Frequently Asked Questions

Mortality increased markedly over the study period, with the age-adjusted mortality rate rising from 6.0 per million in 1999 to a peak of 82.2 per million in 2021, then declining to 66.6 per million in 2023. Across 1999–2023, the study identified 195,180 deaths in adults aged 25 years and older in which both obesity and a psychiatric disorder were listed as multiple causes of death. Joinpoint analysis showed significant increases in 1999–2005 (APC 19.18%), 2005–2014 (APC 9.97%), 2014–2018 (APC 5.04%), and 2018–2021 (APC 20.64%), followed by a significant decline in 2021–2023 (APC −9.35%).

Substance use disorders had the highest mortality burden by a wide margin. The weighted mean age-adjusted mortality rate for mental and behavioral disorders due to psychoactive substance use (F10–F19) was 29.77 per million (95% CI, 29.61–29.91), including 4.90 per million (95% CI, 4.84–4.95) for alcohol use disorder.

• Mood disorders (F30–F39): 2.60 per million (95% CI, 2.56–2.64)
• Organic mental disorders (F01–F09): 2.21 per million (95% CI, 2.17–2.24)
• Schizophrenia spectrum disorders (F20–F29): 1.00 per million (95% CI, 0.96–1.03)
• Anxiety disorders (F41): 0.67 per million (95% CI, 0.63–0.69)

Major depressive disorder accounted for a larger mortality burden than bipolar disorder when obesity was also documented. Within mood disorders, the weighted mean age-adjusted mortality rate was 1.95 per million (95% CI, 1.91–1.99) for major depressive disorder (F32–F33) versus 0.68 per million (95% CI, 0.64–0.71) for bipolar disorder (F31). The overall weighted mean age-adjusted mortality rate for mood disorders (F30–F39) was 2.60 per million (95% CI, 2.56–2.64).

Yes. Men had higher mortality than women in both absolute burden and adjusted analyses. From 1999 to 2023, males accounted for 113,240 deaths, or 58% of the total, and had a higher age-adjusted mortality rate than females (40.4 vs 26.5 per million). In multivariable Poisson regression, males had a significantly higher adjusted mortality rate than females (IRR 1.41, 95% CI, 1.30–1.52; P<.001).

Trend patterns also differed by sex. Male mortality rose sharply from 1999 to 2006 and again from 2018 to 2021 (APC 18.66%, P<.05), while female mortality showed a similar but slightly delayed pattern, with its largest increase in 2018 to 2021 (APC 20.87%, P<.05). A permutation test confirmed that male and female trends significantly diverged (P=.000222).

Adults aged 65–74 years had the highest adjusted mortality rate relative to those aged 25–34 years. In the multivariable Poisson model, the incidence rate ratio for ages 65–74 years was 8.34 (95% CI, 7.67–9.06; P<.001) compared with the 25–34-year reference group. The highest absolute number of deaths occurred among adults aged 55–64 years (n=54,931), while the crude mortality rate was greatest in those aged 65–74 and 75–84 years.

Yes. The study found substantial racial and ethnic disparities, with non-Hispanic decedents having more than twice the adjusted mortality rate of Hispanic decedents. In adjusted analyses, non-Hispanic decedents had an IRR of 2.66 (95% CI, 2.46–2.87; P<.001) compared with Hispanic decedents.

Using American Indian or Alaska Native decedents as the reference group, adjusted mortality was lower among Asian or Pacific Islander decedents (IRR 0.09, 95% CI, 0.08–0.10; P<.001), Black decedents (IRR 0.73, 95% CI, 0.67–0.79; P<.001), and White decedents (IRR 0.81, 95% CI, 0.74–0.89; P<.001). In absolute counts, White decedents accounted for the largest number of deaths (166,176), while American Indians or Alaska Natives showed sustained growth over time with an average APC of 10.46% (P<.05).

Yes. Rural areas had steep early increases, and the Midwest had the highest adjusted regional mortality rate. Rural micropolitan areas had 31,689 deaths and noncore areas had 21,721 deaths, together accounting for more than 27% of all deaths. These areas showed the steepest early age-adjusted mortality increases: APC 23.01% in micropolitan counties and APC 20.05% in noncore counties, both P<.05.

After adjustment, the Midwest had higher mortality than the Northeast (IRR 1.21, 95% CI, 1.10–1.33; P<.001). By contrast, differences for the South (IRR 0.92, 95% CI, 0.85–1.01; P=.078) and West (IRR 1.07, 95% CI, 0.97–1.17; P=.183) were not statistically significant, and no significant differences were detected across urbanization categories after adjustment.

The highest state-level age-adjusted mortality rates were in Vermont, Wyoming, West Virginia, and North Dakota. The reported age-adjusted mortality rates were 216.7 per million in Vermont, 192.2 in Wyoming, 178.9 in West Virginia, and 178.8 in North Dakota.

These rate-based hotspots differed from the states with the largest absolute numbers of deaths, which were Texas (n=9,801), California (n=9,708), and New York (n=8,494), followed by Florida (n=7,973), Ohio (n=6,296), and Pennsylvania (n=5,644).

The ARIMA model projected that mortality will rise again, reaching 74.2 per million by 2030. Specifically, the forecast estimated an age-adjusted mortality rate of 55.1 per million in 2025 (95% CI, 51.4–59.0) and 74.2 per million in 2030 (95% CI, 66.1–82.8).

The best-fit model was ARIMA (0,1,0) with drift. It was trained on 1999–2018 data, validated on 2019–2023 data, and showed adequate diagnostic performance, including a root mean square error of 13.31 and no significant assumption violations on the reported Ljung-Box, Shapiro-Wilk, and Durbin-Watson tests.

This was a retrospective, population-based analysis of US death certificate data, so it describes mortality patterns but cannot establish causality. The study examined adults aged 25 years and older from 1999 through 2023 using CDC WONDER multiple-cause-of-death data. It included decedents in whom obesity (ICD-10 E66) and a psychiatric disorder (ICD-10 F01–F99) were both documented as contributing causes of death.

The investigators used age-adjusted mortality rates, joinpoint regression for temporal trends, multivariable Poisson and negative binomial modeling for disparities, and ARIMA forecasting through 2030. Because the data come from death certificates, the findings depend on documentation and coding practices rather than direct clinical assessment.

The main limitations are underreporting and misclassification in death certificate data, inability to determine causality, and limited clinical detail. The authors noted that psychiatric conditions and obesity may be underdocumented on death certificates, and ICD-10 codes may not fully capture psychiatric subtype or severity. CDC WONDER also did not allow cross-tabulation of cause-specific mortality by psychiatric subtype.

The dataset lacked information on treatment history, medication use, socioeconomic status, and access to care, which limits interpretation of mechanisms. The ARIMA projections are based on historical trends and could change with future policy shifts or medical innovations, and some subgroup analyses were limited by suppression of cells with fewer than 20 deaths.