Analyzing excess mortality from cancer among individuals with mental illness


  • The results were presented in part at the American Society of Clinical Oncology annual meeting in Chicago, Illinois, July 2012.



The objective was to compare patterns of site-specific cancer mortality in a population of individuals with and without mental illness.


This was a cross-sectional, population-based study using a linked data set comprised of death certificate data for the state of Ohio for the years 2004-2007 and data from the publicly funded mental health system in Ohio. Decedents with mental illness were those identified concomitantly in both data sets. We used age-adjusted standardized mortality ratios (SMRs) in race- and sex-specific person-year strata to estimate excess deaths for each of the anatomic cancer sites.


Overall, there was excess mortality from cancer associated with having mental illness in all the race/sex strata: SMR, 2.16 (95% CI, 1.85-2.50) for black men; 2.63 (2.31-2.98) for black women; 3.89 (3.61-4.19) for nonblack men; and 3.34 (3.13-3.57) for nonblack women. In all the race/sex strata except for black women, the highest SMR was observed for laryngeal cancer, 3.94 (1.45-8.75) in black men and 6.51 (3.86-10.35) and 6.87 (3.01-13.60) in nonblack men and women, respectively. The next highest SMRs were noted for hepatobiliary cancer and cancer of the urinary tract in all race/sex strata, except for black men.


Compared with the general population in Ohio, individuals with mental illness experienced excess mortality from most cancers, possibly explained by a higher prevalence of smoking, substance abuse, and chronic hepatitis B or C infections in individuals with mental illness. Excess mortality could also reflect late-stage diagnosis and receipt of inadequate treatment. Cancer 2013;119:2469-2476. © 2013 American Cancer Society.


Mental illness (MI) is an important public health problem in the United States and globally. In the United States alone, 1 in 4 adults (26.2%) suffers from a diagnosable MI in a given year.[1] MI is associated with poor physical health.[2] Patients with severe and persistent MI have many unattended medical needs attributed in part to fewer contacts with primary health care providers.[3] Reduced life expectancy and high rates of premature death have been documented among individuals with serious MI, with these individuals dying as much as 15-25 years younger than the general population.[4, 5] As such, patients with MI constitute a vulnerable population, given their economic disadvantage,[6] difficulty obtaining medical insurance,[7] associated cognitive limitations, and lack of knowledge about their health and the health care system. At the same time, nonpsychiatric physicians are typically poorly prepared to provide care to this complex population.[8] Together, these factors may negatively affect their access to and benefit from medical care,[9] which may impact both the risk for specific cancers and the process of screening, diagnosing, and treating cancer in this population.

Relative to cancer outcomes, a number of studies have documented cancer-related disparities associated with MI, including being diagnosed at an unknown stage of cancer,[10] advanced-stage cancer,[11] receipt of cancer treatment,[10, 12, 13] and survival.[10, 12] Not surprisingly, therefore, excess mortality from cancer has been reported in MI patients compared with their non-MI counterparts,[14-16] although results have also varied across studies, cancer sites, and demographic and psychiatric subgroups of the population.[17, 18]

In this study we sought to detail the broad category of deaths from malignant neoplasms into specific anatomic sites across age, sex, and race strata and to compare the death rates of individuals with mental illness with those of the general Ohio population, using the person-year approach to account for the length of follow-up period. Understanding the extent to which specific cancer diagnoses differ among individuals with and without MI holds both clinical and public health relevance. In particular, the findings will help to develop targeted preventive strategies and to improve clinical care for this vulnerable group.


Data Sources

This was a cross-sectional, population-based study using a linked database composed of death certificate data for the state of Ohio for the years 2004-2007, the Multi-Agency Community Services Information System (MACSIS), and the Patient Care System (PCS).

MACSIS is an automated payment and management information system for the publicly funded outpatient mental health services. It includes patient demographics, billing charges, service dates, and diagnostic codes based on the Diagnostic and Statistical Manual of Mental Disorders, 4th edition text revision (DSM-IV TR).[19]

The PCS has data from regional psychiatric hospitals across the state. It includes dates of admission and discharge, patient demographics, patient identifiers, and a DSM-IV-based diagnosis.

The Ohio death certificate file has data for all Ohio decedents. The underlying causes of death are coded based on the International Classification of Diseases, Clinical Modifications, 10th edition (ICD-10 CM).

Linkage of the 3 data sets was performed using a deterministic, multi-step process based on personal identifiers, including Social Security number (SSN), name, date of birth, and sex, consistent with other studies.[20-22] Successful identification of a decedent in the death certificates and the MACSIS and/or the PCS files implied that the decedent had been a recipient of Ohio's publicly funded mental health system, therefore an individual with mental illness.

All sensitive identifiers, including SSN, name, and date of birth, were removed from the version of the analytical file that was shared with investigators outside the state mental health agency.

This study was considered exempt by the institutional review boards of the Ohio Department of Health and Case Western Reserve University.

Study Variables

The cause of death was determined using ICD-10 CM codes in the death certificate records, and deaths due to cancer (C00-C97) were parsed out (Table 1). Cancers were then categorized based on the anatomic cancer site by using the definitions provided in the SEER Training Modules of the National Cancer Institute[23] and the National Center for Health Statistics (NCHS) of the US Department of Health & Human Services classification.[24]

Table 1. ICD-10 Codes for Specific Cancer Sites
 Cancer Anatomic SiteICD-10 CM
  1. a

    Secondary cancer means that the anatomic site for the primary cancer was not listed in the death certificate, indicating that the patient died of cancer metastasis.

1.Upper gastrointestinal tractC00-C16
2.Colorectal cancerC18-21
3.Hepatobiliary cancerC22
4.Pancreatic cancerC25
5.Lung tissue cancerC33-C34
6.Cancer of the larynxC32
7.Skin cancerC43-C44
8.Breast cancerC50
9.Cervical and uterine cancerC53-C55
10.Ovarian cancerC56
11.Cancer of the prostateC61
12.Urinary bladder cancerC67
13.Kidney and renal pelvis cancerC64-C65
14.Brain and other central nervous system (CNS) tissueC70-C72
15.Cancer of the lymphoid and hemopoiesis tissuesC81-C96
16.Secondarya and ill-defined cancersC76-C80
17.All other cancersC17, C23-C24, C26-C31, C37-C41, C44-C49, C51-C52, C57-C60, C62-C63, C66, C68-C69, C73-C75, and C97

To avoid errors that might have arisen at data entry (eg, prostate cancer in a female decedent), we restricted the analysis for breast, ovarian, and cervical and uterine cancer to females, whereas prostate cancer analysis was done only for males. We recognize that breast cancer rarely occurs in men.[25]

The main variable of interest was the absence or presence of mental illness (0/1), which we coded as 1 when a decedent was identified in either MACSIS or the PCS files and in the death certificate file.

Other variables of interest for this study were decedent demographics, that is, year of birth, race, and sex, all retrieved from the death certificate data set. Age was classified into 5 categories as follows: 1-14, 15-24, 25-44, 45-64, and 65+ years. Racial groups were dichotomized into blacks versus nonblacks. Blacks comprised African Americans and persons of African descent, whereas nonblacks included whites and other racial minorities whose numbers were too small to maintain as separate categories for statistical analyses.

Study Population

Our study population comprised all Ohio residents 1 year of age or older who died of cancer in the years 2004-2007 (n = 101,689). Of those, 4284 were also identified through the above-referenced MACSIS and PCS files. However, 2303 of the 4284 never had claims in the MACSIS or the PCS despite being enrolled in the system, and possibly because of deficiencies in the system, data on their enrollment as well as mental health diagnosis were largely absent. Consequently, we opted to group these 2303 decedents with the general population. Thus, our study population included 101,689 Ohio decedents with cancer, 1981 of whom had documented mental illness, and we termed this group the mental illness decedents.


Because persons seeking care from the publicly funded mental health system were not followed all through the 4-year study period, we used the person-year approach in our analysis. The total number of person-years corresponding to the 604,771 individuals with mental illness was 544,760. Based on the 11.3 million individuals residing in Ohio (our reference group), the number of person-years was 45.4 million.

For the descriptive analysis, we obtained the proportion of decedents per category among those who had mental illness (MI) and among the entire Ohio population. All descriptive analyses were conducted using SAS version 9.2 (Cary, NC).

In addition to the descriptive analysis, using the indirect standardization method, we obtained the age-, race- and sex-specific standardized mortality ratios (SMRs) by dividing the observed by the expected deaths. The stratum-specific expected number of deaths was obtained by multiplying the number of person-years in a stratum by the crude mortality rate in the Ohio population (from the 2000 US Census) in that stratum. Confidence intervals (CIs) for the SMRs were calculated using version 4.11.19 of the online SMR calculator by the Emory University School of Public Health.[26]


Table 2 shows the distribution of lives (in person-years) for individuals with MI and the total Ohio population in age, race, and sex strata. Compared with the total Ohio population, we noted a markedly greater representation of younger individuals among those with MI.

Table 2. Distribution of Total Ohio Population and Individuals Enrolled in Publicly Funded Mental Health System During the Study Period (in person-years)
AgeBlack Men, n (% of Total)Black Women, n (% of Total)Nonblack Men, n (% of Total)Nonblack Women, n (% of Total)
categoryMIaTotal OhioMIaTotal OhioMIaTotal OhioMIaTotal Ohio
  1. a

    MI, individuals with mental illness enrolled in the publicly funded mental health system.

1-1423,890 (36.49)713,772 (28.97)12,176 (19.33)689,036 (25.14)57,797 (29.69)4,194,868 (21.40)32,739 (14.77)3,998,672 (19.39)
15-2412,046 (18.40)391,436 (15.89)11,264 (17.89)401,112 (14.63)37,087 (19.06)2,727,660 (13.93)38,163 (17.21)2,662,976 (12.91)
25-4414,699 (22.45)720,132 (29.22)19,581 (31.09)816,168 (29.77)53,944 (27.72)5,868,764 (29.97)74,666 (33.68)5,895,776 (28.59)
45-6413,964 (21.33)446,148 (18.10)17,684 (28.09)535,524 (19.54)42,166 (21.67)4,552,032 (23.24)65,787 (29.67)4,767,456 (23.12)
65+867 (1.32)192,580 (7.82)2,260 (3.59)299,320 (10.91)3,631 (1.87)2,241,656 (11.45)10,350 (4.67)3,297,472 (15.99)
Total65,466 (100.0)2,464,068 (100.0)62,964 (100.0)2,741,160 (100.0)194,624 (100.0)19,584,980 (100.0)221,705 (100.0)20,622,352 (100.0)

Among those who had cancer (n = 101,689), 1981 (1.95%) also had an MI diagnosis (Table 3). The mean age at death among those with MI and those without MI was 60.34 years (standard deviation, 13.8 years), and 71.51 years (standard deviation, 13.5 years), respectively.

Table 3. Distribution of Decedents by Demographics and by Cancer Causes of Death Among Those Identified With Mental Illness and All Other Ohioans
Demographics and Cause of Death CategoriesDecedents with mental illness (n = 1981), %Total Ohio (N = 101,689), %
Age (y)  
Cancer causes of death  
Upper gastrointestinal tract5.605.71
Lung tissue34.0229.76
Cervical and uterine4.414.03
Urinary bladder1.822.53
Kidney and renal pelvis2.272.27
Brain and other CNS2.522.17
Lymphoid tissue7.529.74
Secondary or ill defined6.716.47
All other cancers3.434.46

More than half the MI decedents were in the 45-64 age group, whereas the majority (71.1%) of decedents in the general population was 65 years or older. Although sex distribution was about equal in the general population, there were more females (56.0%) among the MI individuals.

Blacks represented 20.29% and 10.09% of those who had MI and the general population, respectively.

Among those with MI and the general population, the 3 leading causes of deaths were: 1) cancer of the lungs (34.0% for MI and 29.8% for the general population), 2) breast cancer (17.4% for MI and 15.4% for the general population), and 3) colorectal cancer (8.43% for MI and 9.74% for the general population). These descriptive statistics are detailed in Table 3.

All-Cancer Mortality

Overall there was excess mortality from cancer associated with having MI across all of the race/sex strata. The highest age-adjusted SMR was observed in nonblack men SMR (3.89; 95% CI, 3.61-4.19), followed by nonblack women (3.34 [3.13-3.57]), black women (2.63 [2.31-2.98]), and black men (2.16 [1.85-2.50]].

Site-Specific Cancer Mortality

Figure 1A-D shows the cancer-specific SMRs within race/sex strata. We note statistically significant higher SMRs for every anatomic cancer site in nonblack men and women and for most cancer sites in black men and women. Failure to reach statistical significance for some cancer sites (cancer of the upper gastrointestinal tract, hepatobiliary cancer, and prostate cancer in Black men and ovarian cancer in black women) may have been due to small numbers.

Figure 1.

Age-adjusted standardized mortality rates (SMRs) by race/sex categories. A: SMRs for Black Men; B: SMRs for Black Women; C: SMRs for nonblack Men; D: SMRs for nonblack women. The number in parentheses indicates the observed number of observed deaths in that race/sex stratum and for that cancer site. White bars are for SMRs that were not statistically significant at P < .05. All remaining SMRs were significant at P < .05.

Except among Black women, the highest SMR with statistical significance was observed for laryngeal cancer (SMR, 3.94 [1.45-8.75] in black men, 6.51 [3.86-10.35] in nonblack men, and 6.87 [3.01-13.60] in nonblack women). In Black women, the highest SMR was noted for the kidney and renal pelvis (8.54 [4.49-14.85]). SMRs for cancers of the urinary tract were also considerably elevated in nonblack women and men.

More than 5-fold excess mortality from hepatobiliary cancer was observed, especially in nonblack men and black women (SMR, 5.39 [3.75-7.52] and 5.26 [2.56-9.65], respectively).

Finally, we observed SMRs elevated 3-5 times for cancer of the lung, trachea, and bronchus (SMR, 4.33 [3.82-4.88] in nonblack men, 4.22 [3.75-4.72] in nonblack women, and 2.99 [2.33-3.78] in black women).


To our knowledge, this is the first study to document excess mortality from cancer associated with having mental illness by differentiating the broad category of death from malignant neoplasms into specific sites of malignancies across age, sex, and racial strata. The excess cancer-related mortality among individuals with MI demonstrated striking racial variation. We believe these findings may be partially explained by known epidemiological risk factors for these specific malignancies. Beyond addressing these factors, however, these findings carry several key health care messages for this population and their health care providers, especially given the low likelihood of persons with MI to be diagnosed with early-stage disease and/or to receive guideline care.[10, 27] Indeed, there seems to be insufficient awareness of medical comorbidities, leading to underdiagnosis of such conditions[28] because of the underreporting of physical problems by the patient, the societal stigma, and/or physicians' perception of physical complaints as psychosomatic symptoms, thus overlooking coexisting medical conditions in patients with MI.[8]

The observed SMRs for laryngeal cancer are biologically plausible, given the synergistic impact of the 2 primary risk factors, alcohol and tobacco use,[29] both of which are highly prevalent among persons with MI, and the poorer efficacy of smoking cessation and alcohol treatment programs in this population.[30, 31]

Compared with the general population, persons with MI have a higher prevalence of hepatitis B, hepatitis C, and alcohol and other substance abuse disorders,[32] all of which are risk factors for hepatobiliary cancer. This may explain the excess mortality from hepatobiliary cancer observed in all the race/sex strata, albeit at borderline statistical significance in black men, because of small numbers. What remains unclear is the origin for the differences in excess hepatobiliary cancer-related mortality across the race/sex strata. Potential explanations include race and sex differences in the prevalence of risk factors, differences in synergism between these risk factors, and competing causes of death.

Risk factors for cancers of the larynx, lung, bronchus, and trachea primarily involve direct or passive exposure to tobacco smoke. Persons with chronic MI smoke 44% of all cigarettes, reflecting both a high prevalence of use and a heavy consumption among those who do smoke.[33] Thus, the observed excess mortality of respiratory tract cancers may be largely attributable to the tobacco use patterns of persons with serious MI.

Although competing causes of death may be a possible explanation, the finding that excess mortality from respiratory tract cancer was not observed at a similar magnitude across race/sex strata remains puzzling. In the general population, a case-control study that examined sex-race differences in the risk of lung cancer risk associated with cigarette smoking showed that for a given level of smoking, blacks were at higher risk than whites of developing lung cancer.[34] Studies also reported that although African Americans begin smoking later in life[35, 36] than whites, their rates of cessation are lower,[35-37] and they use brands with higher tar yields.[38, 39] Given this evidence, we would have expected a greater magnitude of excess mortality to be observed among blacks.

Excess mortality from urinary tract cancers associated with having mental illness was more accentuated in women than in men. Similar that with respiratory tract malignancies, the most important risk factor associated with both kidney cancer and bladder cancer is tobacco use.[40] Additional risk factors for kidney cancer include obesity, hypertension, chronic hepatitis C infection, and sickle cell disease.[41, 42] Thus, the excess mortality associated with kidney cancer in black women with mental illness might be explained by a higher prevalence of tobacco use, obesity, hypertension, or chronic hepatitis C compared with the general population.

However, results from this study should be interpreted in light of the following limitations. First, we looked at individuals with MI in general; patterns of excess mortality may in fact differ between specific psychiatric diagnoses.[43] Second, we acknowledge limitations with the accuracy of the underlying causes of death on death certificates.[44] Third, we recall that individuals who were enrolled in Ohio's publicly funded mental health system but for whom we have no claims with a documented MI diagnosis were accounted for in the general population, rather than among those with mental illness. This approach was deemed appropriate because it was not possible to obtain person-year data on this subgroup of the population because of how the data were captured in the state mental health agency's information system. However, regardless of whether they actually received mental health services that were covered by the agency or through another party, this group of decedents in all likelihood indeed suffered from MI, as they were enrolled in the system. As a result, this grouping may have biased our results, at least to some extent, most likely leading to attenuated SMRs reported for those with mental illness. From an analytical standpoint, we note that compared with our previous study,[45] which accounted for individuals enrolled in the publicly funded mental health system during the 4-year period for any length of time, our use of person-years in the current study yielded considerably higher SMRs.

Finally, we note that this study was of a cross-sectional nature; therefore, we cannot infer a causal relationship between MI and mortality.


Among individuals who died of cancer, those with MI died an average of 10 years earlier than those without MI.

For many patients with MI, the psychiatrist is often the physician of first contact and may remain the only treating physician.[46] Our findings support the necessity for persons with MI to also have a primary care physician, whose focus is on preventive health care, including primary prevention and screening for early detection of disease, cancer risk assessment and screening, and health behavior assessment and counseling. Our findings also suggest a need for more intensive and coordinated efforts of medical, behavioral, and public health systems efforts in addressing 1) smoking cessation, 2) alcohol and substance abuse prevention and treatment, and 3) screening for and treatment of chronic hepatitis. Further investigation is necessary to understand and address the observed racial variation in cancer mortality and the excess mortality from urinary tract cancers among women with MI. Ameliorating these observed disparities in cancer site-specific mortality among persons with MI will likely require interventions along the entire cancer control continuum.


Drs. Siran M. Koroukian and Jackson Musuuza were supported by a grant from the Ohio Department of Mental Health. Dr. Siran M. Koroukian is also supported by the Clinical and Translational Science Collaborative of Cleveland, UL1TR000439, from the National Center for Advancing Translational Sciences (NCATS) component of the National Institutes of Health and NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.


The authors made no disclosures.