Service Use in the Month and Year Prior to Suicide Among Adults Enrolled in Ohio Medicaid

Suicide is among the leading causes of death in the United States, claiming more than twice as many lives as homicide (1). After cancer and heart disease, suicide accounts for more years of life lost than any other cause of death (2). Consequently, a specific goal of the National Strategy for Suicide Prevention is the development of programs aimed at improving the ability of general medical and mental health care providers to identify and treat persons at risk of suicide (3). Toward this end, information about health services received by persons who later die by suicide is needed, including their characteristics and patterns of service use before suicide.

Findings from a small but growing number of studies of U.S. samples suggest that contact with health services in the year before suicide is common, particularly with general medical rather than mental health providers. A review of European, Australian, and American studies conducted between 1954 and 1997 found that an average of 77% of suicide decedents made a primary care visit and 32% had a mental health service visit within a year of death (4). High rates of service use in the year before suicide have been reported in recent U.S. studies of male Veterans Health Administration (VHA) patients with substance abuse disorders (94.6%) (5), VHA patients in two geographic areas (90%) (6), and a large population of health maintenance organization enrollees in a mental health research network (83%) (7). Most visits were to primary care or general medical specialty providers. Approximately half of decedents made a health care visit within a month of suicide (6,7).

To our knowledge, no studies have examined health service contacts of suicide decedents in a Medicaid population. It is important to address this gap for several reasons. First, Medicaid is the primary health insurer for poor or disabled individuals in the United States, covering over 66 million Americans—more than one in every five—at some point during the year (8). Second, although suicide prevalence among Medicaid recipients is unknown, rates of mental illness in this population exceed national estimates (48% among Medicaid enrollees versus 31% nationally). Discontinuities in coverage, low literacy, language barriers, lack of transportation, unstable housing, unstable employment, and poverty all place Medicaid recipients at heightened risk of suicide (9). Third, because Medicaid recipients experience many social and economic challenges that affect their health and ability to navigate the health care system, it is possible that clinical profiles and service use patterns prior to suicide may differ from those reported in other studies.

This study aimed to inform suicide prevention efforts by describing the clinical profiles and health care utilization of Ohio Medicaid enrollees who died by suicide. The specific goals were to estimate the condition-specific incidence of suicide among adult Medicaid enrollees, identify the type of visits (general medical or mental health) and the timing and frequency of visits in the year before suicide, compare the characteristics of decedents who did and did not seek care in the month and year prior to suicide, and examine factors associated with type of visits (general medical only, mental health only, or both) among decedents who sought services within a month of suicide.

Methods

Data and Sample

Data for this study came from two sources: Medicaid claims files and death certificate files. The Ohio Department of Medicaid and the Ohio Department of Mental Health and Addiction Services provided the claims files. The Ohio Department of Health provided the death certificate files. Death certificate records from January 1, 2008, through December 31, 2013, included Social Security number (SSN), last name, first name, date of birth, race-ethnicity, county of residence, and ICD-10 codes of all causes of death (X60–X84, Y87.0, and *U03). Medicaid claims data included demographic and clinical characteristics, program eligibility basis, and detailed services information. Medicaid and death certificate records were linked by using a deterministic, multistep algorithm based on combinations of SSN, last name, first name, birth month, and birth year (10). This study was exempt from institutional review board approval because it did not involve human subjects. A total of 1,338 Ohio Medicaid enrollees ages 19 to 65 died by suicide in the study period. These deaths represent 19% of suicide deaths in that age group in the state for the same period.

Measures

Service utilization.

Data were extracted from Medicaid claims files for each of the 1,338 individuals in the year prior to suicide and included general medical, mental health, and substance abuse treatment visits to outpatient programs, inpatient facilities, or emergency rooms. ICD-9-CM codes were linked with encounters, allowing for categorization of reasons for visits by primary diagnosis, which were grouped into two categories: general medical visits and mental health or substance abuse treatment visits (ICD-9-CM 290–319). Because substance abuse treatment visits were infrequent (for example, 7% of the sample [N=90] had a substance abuse treatment visit within a year of suicide), they were included with mental health visits in all analyses.

Demographic characteristics.

Demographic characteristics of the 1,338 decedents included age (19–29, N=314, 24%; 30–49, N=684, 51%; and 50–65, N=340, 25%), race-ethnicity (non-Hispanic white, N=1,160, 87%; and nonwhite, N=178, 13%), sex (male, N=862, 64%), and area of residence categorized according to the Office of Management and Budget metropolitan statistical area designation (urban, N=1,000, 75%, and rural, N=338, 25%).

Clinical characteristics.

Medicaid claims data capture up to nine ICD-9-CM diagnoses per visit. These were used to create variables for presence or absence at any time over the year preceding death of the following conditions: any psychiatric diagnosis, excluding substance use disorder diagnoses (N=804, 60%); any substance use disorder (N=553, 41%); and any chronic general medical condition according to the Charlson Comorbidity Index (N=547, 41%) (11). Comorbidity variables included any co-occurring mental and substance use disorder (N=447, 33%), any mental disorder and chronic medical condition (N=449, 34%), any substance use disorder and chronic medical condition (N=312, 23%), and all three conditions (N=274, 21%). Medicaid eligibility status was an additional clinical characteristic; over half of the sample (N=711, 53%) qualified for Medicaid on the basis of disability. Specific psychiatric diagnoses examined in group comparison and analytic models included depression (N=491, 37%), anxiety disorder (N=377, 28%), bipolar disorder (N=322, 24%), schizophrenia or psychosis (N=182, 14%), and other psychiatric disorder (N=317, 24%). However, specific substance use disorders (such as heroin or cocaine abuse) and chronic general medical conditions (such as diabetes) were insufficiently frequent to examine individually.

Analyses

Condition-specific suicide incidence was calculated for disorder categories (psychiatric, substance use, chronic general medical, and combinations) by dividing the total number of decedents with the condition in each year by the total number of Medicaid enrollees with the condition. Participant demographic and clinical characteristics were documented across groups (in the 30 days prior to suicide, any visit or no visit; visit within 30 days or visit within 31–365 days; only a mental health visit; only a general medical visit; or both types of visit) by using counts and percentages. The p values for the groups (visit within 30 days versus visit within 31–365 days) were based on either logistic regression (two-level characteristic) or multinomial logistic regression (three or more levels) where group was the independent variable. These logistic regression analyses were controlled for by the number of visits (both mental health and general medical) to adjust for the differences in visits over the course of a year between the two groups. For the groups (only mental health visit, only general medical visit, both visit types in the month prior to suicide), p values were based on a chi-square analysis. A multivariate multinomial logistic regression analysis examined decedent characteristics and the likelihood that, in the month prior to suicide, only mental health visits or both mental health and general medical visits occurred, compared with only general medical visits. All analyses were conducted with SAS 9.4 software (12).

Results

Suicide Incidence Rates

Overall, there were 18.9 suicides per 100,000 Ohio Medicaid enrollees (Table 1). Incidence was lowest among enrollees who had only one type of condition, with the exception of those who had co-occurring general medical conditions and substance use disorders, and highest among those who had multiple comorbidities. Rates ranged from 12.4 per 100,000 enrollees among those who had only general medical conditions to 41.7 per 100,000 among those with all three conditions: general medical, mental health, and substance use. Among enrollees with disability as the basis of eligibility, the incidence rate was 31.4 per 100,000.

Service Use Prior to Suicide

Most Medicaid enrollees who were suicide decedents (83%) had at least one health care visit in the year prior to suicide (Table 2); half (50%) had a visit within 30 days of death and about a quarter (27%) within one week. In the year prior to suicide, 79% (N=1,063) of decedents had at least one general medical visit and 56% (N=745) had a mental health visit. In the week prior to suicide, 18% of decedents (N=246) made a general medical visit and 14% (N=196) made a mental health visit. Among those with any visits, most made general medical and mental health visits (63%; N=701 of 1,107) in the year prior to suicide, approximately one-third made a general medical visit only (33%; N=362 of 1,107), and very few made a mental health visit only (4%; N=44 of 1,107).

Decedents who used services typically made multiple visits over the course of the year prior to suicide, with the median number of visits ranging from one in the seven days prior to suicide to 16 in the year prior to suicide. Most visits by decedents took place in outpatient settings (N=872 of 1,107, 79%), followed by emergency rooms (N=181 of 1,107, 16%) and inpatient settings (N=54 of 1,107, 5%).

Characteristics of Decedents by Receipt and Timing of Care

Compared with decedents who did not have any health care visit, those who had a visit within a year of suicide were more likely to be female, to live in an urban location, to be eligible for Medicaid via disability, and to be continuously enrolled in Medicaid. No significant differences between those with and without visits were found by age or race-ethnicity. [A table presents more detail in an online supplement.] A clinical profile for the 231 decedents who had no visits cannot be determined because this study relied on claims information to determine service use patterns.

Decedents who had any visit in the month prior to suicide were compared with those who had visits in the year but not in the month prior to suicide (Table 3). Logistic regression analyses that were adjusted for the number of visits (30 days, median=26.5, interquartile range [IQR]=12–49; 31–365 days, median=6, IQR=3–15) indicated that decedents who had a visit in the month prior to suicide were significantly more likely to have a diagnosis of a mental disorder, a substance use disorder, a chronic general medical condition, or some combination; they were also more likely to be eligible for Medicaid through disability than for other reasons. Of the demographic characteristics examined, age and sex differentiated the groups, with older individuals (age groups 30–49 and 50–65) and females being significantly more likely to have seen a general medical or mental health professional in the month prior to suicide, compared with decedents ages 19–29 or males, respectively.

Factors Associated With Type of Visit in the Month Before Suicide

Table 4 presents results of the analysis examining associations between decedent characteristics and visit types (both general medical and mental health, general medical only, and mental health only) for those who had any visit within a month of suicide. Age, area of residence, and some clinical characteristics differed significantly across the three groups. The group with only mental health visits was significantly more likely than the other two groups to be younger (ages 19–29) and to live in urban areas but less likely to have a chronic general medical condition. The group with only general medical visits was significantly less likely than the other two groups to have a mental or substance use disorder. Individuals with co-occurring conditions were significantly more likely than those without such conditions to have made both general medical and mental health visits in the month before suicide.

Multivariate multinomial logistic regression analysis was used to examine the adjusted association between patient characteristics and the likelihood of only mental health visits or both mental health and general medical visits, compared with only general medical visits, in the month before suicide (Table 5). Characteristics that significantly differentiated those who had both types of visits from those who had only general medical visits were psychiatric diagnoses of depression, bipolar disorder, schizophrenia, and other mental disorder, with odds ranging from 2.05 to 3.34. Compared with younger decedents (ages 19–29), older decedents (adjusted odds ratio [AOR]=.47 for 30–49; AOR=.39 for 50–65) were significantly less likely to make mental health visits only, as were those living in rural areas (AOR=.40 versus urban residence). Compared with those who had only general medical visits, those who had only mental health visits were more likely to have a diagnosis of depression (AOR=2.71), bipolar disorder (AOR=2.81), or schizophrenia (AOR=5.56) and less likely to have a chronic general medical condition (AOR=.33).

Discussion

In this study of Medicaid-enrolled adult suicide decedents in Ohio, the suicide rate of 18.9 per 100,000 is higher than that in the general U.S. population (12.6 per 100,000) and in Ohio (16.3 per 100,000) (13). The condition-specific suicide rates underscore that Medicaid enrollees with psychiatric disorders, especially in combination with substance use disorders and chronic general medical conditions, are a group at particularly high risk of suicide.

The findings regarding the timing and type of health care visits prior to suicide are broadly consistent with those of other studies (5–7,14). Eighty-three percent of decedents had a health care visit within one year of death, 50% had a visit within one month, and 27% within one week. In the year prior to suicide, 79% of the sample had general medical visits and 55% had mental health visits. Most of these visits (79%) were in outpatient settings. Therefore, health care visits, whether for general medical or behavioral health conditions, are opportunities to identify and intervene with individuals at risk of suicide. Although universal screening for suicidality in ambulatory and general medical settings is difficult to justify in the absence of a responsive system of care, general medical and behavioral health professionals might productively direct case-finding efforts toward Medicaid enrollees with known disability, psychiatric disorders, substance use disorders, and chronic general medical illness, particularly those with co-occurring disorders. Tools, such as the Columbia Suicide Severity Rating scale (15) or the Patient Health Questionnaire (16), might prove useful, and state Medicaid programs should explore incentives to facilitate case finding in these populations at risk of suicide.

Data on the frequency of visits indicate that a portion of decedents were intensive service users, and these individuals had severe clinical profiles. Decedents who had more problems, as measured by all types of co-occurring conditions, were more likely than those without co-occurring conditions to have made visits within a month of suicide. In addition, decedents who received care in the month prior to suicide tended to be seen in settings consistent with their clinical profiles: those with serious psychiatric disorders were more likely than those without such disorders to make mental health visits, and those with chronic general medical conditions were more likely than those without such conditions to make general medical visits. Taken together, these results suggest that it is feasible to develop algorithms to identify high-risk Medicaid enrollees on the basis of co-occurring psychiatric, substance use, and chronic general medical conditions and a history of at least monthly health care encounters of any type.

Medicaid programs may want to consider incorporating suicide-specific initiatives into their existing efforts to manage the care of high-cost and vulnerable participants. Alternatives include availability of peer supporters to help coordinate general medical and behavioral health care and better access to and sharing of electronic health records of all general medical and behavioral health care providers involved in the client’s treatment. In addition, Medicaid programs may want to consider adopting VHA protocols that mandate the use of a standard safety plan or annual suicide risk assessments (6) for recipients identified as high risk.

Given the relationship between comorbid disorders and acute suicide risk, improvements in the detection and documentation of mental illness among persons with chronic general medical conditions may be necessary—particularly in primary care, where mental health and suicide screenings typically do not occur (17–19). Individuals with chronic general medical conditions may not perceive a need for mental health services, may have a negative view of such services, or may lack encouragement from family or friends (14,20). Further justification for these improvements comes from discrepancies between autopsy studies, which have estimated that 90% of those who die by suicide have a mental disorder (21,22), and the absence of psychiatric disorders in case records of suicide decedents.

Slightly less than one-fifth (17%) of the decedents in the sample did not use any services in the year before suicide. Outreach and intervention approaches must be tailored accordingly. For example, the nonutilizers in this study were more likely to reside in rural areas and to have discontinuous Medicaid enrollment. Automated reminders of the importance of annual health care visits or mobile apps to help individuals identify providers who are accepting new patients or Medicaid patients could facilitate care, especially for rural residents. Because of shortages of primary and behavioral health care providers in rural areas, investments in mobile clinics and incentives for providers to offer telehealth services should be considered (23). Finally, disenrollment notifications should provide contact information for free or low-cost resources, such as suicide hotline numbers.

This study had several limitations. First, as with all studies that use data from death certificates, it is possible that the number of suicides was underestimated. Second, data were from a single state’s Medicaid program, and results may not apply to other state Medicaid programs given the difference in services and reimbursement options. Also, findings may not be applicable to decedents with other payment sources, such as Medicare and private insurance. Nevertheless, findings about the timing and type of care mirror those of other studies, suggesting broad generalizability. At the same time, the findings reinforced the fact that variations in care on the basis of individual characteristics are likely to differ by payment source. Because of the unique role of Medicaid in the United States, replication of this study with Medicaid data from states that have different treatment landscapes and Medicaid reimbursement levels is warranted. Finally, the study did not include decedents enrolled in Medicaid since 2014, when Ohio expanded Medicaid eligibility under the Affordable Care Act. Medicaid eligibility and service reimbursement rules are in a state of flux. The impact of changes in eligibility and service reimbursement on suicide should be examined in a future study.

Conclusions

This study provided detailed information on the timing and type of health encounters of adult suicide decedents in Ohio Medicaid and a profile of their psychiatric, substance use, and general medical conditions. Findings highlight the substantial public health significance of using Medicaid enrollment to designate a “virtual boundary” around a subpopulation of U.S. health care consumers for purposes of targeted suicide prevention and intervention.