Article

March 2, 2023

Suicide Risk Among Individuals Diagnosed with Cancer in the United States, 2000-2016

Title:

Suicide Risk Among Individuals Diagnosed with Cancer in the United States, 2000-2016

Authors: Xin Hu, MSPH1; Jiemin Ma, PhD2; Ahmedin Jemal, DVM, PhD3; Jingxuan Zhao, MPH3; Leticia Nogueira, PhD3; Xu Ji, PhD4,5; K. Robin Yabroff3, PhD; Xuesong Han, PhD3

Affiliation:

1 Department of Health Policy and Management, Emory University Rollins School of Public Health, Atlanta, Georgia 30322

2 Merck & Co., Inc., Kenilworth, NJ 07033

3 Surveillance and Health Equity Science, American Cancer Society, Atlanta, Georgia 30144

4 Department of Pediatrics, Emory University School of Medicine, Atlanta, Georgia 30322

5 Aflac Cancer & Blood Disorders Center, Children's Healthcare of Atlanta, Atlanta, Georgia 30322

Corresponding Author:

Xuesong Han, PhD

Surveillance and Health Equity Science

American Cancer Society, 3380 Chastain Meadows Pkwy NW, Suite 200, Kennesaw, GA 30144, USA

E-mail: xuesong.han@cancer.org

Telephone: 404.929.6813

Fax numbers: 404.321.4669

Date of Revision: Nov 19th, 2022

Word Count: 3341

Number of Tables: 2

Number of Figures: 2

Number of References:45

Acknowledgments

We gratefully acknowledge the contributions of the state and regional cancer registry staff for their work in collecting the NAACCR CiNA data used in this study.

Funding

None.

Disclosures

Xuesong Han and Jingxuan Zhao have received grant from AstraZeneca for research outside of the current study. Xin Hu has received a predoctoral fellowship grant from PhRMA foundation outside of the current study. K. Robin Yabroff serves on the Flatiron Health Equity Advisory Board.

This study was presented at the 2022 American Society of Clinical Oncology (ASCO) Annual Meeting and at the NAACCR Summer Forum 2022.

Author Contributions

All authors conceptualized the study, interpreted data, and reviewed and revised the manuscript.

Xin Hu and Xuesong Han designed the study, administered the project, and drafted the initial manuscript. Xin Hu carried out the analyses.

All authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

Data availability

The data underlying this article were provided by the American Association of Central Cancer Registries (NAACCR) by permission. The data cannot be shared publicly per the Data Use Agreement. The NAACCR CiNA Public Use Data Set with limited number of variables is available through application at https://www.naaccr.org/cina-public-use-data-set/. Xin Hu and Xuesong Han had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Key Points (100/100 limit)

Question:

What is the suicide risk associated with cancer in the US, and what are sociodemographic and clinical risk factors for suicide among individuals diagnosed with cancer?

Findings:

In this population-based cohort of 16.8 million individuals with cancer from 43 states, suicide risk was 26% higher (statistically significant) compared to the general population, with insurance status and ethnicity contributing to the elevated risk. Higher suicide risk was seen among individuals with poor prognosis cancers within 2 years of diagnosis and among cancers prone to long-term quality-of-life impairments after 2 years.

Meaning:

Timely symptom management and targeted psychosocial interventions are warranted for suicide prevention in individuals diagnosed with cancer.

Tweet (246/257 limit)

Individuals diagnosed with cancer had 26% higher risk of suicide death than general population in 2000-2016 in the US.Insurance status and ethnicity contributed to elevated risks. Suicide risk differed by time since diagnosis and cancer type.

ABSTRACT (349/350 limit)

Importance: Individuals diagnosed with cancer have elevated suicide risk compared to the general population. National estimates of suicide risks among individuals with cancer are lacking in the US and knowledge about risk factors is limited.

Objective: To provide contemporary estimates of suicide risks associated with cancer, and to identify sociodemographic and clinical factors associated with suicide risks among individuals diagnosed with cancer.

Design: A population-based cohort of individuals diagnosed with cancer in 2000-2016 from 43 states were followed through December 31, 2016. Standardized Mortality Ratios (SMR) were calculated based on attained age at death, sex, and race/ethnicity to compare suicide risks in the cancer cohort vs. the general US population. Cox proportional hazard models were fitted to identify cancer-specific risk factors of suicide among the cancer cohort. Analyses were conducted from October 27, 2020 to May 13, 2022.

Exposure: Diagnosis of cancer.

Results: Among a total of 16,771,397 individuals with cancer, 51.5% were male, 78.4% were non-Hispanic White, and 20,792 (0.3%) died from suicide. The overall SMR for suicide was 1.26 (95%CI=1.24-1.28), with a decreasing trend (from SMR=1.67 in 2000 to SMR=1.16 in 2016). Compared to the general population, elevated suicide risk was observed in cancer cohort across all sociodemographic groups, with particularly high SMRs among Hispanics, Medicaid-insured, Medicare-insured aged ≤64 years, or uninsured. Moreover, the highest SMR was observed in the first 6 months following cancer diagnosis (SMR[95%CI] =7.19[6.97-7.41]). Among individuals diagnosed with cancer, relatively higher suicide risks (i.e., Hazard Ratios) were observed for poor prognosis cancer types with high symptom burdens in the first two years following diagnosis, including cancers of oral cavity & pharynx, esophagus, stomach, brain, pancreas, and lung. After two years, individuals with cancers subject to long-term quality-of-life impairments, such as oral cavity & pharynx, female breast, uterine, bladder, and leukemia, had higher suicide risks.

Conclusions and Relevance: In this cohort study of individuals with cancer, elevated suicide risks remained despite a decreasing trend during the past two decades. Suicide risk varied by sociodemographic and clinical factors. Timely symptom management and targeted psychosocial interventions are warranted for suicide prevention in individuals diagnosed with cancer.

INTRODUCTION

Suicide is a leading cause of death worldwide, with more than 700,000 persons dying by suicide every year globally.1 In the US, over 45,000 people died from suicide in 2020, corresponding to a rate of 14.0 per 100,000 person-years.2 Previous studies, mostly from Europe and North America, reported up to 11 times higher risk of suicide death among individuals diagnosed with cancer compared to the general population.3-9 While cancer remains the main cause of death among individuals diagnosed with cancer, the elevated suicide risk is concerning and potentially preventable.

Previously reported risk factors for suicide among individuals with cancer included being male, older ages, advanced stage at diagnosis, and rural residence.3-7 However, estimates in the US have been limited to the Surveillance, Epidemiology, and End Results Program (SEER) registries from 13 states or fewer and were not able to examine the contribution of some unique factors of the US population to increased suicide risks, such as state of residence, insurance coverage, ethnicity, and county-level socioeconomic status (SES). Unlike most of the developed countries, there is not a universal healthcare system in the US, and health insurance coverage is a strong determinant of health care access and health outcomes.10-12 In the US, the federal Medicare program insures Americans 65 years or older and certain younger people with disabilities or specific conditions; most Americans under age 65 years receive private health insurance through employers; and the federally aided, state-operated Medicaid program provides insurance coverage for low-income people with income eligibility varying by state.11 Although the Affordable Care Act (ACA) significantly expanded insurance coverage in the US in the last decade, about 30 million Americans lacked health insurance coverage in 2020.13Moreover, few studies examined temporal trends in suicide risk among individuals with cancer, although the suicide rate has risen dramatically in the US during the past two decades. To fill the knowledge gap, this study aims to provide estimates of suicide death risks and to examine sociodemographic and clinical characteristics associated with suicide risks. We used recently released comprehensive data from population-based cancer registries in 43 states in the US. Evidence from this study can better inform future efforts to improve psychosocial care and symptom management among individuals diagnosed with cancer.

METHODS

Data

We used the Cancer Incidence in North America (CiNA) Survival dataset compiled by the North American Association of Central Cancer Registries (NAACCR), which contained data from Centers for Disease Control and Prevention (CDC) National Program of Cancer Registries (NPCR) and the National Cancer Institute (NCI) SEER Registries and were certified by NAACCR’s high-quality data standards.14 We identified 16,954,604 individuals of all ages whose first primary malignant cancer was diagnosed in 2000-2016 from 43 population-based state cancer registries that agreed to participate in this study and provided data usable for survival analysis. NAACCR survival dataset does not include individuals with unknown age or sex, or cancer diagnoses reported on death certificate or autopsy only. We further excluded individuals alive with survival time missing (n=11,109, 0.1%), or unknown race or ethnicity (n=172,098, 1.0%). Eligible individuals were followed through the date of last contact, death, or December 31, 2016, whichever came first. Mortality data for the US general population in 2000-2016 were acquired from the National Center for Health Statistics.15 The study was based on de-identified data and exempt from review by the NAACCR Institutional Review Board. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline (www.equator-network.org/reporting-guidelines/strobe/) for this cohort study.

Measures

For individuals who died during the study period, cause of death was identified using the International Classification of Diseases, Tenth Revision (ICD-10) codes. Suicide was identified using codes U03 (Suicide terrorism), X60-X84 (Intentional self-harm), and Y87.0 (Sequelae of intentional self-harm).

Individual-level covariates were abstracted from medical records and included sex, age at diagnosis (in 5-year intervals), race/ethnicity, year of diagnosis, state of residence, SEER summary stage, and known primary payer at diagnosis (i.e., insurance type), which was categorized as private, Medicaid, Medicare for individuals aged ≤64 years, Medicare for individuals aged ≥65 years, Veteran Affairs (VA) or Indian/Public Health Service (PHS) insurance, and uninsured (see eTable1 for the availability of insurance information). County-level covariates from the American Community Surveys, including rurality and percent of persons below poverty, were available for individuals diagnosed in 2005 or later, and were included in subgroup analyses for individuals diagnosed in 2005-2016.

Statistical Analyses

Individuals’ sociodemographic and clinical characteristics were described for the entire cohort, those who died within the study period, and those who died of suicide.

To estimate suicide risk associated with cancer and identify risk-factors, we conducted two separate sets of analyses. First, standardized mortality ratios (SMRs), reflecting the risk of suicide among individuals with cancer relative to the general population, were calculated as the observed number of suicides divided by the expected number of suicides in the cancer cohort.16 The expected number of suicides was calculated by multiplying the sex-, age-, and race/ethnicity-specific suicide rate in the study period 2000-2016 from the general population by total person-years of the corresponding sex, age, race/ethnicity group of the cancer cohort. We used attained age at death for standardization, calculated from age at diagnosis and survival time in months in the cancer cohort assuming a uniform distribution within each 5-year age group. Sensitivity analyses assuming the smallest age or the largest age within each interval showed similar results (eTable2). Overall SMR and SMRs stratified by sociodemographic including year of death and clinical characteristics were generated.

Second, to identify cancer-specific risk factors for suicide among individuals with cancer, we estimated Hazard Ratios (HR) using Cox regression controlling for other causes of death as competing risks.17 Unlike SMR, where the suicide risk is interpreted relative to the general population, an HR indicates the suicide risk relative to the reference group among individuals with cancer. Covariates were selected based on a priori knowledge and included year of diagnosis, age at diagnosis, sex, race/ethnicity, state, primary payer at diagnosis, cancer stage, and cancer site as fixed effects. County-level rurality and poverty were included in subgroup analyses among those diagnosed in 2005-2016. After testing the proportional hazards assumption for each variable using visual examination of Cumulative Incidence Function, we included age at diagnosis, cancer stage, and cancer site as time-dependent variables in the extended models, for which HRs for the first 2 years and after 2 years of follow-up were generated.18,19

RESULTS

Sample characteristics and overall SMR

The analytical cancer cohort consisted of 16,771,397 individuals with 93,476,318 person-years during 2000-2016. Most of the cohort were male (51.5%), non-Hispanic White (78.4%), and residing in metropolitan areas (80.8%). Among those whose health insurance information was available, the majority had private insurance (42.3%) or Medicare aged≥65 years (40.4%). The most common cancer sites were prostate (15.3%), female breast (14.9%), lung (13.2%), and colon and rectum (9.6%); 11.3% had multiple cancer diagnoses at the end of follow-up.There were 7,972,782 individuals (47.5%) who died during the study period, with 20,792 (0.3%) suicide deaths. Among individuals with cancer who died by suicide, 43.5% of the deaths occurred within 2 years of cancer diagnosis and 20.1% occurred within 6 months of diagnosis. The overall SMR for suicide was 1.26 (95%CI=1.24-1.28). (Table 1)

Suicide risks of cancer cohort vs. general population by sociodemographic characteristics

Over the study period, the age-standardized SMR (95%CI) by state ranged from 0.47 (0.28-0.65) in Wyoming to 1.77 (1.39-2.14) in Alaska. (Figure 1) Individuals with cancer in Alaska, North Dakota, Nebraska and New Mexico had the highest SMR for suicide compared to the general US population.

SMRs by year of death were plotted to examine the time trend in suicide risks associated with cancer. (Figure 2A) Although individuals diagnosed with cancer had consistently higher suicide risks in all years, elevated risks [SMR (95%CI)] declined from 1.67 (1.47-1.88) in 2000 to 1.15 (1.09-1.21) in 2010 and then leveled off with small fluctuations, with an SMR=1.16 (95%CI=1.11-1.21) in 2016.

Individuals diagnosed with cancer had significantly higher suicide risks than the general population in all attained age groups older than 40 years. There was an increasing trend of SMRs by attained age up to 65-69 years old (SMR=1.44, 95%CI=1.39-1.50) followed by a decreasing trend among age groups older than 69 years. (Table 1 & Figure 2B)

Compared to their counterparts among the general population, we found higher SMR among individuals diagnosed with cancer who were male (1.29, 95%CI=1.27-1.30), Hispanic (1.48, 95%CI=1.38-1.58), American Indian/Alaska Native/Asian/Pacific Islander (AIANAPI) (1.79, 95%CI=1.63-1.95), insured with Medicaid (1.72, 95%CI=1.61-1.84), Medicare aged ≤64 years (1.94, 95%CI=1.80-2.07), VA or Indian/PHS (1.89, 95%CI = 1.72-2.05), uninsured (1.66, 95%CI=1.53-1.80), and those residing in rural areas (1.39, 95%CI=1.24-1.54, Table 1).

Suicide risks of cancer cohort vs. general population by clinical characteristics

Suicide risk in the cancer cohort was substantially higher than the general population within the first 6 months of diagnosis [SMR(95%CI) =7.19(6.97-7.41)] and decreased with longer time since cancer diagnosis, but remained higher than the general population until 5 years after diagnosis [SMR(95%CI)=0.94(0.91-0.97) for 5-9 years since diagnosis] (Table 1, Figure 2C).

Suicide risks were higher among individuals diagnosed with cancer compared to the general population across all cancer stages, with the highest SMR(95% CI) for distant stage [1.90(1.83-1.96)], followed by regional stage [1.46(1.41-1.50)] and in situ/local stages [1.04(1.02-1.06); Table 1]. Across cancer sites, the highest SMRs were observed among individuals diagnosed with cancers of lung [2.34(2.25-2.44)], oral cavity & pharynx [2.42(2.28-2.56)], pancreas [2.50(2.24-2.76)], esophagus [3.15(2.82-3.47)], and stomach [2.32(2.07-2.58)]. In contrast, individuals diagnosed with prostate [0.91(0.89-0.94)] and thyroid [0.82(0.73-0.91)] cancers showed lower suicide risks than the general population.

Risk factors for suicide among individuals diagnosed with cancer

The Cox regressions showed that, compared to individuals diagnosed with cancer in 2000, individuals diagnosed in recent years had significantly lower suicide risks (HRs <1.0 after 2009, eFigure 1A). Compared to individuals residing in the most populous state California, individuals residing in New Mexico [HR (95%CI)=1.68 (1.48-1.91)], Nevada [1.59 (1.43-1.78)], and Alaska [1.47 (1.19-1.83)] had the highest suicide risks (eFigure 1B). Individuals who were male (vs. female), non-Hispanic White (vs. other racial/ethnic minorities) and living in non-metropolitan (vs. metropolitan) counties and counties of high (vs. low) poverty also had higher suicide risks. (Table 2 and eTable 3) Compared to privately insured individuals, individuals ≤64 years old insured with Medicare [1.48 (1.37-1.60)], VA or Indian/PHS [1.34 (1.22-1.47)], Medicaid [1.23 (1.14-1.33)], and uninsured [1.27 (1.16-1.38)] had higher suicide risks (Table 2).

Suicide risks associated with age at diagnosis, cancer stage, and cancer site were time-dependent. Specifically, within the first 2 years of cancer diagnosis, older age at diagnosis and more advanced cancer stage at diagnosis were associated with higher suicide risks; while after 2 years of diagnosis, diagnosis age of 25-49 years and in situ/local stage at diagnosis were associated with higher suicide risks. (eFigure 1C and Table 2) Using colorectal cancer (prevalent in both males and females) as the reference group, higher suicide risks [HR (95%CI)] were seen for the following cancer sites during the first 2 years of diagnosis: oral cavity & pharynx [2.07 (1.87-2.29)], esophagus [2.13 (1.88-2.41)], stomach [1.70 (1.48-1.94)], brain [1.38 (1.18-1.61)], lung [1.31 (1.21-1.42)] and pancreas [1.27 (1.12-1.43)], while the lower suicide risks were seen for prostate [0.64 (0.59-0.70)], leukemia [0.73 (0.63-0.85)], liver [0.78 (0.66-0.93)] and melanoma [0.84 (0.74-0.95)]. In 2 or more years of follow up after diagnosis, individuals with oral cavity & pharynx cancer still had the highest suicide risk [1.51 (1.36-1.68)], followed by leukemia [1.49 (1.25-1.77)], female breast cancer [1.24 (1.11-1.38)], uterine cancer [1.17 (1.01-1.36)], bladder cancer [1.13 (1.03-1.24)] and non-Hodgkin lymphoma [1.12 (0.996-1.25)]; while thyroid, pancreas, brain, lung, esophagus, stomach and kidney cancers showed relatively lower suicide risks compared to colorectal cancer. (Table 2)

DISCUSSION

This study provides national estimates of suicide risks associated with cancer and identifies risk factors of suicide among individuals diagnosed with cancer using data from 43 population-based state cancer registries in 2000-2016 in the US.We found that the elevated suicide risk associated with cancer decreased during the study period, coinciding with increased use of psychosocial and palliative care and advances in symptom management.20-22 However, the suicide risk among individuals with cancer remained higher than the general population in all years. Geographic, racial/ethnic, socioeconomic, and clinical characteristics, some of which are modifiable, contributed to the elevated suicide risks. More specifically, the highest suicide risk occurred in the first six months following diagnosis, during which individuals diagnosed with cancer bore more than seven times the suicide risk of the general population. Cox regression analysis among individuals diagnosed with cancer showed that older age, distant stage, and cancer types with poor prognosis and high symptom burden23 (e.g., cancers of oral cavity & pharynx, esophagus, stomach, brain, pancreas and lung) had higher risks of suicide in the first 2 years of diagnosis. After two years, individuals with oral cavity & pharynx cancers and other cancer types subject to long-term quality of life impairment (e.g., female breast cancer, uterine cancer, leukemia) had higher suicide risks. These findings can inform clinical practice and treatment guidelines to better address patient psychosocial needs and symptom management, including palliative care.

Both the SMR estimates from comparisons with the general population and the HR estimates from comparisons only among individuals diagnosed with cancer showed a decreasing trend of suicide risks during 2000-2016. Our analysis using a defined cancer cohort confirms a previous finding based on death certificate data, which also reported a declining cancer-related suicide rate in the past two decades.24 Moreover, our data supplements death certificates with richer information on key socially and clinically relevant characteristics, including health insurance coverage, and follow-up time. These decreasing trends coincide with the greater incorporation of psychosocial and palliative care into oncological care.20-22 Similarly, continuous efforts in education and training of oncologists, psycho-oncologists, and palliative care specialists, and development of clinical guidelines have been put into place in the past decades.25-27

This is the first time the NAACCR CiNA data, which comprises the largest cohort of individuals newly diagnosed cancer from 43 states, has been used to examine suicide risk.We filled critical gaps in the literature, besides confirming the risk factors identified by previous studies (e.g. male, older age, living in poor areas) in limited geographic areas3-5 . First, with wider geographic areas, we identified several states with high suicide risk—Alaska, Nevada and New Mexico—some of which were non-SEER states and not previously studied. These results may be related to the suicide clusters among American Indian and Alaska Native communities.28 The political, social, culture and economic environment may also contribute to the suicide risk in an area and merit further research.29 For example, Medicaid expansion under the Affordable Care Act was recently found to be associated with decreased suicide rates among both the general population and individuals diagnosed with cancer.30,31 Second, Hispanic and AIANAPI individuals with cancer were at substantially higher suicide risks than their peers without cancer, which points to potential barriers to healthcare resources, structural racism, and difficulties navigating healthcare systems among this population.32,33 This underscores the importance of increasing diversity, language, and cultural competence among healthcare professionals, improving health insurance coverage, and tailored psychosocial support for Hispanic and AIANAPI individuals diagnosed with cancer. Lastly, we characterized suicide risks associated with cancer by insurance type. Besides the uninsured, Medicare beneficiaries ≤ 65 years, whose eligibility for coverage is based on certain medical conditions or disability; Medicaid-insured, whose eligibility for coverage is mostly based on low-income 34,35 and individuals with VA or Indian/PHS coverage, two populations at increased risks of mental health conditions and suicide,36,37 all had substantially higher suicide risks than the general population. Hence, expanding insurance coverage to uninsured population and ensuring comprehensive coverage for cancer care and mental health services, including but not limited to suicide screening and prevention, as well as symptom management and palliative care by all Medicare, Medicaid, VA, and Indian Health Service/PHS programs is crucial.

Our in-depth analyses showed that suicide risks associated with stage and cancer type are time-dependent. During the first 2 years following the cancer diagnosis, when most individuals undergo active treatments, late-stage and cancer types with poor prognosis and heavy symptom burdens had the highest suicide risks (e.g., cancers of the lung, oral cavity & pharynx, esophagus, pancreas and stomach), which is consistent with prior research.3,59 After 2 years, when many individuals transition to survivorship care, those with cancer types that are typically subject to physical and psychological long-term and late effects, functional impairments, and poor quality of life (e.g., female breast cancer, uterine cancer, leukemia) had higher suicide risks. These findings have implications for oncology care in both early and later survivorship phases. Suicide screening and prevention should be prioritized among individuals recently diagnosed with fatal cancers and also be incorporated into long-term survivorship care. Considering the dynamic pattern of care delivery with time since cancer diagnosis, our findings also point to the importance of multi-level interventions from federal and state government-level healthcare reforms ensuring adequate access to care to provider-level engagement, including oncologists, psycho-oncologists, primary care physicians, mental health professionals, and social workers throughout the cancer care continuum.

Our estimated SMRs were smaller than previous studies overall and for different subgroups.3,5,7 This could be due to our inclusion of more recent data and decreasing trends of suicide among individuals with cancer, better representation of individuals with cancer in the US (e.g., we included Maryland and Rode Island which had relatively low suicide rates and were not included in previous SEER estimates), and improvement in risk standardization using attained age at death. Many previous studies used age at diagnosis among the cancer cohort, but attained age at death among the general population for age standardization.3,5,38,39 This could lead to biased estimates because suicide could occur years after cancer diagnosis given the improvement in cancer survival in the past decades40 and the fact that suicide rates vary substantially by age group.41,42 Another advantage of our study is that we used robust analytic models to estimate risk factors among the cancer cohort, such as including other causes of death as competing risks and time-varying interaction terms allowing different hazards in the period just following diagnosis and long-term survivorship.

Limitations

Our study has several limitations. First, exact attained age at death and age at diagnosis were not available and were imputed based on 5-year age intervals. However, our sensitivity analyses using alternative imputation methods (at the minimum and maximum age in the 5-year age intervals) showed similar results (eTable2). Second, the availability of health insurance coverage information varied by state and year (eTable1). However, no systematic patterns were observed and the direction of our estimates by insurance type was as expected. Our estimations by insurance type should be calibrated with more complete data in future studies as the quality of payer data improves.43 Third, we could not estimate differential risks by treatment due to lack of detailed treatment information.Cancer treatments combined with supportive care can improve prognosis while reducing side effects. Examining the associations of cancer treatment(s), including supportive care, and suicide risk, is an important area for future research. Fourth, our data did not include selected cancer registries declining participation or data being unusable for survival analysis. However, cancer cases from the 43 states of our study represent over 80% cancer incidence in the nation.44 Lastly, the cause of death from death certificates may be subject to reporting bias.45 However, we do not expect such bias to differ between the general population and individuals with cancer.

CONCLUSION

In this cohort study of individuals diagnosed with cancer from 43 US states during the past two decades, we found that suicide risks decreased among individuals with cancer but remained higher compared to the general population. Geographic, racial/ethnic, socioeconomic, and clinical characteristics, some of which are modifiable, contributed to the elevated suicide risks among individuals diagnosed with cancer. Screening and tailored social and psych-oncological interventions are needed for suicide prevention in this vulnerable population. These require joint efforts by federal and state governments, as well as healthcare institutions and providers, to ensure comprehensive health insurance coverage for psycho-oncological, psychosocial and palliative care, development of appropriate clinical guidelines for suicide risk screening, and inclusion of suicide prevention in survivorship care plans.

REFERENCES

1.World Health Organization. Suicide worldwide in 2019: Global health estimates. 16 June 2021. https://www.who.int/publications/i/item/9789240026643. Accessed November 10, 2022. In.

2.Suicide and Self-Harm Injury. 2022; https://www.cdc.gov/nchs/fastats/suicide.htm. Accessed Jan 22nd, 2022.

3.Misono S, Weiss NS, Fann JR, Redman M, Yueh B. Incidence of suicide in persons with cancer. J Clin Oncol. 2008;26(29):4731-4738.

4.Suk R, Hong Y-R, Wasserman RM, et al. Analysis of Suicide After Cancer Diagnosis by US County-Level Income and Rural vs Urban Designation, 2000-2016. JAMA Network Open. 2021;4(10):e2129913-e2129913.

5.Zaorsky NG, Zhang Y, Tuanquin L, Bluethmann SM, Park HS, Chinchilli VM. Suicide among cancer patients. Nature Communications. 2019;10(1):207.

6.Anguiano L, Mayer DK, Piven ML, Rosenstein D. A literature review of suicide in cancer patients. Cancer Nurs. 2012;35(4):E14-26.

7.Robson A, Scrutton F, Wilkinson L, MacLeod F. The risk of suicide in cancer patients: a review of the literature. Psychooncology. 2010;19(12):1250-1258.

8.Heinrich M, Hofmann L, Baurecht H, et al. Suicide risk and mortality among patients with cancer. Nature Medicine. 2022;28(4):852-859.

9.Henson KE, Brock R, Charnock J, Wickramasinghe B, Will O, Pitman A. Risk of Suicide After Cancer Diagnosis in England. JAMA Psychiatry. 2019;76(1):51-60.

10.Halpern MT, Ward EM, Pavluck AL, Schrag NM, Bian J, Chen AY. Association of insurance status and ethnicity with cancer stage at diagnosis for 12 cancer sites: a retrospective analysis. Lancet Oncol. 2008;9(3):222-231.

11.Ward E, Halpern M, Schrag N, et al. Association of insurance with cancer care utilization and outcomes. CA Cancer J Clin. 2008;58(1):9-31.

12.Zhao J, Han X, Nogueira L, et al. Health insurance status and cancer stage at diagnosis and survival in the United States. CA Cancer J Clin. 2022;72(6):542-560.

13.Finegold K, Conmy A, Chu RC, Bosworth A, Sommers BD. Trends in the US uninsured population, 2010-2020. Published online February 11, 2021. https://aspe.hhs.gov/sites/default/files/private/pdf/265041/trends-in-the-us-uninsured.pdf. Accessed November 10, 2022. 2021.

14.Weir HK, Johnson CJ, Mariotto AB, et al. Evaluation of North American Association of Central Cancer Registries’ (NAACCR) Data for Use in Population-Based Cancer Survival Studies. JNCI Monographs. 2014;2014(49):198-209.

15.Surveillance, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov) SEER*Stat Database: Mortality - All COD, Aggregated With State, Total U.S. (1969-2019) <Katrina/Rita Population Adjustment>. National Cancer Institute, DCCPS, Surveillance Research Program. Accessed Feb 25 2022.

16.Breslow NE, Day NE. Statistical methods in cancer research. Volume II--The design and analysis of cohort studies. IARC Sci Publ. 1987(82):1-406.

17.Lunn M, McNeil D. Applying Cox regression to competing risks. Biometrics. 1995;51(2):524-532.

18.Bellera CA, MacGrogan G, Debled M, de Lara CT, Brouste V, Mathoulin-Pélissier S. Variables with time-varying effects and the Cox model: Some statistical concepts illustrated with a prognostic factor study in breast cancer. BMC Medical Research Methodology. 2010;10(1):20.

19.Tyler Smith BS. Survival Analysis And The Application Of Cox's Proportional Hazards Modeling Using SAS. SAS.

20.Institute of Medicine Committee on Psychosocial Services to Cancer Patients/Families in a Community S. The National Academies Collection: Reports funded by National Institutes of Health. In: Adler NE, Page AEK, eds. Cancer Care for the Whole Patient: Meeting Psychosocial Health Needs. Washington (DC): National Academies Press (US)

21.Teno JM, Gozalo PL, Bynum JP, et al. Change in end-of-life care for Medicare beneficiaries: site of death, place of care, and health care transitions in 2000, 2005, and 2009. JAMA. 2013;309(5):470-477.

22.Rubens M, Ramamoorthy V, Saxena A, et al. Palliative Care Consultation Trends Among Hospitalized Patients With Advanced Cancer in the United States, 2005 to 2014. American Journal of Hospice and Palliative Medicine®. 2018;36(4):294-301.

23.van den Beuken-van Everdingen MH, de Rijke JM, Kessels AG, Schouten HC, van Kleef M, Patijn J. Prevalence of pain in patients with cancer: a systematic review of the past 40 years. Ann Oncol. 2007;18(9):1437-1449.

24.Han X, Hu X, Zhao J, Ma J, Jemal A, Yabroff KR. Trends of Cancer-Related Suicide in the United States: 1999-2018. J Natl Cancer Inst. 2021;113(9):1258-1262.

25.Ferris FD, Bruera E, Cherny N, et al. Palliative Cancer Care a Decade Later: Accomplishments, the Need, Next Steps—From the American Society of Clinical Oncology. Journal of Clinical Oncology. 2009;27(18):3052-3058.

26.Holland JC. Psycho-oncology: Overview, obstacles and opportunities. Psycho-Oncology. 2018;27(5):1364-1376.

27.Pirl WF, Greer JA, Gregorio SW, et al. Framework for planning the delivery of psychosocial oncology services: An American psychosocial oncology society task force report. Psychooncology. 2020;29(12):1982-1987.

28.Substance Abuse and Mental Health Services Administration. Suicide Clusters within American Indian and Alaska Native Communities: A review of the literature and recommendations. HHS Publication No. SMA17-5050. Rockville, MD: Center for Mental Health Services, Substance Abuse and Mental Health Services Administration, 2017. https://store.samhsa.gov/sites/default/files/d7/priv/sma17-5050.pdf. Accessed November 10, 2022.

29.Stack S. Contributing factors to suicide: Political, social, cultural and economic. Prev Med. 2021;152(Pt 1):106498.

30.Barnes JM, Graboyes EM, Adjei Boakye E, et al. The Affordable Care Act and suicide incidence among adults with cancer. J Cancer Surviv. 2022.

31.Patel H, Barnes J, Osazuwa-Peters N, Bierut LJ. Association of State Medicaid Expansion Status With Rates of Suicide Among US Adults. JAMA Netw Open. 2022;5(6):e2217228.

32.Fernandez A, Schillinger D, Grumbach K, et al. Physician language ability and cultural competence. An exploratory study of communication with Spanish-speaking patients. J Gen Intern Med. 2004;19(2):167-174.

33.Ellis KR, Black KZ, Baker S, et al. Racial Differences in the Influence of Health Care System Factors on Informal Support for Cancer Care Among Black and White Breast and Lung Cancer Survivors. Fam Community Health. 2020;43(3):200-212.

34.Riley GF, Lubitz JD, Zhang N. Patterns of health care and disability for Medicare beneficiaries under 65. Inquiry. 2003;40(1):71-83.

35.Yabroff KR, Zhao J, Zheng Z, Rai A, Han X. Medical Financial Hardship among Cancer Survivors in the United States: What Do We Know? What Do We Need to Know? Cancer Epidemiology Biomarkers & Prevention. 2018;27(12):1389-1397.

36.VA National Suicide Data Report 2005–2016. Office of Mental Health and Suicide Prevention, U.S. Department of Veterans Affairs;2018.

37.Leavitt RA EA, Sheats K, Petrosky E, Ivey-Stephenson A, Fowler KA. . Suicides Among American Indian/Alaska Natives — National Violent Death Reporting System, 18 States, 2003–2014. MMWR Morb Mortal Wkly Rep. 2018;67:237–242.

38.Fang F, Keating NL, Mucci LA, et al. Immediate risk of suicide and cardiovascular death after a prostate cancer diagnosis: cohort study in the United States. J Natl Cancer Inst. 2010;102(5):307-314.

39.Turaga KK, Malafa MP, Jacobsen PB, Schell MJ, Sarr MG. Suicide in patients with pancreatic cancer. Cancer. 2011;117(3):642-647.

40.Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer Statistics, 2021. CA: A Cancer Journal for Clinicians. 2021;71(1):7-33.

41.Stanley IH, Hom MA, Rogers ML, Hagan CR, Joiner TE. Understanding suicide among older adults: a review of psychological and sociological theories of suicide. Aging & Mental Health. 2016;20(2):113-122.

42.Garnett MF CS, Stone DM. Suicide mortality in the United States, 2000–2020. NCHS Data Brief, no 433. . Hyattsville, MD: National Center for Health Statistics. 2022.

43.Sherman RL, Williamson L, Andrews P, Kahn A. Primary Payer at DX: Issues with Collection and Assessment of Data Quality. J Registry Manag. 2016;43(2):99-100.

44.American Cancer Society. Cancer Facts & Figures 2022. Atlanta: American Cancer Society; 2022. https://www.cancer.org/research/cancer-facts-statistics/all-cancer-facts-figures/cancer-facts-figures-2022.html. Accessed November 10, 2022.

45.Wissing MD, Greenwald ZR, Franco EL. Improving the reporting of cancer-specific mortality and survival in research using cancer registry data. Cancer Epidemiol. 2019;59:232-235.

Tables and Figures

Table 1. Characteristics of Study Cohort and Standardized Mortality Ratios for Suicide

Table 2. Risk of suicide associated with individuals’ characteristics among cancer cohort, 2000-2016

Figure 1. Standardized Mortality Ratios by State, 2000-2016

Notes: Standardized Mortality Ratios standardized by age group only because of the small population in certain states.

Figure 2. SMRs by Year of Death, Age at Death, and Time Since Diagnosis, 2000-2016

By Year of Death

By Age at Death

By Time Since Diagnosis

Notes:SMR stands for standard mortality ratio, and is standardized by age at death, sex, and race/ethnicity. SMR for year at death was also standardized by year.

Table 1. Characteristics of Study Cohort and Standardized Mortality Ratio for Suicide

Individuals with cancer, No. (%)

SMRd (95% CI)

Total Cancer (N=16,771,397)

Total Death (n=7,972,782)

Total Suicide Death (n=20,792)

Overall

16771397 (100.0)

7972782 (100.0)

20792 (100.0)

1.26 (1.24-1.28)

Year of Death

2000

149464 (0.9)

149464 (1.9)

247 (1.2)

1.67 (1.47-1.88)

2001

268362 (1.6)

268362 (3.4)

457 (2.2)

1.64 (1.49-1.79)

2002

326604 (1.9)

326604 (4.1)

600 (2.9)

1.51 (1.39-1.63)

2003

369459 (2.2)

369459 (4.6)

715 (3.4)

1.48 (1.37-1.59)

2004

395881 (2.4)

395881 (5.0)

792 (3.8)

1.37 (1.28-1.47)

2005

428200 (2.6)

428200 (5.4)

883 (4.2)

1.31 (1.22-1.39)

2006

450335 (2.7)

450335 (5.6)

946 (4.5)

1.24 (1.16-1.32)

2007

470632 (2.8)

470632 (5.9)

1144 (5.5)

1.31 (1.24-1.39)

2008

494666 (2.9)

494666 (6.2)

1174 (5.6)

1.18 (1.11-1.25)

2009

511953 (3.1)

511953 (6.4)

1397 (6.7)

1.29 (1.22-1.35)

2010

529337 (3.2)

529337 (6.6)

1367 (6.6)

1.15 (1.09-1.21)

2011

547771 (3.3)

547771 (6.9)

1562 (7.5)

1.21 (1.15-1.26)

2012

566240 (3.4)

566240 (7.1)

1585 (7.6)

1.13 (1.08-1.19)

2013

587680 (3.5)

587680 (7.4)

1809 (8.7)

1.19 (1.13-1.24)

2014

603425 (3.6)

603425 (7.6)

1927 (9.3)

1.17 (1.12-1.23)

2015

629898 (3.8)

629898 (7.9)

2093 (10.1)

1.21 (1.16-1.27)

2016

642875 (3.8)

642875 (8.1)

2094 (10.1)

1.16 (1.11-1.21)

Alive

8798615 (52.5)

Attained age at death

00-24 years

41651 (0.2)

41651 (0.5)

106 (0.5)

1.05 (0.85-1.25)

25-39 years

109513 (0.7)

199731 (2.5)

557 (2.7)

1.06 (0.85-1.25)

40-49 years

342521 (2.0)

342521 (4.3)

1302 (6.3)

1.09 (1.03-1.15)

50-54 years

405283 (2.4)

405283 (5.1)

1411 (6.8)

1.13 (1.07-1.19)

55-59 years

607509 (3.6)

607509 (7.6)

1989 (9.6)

1.17 (1.12-1.22)

60-64 years

791876 (4.7)

791876 (9.9)

2461 (11.8)

1.31 (1.26-1.36)

65-69 years

944816 (5.6)

944816 (11.9)

2866 (13.8)

1.44 (1.39-1.50)

70-74 years

1058651 (6.3)

1058651 (13.3)

2926 (14.1)

1.38 (1.33-1.43)

75-79 years

1136998 (6.8)

1136998 (14.3)

2965 (14.3)

1.39 (1.34-1.44)

80-84 years

1103302 (6.6)

1103302 (13.8)

2345 (11.3)

1.26 (1.21-1.31)

85+ years

1430662 (8.5)

1430662 (17.9)

1864 (9.0)

1.07 (1.03-1.12)

Alive

8798615 (52.5)

Sex

Male

8645631 (51.5)

4329718 (54.3)

17572 (84.5)

1.29 (1.27-1.30)

Female

8125766 (48.5)

3643064 (45.7)

3220 (15.5)

1.14 (1.10-1.18)

Race/Ethnicity

Hispanic

1255786 (7.5)

505830 (6.3)

785 (3.8)

1.48 (1.38-1.58)

Non-Hispanic AIANAPIa

588206 (3.5)

245128 (3.1)

475 (2.3)

1.79 (1.63-1.95)

Non-Hispanic Black

1778132 (10.6)

896830 (11.2)

632 (3.0)

1.15 (1.06-1.24)

Non-Hispanic White

13149273 (78.4)

6324994 (79.3)

18900 (90.9)

1.25 (1.23-1.27)

Region

Northeast

2212373 (13.2)

1044950 (13.1)

2014 (9.7)

1.25 (1.20-1.30)

Midwest

3128127 (18.7)

1507915 (18.9)

3529 (17.0)

1.25 (1.21-1.30)

South

7279094 (43.4)

3540101 (44.4)

9222 (44.4)

1.21 (1.18-1.23)

West

4151803 (24.8)

1879816 (23.6)

6027 (29.0)

1.18 (1.15-1.21)

Insurance typeb

Private

4914979 (29.3)

1523193 (19.1)

5450 (26.2)

1.08 (1.05-1.11)

Medicaid

780220 (4.7)

407944 (5.1)

851 (4.1)

1.72 (1.61-1.84)

Medicare ≤64 years old

540921 (3.2)

261066 (3.3)

809 (3.9)

1.94 (1.80-2.07)

Medicare 65+ years old

4690270 (28.0)

2742090 (34.4)

5790 (27.8)

1.42 (1.38-1.46)

VA or Indian/PHS

233349 (1.4)

120746 (1.5)

511 (2.5)

1.89 (1.72-2.05)

Uninsured

463664 (2.8)

228741 (2.9)

595 (2.9)

1.66 (1.53-1.80)

Unknown/Missing

5147994 (30.7)

2689002 (33.7)

6786 (32.6)

1.17 (1.14-1.20)

County-level Ruralityc

Metro

9800370 (80.8)

3935950 (79.2)

9800 (77.9)

1.29 (1.27-1.32)

Non-Metro Urban

1823072 (15.0)

824569 (16.6)

2233 (17.8)

1.34 (1.28-1.40)

Non-Metro Rural

243729 (2.0)

113206 (2.3)

321 (2.6)

1.39 (1.24-1.54)

Unknown/Missing

265043 (2.2)

94081 (1.9)

219 (1.7)

County-level povertyc

< 5.0%

601585 (5.0)

208402 (4.2)

493 (3.9)

1.24 (1.13-1.35)

5.0% - 9.99%

3701312 (30.5)

1430150 (28.8)

3704 (29.5)

1.42 (1.38-1.47)

10.0% - 19.99%

6684792 (55.1)

2828252 (56.9)

7306 (58.1)

1.35 (1.32-1.38)

20.0% +

879482 (7.2)

406921 (8.2)

851 (6.8)

1.35 (1.26-1.44)

Unknown

265043 (2.2)

94081 (1.9)

219 (1.7)

Stage

In situ/Local

7927770 (47.3)

2307856 (28.9)

10787 (51.9)

1.04 (1.02-1.06)

Regional

3605880 (21.5)

1734455 (21.8)

4397 (21.1)

1.46 (1.41-1.50)

Distant

3833893 (22.9)

2932044 (36.8)

3794 (18.2)

1.9 (1.83-1.96)

NA/Unknown/Missing

1403854 (8.4)

998427 (12.5)

1814 (8.7)

1.62 (1.54-1.69)

Multiple Primary

One

14872585 (88.7)

7074007 (88.7)

18631 (89.6)

1.35 (1.34-1.37)

Two or More

1898812 (11.3)

898775 (11.3)

2161 (10.4)

0.79 (0.76-0.82)

Time since Diagnosis at Death

0-5 months

2633423 (33.0)

4188 (20.1)

7.19 (6.97-7.41)

6-11 months

1080065 (13.5)

1987 (9.6)

5.6 (5.35-5.84)

12-23 months

1204531 (15.1)

2874 (13.8)

4.18 (4.03-4.33)

24-35 months

689630 (8.6)

2145 (10.3)

3.09 (2.96-3.22)

3-4 years

852473 (10.7)

3083 (14.8)

1.98 (1.91-2.05)

5-9 years

1086184 (13.6)

4688 (22.5)

0.94 (0.91-0.97)

≥10 years

426476 (5.3)

1827 (8.8)

0.24 (0.23-0.25)

Primary Site

Female breaste

2500492 (14.9)

667170 (8.4)

1187 (5.7)

1.03 (0.97-1.09)

Prostate

2559281 (15.3)

730459 (9.2)

5083 (24.4)

0.91 (0.89-0.94)

Colon and rectum

1611161 (9.6)

851611 (10.7)

1992 (9.6)

1.25 (1.20-1.31)

Lung cancer and bronchus

2213830 (13.2)

1852179 (23.2)

2462 (11.8)

2.34 (2.25-2.44)

Female Genital

1008235 (6.0)

394540 (4.9)

431 (2.1)

1.11 (1.01-1.22)

Oral cavity& pharynx

409033 (2.4)

195718 (2.5)

1157 (5.6)

2.42 (2.28-2.56)

Kidney and renal pelvis

549636 (3.3)

221023 (2.8)

673 (3.2)

1.14 (1.05-1.23)

Melanoma

674205 (4.0)

162744 (2.0)

1017 (4.9)

1.03 (0.97-1.09)

Non-Hodgkin lymphoma

693316 (4.1)

314827 (3.9)

826 (4.0)

1.17 (1.09-1.25)

Thyroid

439331 (2.6)

38511 (0.5)

306 (1.5)

0.82 (0.73-0.91)

Pancreas

425278 (2.5)

375122 (4.7)

354 (1.7)

2.5 (2.24-2.76)

Liver and intrahepatic bile duct

268348 (1.6)

213563 (2.7)

189 (0.9)

1.59 (1.36-1.82)

Bladder

700303 (4.2)

329822 (4.1)

1396 (6.7)

1.27 (1.21-1.34)

Esophagus

170794 (1.0)

139525 (1.8)

365 (1.8)

3.15 (2.82-3.47)

Leukemia

472297 (2.8)

246433 (3.1)

476 (2.3)

1.17 (1.06-1.27)

Brain and other nervous system

253327 (1.5)

169642 (2.1)

251 (1.2)

1.67 (1.47-1.88)

Stomach

239489 (1.4)

177313 (2.2)

317 (1.5)

2.32 (2.07-2.58)

Other

1583041 (9.4)

892580 (11.2)

2310 (11.1)

1.62 (1.55-1.68)

Notes:

a Non-Hispanic AIANAPI stands for Non-Hispanic American Indian/Alaska Native, and Non-Hispanic Asian or Pacific Islander.

b Private insurance includes private fee-for-service and managed care, TRICARE, and military insurance; Medicaid, no Medicare includes traditional Medicaid and managed care Medicaid, and other not specified insurance; Medicare includes traditional fee-for-service insurance, managed care insurance, with supplemental coverage, and dual eligible insurance; VA or Indian/PHS insurance stands for Veteran Affairs, Indian/Public Health Service Insurance.

cCounty-level data only available in 2005 and after. N= 11,867,171. County-level rurality were coded with Rural-Urban Continuum Codes developed by the United States Department of Agriculture (https://seer.cancer.gov/seerstat/variables/countyattribs/ruralurban.html). County-level poverty were coded with percent of persons below poverty data from the American Community Survey 2007-2011 (https://seer.cancer.gov/seerstat/variables/countyattribs/static.html#07-11 ).

d SMR stands for standard mortality ratio, and is standardized by attained age, sex, and race/ethnicity. SMRs for year of death, region, county-level rurality, and county-level poverty subcategories were also standardized by year, region, county-level rurality, and county-level poverty respectively.

e Male breast cancer was grouped into “Other” category.

Table 2. Risk of Suicide Associated with Individuals’ Characteristics among Cancer Cohort, 2000-2016

Univariable Regression (n=16,771,397)

Multivariable Regression (n=16,771,397)

No time-interaction characteristics

HR (95% CI)

P-value

HR (95% CI)

P-value

Sex

Female

Ref

Male

5.06 (4.87-5.25)

<.001

5.52 (5.21-5.85)

<.001

Race/Ethnicity

Hispanic

0.46 (0.43-0.50)

<.001

0.43 (0.40-0.47)

<.001

Non-Hispanic AIANAPIa

0.60 (0.55-0.66)

<.001

0.60 (0.54-0.66)

<.001

Non-Hispanic Black

0.25 (0.23-0.27)

<.001

0.25 (0.23-0.28)

<.001

Non-Hispanic White

Ref

Insurance typeb

Private

Ref

Medicaid

0.99 (0.92-1.06)

.78

1.23 (1.14-1.33)

<.001

Medicare ≤64 years old

1.38 (1.29-1.49)

<.001

1.48 (1.37-1.60)

<.001

Medicare 65+ years old

1.12 (1.08-1.16)

<.001

1.14 (1.09-1.20)

<.001

VA or Indian/PHS

1.99 (1.82-2.18)

<.001

1.34 (1.22-1.47)

<.001

Uninsured

1.14 (1.04-1.24)

.003

1.27 (1.16-1.38)

<.001

Unknown/Missing

1.07 (1.03-1.11)

<.001

1.06 (1.02-1.11)

.004

County-level Ruralityc

Metro

Ref

NonMetro Urban

1.22 (1.16-1.27)

<.001

1.09 (1.03-1.14)

.001

NonMetro Rural

1.31 (1.17-1.46)

<.001

1.14 (1.02-1.28)

.026

County-level Povertyc

< 5.0%

Ref

5.0% - 9.99%

1.22 (1.11-1.34)

<.001

1.07 (0.97-1.18)

.17

10.0% - 19.99%

1.33 (1.21-1.45)

<.001

1.13 (1.02-1.25)

.017

20.0% +

1.17 (1.05-1.31)

.005

1.12 (0.99-1.27)

.073

Time-interacted characteristics

First 2 years

After 2 years

First 2 years

After 2 years

HR (95% CI)

P-value

HR (95% CI)

P-value

HR (95% CI)

P-value

HR (95% CI)

P-value

Stage

In situ/Local

Ref

Regional

1.41 (1.34-1.48)

<.001

0.68 (0.65-0.71)

<.001

1.29 (1.22-1.36)

<.001

0.85 (0.81-0.90)

<.001

Distant

1.63 (1.55-1.71)

<.001

0.33 (0.31-0.36)

<.001

1.32 (1.25-1.41)

<.001

0.41 (0.38-0.45)

<.001

NA/Unknown/Missing

1.78 (1.68-1.89)

<.001

0.54 (0.50-0.59)

<.001

1.39 (1.30-1.49)

<.001

0.66 (0.61-0.73)

<.001

Primary Site

Colon and rectum

Ref

Female breaste

0.28 (0.25-0.32)

<.001

0.47 (0.42-0.51)

<.001

1.06 (0.93-1.20)

.40

1.24 (1.11-1.38)

<.001

Prostate

0.95 (0.87-1.03)

.20

2.05 (1.91-2.19)

<.001

0.64 (0.59-0.70)

<.001

1.09 (1.01-1.17)

.02

Lung cancer and bronchus

1.54 (1.43-1.66)

<.001

0.43 (0.38-0.48)

<.001

1.31 (1.21-1.42)

<.001

0.53 (0.47-0.60)

<.001

Uterine corpus

0.32 (0.27-0.37)

<.001

0.39 (0.34-0.45)

<.001

1.18 (1.00-1.39)

.05

1.17 (1.01-1.36)

.04

Oral cavity& pharynx

2.65 (2.40-2.93)

<.001

2.19 (1.98-2.43)

<.001

2.07 (1.87-2.29)

<.001

1.51 (1.36-1.68)

<.001

Kidney and renal pelvis

1.00 (0.89-1.13)

.98

1.10 (0.97-1.24)

.14

0.99 (0.87-1.12)

.84

0.86 (0.76-0.97)

.01

Melanoma

0.86 (0.76-0.97)

.02

1.62 (1.47-1.78)

<.001

0.84 (0.74-0.95)

.006

1.05 (0.95-1.16)

.33

Non-Hodgkin lymphoma

0.97 (0.87-1.09)

.65

1.00 (0.90-1.12)

.95

0.93 (0.82-1.04)

.19

1.12 (0.996-1.25)

.06

Thyroid

0.43 (0.35-0.52)

<.001

0.79 (0.68-0.92)

.002

0.90 (0.74-1.10)

.31

0.91 (0.78-1.06)

.21

Pancreas

1.35 (1.20-1.52)

<.001

0.17 (0.11-0.28)

<.001

1.27 (1.12-1.43)

<.001

0.25 (0.15-0.40)

<.001

Liver & intrahepatic bile duct

0.88 (0.74-1.05)

.15

0.38 (0.27-0.54)

<.001

0.78 (0.66-0.93)

.006

0.32 (0.23-0.46)

<.001

Bladder

1.41 (1.27-1.56)

<.001

1.81 (1.65-1.98)

<.001

1.09 (0.98-1.21)

.13

1.13 (1.03-1.24)

.009

Esophagus

3.17 (2.80-3.59)

<.001

0.66 (0.51-0.86)

.002

2.13 (1.88-2.41)

<.001

0.55 (0.42-0.71)

<.001

Leukemia

0.83 (0.72-0.95)

.008

0.86 (0.74-1.00)

.05

0.73 (0.63-0.85)

<.001

1.49 (1.25-1.77)

<.001

Brain and other nervous system

1.12 (0.96-1.31)

.15

0.58 (0.46-0.73)

<.001

1.38 (1.18-1.61)

<.001

0.41 (0.33-0.52)

<.001

Stomach

1.85 (1.62-2.11)

<.001

0.52 (0.39-0.68)

<.001

1.70 (1.48-1.94)

<.001

0.58 (0.44-0.76)

<.001

Other

1.40 (1.29-1.52)

<.001

1.07 (0.98-1.17)

.14

1.34 (1.23-1.46)

<.001

0.98 (0.89-1.07)

.58

Notes:Cox proportional hazard models controlled for other cause of death as competing risks. Time interaction terms were included for factors where proportional hazard assumption was not met. Models also adjust for age at diagnosis, NAACCR Registry, and year of diagnosis, of which results are shown in eFigure 1.

a Non-Hispanic AIANAPI stands for Non-Hispanic American Indian/Alaska Native, and Non-Hispanic Asian or Pacific Islander.

cEstimates for county-level rurality and poverty came from separate models among individuals diagnosed with cancer in 2005-2016 (n=11,867,171), because county-level data were only available in 2005 and after. Estimates for the full model among this sub-cohort are provided in Appendix S3. County-level rurality were coded with Rural-Urban Continuum Codes developed by the United States Department of Agriculture (https://seer.cancer.gov/seerstat/variables/countyattribs/ruralurban.html). County-level poverty were coded with percent of persons below poverty data from the American Community Survey 2007-2011 (https://seer.cancer.gov/seerstat/variables/countyattribs/static.html#07-11).

e Male breast cancer was grouped into “Other” category.