Chapter 6.3 · Two despairs

Overdose, alcohol and suicide

Across counties, suicide death rates are only weakly related to overdose and homicide rates (population-weighted correlations of +0.13 and −0.02), and among the larger counties where CDC WONDER publishes every year, overdose exceeds alcohol-induced deaths in 479 of 545 and the population-weighted correlation of county deprivation is +0.42 with suicide but +0.13 with alcohol-induced death rates.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 9 October 2026

Suicide, overdose and homicide death rates are highest in different counties, and suicide is the most separate. Across the 1,899 counties with a stable modeled rate for all three in 2019 to 2023, which hold 93.5% of residents, the county overdose rate has a population-weighted correlation of +0.13 (95% CI 0.05 to 0.21) with the suicide rate, and the suicide rate has a correlation of −0.02 (95% CI −0.10 to 0.06) with the homicide rate. Overdose and homicide rates are more closely related, with a correlation of +0.48 (95% CI 0.35 to 0.58). Counting every county once, the three correlations are smaller and in the same order: +0.19, +0.02 and +0.27.

Drug and alcohol deaths separate as well, but the counts behind that result come from a selected set of counties. CDC WONDER publishes a county's count in every year only when the count is zero or at least 10 each year, which favors larger counties with high rates for their size: among counties of 30,000 to 60,000 residents, the mean overdose rate in CDC's modeled panel is 50.9 (95% CI 46.1 to 56.4) in counties with a published count in every year and 22.5 (95% CI 21.5 to 23.4) in the rest. In the 545 such counties where each cause has at least 20 deaths, the overdose death rate is higher than the alcohol-induced death rate in 479 counties and lower in 66, and across the 569 counties that publish both counts every year the two rates have a correlation of +0.25 (95% CI 0.16 to 0.35) (+0.30 (95% CI 0.20 to 0.43) counting every county once). Among 525 of these counties, deprivation, measured with the Area Deprivation Index, has a population-weighted correlation of +0.42 (95% CI 0.30 to 0.51) with the suicide rate and +0.35 (95% CI 0.24 to 0.44) with the overdose rate, against +0.13 (95% CI 0.04 to 0.22) with the alcohol-induced rate. In all counties with a modeled rate, the population-weighted correlations with deprivation are +0.46 (95% CI 0.37 to 0.53) for suicide and +0.22 (95% CI 0.15 to 0.28) for overdose, and counting every county once they are +0.01 (95% CI −0.04 to 0.05) and −0.01 (95% CI −0.04 to 0.03).

Case and Deaton found that mortality among middle-aged white non-Hispanic Americans rose between 1999 and 2013, mostly through drug and alcohol poisonings, suicide and chronic liver disease,[1] and later grouped suicide, drug overdose and alcoholic liver disease as deaths of despair.[2] The grouping treats three kinds of death as one problem. This chapter asks whether they share an address, using death rates for counties. A county rate is deaths per hundred thousand residents, crude and not adjusted for age, and it says nothing about who died or why. Whether people in these places feel despair is a person-level question, taken up in Chapter 7.2, Life evaluation and daily feelings.

Suicide, overdose and homicide rates peak in different counties

The first three figures use the county file of CDC's injury and violence dashboard, which carries modeled estimates of death rates and is the only source in this chapter that also covers homicide. A model borrows strength across counties where deaths are few, so the maps can show counties that CDC WONDER's public counts leave blank. A county is drawn when CDC publishes a stable rate in each year from 2019 to 2023, and its rate is the mean of the five annual rates. That leaves 2,714 counties with an overdose rate, 2,273 with a suicide rate and 2,807 with a homicide rate, holding 97.7%, 94.9% and 97.6% of residents. CDC's file has 3,135 counties and leaves out Connecticut's. The suicide map has the fewest counties because CDC marks more small-county suicide rates unstable.

Figure 1

Across counties, the population-weighted correlation of suicide rates is +0.13 with overdose rates and −0.02 with homicide rates, against +0.48 between overdose and homicide rates

CDC modeled death rates per 100,000 residents, mean of annual rates for 2019 to 2023, by county: overdose, suicide and homicide in three views

Loading county values

Overdose deaths per 100,000 residents, 2019 to 2023 average

Values outside 0.0 to 53.9 take the end colors.

No data (3,143 counties)

Select a class to show only those places.

Find a county

Hover or tap a county for its value, or search by name. Zoom with the buttons on the map, then drag to move it.

Counties toward the high end of the key have higher overdose death rates, and 68.0% of the counties in the highest tenth are in the South, which holds 45.7% of counties with a rate. Across the 1,899 counties with a rate for all three kinds of death, the correlations are +0.13 between overdose and suicide, +0.48 between overdose and homicide and −0.02 between suicide and homicide when counties are weighted by population, and +0.19, +0.27 and +0.02 when every county counts once. The buttons switch the map to suicide and to homicide.Note: Each map plots a county's mean annual modeled rate for 2019 to 2023. A county is drawn when CDC publishes a stable rate in each of the five years; gray hatched counties have a year in which CDC marks the rate unstable (fewer than 10 deaths and no stable model) or, for two Alaska areas that did not exist in 2019, no row, and the tooltip says which. The overdose map has 2,714 counties with a rate, the suicide map 2,273 and the homicide map 2,807, and they hold 97.7%, 94.9% and 97.6% of residents.Note: These are CDC modeled estimates, which borrow strength across counties where deaths are few, so the WONDER floor of 20 pooled deaths does not apply here; it applies to the maps of WONDER counts in Figure 4. CDC publishes counts under 10 as ranges, and the tooltip gives the lowest and highest five-year totals those ranges allow. Overdose is unintentional or undetermined-intent drug overdose, so it leaves out overdoses recorded as suicides; suicide and homicide include every method. Rates are crude, not age-adjusted. The panel's 2024 values are provisional and left out.Note: Connecticut's eight counties are gray because CDC's county file does not carry them. Puerto Rico is not drawn. Alaska is drawn.Source: CDC Mapping Injury, Overdose, and Violence county file, 2019-2023; modeled estimates, crude rates

The counties with the highest rates lie in different places. In the highest tenth of overdose rates, 68.0% of the counties are in the South, which holds 45.7% of counties with an overdose rate. The South also holds 85.8% of the highest tenth for homicide, against 47.7% of counties with a homicide rate. The highest tenth for suicide lies elsewhere, with 50.0% of its counties in the West, which holds 13.7% of counties with a suicide rate.

The correlations in the opening paragraph give the same picture county by county. The suicide rate shares little with the overdose rate or the homicide rate, and the interval for its correlation with the homicide rate includes zero under both weightings. Overdose and homicide share more, but much less when every county counts once than when large counties carry their weight. The panel counts only unintentional and undetermined-intent drug poisonings as overdose, so an overdose recorded as a suicide appears in the suicide rate and not in the overdose rate. Homicide counts every method. Gun homicide is the subject of Chapter 6.2, Gun homicide and the causes of early death.

Homicide, suicide and overdose rates are higher in the most deprived tenth of counties, but homicide and overdose do not rise steadily

Rank counties by the Area Deprivation Index of their residents and cut the ranking into tenths with equal numbers of counties (Figure 2). A county's rank is the population-weighted mean of the national ranks of its tracts. Weighted by population, the homicide rate is 4.8 (95% CI 4.2 to 5.4) deaths per hundred thousand residents in the least deprived tenth and 13.6 (95% CI 11.1 to 16.2) in the most deprived, a ratio of 2.83 (95% CI 2.24 to 3.58). Suicide goes from 12.2 (95% CI 11.3 to 13.2) to 17.4 (95% CI 15.7 to 19.4), a ratio of 1.43 (95% CI 1.25 to 1.63), and the line is level over the more deprived tenths. Overdose goes from 23.8 (95% CI 22.2 to 25.6) to 31.1 (95% CI 25.7 to 36.5), a ratio of 1.30 (95% CI 1.07 to 1.55).

Homicide and overdose do not climb steadily. The point estimates of the homicide line fall in 4 of the nine steps between tenths, and those of the overdose line in 4. The overdose line reaches 36.1 (95% CI 30.6 to 42.0) in tenth 4, which is 5.0 (95% CI −2.6 to 13.2) above the most deprived tenth. The interval for that difference includes zero, and a peak chosen from ten noisy means overstates its lead, so the data do not show that overdose rates are higher in the middle of the ranking than at the deprived end. Across all counties with a rate, the population-weighted correlation with deprivation is +0.46 (95% CI 0.37 to 0.53) for suicide, +0.30 (95% CI 0.25 to 0.36) for homicide and +0.22 (95% CI 0.15 to 0.28) for overdose.

Figure 2

Weighted by population, homicide rates in the most deprived tenth of counties are 2.83 times those in the least deprived, suicide rates 1.43 times and overdose rates 1.30 times, but the homicide and overdose lines do not rise steadily

CDC modeled death rates per 100,000 residents by tenth of counties ranked by Area Deprivation Index, mean of annual rates for 2019 to 2023, county rates weighted by population

Show data

Data for Weighted by population, homicide rates in the most deprived tenth of counties are 2.83 times those in the least deprived, suicide rates 1.43 times and overdose rates 1.30 times, but the homicide and overdose lines do not rise steadily
cause	conf_high	conf_low	decile	estimate	n_units	weighting
Overdose	25.58	22.17	1	23.84	302	Population-weighted
Overdose	29.93	24.57	2	27.28	296	Population-weighted
Overdose	34.96	26.90	3	30.74	289	Population-weighted
Overdose	42.01	30.60	4	36.13	283	Population-weighted
Overdose	35.49	27.61	5	31.86	270	Population-weighted
Overdose	31.25	24.58	6	27.90	268	Population-weighted
Overdose	35.13	27.89	7	31.39	270	Population-weighted
Overdose	33.13	21.71	8	27.17	260	Population-weighted
Overdose	30.12	23.86	9	26.83	248	Population-weighted
Overdose	36.49	25.66	10	31.09	215	Population-weighted
Suicide	13.16	11.31	1	12.18	267	Population-weighted
Suicide	15.20	13.23	2	14.15	255	Population-weighted
Suicide	17.11	14.83	3	15.83	247	Population-weighted
Suicide	17.04	15.09	4	16.03	239	Population-weighted
Suicide	17.94	16.04	5	16.91	243	Population-weighted
Suicide	18.76	16.57	6	17.63	225	Population-weighted
Suicide	19.09	17.60	7	18.32	214	Population-weighted
Suicide	18.94	14.20	8	16.62	211	Population-weighted
Suicide	18.83	17.33	9	18.10	183	Population-weighted
Suicide	19.41	15.71	10	17.37	179	Population-weighted
Homicide	5.44	4.19	1	4.82	301	Population-weighted
Homicide	9.17	5.65	2	7.31	292	Population-weighted
Homicide	9.02	6.40	3	7.86	281	Population-weighted
Homicide	14.24	8.53	4	11.44	282	Population-weighted
Homicide	12.33	7.48	5	9.89	277	Population-weighted
Homicide	9.99	6.87	6	8.45	275	Population-weighted
Homicide	10.00	6.19	7	7.86	275	Population-weighted
Homicide	7.73	5.66	8	6.63	280	Population-weighted
Homicide	10.31	7.40	9	8.79	273	Population-weighted
Homicide	16.17	11.14	10	13.63	257	Population-weighted
Overdose	23.25	21.05	1	22.19	302	Each county counted once
Overdose	25.61	22.63	2	24.07	296	Each county counted once
Overdose	25.72	22.45	3	24.04	289	Each county counted once
Overdose	27.06	22.97	4	25.01	283	Each county counted once
Overdose	24.59	20.70	5	22.56	270	Each county counted once
Overdose	23.20	19.55	6	21.40	268	Each county counted once
Overdose	26.06	21.64	7	23.81	270	Each county counted once
Overdose	25.14	20.33	8	22.68	260	Each county counted once
Overdose	23.62	19.26	9	21.33	248	Each county counted once
Overdose	26.39	19.55	10	22.96	215	Each county counted once
Suicide	16.12	14.57	1	15.34	267	Each county counted once
Suicide	18.19	16.75	2	17.46	255	Each county counted once
Suicide	18.63	17.09	3	17.85	247	Each county counted once
Suicide	18.27	16.91	4	17.58	239	Each county counted once
Suicide	18.45	16.62	5	17.46	243	Each county counted once
Suicide	18.36	16.64	6	17.49	225	Each county counted once
Suicide	18.40	16.95	7	17.66	214	Each county counted once
Suicide	17.53	15.88	8	16.72	211	Each county counted once
Suicide	16.89	15.09	9	16.04	183	Each county counted once
Suicide	15.33	13.24	10	14.22	179	Each county counted once
Homicide	3.74	3.11	1	3.41	301	Each county counted once
Homicide	4.88	3.94	2	4.34	292	Each county counted once
Homicide	4.76	3.86	3	4.29	281	Each county counted once
Homicide	5.48	4.14	4	4.79	282	Each county counted once
Homicide	5.48	4.17	5	4.79	277	Each county counted once
Homicide	5.10	3.90	6	4.48	275	Each county counted once
Homicide	6.06	4.63	7	5.31	275	Each county counted once
Homicide	5.55	4.23	8	4.91	280	Each county counted once
Homicide	7.24	5.54	9	6.36	273	Each county counted once
Homicide	10.86	8.16	10	9.47	257	Each county counted once

Each ratio compares the most deprived tenth with the least deprived. The point estimates of the homicide and overdose lines fall in 4 and 4 of the nine steps between tenths. The overdose line reaches 36.1 in tenth 4, which is 5.0 above the most deprived tenth, a difference whose interval includes zero. Each line plots the population-weighted mean of county rates in a tenth of counties, with 95% intervals from a bootstrap over counties. The tenths hold equal numbers of counties, so the most deprived tenth is mostly small rural counties. The second view counts every county once.Note: The deprivation rank of a county is the population-weighted mean of the Area Deprivation Index national ranks of its tracts, so a county ranked as least deprived can contain deprived tracts. The index understates deprivation in high-cost areas.Note: Counties without a stable rate in all five years are left out of that cause's line: 2,714 counties for overdose, 2,273 for suicide and 2,807 for homicide, out of 3,120 counties with a deprivation rank. The suicide line is thinnest in the most deprived tenth, where CDC marks the most small-county rates unstable. Connecticut is not in CDC's county file.Note: The weighted and unweighted views differ for suicide and overdose: counting every county once, the ratio of the most deprived tenth to the least deprived is 0.93 for suicide and 1.03 for overdose, against 2.78 for homicide, so the weighted rise comes from the larger counties.Source: CDC Mapping Injury, Overdose, and Violence county file, 2019-2023, and Area Deprivation Index 2023; modeled estimates, crude rates, population-weighted and unweighted

Counting every county once changes the picture for two of the three. Homicide is 3.4 (95% CI 3.1 to 3.7) in the least deprived tenth and 9.5 (95% CI 8.2 to 10.9) in the most deprived, a ratio of 2.78 (95% CI 2.34 to 3.30), close to the weighted ratio. Suicide is 15.3 (95% CI 14.6 to 16.1) in the least deprived tenth and 14.2 (95% CI 13.2 to 15.3) in the most deprived, a ratio of 0.93 (95% CI 0.85 to 1.02), and overdose is 22.2 (95% CI 21.1 to 23.2) and 23.0 (95% CI 19.5 to 26.4), a ratio of 1.03 (95% CI 0.88 to 1.20). Neither ratio's interval excludes equal rates in the two tenths. The correlation with deprivation is −0.01 (95% CI −0.04 to 0.03) for overdose and +0.01 (95% CI −0.04 to 0.05) for suicide, against +0.18 (95% CI 0.16 to 0.21) for homicide. The weighted rise in suicide and overdose comes from the larger counties, and the small counties that make up most of the most deprived tenth do not show it.

Three features of the tenths limit what they show. The tenths cover the 3,120 counties with a deprivation rank and hold equal numbers of counties, so the most deprived tenth is mostly small rural counties, and a county without a stable rate for a kind of death is left out of that line. Those counties are also small ones, and the suicide line has the fewest counties in its most deprived tenth for that reason. The Area Deprivation Index is built from American Community Survey measures of income, education, employment and housing, and it understates deprivation in high-cost cities (Chapter 10.2, The anatomy of a deprivation index), so a high-cost county can sit in a less deprived tenth than its tracts would suggest. The homicide correlation with deprivation here differs from the one in Chapter 6.2, Gun homicide and the causes of early death, which uses County Health Rankings' homicide rates, available only for counties with enough homicides, and a different set of counties.

Suicide is higher in more urban nonmetro counties than in large metro counties, and homicide is higher in large metro counties

Figure 3 groups counties by the 2023 Rural-Urban Continuum Code. Weighted by population, suicide is 19.7 (95% CI 19.2 to 20.2) deaths per hundred thousand residents in nonmetro counties with an urban population of 5,000 or more (more urban nonmetro counties) and 12.2 (95% CI 11.6 to 12.9) in large metro counties, those in metro areas of 1 million or more. The ratio is 1.6 (95% CI 1.5 to 1.7). Homicide runs the other way, 7.6 (95% CI 6.7 to 8.7) in large metro counties against 4.6 (95% CI 4.2 to 5.1) in nonmetro counties with a smaller urban population (least urban nonmetro counties), a ratio of 1.6 (95% CI 1.4 to 1.9). The overdose rate is 27.8 (95% CI 25.7 to 30.0) in large metro counties and 21.8 (95% CI 20.2 to 23.5) in least urban nonmetro counties, a ratio of 1.3 (95% CI 1.1 to 1.4).

Figure 3

Weighted by population, suicide rates in more urban nonmetro counties are 1.6 times those in large metro counties, and homicide rates in large metro counties are 1.6 times those in the least urban nonmetro counties

CDC modeled death rates per 100,000 residents by 2023 Rural-Urban Continuum Code group of county, mean of annual rates for 2019 to 2023, county rates weighted by population

Overdose
Suicide
Homicide

Show data

Data for Weighted by population, suicide rates in more urban nonmetro counties are 1.6 times those in large metro counties, and homicide rates in large metro counties are 1.6 times those in the least urban nonmetro counties
cause	conf_high	conf_low	county_type	estimate	n_units	weighting
Overdose	30.00	25.67	Large metro	27.79	428	Population-weighted
Overdose	29.93	26.91	Smaller metro	28.41	702	Population-weighted
Overdose	27.72	25.21	Nonmetro, more urban	26.45	851	Population-weighted
Overdose	23.48	20.20	Nonmetro, least urban	21.77	733	Population-weighted
Suicide	12.88	11.57	Large metro	12.21	387	Population-weighted
Suicide	17.15	16.00	Smaller metro	16.58	629	Population-weighted
Suicide	20.19	19.23	Nonmetro, more urban	19.69	668	Population-weighted
Suicide	18.54	17.01	Nonmetro, least urban	17.74	589	Population-weighted
Homicide	8.69	6.65	Large metro	7.62	431	Population-weighted
Homicide	7.29	6.15	Smaller metro	6.69	704	Population-weighted
Homicide	6.30	5.24	Nonmetro, more urban	5.76	846	Population-weighted
Homicide	5.13	4.19	Nonmetro, least urban	4.65	826	Population-weighted
Overdose	29.92	26.84	Large metro	28.34	428	Each county counted once
Overdose	27.02	24.61	Smaller metro	25.81	702	Each county counted once
Overdose	24.96	22.88	Nonmetro, more urban	23.88	851	Each county counted once
Overdose	17.23	14.75	Nonmetro, least urban	15.96	733	Each county counted once
Suicide	15.03	14.14	Large metro	14.59	387	Each county counted once
Suicide	17.86	17.08	Smaller metro	17.47	629	Each county counted once
Suicide	20.02	19.15	Nonmetro, more urban	19.58	668	Each county counted once
Suicide	14.99	13.69	Nonmetro, least urban	14.34	589	Each county counted once
Homicide	6.87	5.62	Large metro	6.20	431	Each county counted once
Homicide	6.34	5.50	Smaller metro	5.90	704	Each county counted once
Homicide	6.00	5.08	Nonmetro, more urban	5.53	846	Each county counted once
Homicide	3.89	3.16	Nonmetro, least urban	3.53	826	Each county counted once

Weighted by population, suicide rates are higher in each of the three smaller types of county than in large metro counties, homicide rates are lower at each step from large metro counties to the least urban nonmetro counties, and the overdose rate in large metro counties is 1.3 times the rate in the least urban nonmetro counties. Each dot is the population-weighted mean of county rates with a 95% interval from a bootstrap over counties. The second view counts every county once.Note: The groups follow the Rural-Urban Continuum Code: large metro is counties in metro areas of 1 million or more, smaller metro is the other metro counties, more urban nonmetro counties have an urban population of 5,000 or more, and least urban nonmetro counties have fewer. Connecticut is not in CDC's county file.Note: Rates are crude, not age-adjusted. Age structure differs between rural and urban counties, and suicide rates differ by age, so part of a difference between types of county can reflect age structure and not the same risk at every age.Source: CDC Mapping Injury, Overdose, and Violence county file, 2019-2023, and USDA Rural-Urban Continuum Codes 2023; modeled estimates, crude rates, population-weighted and unweighted

Counting every county once, suicide is 19.6 (95% CI 19.2 to 20.0) in more urban nonmetro counties against 14.6 (95% CI 14.1 to 15.0) in large metro counties, a ratio of 1.3 (95% CI 1.3 to 1.4), and homicide is 6.2 (95% CI 5.6 to 6.9) in large metro counties against 3.5 (95% CI 3.2 to 3.9) in least urban nonmetro counties, a ratio of 1.8 (95% CI 1.5 to 2.0), so both contrasts hold. Two statements do not. Suicide is no higher in least urban nonmetro counties than in large metro counties when every county counts once, and the overdose ratio is 1.8 (95% CI 1.6 to 2.0), with 28.3 (95% CI 26.8 to 29.9) in large metro counties and 16.0 (95% CI 14.8 to 17.2) in least urban nonmetro counties, a larger gap than the weighted one. Overdose rates are lower in least urban nonmetro counties under both weightings, and they differ more by county type than the weighted figures suggest.

These are crude rates. Age structure differs between rural and urban counties and suicide rates differ by age, so part of a gap between county types can reflect who lives in the county and not a different risk at the same age. The panel's rates are also modeled for small counties, which can borrow strength from neighboring areas, so a nonmetro county's rate may partly reflect its neighbors.

Overdose deaths exceed alcohol-induced deaths in most counties where both can be mapped

Alcohol-induced deaths appear in this chapter's data only through CDC WONDER, which counts deaths by the underlying cause on the death certificate. Here drug means drug poisoning of any intent (unintentional, suicide, homicide or undetermined), alcohol-induced means WONDER's category of alcohol-induced deaths, which includes alcoholic liver disease and alcohol poisoning, and suicide means intentional self-harm by any method. A suicide by drug poisoning is in both the drug count and the suicide count. All three are crude rates per hundred thousand residents.

WONDER withholds any county-year count above zero and below 10, and it publishes a count of zero, so a rate pooled over the published years of a county is biased. This chapter does not impute, bound or back out the withheld counts. It uses the counties whose count is published in every year, which it calls complete counties, for which the rate pooled over 2018 to 2024 is exact. Complete counties are selected on the outcome as well as on size, so the maps and gradient of WONDER counts describe these counties and not counties in general. Chapter 10.5, Suppressed counties in death data covers how many county-years are withheld, which counties lose their counts and what the selection does to comparisons.

Figure 4 maps the ratio of the two rates, on a log scale, for the 545 counties that are complete for both and have at least 20 pooled deaths of each. They hold 76.6% of residents outside Connecticut. The overdose rate is the higher one in 479 of them and the alcohol-induced rate in 66. Each class of the key is a doubling of the ratio: the overdose rate is at least 2 times the alcohol-induced rate in 234 mapped counties, and the alcohol-induced rate is at least 2 times the overdose rate in 6. The counties where alcohol is higher are mostly in the West and Midwest, which hold 78.8% of them. Alcohol is the higher rate in 28.0% of mapped counties in the West and 16.4% in the Midwest, against 6.5% in the South and 0.0% in the Northeast.

Figure 4

Overdose deaths exceed alcohol-induced deaths in 479 of 545 mapped counties, and the 66 counties where alcohol is higher lie mostly in the West and Midwest

Crude alcohol-induced death rate relative to crude overdose death rate, log base 2 of the ratio, counties with a published count in every year and at least 20 pooled deaths of each cause, 2018 to 2024

Loading county values

Alcohol-induced death rate relative to overdose death rate (log base 2 of the ratio; each class is a doubling)

More overdose deathsMore alcohol-induced deaths

Values outside −3 to +3 take the end colors.

No data (3,143 counties)

Select a class to show only those places.

Find a county

Hover or tap a county for its value, or search by name. Zoom with the buttons on the map, then drag to move it.

Violet counties have a higher overdose than alcohol-induced death rate and whiskey counties the reverse. Each class of the key is a doubling of the ratio, and the outermost classes hold counties where one rate is at least 4 times the other. The overdose rate is at least 2 times the alcohol-induced rate in 234 mapped counties, and the alcohol-induced rate is at least 2 times the overdose rate in 6. Alcohol is the higher rate in 28.0% of mapped counties in the West and 16.4% in the Midwest, against 6.5% in the South and 0.0% in the Northeast. Across the 569 counties whose two counts are published in every year, the overdose and alcohol-induced rates have a correlation of +0.25 weighted by population and +0.30 counting every county once.Note: A county is drawn when WONDER publishes its drug and alcohol counts in every year from 2018 to 2024 and it has at least 20 pooled deaths of each. WONDER withholds county-year counts above zero and below 10, so a rate built from the published years of a county with withheld years would leave out its low-count years; the map uses only counties with no withheld year. That leaves 545 of the counties, which hold 76.6% of residents outside Connecticut, and the tooltip gives the reason for every gray county.Note: Those counties are selected on the outcome as well as on size. Among counties of 30,000 to 60,000 residents, the mean overdose rate in CDC's modeled panel is 50.9 in counties whose WONDER overdose count is published in every year and 22.5 in the rest: a county whose deaths reach 10 or more in every year has a high rate for its size. The map describes those counties and not counties in general.Note: Rates are crude, not age-adjusted. Drug deaths are drug poisonings of any intent (unintentional, suicide, homicide and undetermined), alcohol-induced deaths are WONDER's alcohol-induced category, and the county ratio compares counts and populations from the same WONDER query. Connecticut's counties have no population in WONDER for 2022 to 2024, so they are gray. Valdez-Cordova AK is coded by its old code and draws on both successor areas; Kusilvak AK and Oglala Lakota SD are drawn under their current codes. Puerto Rico is not drawn.Source: CDC WONDER underlying cause of death, 2018-2024; crude rates, counties with a published count in every year

Several of the counties with the highest alcohol-induced rates have large American Indian and Alaska Native populations. Of the ten mapped counties with the highest alcohol-induced rate, 6 are counties where American Indian or Alaska Native residents are at least 10% of residents, by County Health Rankings' figures from the American Community Survey. Only 10 of the 133 counties in WONDER of that kind are mapped, and alcohol is the higher rate in 6 of those. Their pooled alcohol-induced rate is 57.3 deaths per hundred thousand residents, against 13.6 in the other mapped counties. The cutoff matters. At 5% instead, there are 27 such counties and alcohol is the higher rate in 8 of them. These county totals are not divided by race, so the comparison describes counties and not the people who died.

Among counties with published counts in every year, population-weighted correlations with deprivation are higher for suicide and overdose rates than for alcohol-induced rates

Figure 5 repeats the gradient for the WONDER rates, in the 525 counties whose drug, alcohol and suicide counts are published in every year, which hold 76.0% of residents outside Connecticut. The tenths are tenths of these counties, which are selected as described above, so they do not line up with the tenths of Figure 2. The correlation of the county rate with the deprivation rank, weighted by person-years, is +0.42 (95% CI 0.30 to 0.51) for suicide, +0.35 (95% CI 0.24 to 0.44) for overdose and +0.13 (95% CI 0.04 to 0.22) for alcohol-induced deaths. Counting every county once, the correlations are +0.28 (95% CI 0.19 to 0.36), +0.26 (95% CI 0.19 to 0.34) and +0.15 (95% CI 0.05 to 0.22). From the same bootstrap draws, the weighted correlation for overdose is higher than the one for alcohol by +0.21 (95% CI 0.09 to 0.33), and the one for suicide by +0.28 (95% CI 0.18 to 0.39). In the least deprived tenth the rates are 22.0 (95% CI 20.0 to 24.8), 12.5 (95% CI 10.8 to 14.0) and 9.9 (95% CI 9.1 to 11.1) for overdose, alcohol-induced deaths and suicide. In the most deprived tenth they are 32.3 (95% CI 25.2 to 40.0), 17.6 (95% CI 13.9 to 22.7) and 16.5 (95% CI 14.2 to 18.9). The overdose rate is higher in the most deprived tenth than in the least deprived by 10.3 (95% CI 2.5 to 18.3) deaths per hundred thousand residents and the suicide rate by 6.6 (95% CI 4.0 to 9.2), differences between the two tenths' pooled rates with intervals from bootstrap draws that the two tenths share.

Figure 5

Among counties with published counts in every year, the population-weighted correlation of deprivation is +0.42 with suicide and +0.35 with overdose death rates, against +0.13 with alcohol-induced death rates

Crude death rates per 100,000 residents by tenth of counties ranked by Area Deprivation Index, counties with a published count in every year for all three kinds of death, 2018 to 2024, weighted by person-years

Tenth of the counties in the figure by Area Deprivation Index (1 = least deprived, 10 = most deprived)

Show data

Data for Among counties with published counts in every year, the population-weighted correlation of deprivation is +0.42 with suicide and +0.35 with overdose death rates, against +0.13 with alcohol-induced death rates
cause	conf_high	conf_low	decile	estimate	n_units
Overdose	24.80	20.04	1	21.99	52
Overdose	27.34	20.94	2	24.01	53
Overdose	31.36	23.49	3	27.64	52
Overdose	29.42	23.00	4	26.06	53
Overdose	30.54	22.71	5	27.06	52
Overdose	39.78	26.38	6	32.06	53
Overdose	41.98	27.94	7	34.55	52
Overdose	45.37	30.72	8	37.26	53
Overdose	39.62	29.73	9	34.91	52
Overdose	40.00	25.20	10	32.29	53
Alcohol-induced causes	14.04	10.78	1	12.54	52
Alcohol-induced causes	15.52	11.99	2	13.76	53
Alcohol-induced causes	16.56	12.24	3	14.44	52
Alcohol-induced causes	14.22	10.93	4	12.30	53
Alcohol-induced causes	15.58	12.07	5	13.42	52
Alcohol-induced causes	17.39	11.97	6	14.30	53
Alcohol-induced causes	15.22	12.22	7	13.48	52
Alcohol-induced causes	16.86	14.36	8	15.57	53
Alcohol-induced causes	16.05	13.60	9	14.71	52
Alcohol-induced causes	22.74	13.88	10	17.57	53
Suicide	11.05	9.07	1	9.90	52
Suicide	14.56	11.76	2	13.07	53
Suicide	16.09	12.55	3	14.35	52
Suicide	14.53	12.38	4	13.27	53
Suicide	15.88	11.98	5	13.72	52
Suicide	16.51	13.24	6	14.61	53
Suicide	16.50	14.06	7	15.08	52
Suicide	17.56	15.05	8	16.27	53
Suicide	16.99	14.71	9	15.73	52
Suicide	18.93	14.24	10	16.50	53

The suicide and overdose lines are higher in the most deprived tenth than in the least deprived, by 6.6 and 10.3 deaths per hundred thousand residents (differences of pooled rates, with intervals from draws shared by the two tenths), while the alcohol-induced line is level until the most deprived tenth, where it is 17.6 against 12.5 in the least deprived. Each line plots the pooled crude rate of a tenth of the 525 counties in the figure, with 95% intervals from a bootstrap over counties. The population-weighted correlation with deprivation is higher for overdose than for alcohol by +0.21 and higher for suicide than for alcohol by +0.28. Counting every county once, the correlations are +0.28 for suicide, +0.26 for overdose and +0.15 for alcohol-induced deaths.Note: The figure uses the 525 counties whose drug, alcohol and suicide counts WONDER publishes in every year from 2018 to 2024, which hold 76.0% of residents outside Connecticut. WONDER withholds county-year counts above zero and below 10, which would drop the low-count years of a small county and bias its rate. The tenths are tenths of these counties ranked by deprivation, with equal numbers of counties, so the most deprived tenth here is not the most deprived tenth of all counties.Note: These counties are selected on the outcome as well as on size (see Figure 4), so the correlations describe them and not counties in general. In all counties with a modeled rate, the population-weighted correlation with deprivation is +0.46 for suicide and +0.22 for overdose (Figure 2), and +0.01 and −0.01 counting every county once.Note: Rates are crude, not age-adjusted. A suicide by drug poisoning is counted in both the overdose and the suicide rate. The most deprived tenth includes counties with large American Indian and Alaska Native populations, where alcohol-induced death rates are high.Source: CDC WONDER underlying cause of death, 2018-2024, and Area Deprivation Index 2023; crude rates, population-weighted (person-years)

The overdose line here keeps rising into more deprived tenths than the panel's line in Figure 2, which is a different set of counties ranked in a different way. In all counties with a modeled rate the population-weighted correlation with deprivation for overdose is +0.22 (95% CI 0.15 to 0.28), below the +0.35 (95% CI 0.24 to 0.44) here, so the selection matters for the overdose result. The alcohol-induced line is level until the most deprived tenth, and the jump there depends on a few counties. 5 of the tenth's counties are ones where American Indian or Alaska Native residents are at least 10% of residents, and without them the tenth's rate is 14.4 (95% CI 12.1 to 16.6), in line with the other tenths.

The panel and WONDER give the same county rates where both can be compared. Across counties whose WONDER counts are published in every year, the county rate in the two sources has a correlation of +0.995 (95% CI 0.993 to 0.996) for overdose and +0.995 (95% CI 0.994 to 0.996) for suicide, although WONDER's overdose count includes overdoses recorded as suicides and the panel's does not. In those counties the panel's rates are modeled in 0.6% of overdose county-years and 0.0% of suicide county-years, against 44.9% and 54.0% in all counties, so the agreement checks the processing of the same death records where the panel's rates are direct and does not validate the modeled rates for small counties behind the first three figures.

The maps do not show change over time, which the chapter on mortality trends covers

The paths of the national overdose, alcohol-induced and suicide rates from 2018 to 2024, the size of the overdose peak and the fall since, and the check of WONDER's crude series against CDC's modeled panel, which gives a different size for the rise, are in Chapter 6.4, Mortality trends, 2018 to 2024. A study of death records describes the fall in overdose deaths from 2023 to 2024 as concentrated in deaths involving illicit fentanyl, while deaths involving stimulants without fentanyl and deaths involving xylazine continued to make up a growing share of overdose deaths.[3] That study is descriptive, and a fall in a national rate says nothing about which counties it reached. This chapter does not map change over time.

Explaining the fall is a causal question, and the maps cannot answer it. One policy lever has been studied with designs that compare states by when they adopted it.

Causal evidenceDifference-in-differences, staggered state adoption

Naloxone access laws and opioid deaths

Two studies use the staggered timing of state laws that widened access to naloxone, the drug that reverses an opioid overdose. Abouk, Pacula and Powell compared states before and after they adopted each type of law with states that had not adopted it, using state and time fixed effects and the years since adoption, for 2005 to 2016.[4] Laws that gave pharmacists direct authority to dispense naloxone were followed by fewer fatal opioid overdoses, and laws of other types showed no association with deaths. Doleac and Mukherjee used the same variation in timing and found more opioid-related emergency room visits and more opioid-related theft, with no net reduction in opioid-related deaths.[5] The two studies disagree about the same kind of policy, so the evidence on naloxone laws is mixed, and both studies end well before 2024, so neither can say how much of the fall after 2023 naloxone, treatment or changes in the drug supply account for.

What these data cannot show

Every figure describes counties. A county rate is an average over all of its residents, so it cannot say who died, and a correlation across counties does not describe the people in them (ecological inference). The data do not measure despair. Deaths from suicide, overdose and alcohol are counted, and whether the people who live in these counties feel despair is a person-level question for Chapter 7.2, Life evaluation and daily feelings. Survey measures of drinking are in Chapter 5.2, Health behaviors, person by person.

The WONDER figures cover a selected set of counties. Counties whose counts are published in every year hold most residents but are a minority of counties, they are larger, and they have higher rates than other counties of similar size. Most counties where American Indian or Alaska Native residents are at least 10% of residents are outside the maps, as are other small counties with high rates. Withheld counts could hold up to the shares given in the methods note, and no method in this chapter recovers them. North Carolina's drug-poisoning count for 2023 in WONDER predates the state's amended death certificates and is low. It is 15.8% of the published deaths of the 48 complete North Carolina counties in the window. Raising each of their published 2023 counts by 10% to 19% raises their pooled overdose rate by 1.6% to 3.0%. At the larger increase the correlation of overdose with deprivation is +0.35 (95% CI 0.24 to 0.44) and the overdose gap between the most and least deprived tenths is 10.4 (95% CI 2.6 to 18.4), and no comparison in this chapter changes direction. The modeled panel covers nearly every county but borrows strength across them, so its rate for a small county is partly its neighbors' rate, and its agreement with WONDER is checked only where its rates are direct. Alcohol-induced deaths are deaths whose underlying cause is alcohol, such as alcoholic liver disease, and they leave out deaths from other conditions in which alcohol was involved.

Several results depend on the weighting. Weighted by population, suicide and overdose rates are higher in the most deprived tenth of counties and suicide rates are higher outside large metro counties, and counting every county once the first holds for homicide only and the second for smaller metro and more urban nonmetro counties only. Rates are crude, so a county with an older population has a higher rate for the same risk at each age, and comparisons of rural and urban counties carry that difference. The Area Deprivation Index understates deprivation in high-cost cities. The tenths of counties are relative to the counties in each figure. The three kinds of death also overlap a little: a suicide by drug poisoning is counted as both a suicide and a drug death. Outside the box above, nothing in this chapter supports a claim about what lowers any of these death rates.

Data sources

CDC injury and violence mappingCDC Mapping Injury, Overdose, and Violence county file: modeled overdose, suicide and homicide death rates, annual 2019 to 2023Years: 2019-2024 for counties, 2022-2024 for tracts (data as of 2026-05-14) · Weights: 2019-2023 ACS residents of each county, and every county once · License: Public domain (CDC, U.S. Department of Health and Human Services)
CDC WONDER underlying cause of deathCDC WONDER underlying cause of death: drug poisoning (overdose), alcohol-induced causes and suicide, counties 2018 to 2024Years: 2018-2024 · Weights: none (crude rates from counts); WONDER county population · License: Public domain (NCHS, CDC); WONDER data-use restrictions apply
American Community SurveyTotal population of 2020 tracts, 2019-2023, summed to counties as weights for the panel's county ratesYears: 2019-2023 (5-year) · Weights: none (full count) · License: Public domain (U.S. Census Bureau)
Area Deprivation IndexNeighborhood Atlas Area Deprivation Index 2023, tract national rank averaged to countiesYears: 2023 release (version 4.0.1; built on ACS 2019-2023) · Weights: 2019-2023 tract population · License: Free for non-commercial use; cite the Neighborhood Atlas and Kind and Buckingham
Rural-Urban Continuum CodesUSDA Rural-Urban Continuum Codes 2023, countyYears: 2023 · Weights: none (classification) · License: Public domain (U.S. federal)
County Health RankingsCounty Health Rankings 2024, share of residents who are American Indian or Alaska Native (Census Population Estimates 2022)Years: 2024 release · Weights: none (published county measure) · License: CC BY 4.0 (University of Wisconsin Population Health Institute)

All data sources

Methods

The chapter compares death rates across counties from two sources: the CDC injury and violence panel, which carries modeled overdose, suicide and homicide rates for 2019 to 2023, and CDC WONDER counts for 2018 to 2024. WONDER withholds small nonzero counts, so the WONDER results use only counties whose counts are published in every year; these are larger counties, selected on the outcome as well as on size, with higher rates than other counties of similar size. Statistics across counties are computed twice, weighted by residents and counting every county once, and several results differ between the two. The panel and WONDER are compared only in the counties that are complete in WONDER.

How the panel's county rates are built

Two sources serve different jobs. The CDC injury and violence panel (county file, data as of May 2026) carries modeled overdose, suicide and homicide death rates for every county except Connecticut's.

A county's rate is the mean of its annual rates for 2019 to 2023 and is kept only if CDC publishes a stable rate in each of the five years; 2024 is provisional and left out.

Overdose in the panel is unintentional or undetermined-intent drug overdose; suicide and homicide include every method.

Figures explained here: Figure 1, Figure 2 and Figure 3

Why the WONDER results use only complete counties

CDC WONDER underlying cause of death supplies drug poisoning (overdose, any intent), alcohol-induced causes and suicide, all crude rates per 100,000 residents.

WONDER withholds a county-year count above zero and below 10 and publishes a count of 0, so the published years of a county leave out its low-count years and keep its zero years; a rate pooled over those years is biased. Nothing is imputed, bounded or derived from totals.

The chapter instead uses counties whose count is published in every year 2018 to 2024 (complete counties), for which the pooled rate is exact, and it reports how many counties and how many residents that leaves.

Complete counties are larger counties and are selected on the outcome as well as on size: among counties of 30,000 to 60,000 residents the panel's mean overdose rate is 50.9 in complete counties and 22.5 in the rest.

Withheld county-years are 47.3% of county-years for overdose, 57.9% for alcohol-induced causes and 57.4% for suicide; at 9 deaths each, they could hold at most 13.7%, 29.0% and 28.3% of the deaths in the county files.

Figures explained here: Figure 4 and Figure 5

How counties are ranked, grouped and mapped

County deprivation is the mean of the Area Deprivation Index national ranks of a county's tracts, weighted by 2019-2023 ACS tract population, with Connecticut tracts assigned to the eight legacy counties; tenths of counties hold equal numbers of counties.

County types follow the 2023 Rural-Urban Continuum Code: large metro (code 1), smaller metro (2 and 3), more urban nonmetro with an urban population of 5,000 or more (4 to 7) and least urban nonmetro with fewer (8 and 9).

Connecticut's counties have no WONDER population for 2022 to 2024 and are left out, and the shares of residents that the maps and the gradient hold divide by the residents of the other 49 states and DC.

WONDER codes Valdez-Cordova AK, Kusilvak AK and Oglala Lakota SD by their old codes, so tracts are merged to those codes for deprivation and county type, and County Health Rankings rows are reached through the successor codes. Maps need at least 20 pooled deaths.

The share of residents who are American Indian or Alaska Native comes from County Health Rankings (Census Population Estimates 2022).

Figures explained here: Figure 1, Figure 2, Figure 3, Figure 4 and Figure 5

How counties are weighted and the two sources compared

Statistics across counties are computed twice, weighted by 2019-2023 ACS residents and counting every county once, and both are reported where they differ.

The panel and WONDER are compared only in complete counties, where the panel's rates are modeled in 0.6% of overdose county-years and 0.0% of suicide county-years, against 44.9% and 54.0% in all counties.

Figures explained here: Figure 2, Figure 3 and Figure 5

How North Carolina's low 2023 drug count is handled

North Carolina's 2023 drug count in WONDER predates amended death certificates and is low; it is 15.8% of the published deaths of its 48 complete counties in the window.

Raising each of their published 2023 counts by 10% to 19% (a withheld cell is never changed) raises their pooled overdose rate by 1.6% to 3.0%, and every comparison asserted in this build holds at both increases.

Figures explained here: Figure 4 and Figure 5

How uncertainty is measured

Intervals for means, ratios, correlations and differences are 95% percentile intervals from a bootstrap that resamples whole counties (2,000 replicates, seed 12345), with shared draws for ratios and differences.

References

Case, Anne; Deaton, Angus (2015). Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century. Proceedings of the National Academy of Sciences 112(49): 15078-15083. doi:10.1073/pnas.1518393112
Case, Anne; Deaton, Angus (2020). Deaths of Despair and the Future of Capitalism. Princeton University Press.
Friedman, Joseph R.; Palamar, Joseph J.; Ciccarone, Daniel; Gaines, Tommi L.; Borquez, Annick; Shover, Chelsea L.; Strathdee, Steffanie A. (2026). Charting the decline of the fourth wave: US overdose deaths by race, ethnicity and substance involvement. Addiction 121(9): 2552-2558. doi:10.1111/add.70472
Abouk, Rahi; Pacula, Rosalie Liccardo; Powell, David (2019). Association between state laws facilitating pharmacy distribution of naloxone and risk of fatal overdose. JAMA Internal Medicine 179(6): 805-811. doi:10.1001/jamainternmed.2019.0272
Doleac, Jennifer L.; Mukherjee, Anita (2022). The effects of naloxone access laws on opioid abuse, mortality, and crime. Journal of Law and Economics 65(2): 211-238. doi:10.1086/719588

How to cite

@misc{singh2026overdosealcoholandsuicide,
  author = {Singh, Vishal},
  title = {Overdose, alcohol and suicide},
  year = {2026},
  url = {https://vishalsingh.org/zipcode/overdose-alcohol-and-suicide},
  note = {In ZIP Code Destiny. Published 9 October 2026.}
}

Same question at another grain

These chapters take up the same axis with a different unit of analysis.

Despair
- Chapter 7.2, Life evaluation and daily feelings Individuals

Reading paths

Outcomes that do not follow the gradient