control, Mapping Injury, Overdose, and Violence (MIOV), county dataset (data.cdc.gov psx4-wq38), based on National Center for Health Statistics National Vital Statistics System death records, as retrieved 2026-10-24","source_url":"https://www.cdc.gov/injury-violence-data/data-vis/index.html","endpoint":"/api/v1/health/homicide_deaths","columns":[{"name":"year","type":"integer","role":"year","nullable":false,"label":"Year","short_description":"","description":"Calendar year of death (2019-2024), final as CDC publishes it. CDC's provisional trailing 12 months (June 2025 to May 2026) are not included.","valid_values":null,"example":2024,"unit":null,"value_min":null,"value_max":null,"null_meaning":null,"key_metric_grain_contributor":false,"metric_component":null},{"name":"county_fips","type":"string","role":"fk_county","nullable":false,"label":"County FIPS","short_description":"","description":"5-digit county FIPS code as published (FK to the counties dimension): the decedent's county of residence. Every row is a county; CDC publishes no Georgia total in this dataset.","valid_values":null,"example":"13121","unit":null,"value_min":null,"value_max":null,"null_meaning":null,"key_metric_grain_contributor":false,"metric_component":null},{"name":"tract_geoid","type":"string","role":"fk_tract","nullable":true,"label":"Tract GEOID","short_description":"","description":"Always NULL: this dataset is by county. Kept for the health domain's standard key shape.","valid_values":null,"example":null,"unit":null,"value_min":null,"value_max":null,"null_meaning":"Always NULL (county rows only).","key_metric_grain_contributor":false,"metric_component":null},{"name":"method","type":"string","role":"categorical","nullable":false,"label":"Method","short_description":"'all' is every homicide; 'firearm' is homicides by firearm, a subset of 'all'. Never add the two.","description":"Method of homicide, as CDC groups it: 'all' (every homicide death, any method: the total) or 'firearm' (homicides by firearm, which are also counted in 'all'). Never add the two. No other method is published, and a non-firearm figure ('all' minus 'firearm') is not served: it would need, or narrow, the counts CDC withholds.","valid_values":["all","firearm"],"example":"firearm","unit":null,"value_min":null,"value_max":null,"null_meaning":null,"key_metric_grain_contributor":true,"metric_component":null},{"name":"death_rate_per_100k","type":"number","role":"metric","nullable":true,"label":"Homicide Death Rate (per 100,000)","short_description":"Homicide deaths per 100,000 residents a year (crude); an estimate from CDC's model where the county had 1-9 deaths.","description":"Homicide deaths per 100,000 residents of the county in the year, crude (all ages, not age-adjusted), as CDC publishes it (rounded to 0.1). With 0 or 10 or more deaths it is the observed rate (deaths / population x 100,000; 0 with no deaths). With 1 to 9 deaths it is an estimate from CDC's Bayesian model, which borrows strength from neighboring counties to steady small counts and protect privacy (rate_modeled true), with a 95% credible interval: not deaths / population, and never 0. Rates never add or average across counties, and no population is published to weight them. The highest is a model estimate, 47.0 (Terrell County, 13273, 2023, firearm homicides: 1 to 9 deaths, 95% credible interval 21.3-103.7); the highest observed rate is 37.8 (Bibb County, 13021, 2022: 59 deaths). Both preserved as published: a few deaths in a small county give a high rate. A modeled firearm rate can exceed the same county's modeled all-homicide rate (separate model fits; see limitations). Fulton County, 2024: 16.8 (181 deaths).","valid_values":null,"example":16.8,"unit":"rate_per_population","value_min":null,"value_max":null,"null_meaning":"No rate: the county had 1 to 9 deaths (rate_modeled true) and CDC flags the modeled rate as unstable and does not publish it (-999 in the source). Never a zero.","key_metric_grain_contributor":false,"metric_component":null},{"name":"death_rate_per_100k_ci_lower","type":"number","role":"metric","nullable":true,"label":"Modeled Rate 95% Credible Interval Lower Bound","short_description":"","description":"Lower bound of the 95% credible interval of death_rate_per_100k, as CDC publishes it, for modeled rates only (rate_modeled true): under CDC's model, the true rate lies in the interval with 95% probability. Wide for small counties.","valid_values":null,"example":null,"unit":"rate_per_population","value_min":null,"value_max":null,"null_meaning":"No interval: the rate is observed (rate_modeled false: 0 or 10 or more deaths), and CDC publishes credible intervals only for modeled rates; or the county had 1 to 9 deaths (rate_modeled true) and CDC flags the modeled rate as unstable and does not publish it or its interval (-999 in the source).","key_metric_grain_contributor":false,"metric_component":null},{"name":"death_rate_per_100k_ci_upper","type":"number","role":"metric","nullable":true,"label":"Modeled Rate 95% Credible Interval Upper Bound","short_description":"","description":"Upper bound of the 95% credible interval of death_rate_per_100k, as CDC publishes it, for modeled rates only (rate_modeled true).","valid_values":null,"example":null,"unit":"rate_per_population","value_min":null,"value_max":null,"null_meaning":"No interval: the rate is observed (rate_modeled false: 0 or 10 or more deaths), and CDC publishes credible intervals only for modeled rates; or the county had 1 to 9 deaths (rate_modeled true) and CDC flags the modeled rate as unstable and does not publish it or its interval (-999 in the source).","key_metric_grain_contributor":false,"metric_component":null},{"name":"num_deaths","type":"integer","role":"metric","nullable":true,"label":"Deaths","short_description":"","description":"Number of homicide deaths among residents of the county in the year, as CDC publishes it: 0, or 10 and more; CDC withholds 1 to 9 (NULL here, rate_modeled true). A published 0 is a real zero. Adds across counties only where none is NULL, so a sum over counties with a NULL is an undercount: there is no Georgia total in this dataset. Firearm homicides are also counted in 'all': never add the two methods.","valid_values":null,"example":181,"unit":"count","value_min":null,"value_max":null,"null_meaning":"Withheld by CDC: the county had 1 to 9 deaths in the year (shown as '1-9'); its rate is CDC's modeled estimate (rate_modeled true). Never a zero, and never backed out of a total.","key_metric_grain_contributor":false,"metric_component":null},{"name":"rate_modeled","type":"boolean","role":"categorical","nullable":false,"label":"Rate Is Modeled","short_description":"True when the rate is CDC's model estimate (1-9 deaths, count withheld).","description":"CDC's modeled-rate flag (rate_m): true where the county had 1 to 9 deaths, so num_deaths is withheld and death_rate_per_100k is an estimate from CDC's Bayesian model with a 95% credible interval (rate and interval NULL where CDC flags the modeled rate as unstable and does not publish it, -999 in the source); false where the rate is observed (0 or 10 or more deaths). Describes the row, not part of its key.","valid_values":null,"example":false,"unit":null,"value_min":null,"value_max":null,"null_meaning":null,"key_metric_grain_contributor":false,"metric_component":null}],"key_metric":"death_rate_per_100k","numeric_filter_columns":["death_rate_per_100k","death_rate_per_100k_ci_lower","death_rate_per_100k_ci_upper","num_deaths"],"usage":"Star-schema fact table at county grain, calendar years. Join the counties dimension on county_fips; tract_geoid is always NULL. Pin method='all' for every homicide or 'firearm' for firearm homicides (a subset of 'all': never add the two). Where rate_modeled is true the rate is CDC's model estimate (1-9 deaths, count withheld), not deaths / population: read it with its 95% credible interval, and treat two counties whose intervals overlap as not clearly different. Rates never add or average across counties or years, and no population is published, so a combined rate for several counties cannot be computed from this table. num_deaths adds across counties or years only where none is NULL (withheld): a sum with NULLs is an undercount. Read directly with DuckDB over Parquet.","limitations":"Homicide deaths by the victim's county of residence, not where the homicide happened, calendar years only. Counties with 1 to 9 deaths get a CDC model estimate in place of an observed rate (rate_modeled true), so many county rates are estimates. Rates are crude (deaths per 100,000 residents of all ages), not age-adjusted, so places with older or younger populations are not directly comparable. Where a county had 1 to 9 deaths CDC withholds the count (shown as '1-9': num_deaths NULL, never a zero, never backed out of a total) and publishes instead an estimate from its Bayesian model, which borrows strength from neighboring counties (rate_modeled true), with a 95% credible interval; a modeled rate is not deaths divided by population, and where CDC flags the modeled rate as unstable and does not publish it (-999 in the source) the rate and interval are NULL. Here 61% of rows have 1 to 9 deaths (18 of them with no rate at all): most counties' figures are model estimates, so compare them by their credible intervals. 'firearm' homicides are a subset of 'all' (never add the two) and are also counted in firearm_deaths (firearm deaths of every intent), so never add figures across the MIOV topics. This county dataset has no Georgia row: a sum of counties with a withheld count is an undercount (CDC publishes Georgia's totals in its separate MIOV state dataset, data.cdc.gov fpsi-y8tj, which is not part of this table). Not included: CDC's provisional trailing 12 months (June 2025 to May 2026), which CDC revises and rolls forward monthly; every row here is a calendar year. A modeled firearm rate and a modeled all-homicide rate come from separate model fits, so a firearm estimate can exceed the all-homicide estimate for the same county and year: in this release one does (Terrell County, 2023: 47.0 against 45.5, both from 1 to 9 deaths, the firearm figure inside the all-homicide estimate's credible interval). Observed firearm rates never exceed observed all-homicide rates. Not comparable with police crime counts, which count homicides where they happen and by their own definitions. CDC's data dictionary (not part of this source file) gives the cause-of-death codes; Georgia DPH's OASIS also publishes assault (homicide) deaths, with age-adjusted rates and demographics, from the state's own records and population estimates: compare places within one source only. No demographics, methods other than firearm, or nonfatal assaults in this table.","null_semantics":{"suppressed_to_null":true,"zero_is_real":true},"available_years":[2019,2020,2021,2022,2023,2024],"schema_hash":"7b61e1b6a35b9331c12f31d0044ec02f4c0591bab9695378f9f8a62a2a8b34f0","tags":["health","centers for disease control and prevention, national center for injury prevention and control, mapping injury, overdose, and violence","homicide_deaths"],"granularity_description":"One row per year, county_fips, tract_geoid, method.","example_queries":[{"description":"Latest year (2024), counties detail","query":"SELECT * FROM homicide_deaths WHERE year = 2024 LIMIT 100"},{"description":"Filter by method = firearm","query":"SELECT * FROM homicide_deaths WHERE method = 'firearm' AND year = 2024 LIMIT 100"}],"foreign_keys":[{"column":"county_fips","target_object":"counties","target_columns":["county_fips"],"attribute_columns_added":["county_name"],"scope":"global"},{"column":"tract_geoid","target_object":"tracts","target_columns":["tract_geoid"],"attribute_columns_added":["tract_name"],"scope":"global"}],"filters":[{"param":"year","kind":"year_exact","type":"integer","multi_value":false,"enum_enforced":false,"values":null},{"param":"year_min","kind":"year_range","type":"integer","multi_value":false,"enum_enforced":false,"values":null},{"param":"year_max","kind":"year_range","type":"integer","multi_value":false,"enum_enforced":false,"values":null},{"param":"detail","kind":"detail","type":"string","multi_value":false,"enum_enforced":true,"values":["counties"],"default":"counties"},{"param":"county_fips","kind":"foreign_key","column":"county_fips","type":"string","multi_value":true,"enum_enforced":false,"values":null},{"param":"tract_geoid","kind":"foreign_key","column":"tract_geoid","type":"string","multi_value":true,"enum_enforced":false,"values":null},{"param":"method","kind":"categorical","column":"method","type":"string","multi_value":true,"enum_enforced":true,"values":["all","firearm"],"has_total":true,"total_value":"all","requires_single_value":false,"nullable":false,"null_filter_token":null},{"param":"rate_modeled","kind":"categorical","column":"rate_modeled","type":"boolean","multi_value":true,"enum_enforced":false,"values":null,"has_total":false,"total_value":null,"requires_single_value":false,"nullable":false,"null_filter_token":null}],"filter_hints":[],"non_additive":[{"column":"method","reason":"'firearm' homicides are a subset of 'all' homicides (every firearm homicide is also in 'all'), so a sum across the two counts them twice. Use 'all' for every homicide.","metrics":["num_deaths"]}],"partitions":[],"value_implications":[],"year_meaning":null,"pooled_rates":[],"series_breaks":[],"year_window":null,"pooled_min_count":null,"values_endpoint":"/api/v1/health/homicide_deaths/values"}