Analysis of Causes of Death for All Decedents in Ohio With and Without Mental Illness, 2004–2007

What is killing us? Are these causes of death primarily disease related, or are they acts of injury or violence? How does the presence of mental illness influence longevity and cause of death? The quality of health care, including both provider and patient factors, affects cause of death. Quality of health care may be assessed from multiple perspectives, including mortality. Individuals with undiagnosed or untreated mental illness have a higher risk of dying prematurely from unintentional and intentional violence (1–4). Everett and colleagues (5) recommended that public health planners gain better access to mortality information to identify and prioritize policy, resources, and health promotion activities for special populations.

A strong association between mental illness and reduced life expectancy has been documented, with multiple associated factors: social and economic disadvantages, certain demographic characteristics, and unhealthy lifestyle, including lack of exercise, poor nutrition or obesity, smoking, substance abuse, and poor access to routine medical care (2,6–9). Previously published studies in this area have been limited to mental illness population subgroups, including patients with severe mental illness using inpatient, residential, or case management services, and Medicaid enrollees (2,4,10). To address the sample limitations of previous studies, we linked all Ohio deaths (N=438,749) with deaths within the Ohio publicly funded system (N=30,219 deaths identified, including 14,506 with mental health service claims, out of 1 million individuals in the 2004–2007 Ohio Medicaid Eligibility List, within the population of 11.5 million Ohioans). We did not include in this analysis deaths of individuals identified within the Ohio Department of Mental Health (ODMH) databases who did not receive services through ODMH.

Even though risk factors are associated with premature death in the general population, these factors disproportionately affect persons with mental illness through mental illness symptoms and physiology or iatrogenic effects of treatment (6,11,12). A 2006 report of the National Association of State Mental Health Program Directors recommended setting severe mental illness morbidity as a public health disparities priority (12). In 2007, the Substance Abuse and Mental Health Services Administration set a goal “to reduce the 25-year mortality gap for mental health patients by ten years within ten years” (13).

In addition, studies indicate that persons with mental illness are 1.7 to 5.0 times more likely than the general population to die prematurely from chronic health conditions (2,6). This finding is particularly troubling when increased mortality is due to treatable medical conditions and preventable complications. Chronic conditions may be modified by consumer-level interventions, such as health education and training, screening and management of chronic health conditions, monitoring side effects of medications used to treat mental illness, and promoting wellness activities, such as improving the diet (5,6,14). They may also be modified by system-level interventions, including integration of mental health care and primary care, collaboration among treatment staff, and creation of medical homes (5,6,15,16). Roshanaie-Moghaddam and Katon (6) suggested that “new models of care may be necessary to improve general medical and psychological outcomes of patients.”

This retrospective study analyzed mortality in race and sex strata by cause of death among individuals with mental illness who received services through Ohio’s publicly funded mental health care system. Results reported here contribute to understanding excess mortality among decedents with mental illness receiving care in a publicly funded mental health system. Mortality findings can assist policy makers in integrated care planning and identify indicator tracking needed to develop disease and injury prevention strategies in private and public systems, including the Veterans Health Administration system.

Methods

Data sources

This was a population-based cross-sectional study linking 2004–2007 Ohio death certificate data to data from the ODMH Multi-Agency Community Services Information System (MACSIS) and the Patient Care System (PCS). MACSIS is an automated payment management information system for publicly funded outpatient mental health services, which includes patient identifiers, billing charges, service dates, and DSM-IV diagnosis codes. The PCS contains services data for individuals served by regional psychiatric hospitals. The Ohio death certificate files list each underlying cause of death, as coded in the International Classification of Diseases (ICD-10). This study, approved by ODMH, included only decedents and thus was deemed exempt by the institutional review boards of ODMH and Case Western Reserve University.

Linkage algorithm

With an approach described in previous studies (17–19), ODMH service files and death certificate data were linked by year with a deterministic, multistep algorithm based on decedent identifiers, including Social Security number (SSN), first and last names (truncated to the first six letters), and date of birth, as follows: step 1: SSN, last name, first name, and date of birth (month); step 2: SSN, last name, and date of birth (month); step 3: SSN, first name, and date of birth (month); and step 4: last name, first name, and date of birth (month and year).

Of all decedents identified successfully in both the death certificate files and the ODMH databases, 85% were identified through step 1, and 5% were identified through the last step, consistent with other studies using this algorithm.

A decedent identified in the mental illness decedents group meant that the individual was successfully identified in both the ODMH sources and in the death certificate files (N=30,219). Of these 30,219 decedents, we excluded 15,713 whose record was not present in the MACSIS or PCS service file for that year, leaving 14,506 total mental illness decedents.

Study variables

Dependent variables.

Cause of death was determined with the ICD-10 code found on the death certificate record. Deaths were divided into two groups, injury and disease. Injury-related causes of death included accidents (ICD-10 codes V01–X59 and Y40–Y89), injuries of undetermined intent (Y10–Y36), suicide (X60–X84), and homicide (X85–Y09).

Disease-related deaths were categorized as occurring from cancer (ICD-10 codes C00–C97), diabetes mellitus (E10–E14), substance abuse (F100–F199 and F55), mental illness (F20–F54 and F59–F69), nervous system disease (F00–F09 and G00–G99), cardiovascular disease (I00–I99), and respiratory disease (J00–J99). Disease not falling into any of these categories was grouped in “other disease.” Regardless of whether decedents received services through ODMH, mental illness was documented on some death certificates as the cause of death—for example suicide deemed to have resulted from mental illness. Some deaths were determined as caused by substance abuse—for example in the case of overdose or chronic substance misuse. Although we recognize the issue of possible misclassification, we looked for marked differences in demographic characteristics and leading cause of death.

Independent variables.

Demographic characteristics, including age at death, race, and sex variables, were retrieved from death certificates. Age groupings were <15, 15–24, 25–44, 45–64, and ≥65 years. Decedents were identified as male or female and as black or nonblack. Black was defined as African American or as being of African descent. The vast majority of nonblack decedents were white, and the representation of racial-ethnic minorities other than black was too small to conduct stratified analyses.

Analysis

In addition to descriptive analysis, we derived age-, race-, and sex-adjusted standardized mortality ratios (SMRs) by dividing the observed by the expected number of deaths in each stratum. The stratum-specific expected number of deaths was calculated by multiplying the crude mortality rate in the Ohio population in that stratum by the total number of individuals in that stratum who were enrolled in the ODMH system for at least one month during the study period. The age, race, and sex distributions of the Ohio population at large were obtained from the 2000 U.S. census. The cause-specific SMRs were derived similarly but by using the crude and expected deaths in the relevant cause-of-death categories. The 95% confidence intervals for the SMRs were obtained by using an online calculator developed by Emory University Rollins School of Public Health.

The age-standardized death rates and 95% confidence intervals were calculated with a method described by Curtin and Klein (20), based on the 1940 U.S. Standard Population.

All analyses were conducted with SAS version 9.2.

Results

Of the 438,749 Ohio decedents, 14,506 (3.3%) had mental health service claims, and we termed this group the mental illness decedents. The distribution of mental illness decedents and the distribution of all Ohio decedents during the study period are shown in Table 1. Compared with the Ohio population, much smaller proportions of mental illness decedents were in the oldest age group. The proportion of mental illness decedents in the ≥65 age group was less than half that of the general population (31.9% versus 75.1%, respectively). Also, blacks represented a greater proportion of mental illness decedents than of the general population (18.5% versus 11.1%).

The distribution of decedents by cause of death revealed important differences. For example, nearly one in four mental illness decedents (23.8%) died from an injury-related cause of death, compared with one in 16 (6.4%) in the general population. Conversely, nine of ten decedents in the general population but only three of four mental illness decedents died from diseases. Deaths from substance abuse were nearly five times higher among mental illness decedents compared with all Ohio decedents. Deaths from mental illness were more than double the rate for all Ohio decedents.

Table 2 presents the SMRs and 95% confidence intervals by age, race, and sex strata among mental illness decedents. The highest SMR was observed among nonblack males (SMR=2.14), followed by nonblack females (SMR=1.95), black females (SMR=1.24), and black males (SMR=1.00). These statistics indicate that mortality among mental illness decedents was excessive mostly among nonblack males. In contrast, for black males mortality was not higher among mental illness decedents.

We also saw differences in SMR across age groups within each of the race- and sex-specific strata for mental illness decedents. Among black males, excess mortality was observed in the 25–44 and ≥65 age groups. Among black females, SMR was higher among decedents ≥25 years but not in the younger age groups. Among nonblack males and females, mortality was significantly higher in all but the youngest age group.

The age-standardized death rates were also significantly higher in the mental illness group than in the general population for each of the race- and sex-specific strata.

Table 3 presents the cause-specific SMRs and confidence intervals for the race- and sex-specific strata. Mortality of nonblack males and of female mental illness decedents was excessive for every cause of death, with over twice the expected death rate among nonblack males. Mortality also was excessive among black female mental illness decedents in certain disease categories. On the other hand, compared with the Ohio general population, black male mental illness decedents experienced lower mortality for diseases (SMR=.91) but higher mortality for injury- or violence-related deaths (SMR=1.4). Most important, the homicide rate was similar for black males with or without mental illness. Disease-related mortality among black male mental illness decedents was higher for those with diabetes, substance use disorder, mental illness, and other nervous system disorders but not for other diseases.

Table 4 presents the top ten causes of death by age strata for all Ohio decedents, and the top two causes of death for mental illness decedents. Homicide followed by unintentional injuries led deaths for all Ohio decedents age 15–34, whereas homicide followed by suicide led deaths among mental illness decedents for this age group. Chronic disease, especially cardiovascular disease and cancer, constituted the leading causes of death in the older age groups. Cardiovascular disease was the leading cause of death for those with and without mental illness for persons age 35 and older.

Discussion

Through analysis of mortality data, we compared mental illness decedents (N=14,506) with all other Ohio decedents (N=438,749). Consistent with other studies (4,11), these data indicate that mental illness decedents generally had a higher likelihood of dying sooner than other Ohioans. Mental illness decedents showed higher SMRs than other Ohioans for deaths due to substance abuse, mental illness, diabetes, nervous system disease, cardiovascular disease, respiratory disease, and injury-related causes. About 24% of mental illness decedents but only 6.4% of all Ohio deaths resulted from injury, demonstrating a disproportionate burden of violent deaths among those with mental illness, and yet the top cause of death for all persons under 35, with and without mental illness, was by injury: by accidents for those <15 and by homicide for those ages 15–34. It may be that harsh socioeconomic forces, homelessness, or combat and traumatic experiences have increased these violent causes of death for both the mental illness decedents and the other Ohio young adult decedents studied.

Remarkably, for all decedents above age 35, cardiovascular disease was the dominant cause of death. Because mental illness decedents tend to die earlier than persons without mental illness, this would affect the prevalence of specific causes of death among mental illness decedents compared with the general population. Some causes of death occur most often in older age groups and thus would be less prevalent among mental illness decedents. For example, cancer prevalence increases with age and thus may contribute to the lower percentage of cancer deaths among mental illness decedents compared with other Ohioans.

This study divided deaths into those caused by injury or disease. The subgrouping by disease revealed that both the general population and those with mental illness in Ohio died from preventable, treatable conditions. Study findings were supported by the low Ohio rankings in multiple health areas: 42nd overall in health outcomes, 37th in preventable deaths for decedents under age 75, and 44th in Medicare hospital admission for preventable conditions (www.commonwealthfund.org). Primary care–mental health integration, which offers mental health services embedded within primary care teams, remains essential to lower these preventable deaths, regardless of race, socioeconomic group, military or veteran status, or gender.

This study’s mortality findings raise questions regarding race and mental illness. Excessive mortality among mental illness decedents compared with all Ohio decedents was much higher for nonblack than for black mental illness decedents, especially black men. Does the “lower” mortality for black men reflect worse socioeconomic or primary care availability to black men in general, or is there a protective factor benefiting the black man who has mental illness?

In addition, findings from this study suggest that research and health care programs should address modifiable risk factors, such as health screening, education, and treatment programs for epidemic problems such as obesity, smoking, and substance use disorders. Suicide- and violence-related assessment and prevention programs with a focus on mental illness must be created and implemented in public, military, veterans, and private health care sectors. Focused approaches should be developed especially for patients with complex comorbidity, including those with mental illness, considering that only 5% of the U.S. population account for half of the nation’s total health care spending (www.commonwealthfund.org). Findings from this study underscore the need for learning more about specifics of complex comorbidities that lead to increasing health care costs along the path of terminal suffering and death.

Limitations

Although this study reveals critical information, we also found several limitations that may inform future data tracking. We compared causes of death and measures of excess mortality between mental illness decedents and all Ohio decedents only; however, data sources were unavailable to identify the perhaps large numbers of decedents with mental illness who were treated in the private sector, military, and U.S. Department of Veterans Affairs or who were undiagnosed and untreated. Because a decade of significant U.S. military combat operations encompassed the years of this study, knowing which of these decedents served in combat or suffered from mental illness as a result of their military service would be useful information. In addition, our study excluded the large number of individuals (N=15,713) who were enrolled in the ODMH system but did not receive mental health services, some of whom may have unknown mental illness. We do not know which treatment modalities were used. We also must note limitations with the accuracy of the underlying cause of death due to reliance on death certificates. In the overall analysis, we found some variations by year but not large enough to warrant temporal trend analysis, and thus the cross-sectional analysis led only to correlation and not to causality of mental illness in regard to mortality. Finally, because this study was limited to Ohio decedents, findings may not generalize to other states.

Future directions

Future studies could help reduce morbidity through focus on disparities between specific causes of death among those with and without mental illness so that providers could refine their health care practices. Future research encompassing age-specific and geographic analysis of these disparities could lead to more informed health care distribution, such as urban and rural differences, or analysis within age strata.

The remarkable devastation of cardiovascular disease to all persons over age 35 calls for more research on the breakdown of the subgroups within the broad category, as well as for prevention and treatment of cardiovascular disease. In all sectors we need to work to integrate primary care and mental health care. Injury-related mortality threatens all Ohioans, not only those with mental illness, and collaborative research efforts should analyze this category by mechanism of injury, especially self-directed injury. We would do well to improve the accuracy of the identified cause of death—for example, by avoiding listing suicide as accidental death.

Mental illness decedents in this study were enrolled in the publicly funded mental health system for at least one month during the study period. Future studies should consider a side-by-side analysis using length of enrollment in the system along with the person-year approach. Using a person-year approach would likely increase the SMRs for mental illness decedents.

Conclusions

To better analyze causes of death and to identify mental illness diagnoses for individuals not enrolled in the publicly funded mental health system, we must bridge data silo barriers to research in all treatment arenas, including the private sector, the civilian public sector, and the military service sector. Cause of death, especially self-directed violence, among returning military veterans and other combat veterans should be compared with civilian cohorts. Finally, to identify cause-of-death patterns over time, we should complete a longitudinal study of death certificates in future years. Health care strategies for individuals with mental illness should address demographic risk factors associated with both injury- and disease-related deaths. These study findings demonstrate the need to better understand deaths in the United States, in order to transform health care so that people live longer and healthier lives.