Machine Learning–Based Projections of Cancer and Depression Mortality in the United States through 2035

Document Type

Conference Proceeding

Publication Date

5-27-2026

Publication Title

J Clin Oncol

Keywords

moclobemide, adult, autocorrelation, Black person, cancer mortality, conference abstract, controlled study, cross validation, depression, emotional stress, epidemiology, fear, female, forecasting, health disparity, high risk population, human, machine learning, male, malignant neoplasm, mental stress, metropolitan area, mortality, mortality rate, patient compliance, predictive model, quality of life, race, racial disparity, rural area, survival rate, time series analysis, underlying cause of death, United States, urban rural difference, urbanization

Abstract

Background: Clinical depression is a relatively common yet frequently overlooked source of suffering among patients with cancer. Recent research has highlighted a strong association between cancer and depression, where patients often develop depressive symptoms due to profound emotional stress, fear of the future, and the psychological burden of the disease. It is crucial to understand this association to establish integrated therapeutic strategies and for patient well-being, as depression has a substantial influence on treatment compliance, quality of life and survival rates. Furthermore, identifying sociodemographic disparities is vital to addressing health inequities and targeting high-risk populations. Methods: We analyzed U.S. national mortality data from 1999–2024 using the CDC WONDER database, identifying cancer as the underlying cause of death with depression as an associated cause. Data were stratified by sex, race, census region, and urbanization to identify specific population-level trends. Age-adjusted mortality rates (AAMR) were log-transformed and modeled using ARIMA-based machine learning forecasting. Machine learning–driven 10-fold time-series cross-validation was implemented to optimize model performance and generalizability, with residual autocorrelation assessed using the Ljung–Box test. Forecasted trends through 2035 were quantified using annual percentage change (APC) and average annual percent change (AAPC). Results: Between 1999 and 2024, there were 40, 000 deaths with a marked rise of 10, 000 such deaths occurring during the 2021–2024 period. Overall AAMR remained steady initially, followed by increasing mortality trends in females (APC: 5.06%) from 2016 to 2024, contributing to an overall significant rise (AAPC: +1.64%; P < 0.001) while males showed no net significant change (AAPC: +1.03; P=0.12). However, forecast modeling projected a continued increase through 2035, with an estimated rise of 16.3% among females and a decline of 12.3% among males. Regional and racial disparities were significant, particularly for Black African/American, who showed a rise between 2017 and 2024 (APC: 5.51%) and a projected change of 147.4% by 2035. Urban-rural differences also persisted, with medium metropolitan areas experiencing a sharp increase from 2015 to 2020 (APC: +8.71%) while the burden in rural areas rose from 2011 to 2020 (APC: +3.88%), followed by a projected change of 19.4% by 2035. Regionally, the South had an increase from 0.63% APC (1999–2016) to 21.3% through 2035. Conclusions: Cancer mortality among patients with comorbid depression varies significantly by populations. Projections through 2035 predict a rising burden, particularly among females, Black African/American and rural and south areas. These findings highlight the urgent need for targeted, integrated interventions that prioritize these high-risk groups to mitigate future mortality.

Volume

44

Issue

16_Suppl

First Page

e22577

Find It @ Sladen

Share

COinS