Racial-Ethnic Differences in Receipt of Past-Year Health Care Services Among Suicide Decedents: A Case-Control Study

Abstract

Objective:

Suicide remains an urgent public health crisis. Although some sociodemographic characteristics are associated with greater suicide risk in the general population, it is unclear whether individuals utilizing health care in the United States have similar suicide incidence patterns. The authors examined whether race-ethnicity is associated with suicide death among patients seeking health care and investigated health care utilization patterns.

Methods:

Data were collected from electronic health records and government mortality records for patients seeking health care across nine health care systems in the United States. Patients who died by suicide (N=1,935) were matched with patients in a control group (N=19,350) within each health care system.

Results:

Patients who died by suicide were significantly more likely to be White, older, male, living in low-education areas, living in rural areas, or diagnosed as having mental health conditions or were significantly less likely to have commercial insurance (p<0.05). Among most racial-ethnic groups, those who died by suicide had a higher number of past-year mental health, primary care, and total health care visits; for American Indian/Alaska Native patients, the number of health care visits tended to be lower among suicide decedents.

Conclusions:

These findings suggest that higher past-year health care utilization was associated with increased likelihood of suicide death across several racial-ethnic groups. This observation underscores the need for identifying and managing suicide risk in health care settings, including outside of mental health visits, among most racial-ethnic groups.

In the United States, approximately 48,500 persons died by suicide in 2019, making suicide a major public health crisis (1). Suicide contributes to early death and is currently the 10th leading cause of death (2). Rates of suicide have risen approximately 35% from 1999 to 2018 (3), making the need for prevention strategies even more salient.

Although a person’s history of suicide attempts is one of the strongest predictors of future suicidal behavior (4), prediction of suicide remains challenging. Several studies have investigated sociodemographic differences in suicide deaths for community samples. Suicide disproportionately affects those with lower education levels (5), lower income (6), lack of health insurance (7), and diagnosed mental health problems (8). Men are more likely to die by suicide; however, women report suicidal ideation more often and have more suicide attempts (9–11). Regarding race-ethnicity, although non-Hispanic Whites have the greatest number of suicide deaths, the rate of suicide is consistently highest among American Indian/Alaska Natives (12). Other racial-ethnic groups (i.e., Hispanic, Asian, and Black) have comparatively lower rates (12), although there has been a recent increase in suicide death among Black youths (13).

Racial-ethnic differences in rates of psychiatric diagnoses and treatment have also been noted. Asian individuals have the lowest rates of psychiatric diagnoses, whereas American Indian/Alaska Natives have the highest (14). Non-Hispanic Whites were more likely to receive a psychiatric medication, compared with other racial-ethnic groups (14). Among those with severe suicidal ideation or postsuicide attempt, non-Hispanic Black, Asian, Hawaiian/ Pacific Islander, and Hispanic individuals were less likely to receive outpatient mental health services, compared with non-Hispanic Whites (15). The reasons for the lower care utilization among these groups are likely multifactorial, including barriers to access (16), stigma (17), and differences in the perceived need for mental health care (18).

Although some racial-ethnic groups may be at lower risk for suicide death overall, people in these racial-ethnic groups who are seeking care for a mental disorder may be at higher risk for suicide. However, whether this scenario is true is not yet known. Such information would be critical for mental health providers. In addition, it is important to understand factors associated with suicide death specifically within health care systems because a Sentinel Event Alert recommendation and a national patient goal requirement from The Joint Commission advise suicide prevention efforts be conducted within health care settings (19). The purpose of the present study was to investigate racial-ethnic differences in suicide death and to explore whether a documented mental disorder and past-year utilization of health care were associated with increased likelihood of suicide death.

METHODS

Data Source

This study was conducted with a data set that recorded health care utilization in the year before a suicide event. Data were extracted from electronic health records (EHRs) from nine health care systems across the United States that participate in the Mental Health Research Network (MHRN): Henry Ford Health, Essentia Institute of Rural Health, Harvard Pilgrim Health Care Institute, Health-Partners Institute, and Kaiser Permanente of Colorado, Georgia, Hawaii, Northwest, and Washington State. The MHRN is a consortium of health care systems that serve >20 million patients with the goal to address mental health problems; details regarding the participating sites can be found on the MHRN website (https://mhresearchnetwork.org). This collaboration allowed for the rapid examination and dissemination of generalizable findings. Institutional review board approval was obtained from each of the nine health care systems. Insurance claims and EHR data from each of these systems are stored in a virtual data warehouse (VDW). The VDW uses the same definitions for variables across sites, and data are quality checked regularly, allowing for integration of data across the sites.

Measures

The outcome of suicide death was determined from official government mortality records and matched to ICD codes (X60–X84 and Y87.0) in the VDW. Health diagnoses and visits for all patients were captured in the year before the index date (i.e., the date of suicide death). Psychiatric diagnoses were defined by ICD-9-CM codes, which included codes 290–319. Patients were noted to have a psychiatric diagnosis if they had one of these diagnosis codes in the year before the index date. Health care visits were also extracted from the EHRs. The number and type of visits were identified and categorized into subtypes, including primary care visits, mental health visits, and all visits (i.e., both primary care and mental health visits and other specialty care visits).

Race-ethnicity was identified in the VDW. Patients were categorized as Asian, Black/African American, Hawaiian/ Pacific Islander, Hispanic (e.g., Hispanic or Latinx), American Indian/Alaska Native, White, and other (including individuals of multiple races and those who selected “other” to record their race in the EHR).

Other sociodemographic factors extracted from the VDW included age, gender, insurance type, neighborhood income, and neighborhood education level. Geocoded addresses and U.S. Census Bureau block data were used to estimate neighborhood income and education. Individuals were considered to have a low income if they lived in a census area where >20% of residents had an income below the U.S. poverty line (i.e., $22,314, which was based on the weighted average poverty threshold for a family of four in 2010). Individuals were categorized in the low-education group if they lived in a census block where <25% of the residents were college graduates.

Sample

Patients included in the study were seen at one of the participating nine health care systems between 2000 and 2015 and were ages ≥10 years. Patients were included only if their race-ethnicity was documented in the VDW. Race-ethnicity was considered documented when patients responded to the standard questions for an outpatient visit. The original data were from 3,330 individuals who died by suicide and 333,000 individuals in the control group. When a racial-ethnic category was not available in an EHR, its absence could have reflected that the question was not asked or that the patient declined to answer. Thus, not all patients’ race-ethnicity was captured routinely in the EHR and was missing for 42% (N=1,395) of the individuals who died by suicide and 65% (N=216,450) of the individuals in the control group. The remaining sample therefore included 1,935 suicide decedents. To obtain 80% statistical power in analyses, a 1:4 match was needed. Because we had access to a larger sample, we used 1:10 matching to reach higher statistical power to detect any risk associations more precisely. Each suicide decedent was therefore matched with 10 individuals with documented race-ethnicity in the control group (N=19,350), comprising individuals who were randomly selected from the same health care system and who were enrolled in care during the same period but did not die by suicide.

Analyses

We conducted t and chi-square tests to examine descriptive differences between suicide decedents and individuals in the control group on demographic characteristics, psychiatric diagnoses, and health care visit type. Conditional logistic regression adjusted for demographic characteristics was used to investigate whether psychiatric diagnoses were associated with suicide death. The average number of visits (in total and by visit type) was also examined for both groups during the 1-year period before the index date. To examine racial-ethnic differences in suicide death in the context of mental health care utilization, analyses were performed separately for each racial-ethnic group. For health care utilization, analyses were not adjusted for covariates because in clinical practice, a provider would not record all demographic factors when looking at overall health care utilization of a single patient. Analyses were performed with SAS, version 9.4.

RESULTS

Table 1 displays the sociodemographic characteristics and health care visits of the patients in the case group who died by suicide and the matched patients in the control group. A greater proportion of White patients was found in the case group (case vs. control individuals: 86.6% vs. 77.9%), compared with lower proportions of Asian (4.0% vs. 6.3%), Black/African American (3.2% vs. 6.5%), and Hispanic (1.5% vs. 3.1%) patients. Individuals who died by suicide were more likely to be older (age ≥65 years: 27.2% vs. 18.5%), male (75.9% vs. 44.2%), living in a low-education neighborhood (41.7% vs. 39.0%), living in rural areas (36.8% vs. 33.1%), and diagnosed as having mental disorders (58.3% vs. 19.9%), compared with individuals in the control group. Patients who died by suicide were less likely to have commercial insurance (54.4% vs. 65.8%, respectively). For health care visits, compared with the control group, patients in the case group had a significantly greater number of overall visits (28.0 vs. 12.4), primary care visits (7.7 vs. 4.0), and mental health visits (5.3 vs. 0.8) in the year before the index date.

TABLE 1. Sociodemographic characteristics and health care utilization among patients who died by suicide and matched patients in a control group^a.

Characteristic	Control group (N=19,350)		Suicide decedent group (N=1,935)
	N	%	N	%	p
Age in years					<.001
10–17	3,544	18.3	64	3.3
18–39	4,316	22.3	441	22.8
40–64	7,907	40.9	903	46.7
≥65	3,583	18.5	527	27.2
Gender					<.001
Female	10,800	55.8	466	24.1
Male	8,550	44.2	1,469	75.9
Race-ethnicity					<.001
Asian	1,210	6.3	77	4.0
Black/African American	1,251	6.5	62	3.2
American Indian/Alaska Native	113	.6	13	.7
Hawaiian/Pacific Islander	221	1.1	17	.9
Hispanic	605	3.1	29	1.5
Other	882	4.6	62	3.2
White	15,068	77.9	1,675	86.6
Insurance type					<.001
Commercial	12,739	65.8	1,053	54.4
Medicaid	537	2.8	47	2.4
Medicare	3,155	16.3	536	27.7
Private payment	607	3.1	65	3.4
Other or unknown	2,312	11.9	234	12.1
Low-income areab					.16
No	12,809	66.2	1,242	64.2
Yes	1,350	7.0	151	7.8
NA	5,191	26.8	542	28.0
Low-education areac					.01
No	11,294	58.4	1,066	55.1
Yes	7,544	39.0	806	41.7
NA	512	2.6	63	3.3
Rural-urban location					<.001
Rural	6,396	33.1	713	36.8
Urban	12,758	65.9	1,195	61.8
NA	196	1.0	27	1.4
Diagnosed mental health condition					<.001
No	15,500	80.1	806	41.7
Yes	3,850	19.9	1,129	58.3
N of any health care visits					<.001
M±SD, range	12.4±17.3	.0–408.0	28.0±33.4	.0–363.0
Median, IQR	7.0	3.0–15.0	17.0	6.0–38.0
N of primary care visits					<.001
M±SD, range	4.0±5.9	.0–122.0	7.7±10.7	.0–99.0
Median, IQR	2.0	0.0–5.0	4.0	1.0–10.0
N of mental health visits					<.001
M±SD, range	.8±3.6	.0–110.0.	5.3±11.6	.0–143.0.
Median, IQR	.0	0–.0	1.0	0–5.0

^aSource: electronic health records of patients seeking health care in one of nine U.S. health care systems. Individuals who died by suicide were documented in government mortality records, 2000–2015. Each suicide decedent was matched with 10 individuals randomly selected from the same health care system as a control group. The “other” racial group includes individuals of multiple races and those who selected “other” to represent their race. IQR, interquartile range; NA, not available.

^bProportion of individuals living in U.S. Census blocks where >20% have incomes below the U.S. poverty level.

^cProportion of individuals living in U.S. Census blocks where <25% are college graduates.

As shown in Table 2, patients diagnosed as having a mental disorder were more likely to die by suicide. Within some racial-ethnic groups, those with a mental disorder had significantly higher odds of suicide death compared with those without a mental disorder (p<0.001). For example, Hispanic patients diagnosed as having a mental disorder were more likely to die by suicide relative to Hispanic patients without a diagnosed mental disorder (adjusted OR [AOR]=12.34), followed by Black/African American (AOR=8.46), White (AOR=6.79), and Asian (AOR=6.19) patients. We also investigated the associations of race-ethnicity with suicide death stratified by the presence of a diagnosed mental disorder (Table 3). Among those with a mental disorder, no significant differences in suicide death were observed for any racial-ethnic group, compared with White patients. Among those without a mental disorder, Black/African American patients were less likely to die by suicide compared with White patients (AOR=0.45).

TABLE 2. Patients with a mental disorder and adjusted ORs (AORs) for suicide death, by race-ethnicity.

Race-ethnicity	Control group			Suicide decedent group
	Total N	N with mental disorder	%	Total N	N with mental disorder	%	AORa	95% CI
Asian	1,210	134	11.1	77	35	45.5	6.19	3.54–10.82
Black/African American	1,251	189	15.1	62	36	58.1	8.46	4.40–16.29
American Indian/Alaska Native	113	24	21.2	13	5	38.5	—	—
Hawaiian/Pacific Islander	221	36	16.3	17	8	47.1	—	—
Hispanic	605	132	21.8	29	18	62.1	12.34	3.58–42.50
Other	882	180	20.4	62	38	61.3	7.65	3.67–15.96
White	15,068	3,155	20.9	1,675	989	59.0	6.79	6.00–7.69

^aSource: electronic health records of patients seeking health care in one of nine U.S. health care systems, 2000–2015. ORs were adjusted for age, gender, insurance, education, income, and urban-rural location; the reference for each racial-ethnic group was a control group comprising individuals with the same race-ethnicity (AORs were statistically significant at p<0.001 in all comparisons). Individuals who died by suicide were documented in government mortality records. Each suicide decedent was matched with 10 individuals randomly selected from the same health care system as a control group. The “other” racial group includes individuals of multiple races and those who selected “other” to represent their race. Because of small numbers in the American Indian/Alaska Native and Hawaiian/Pacific Islander groups, AORs could not be estimated for these two groups.

TABLE 3. Adjusted ORs (AORs) for suicide death for each racial-ethnic group, by presence of a mental disorder.

Race-ethnicity	Patients with mental disorder (N=4,979)		Patients without mental disorder (N=16,306)
	AORa	95% CI	AORa	95% CI	p
Asian	.79	.51–1.23	.95	.66–1.36	.76
Black/African American	.72	.48–1.08	.45	.29–.70	<.001
American Indian/Alaska Native	.82	.29–2.27	1.90	.88–4.12	.10
Hawaiian/Pacific Islander	.52	.22–1.22	1.04	.49–2.19	.92
Hispanic	.66	.39–1.13	.58	.31–1.09	.09
Other	.71	.48–1.04	.83	.53–1.28	.39

^aORs were adjusted for age, gender, insurance, education, income, and urban-rural location; White was the reference group. Source: electronic health records of patients seeking health care in one of nine U.S. health care systems, 2000–2015. Individuals who died by suicide were documented in government mortality records. Each suicide decedent was matched with 10 individuals randomly selected from the same health care system as a control group. The “other” racial group includes individuals of multiple races and those who selected “other” to represent their race.

Figure 1 shows the number of health care visits in the year before the index date for each racial-ethnic group and by case and control groups (see also Table S1 in the online supplement to this article). Across the racial-ethnic groups, individuals in the case group had a mean number of 13–31 total health care visits during the preceding 1-year period, compared with 9–14 visits for the control group. Patients in the case group had 3–10 primary care visits and 2–6 mental health visits, and patients in the control group had 3–5 primary care visits and 0–1 mental health visits. For several racial-ethnic groups, patients in the case group had a significantly higher number of total health care visits than those in the control group. This was true for Asian (19.4 vs. 9.0 visits), Black/African American (17.9 vs. 11.0 visits), White (29.1 vs. 12.9 visits), and “other” (30.5 vs. 11.2 visits) groups. Individuals who died by suicide among Asian, Black/African American, White, and “other” groups also had more mental health visits than individuals in the control group (Asian: 4.4 vs. 0.4; Black/African American: 3.1 vs. 0.5; White: 5.5 vs. 0.9; and “other”: 6.5 vs. 0.7). White patients who died by suicide had significantly more primary care visits than White patients in the control group (7.8 vs. 4.2), and this was also observed for the patients who reported their race-ethnicity as “other” (10.3 vs. 3.9). Of note, in the American Indian/Alaska Native group, suicide decedents and individuals in the control group had similar numbers of visits in all three categories.

FIGURE 1. Total health care, primary care, and mental health care visits in the year before the index date, by racial-ethnic and suicide decedent and control groups^a.

^a Source: electronic health records of patients seeking health care in one of nine U.S. health care systems. Individuals who died by suicide were documented in government mortality records, 2000–2015. Each suicide decedent was matched with 10 individuals randomly selected from the same health care system as a control group. The “other” racial group includes individuals of multiple races and those who selected “other” to represent their race. Error bars indicate 95% CIs. The vertical broken lines represent the mean number of visits for the suicide decedents and individuals in the control group across racial-ethnic groups.

*p<0.05, **p<0.01, ***p<0.001 for differences within racial-ethnic groups.

Because of the large number of patients who did not have race-ethnicity documented in the EHR, we conducted a sensitivity analysis. We observed that those without a recorded race-ethnicity were 1.24 times more likely to die by suicide than were White patients. Similar to those with documented race-ethnicity, individuals without a recorded race-ethnicity and diagnosed as having a mental disorder were more likely to die by suicide than their counterparts without a mental disorder (AOR=7.32, 95% CI=6.39–8.39, p<0.001). Compared with White patients and among individuals with a mental disorder, those without a recorded race-ethnicity were more likely to die by suicide (AOR=1.84, 95% CI=1.60–2.21, p<0.001). We also observed that among those without a mental disorder, patients without a recorded race-ethnicity were also more likely to die by suicide (AOR=1.95, 95% CI=1.69–2.25, p<0.001). Individuals without a recorded race-ethnicity had fewer total health care visits compared with White, Asian, Black/ African American, and Hispanic individuals (control group: mean±SD=7.1±10.6; case group: mean=16.8±24.0) but had similar numbers of primary care visits (control group: mean=2.5±3.9; case group: mean=4.5±6.6) and mental health visits (control group: mean=0.8±3.9; case group: mean=4.9±11.0).

DISCUSSION

In this study, we examined racial-ethnic differences in suicide death and past-year health care utilization among those seeking health care within nine hospital systems across the United States. We found that sociodemo-graphic factors, such as older age, male gender, White race, no commercial insurance, rural living location, and lower level of education in a patient’s neighborhood, were significantly associated with increased odds of suicide death. These findings are consistent with known associations in the general population (5–7, 9–12, 20).

After adjusting our analysis for sociodemographic factors, we found that the presence of a mental disorder was associated with a higher likelihood of suicide death across all racial-ethnic groups. This observation is consistent with national data suggesting that people with serious psychological distress are more likely to self-report a suicide attempt (21). However, many of those who died by suicide did not have a mental disorder documented in their EHR, and within some racial-ethnic categories (Asian, Hawaiian/ Pacific Islander, and American Indian/Alaska Native), more than half of those who died by suicide did not have a documented mental disorder (Table 2). Notably, although self-reported suicide attempts have continued to increase overall in the U.S. population, this increase has not been accompanied by an increase in mental health–or substance use–related service utilization (21). With respect to the sample for this study, it may be that patients did not report concerns consistent with a mental disorder or that their con cerns were not perceived as warranting a psychiatric diagnosis. Some racial-ethnic groups (i.e., African American and Latinx) are less likely to seek mental health care (22–24) and may therefore be less likely to have a mental disorder documented in their EHR.

In this study, compared with suicide decedents in other racial-ethnic groups, White patients who died by suicide had on average a higher number of mental health, primary care, and total health care visits. This finding has important implications for health care systems and clinicians. It supports the notion that suicide risk needs to be assessed outside the context of a mental health visit and among those who do not have a documented mental disorder (25), especially for some racial-ethnic groups, such as White patients, because such risk may be more easily identified in non–mental health settings. Indeed, health care providers outside of mental health settings frequently have contact with patients who die by suicide. Nearly 30% of decedents had a health care visit in the 7 days before suicide, approximately half were seen within the preceding 30 days, and >90% had a health care visit within the previous year (25). Because half of the patients who die by suicide do not have a psychiatric diagnosis (26), understanding overall health care utilization patterns (i.e., mental health visits, primary care visits, and total health care visits) may be an approach that could be useful in identifying at-risk individuals who do not have a psychiatric diagnosis, especially for patients from racial-ethnic groups who may be more likely to be seen in other care contexts.

Across several racial-ethnic groups (e.g., Asian, Black/African American, White, and other race), those who died by suicide had a higher number of health care visits in the year before suicide compared with individuals in the matched control group. This finding may suggest that individuals at increased risk for suicide are seeking care at higher rates yet may not be identified as being at high suicide risk. That individuals who died by suicide attended more health care visits overall, as well as primary care visits and mental health care visits, compared with those in the control group again supports the recommendation that patients be assessed for suicide risk outside of mental health visits. In addition, clinicians may also consider screening for suicide risk among individuals who are high utilizers of health care. In this study, as the number of past-year health care visits increased, so did the risk for suicide death, except for American Indian/Alaska Natives. In that group, those who died by suicide tended to have fewer health care visits compared with individuals in the control group. Therefore, identifying American Indian/Alaska Natives at suicide risk in health care systems may be especially challenging, and culturally based programs may be more successful at engaging individuals at risk (27). Individuals who are Black/African American or Hispanic are less likely to report suicidal ideation but may be more likely to attempt suicide than individuals who are White (28), and so alternative ways of identifying those who may be at risk for suicide (i.e., high utilizers of health care) may be beneficial for patients from these racial-ethnic groups. In addition, screening for depression and implementation of the Zero Suicide initiative in health care settings outside the context of mental health visits could be useful (29).

This study had several strengths, including using data from a large, multisite, and geographically diverse sample and collected over an extended period and using a matched case-control design. However, a limitation of this research was that race-ethnicity was not recorded for a significant portion of patients, which may have introduced bias. We found that race-ethnicity may not have been missing at random because those without recorded race-ethnicity were more likely to die by suicide. This group also had on average a lower number of total health care visits than most other racial-ethnic groups. It is possible that individuals in this group had fewer opportunities to have race-ethnicity added to their EHRs. Alternatively, those who did not have race-ethnicity recorded could have been from a racial-ethnic group that may be at higher risk for suicide. More work is needed to explore this question. Moreover, our sample did not include uninsured individuals. These individuals would likely have lower health care utilization and may need alternative methods of risk identification. Another limitation was that heterogeneity within the suicide decedent group (e.g., history of suicide attempts, suicidal ideation, and severity of mental health conditions) was not measured. Finally, deaths that were coded as “undetermined” intent were not recorded as suicide deaths in this study. We also note sociodemographic differences in how deaths were labeled, which could have influenced the findings.

CONCLUSIONS

Health care visits may be a critical clinical context in which to assess a patient’s suicide risk and to offer needed support, and such an approach may be especially important for patients from racial-ethnic groups who may be less likely to engage in mental health treatment or to have a documented mental disorder. In addition, risk for death by suicide may be higher among those who frequently use any type of health care, and this utilization pattern may offer opportunities to identify individuals at risk for suicide who are in need of appropriate care.

HIGHLIGHTS.

• Several sociodemographic factors are associated with increased risk for suicide death among individuals seeking health care.

• Patients across all racial-ethnic groups who died by suicide were more likely to have a documented mental disorder, but a substantial proportion did not have a recorded psychiatric diagnosis.

• Across most racial-ethnic groups, patients who died by suicide had higher levels of health care utilization in the year before death.