Although we assume that black and Hispanic suicides experienced at least as much proximate mental and physical illness as white suicides, we expect that medicolegal authorities record less comorbidity on their death certificates. This assumption and expectation stem primarily from comparative survey research on the general US population. A study utilizing the Sample Adult component of the 1998–2003 National Health Interview Surveys showed that blacks and Hispanics report being in worse health, having more physical limitations and annual bed days, less health insurance coverage, fewer physician visits, and more unmet mental healthcare needs than whites [5]. These groups also compare unfavorably on objective measures of healthcare access and management [6]. With the focus more specifically on mental health, an analysis of the 2005–2006 National and Nutrition Examination Survey showed that blacks had a higher prevalence of depression than whites and Hispanics [7]. An analysis of the 2001–2004 National Health Interview Survey indicated no difference in the prevalence of serious psychological distress between blacks and whites [8]. While the prevalence was equivalent between Hispanics and whites at ages 18 through 64 years, it was almost three times higher for Hispanics at ages 65 and older. Limited to a comparison between blacks and whites, the National Survey of American Life revealed 45% excess chronicity in major depressive disorder among blacks [9]. Turning back to racial-ethnic treatment disparities, an analysis of the Collaborative Psychiatric Epidemiology Surveys found that 36% of Latinos and 41% of African-Americans with a past-year depressive disorder had received mental health treatment compared to 60% of non-Latino whites [10].

Our expectation of relative underdocumentation of comorbidity on the death certificates of the minority suicides is partially supported by a records-based study [11]. Subjects were adult enrollees in TennCare, Tennessee's Medicaid-waiver managed care program, a public assistance program which offers equal financial access to all low-income groups [12]. This study found that only 29% of African-American suicides and possible suicides, as compared to 51% of white opposites, used prescribed antidepressants in the year prior to their deaths [11]. The authors interpreted this differential as indirect evidence that serious mood disorders are underdiagnosed or undertreated in African-Americans. However, they acknowledged that African-Americans may have been more likely to receive nonpharmacological therapy than whites, and thus could have been less likely to have antidepressant prescriptions filled. They operationalized their possible suicides as decedents whose deaths were officially classified under injury of undetermined intent, and no subjects had a record of chronic psychosis or serious medical illness.

Meta-analyses which combine studies of predominantly patient samples affirm that both comorbid psychopathology and physical disease are suicide determinants [13-16]. Reinforcement comes from two national multiple cause-of-death (MCOD) analyses, an Australian suicide study [17] and a US replication [18]. Implying high bi-national reliability of MCOD data, comparison of their multivariate results revealed identical suicide-associated comorbidities: depression and mood disorders, schizophrenia, and cancer. By contrast, gender-specific prevalences of comorbid psychopathology were almost three times as high in Australian suicides. Moreover, the prevalence of comorbid physical disease was twice as high in Australian as US male suicides, and there was 78% excess prevalence among Australian female suicides compared to US counterparts. The magnitude of these prevalence gaps was surprising, as was the revelation that only among Australian suicides was comorbid psychopathology more prevalent than comorbid physical disease. While both nations are economically and technologically sophisticated democracies, with a shared suicide rate of 11 per 100,000 population [1,19], life expectancy and healthy life expectancy at birth in Australia exceed US expectancies by approximately three years [20,21]. Indeed, the prevalence gaps and differential weight of comorbid psychopathology indicate that Australian death certificates surpass those of the United States in documenting true comorbidity among suicides. We subsequently ascertained that Australian death investigations are more standardized than those in the United States, and that US coders have less capacity than Australian counterparts to access medicolegal records to help resolve outstanding questions [[22]; personal communication, Robert N. Anderson, National Center for Health Statistics, September 22, 2008].

We employed a comparison group in order to identify suicide-associated comorbidities through multivariate analysis. This group comprised white, black, and Hispanic decedents whose underlying cause of death was unintentional injury (V01–X59), sequelae of transport unintentional injury (Y85), or sequelae of other unintentional injury (Y86). We further confined our study and comparison groups to persons who died at ages 15 years or older in the 50 states and the District of Columbia. Fifteen years was the minimum age cutoff because suicide was rarely reported among children and younger teens [34], who constituted less than 1% of all suicides during the observation period. [1] Relevant given the nature of the racial-ethnic comparisons in this study, Hispanic blacks comprised only 1% of Hispanic suicide subjects. This finding holds independent of gender.

Suicide almost invariably becomes the official underlying cause of death when medicolegal authorities rule a death as suicide in a timely manner [35]. Concordance between a suicidal mention on the death certificate and its assignment as the underlying cause approaches 100% [36]. Unlike deaths attributed to a chronic disease, comorbidity enumerated on the death certificates of suicides is almost necessarily antecedent owing to the typically acute nature of injury mortality. However, antecedence does not imply causality.

Replicating the methods of the prior US MCOD suicide study [18] in a racial-ethnic context, we disaggregated comorbidity into 14 types within two major subheadings: mental/behavioral disorders and physical disease. Our prevalence calculations drew upon all potential comorbid mentions on a death certificate – up to 20. However, we counted only one condition where more than one was associated with a particular type of comorbidity. Each of the 14 types served as a distinct outcome variable in a series of unconditional logistic regression analyses that we performed to identify suicide-associated comorbidities in the MCOD dataset for whites, blacks, and Hispanics, respectively. We adjusted the odds ratios for age (three ordinal categories: 15–34, 35–64, and 65+ years) and gender. Our results provide a preliminary basis for assessing the reliability of the comorbidity data in relation to suicide. In a sensitivity analysis we eliminated our covariates, age and gender, from the model. All variables derived from the MCOD database.

Sampling variability was not an issue in this study since the data represented complete counts. However, to allow for random variation in our cross-tabulated prevalence data on suicides, we conducted two series of Chi-square tests before reporting on differentials and patterns. Our first series were two-by-three tests. For each category of comorbidity, distinguishing presence or absence, we separately assessed statistically significant variation within gender and across racial-ethnic groups and age groups, respectively. These results justified our series of two-by-two tests. In these tests we compared white suicides with black suicides and Hispanic suicides, respectively, within specific comorbidity categories for all ages and also within each of our three age groups for each gender. Then separately for each racial-ethnic group within gender, we similarly compared cause-specific comorbidity prevalences across these age groups. To assess change in prevalence across age, we simply compared the youngest age group with the middle age group and then the middle group with the oldest group. In the prevalence tables, we identified each estimate which was calculated on the basis of a cell size with fewer than 20 cases.

Even the NVDRS is likely to be wanting in the documentation of comorbid psychopathology among suicides. For if we emphasize the absolute instead of the relative, psychological autopsy studies provide compelling evidence that comorbid psychopathology is probably grossly under-documented on the death certificates of suicides irrespective of racial-ethnicity and type of disorder. A meta-analysis of 27 of these in-depth studies of record reviews and interviews with survivors estimated that 87% of suicides had a mental disorder [38]. A systematic review, which included meta-analyses conducted separately on 22 case-control studies and 54 case series, reported 90% prevalence of mental disorders in suicide cases compared to 27% in controls, and 91% prevalence in suicides from the case series [39]. Few of the case-control studies involved US adults.

HIV was not a determinant of suicide in any of the three racial-ethnic groups in our study, even though a 1994 meta-analysis identified this disease as a risk factor [13]. This finding more refinedly affirmed such a finding from the first US MCOD suicide study [18]. Posited by the authors of that study as an ex post facto explanation, receipt of effective antiviral therapy could have altered the mindset of HIV-positives by removing a hitherto imminent and grave threat to survival. Thus, if tenable, minority-majority status may not modify this explanation. To that point, another recent US study attributed a relative reduction in HIV mortality to some closure of the life expectancy gap separating blacks from whites [40]. Nevertheless, blacks and Hispanics continue to bear a disproportionate share of the diminishing national HIV/AIDS burden [41,42]. The absence of HIV as a suicide determinant in these groups may be a marker of success by the healthcare system in delivering appropriate therapy to minority HIV-positives. Epidemiologists and social scientists may further clarify the relationship between HIV and suicide utilizing longitudinal data.

A longitudinal evaluation of race and Hispanic origin on death certificates indicated excellent agreement between self-report and death certificate proxy report for blacks and whites and good agreement for Hispanic subgroups [43]. Nevertheless, our results could have been distorted by racial-ethnic heterogeneity, such as between African-Americans and black Caribbean Americans and native-born and foreign-born Hispanics, respectively [5,9,44,45]. For example, an analysis of the 2001–2002 National Epidemiologic Survey on Alcohol and Related Conditions showed that Mexican-Americans manifested a lower lifetime risk of psychiatric comorbidity than US-born non-Hispanic whites [46]. There was, however, no differentiation of risk among their foreign-born counterparts.

Racial-ethnic variation in the respective prevalences of comorbid psychopathology and physical disease may be real, and simply reflect differences in suicide etiology [47-53]. Alternatively, the data are consistent with our expectation that this variation may be partially or wholly artifactual. Such artifactuality would likely be a function of interactions between family, friends, and acquaintances of decedents with medicolegal authorities, variations in social pressure to withhold psychological information across racial and ethnic groups, and differential fastidiousness with which authorities compile medical records on suicides. However, this research could not assess the merits or relative contributions of these factors. Beyond the highly restricted comparisons, our cross-validation of the documentation of comorbid psychopathology on death certificates of suicides was further constrained by the fact that the NVDRS covers only a minority of states – 14 states in 2004 and 17 currently [54]. Moreover, this system comingles personal reports from family and friends on the proximate mental status of decedents with diagnostic data. Thus, any improvement in the sensitivity of documentation of comorbid psychopathology among suicides, rendered by the NVDRS, likely means diminished specificity.