Over the past two decades, suicide rates have risen steadily across the United States. From 1999 to 2017, the average annual percentage increase in the national suicide rate was between 1% and 2% (
1). Over the same period, there was a corresponding rise in overdoses resulting from prescription and illicit opioid abuse. More than 700,000 people have died from a drug overdose, and nearly 400,000 have died from an opioid-involved overdose, since the start of the opioid crisis (
2).
Deliberate, or suicide, overdoses have accounted for far fewer U.S. deaths than accidental overdoses. Between 2005 and 2016, the rate of suicide from deliberate overdose ranged from 1.80 to 1.93 per 100,000 people, whereas the mortality rate from accidental overdoses ranged from 9.61 to 21.73 per 100,000 people (
3). However, the actual proportion of suicides among U.S. overdose deaths has possibly been underestimated, given that many overdose suicides are misclassified as undetermined or accidental deaths (
4). A recent report suggested that 20%−30% or more of opioid overdose deaths may be suicides (
5).
Emergency departments (EDs) are at the forefront of managing and treating people affected by overdose and suicide. By providing urgent treatment, evaluating patients for suicide risk, and linking patients with postdischarge treatment and support services, EDs are important settings for suicide prevention. This is especially true for patients admitted for deliberate self-harm because they are at high risk for repeat attempts and eventual suicide (
6). Because suicide risk is elevated after discharge from EDs and inpatient hospital settings (
7,
8), the Joint Commission has recommended that hospitals integrate suicide prevention and treatment strategies, such as mental health assessments, discharge planning, and follow-up contacts, into all care settings (
9).
Previous studies have documented racial-ethnic and insurance status differences in receipt of mental health services among patients admitted to the ED for deliberate self-harm. A 2012 study reported that Hispanics were less likely than whites to receive an emergency mental health assessment and that Hispanics and African Americans were less likely than whites to receive follow-up outpatient care in the 30 days after discharge (
10). A similar study reported that privately insured patients were more likely than Medicaid-insured patients to receive a mental health assessment (
11). We are unaware of any studies conducted in an inpatient setting that have evaluated hospital-based mental health services received by patients after admission for deliberate self-harm. Given that sociodemographic factors should not affect access to mental health care, more research is needed to understand which factors may influence receipt of inpatient mental health services by patients who deliberately harm themselves.
In this study, we examined patient and hospitalization characteristics associated with receiving a mental health assessment and disposition to an inpatient psychiatric facility among patients hospitalized for deliberate drug overdose. Our primary objective was to assess potential differences in receipt of inpatient mental health services on the basis of race-ethnicity and insurance status. Deliberate drug overdose is the most common method of self-harm in the United States (
12) and a strong predictor of future completed suicide (
13–
15). Therefore, it is important that all patients hospitalized for a deliberate drug overdose have access to inpatient mental health services as needed. To eliminate potential disparities in care, a better understanding of the effects of race-ethnicity and insurance status on access to such services is required. We hypothesized that non-Hispanic white race-ethnicity and having insurance would be predictive of receiving a mental health assessment and disposition to an inpatient psychiatric facility.
Methods
Data Source
We used 2012–2013 data from the South Carolina All-Payer Healthcare Databases, which have been used in previous studies to examine health care utilization patterns (
16–
18). These databases capture all ED, inpatient, outpatient surgery, and health clinic visits to licensed, nonfederal health care facilities in South Carolina. Data elements include encounter-level demographic, clinical, and nonclinical information and a unique patient identifier that allows patients to be tracked over time and across various care settings. Internal audits of the databases are routinely performed to ensure that data elements are 99% complete and 99.5% accurate, as required by law (
19). The all-payer databases used in this study are exempt from review by the Clemson University institutional review board.
Analytic Sample
To obtain the analytic sample, we used inpatient database records to identify hospitalizations for deliberate drug overdose and linked to ED records to obtain data on recent mental health services. Adults ages 18–64 who had at least one inpatient admission for deliberate illicit or pharmaceutical drug overdose were followed from 60 days before hospitalization through discharge. We included drug overdoses if the primary diagnosis field of each record contained an
ICD-9-CM diagnosis code for “poisonings by drugs, medicinal or biological substances” (960–979) (
20). Overdoses were classified as deliberate if the supplementary classification of external causes of injury and poisoning code (E-code) indicated suicide and self-inflicted poisoning (E950.0–E950.5). Patients who left against medical advice (N=60) or died in the hospital (N=24) were excluded from the study. Because of our interest in assessing inpatient mental health services, we further excluded patients discharged to correctional (N=53) or post–acute care facilities (N=126). Finally, we excluded encounters in January and February 2012 (N=277), given that assessment of receipt of 60-day prior mental health services for these patients was not possible.
Outcomes
Study outcomes included receipt of a mental health assessment and disposition to an inpatient psychiatric facility. We defined the first outcome, mental health assessment, using a general approach by Olfson et al. (
10): presence of a psychiatric diagnosis (
ICD-9-CM diagnosis codes 290–319) in any secondary diagnosis field; evidence of psychological or psychiatric interviews, consultations, evaluations, or testing on the basis of
ICD-9-CM procedure codes in primary or secondary procedure fields (94.0–94.19); or indication of a visit from a psychiatrist before hospital discharge (
10,
20). We defined the second outcome, disposition to inpatient psychiatric facility, as disposition to home or disposition to an inpatient psychiatric facility.
Independent Variables
Independent variables were patient and hospitalization characteristics. Patient characteristics included age in years (18–24, 25–34, 35–44, 45–54, and 55–64), sex, race-ethnicity (non-Hispanic white, non-Hispanic black, and other), and primary insurance payer (uninsured, Medicare, Medicaid, private, and other payer). Hospitalization characteristics included drug ingested before admission, length of stay, and admission or discharge day of week (weekend versus weekday). Drug ingested before admission was categorized into opioids or sedative-hypnotics (
ICD-9-CM diagnosis codes 965.00–965.09, 967, 969.4); drugs with high abuse or misuse liability, including central nervous system stimulants, psychotropic agents, and other central nervous system depressants (968, 969.0–969.3, 969.5–969.9, 970); and other drugs, medicinals, or biologicals (960–964, 965.1–965.9, 966, 971–979). We categorized opioids and sedative-hypnotics together because opioids and benzodiazepines increase risk of respiratory depression, and concomitant use is associated with high risk of overdose death (
21).
Mental Health Services
To account for the potential confounding effects of receipt of recent mental health services on study outcomes, we examined deliberate drug overdoses, psychiatric hospitalizations, and treatment for mental disorders in the 60 days before hospitalization. Deliberate drug overdoses were defined by using the criteria described earlier. Psychiatric hospitalizations were defined as disposition to an inpatient psychiatric facility. Treatment for mental disorder included any encounter with an ICD-9-CM diagnosis code (290–319) in the primary or a secondary diagnosis field. Mental disorders were categorized as schizophrenia and other psychoses (295, 297–298), bipolar disorder (296.00–296.16, 296.40–296.89), depressive disorders (296.20–296.36, 300.4, 311), anxiety disorder (293.84, 300.0, 300.2, 300.3, 308.3, 309.21, 309.81), personality disorder (301), alcohol and drug use disorders (291–292, 303–305), adjustment disorder (309.0, 309.1, 309.22–309.4, 309.82–309.9), and other disorders (defined as any codes between 290 and 319 not listed earlier).
Statistical Analysis
We report descriptive statistics of the sample, stratified by disposition. In bivariable analysis, we compared differences in disposition across characteristics and mental health services variables using chi-square and Mann
-Whitney U tests. Generalized linear models with a binomial distribution and logit link were used to examine the relationship between patient and hospitalization characteristics and receipt of mental health assessment and disposition to an inpatient psychiatric facility. Models were adjusted for independent variables, recent mental health services, and hospital fixed effects, which accounted for hospital differences, such as hospital type and region, known to influence disposition to an inpatient psychiatric facility (
22).
For the second study outcome, we further adjusted the model for presence of a psychiatric diagnosis during hospitalization. Patients could contribute more than one hospitalization for deliberate drug overdose to the sample. Therefore, we obtained adjusted odds ratios (AORs) and 95% confidence intervals (CIs) by using generalized estimating equations to account for clustering due to repeated admissions of the same patient. In sensitivity analyses, we examined whether factors associated with receiving a mental health assessment and disposition to an inpatient psychiatric facility were modified when including only one deliberate drug overdose admission per patient. Statistical analyses were performed using SAS version 9.4. Statistical tests were two-sided with p<.05 indicating statistical significance.
Results
From 2012 to 2013, 2,489 unique patients had 2,686 encounters for deliberate drug overdose in South Carolina. The largest proportion of deliberate drug overdoses were those categorized as other drugs or medicinal or biological substances (N=1,078; 40%), such as anticonvulsants, antibiotics, and antiallergic and antiemetic drugs, followed by opioids or sedative-hypnotics (N=987; 37%), and then drugs with high abuse or misuse liability (N=621; 23%). Most persons in the sample were non-Hispanic white (N=2,250; 84%) and had insurance (N=1,715; 64%). Approximately half the sample was discharged to home (51%); the other half was discharged to an inpatient psychiatric facility (49%;
Table 1). A lower percentage of non-Hispanic whites than non-Hispanic blacks and people of other races-ethnicities were discharged to home (48% versus 61% and 52%, respectively). Persons without insurance were more likely than persons with insurance to be discharged home (63% versus 38% with Medicare, 57% with Medicaid, 39% privately insured, and 24% with other insurance). Notably, patients with Medicaid were more likely to be discharged home than patients with other types of insurance.
Approximately 94% (N=2,514) of patients received a mental health assessment during hospitalization. Non-Hispanic blacks and people of other races-ethnicities had lower odds than non-Hispanic whites of receiving a mental health assessment (non-Hispanic blacks, AOR=0.52; other races, AOR=0.24;
Table 2). Patients hospitalized for overdose from drugs with high abuse or misuse liability had higher odds of receiving a mental health assessment than patients admitted for overdose from opioids or sedative-hypnotics (AOR=1.69). Results were similar in sensitivity analyses that included one admission per patient (a table comparing AORs from the main analyses and sensitivity analyses is available as an
online supplement to this article).
Table 3 presents AORs of the association between patient and hospitalization characteristics and disposition to an inpatient psychiatric facility. Compared with persons ages 18–24 years, those ages 45–54 years were less likely to be discharged to an inpatient psychiatric facility (AOR=0.66). Non-Hispanic blacks had lower odds of disposition to an inpatient psychiatric facility than non-Hispanic whites (AOR=0.60). Insured patients, except those covered by Medicaid, had increased odds of disposition to an inpatient psychiatric facility compared with patients without insurance (Medicare, AOR=3.06; private, AOR=2.78; other, AOR=7.58). In addition, longer length of stay was associated with lower odds of disposition to an inpatient psychiatric facility than to home (AOR=0.97). Results were similar in sensitivity analyses that included one admission per patient (see
online supplement).
Discussion
Most patients admitted to the hospital for a deliberate drug overdose received a mental health assessment (94%). This assessment rate was higher than the mental health assessment rates of around 50%–70% reported in studies of patients admitted to the ED for self-harm (
10,
23). We did not find other studies reporting rates of mental health assessments among patients hospitalized for deliberate drug overdose. Nearly half of the sample were discharged to an inpatient psychiatric facility, and the other half were discharged to home. A higher percentage of patients who received recent mental health services compared with those who did not were discharged to an inpatient psychiatric facility. It is important to note that disposition to an inpatient psychiatric facility does not indicate better quality of care, nor does it reduce the risk of future suicide. Individuals are at high risk of suicide in the period immediately after discharge from an inpatient psychiatric facility and for many years after (
7). Nevertheless, inpatient psychiatric facilities offer a safe, controlled environment for patients in need of stabilization from an acute mental health crisis, and they play a central role in the United States for treating patients at risk of suicide (
8).
Of all the characteristics examined, race-ethnicity had the strongest magnitude of association with receiving a mental health assessment and disposition to an inpatient psychiatric facility. Both non-Hispanic blacks and people of other racial-ethnic groups were less likely than non-Hispanic whites to receive a mental health assessment. Only non-Hispanic blacks were less likely than non-Hispanic whites to be discharged to an inpatient psychiatric facility. It is unclear whether the disparities we observed are due to implicit provider bias, cultural differences in patient preferences for care, socioeconomic differences between groups, or other clinically relevant factors. Research conducted in ED settings has found that psychiatric admission decisions vary across institutions (
24–
26) and may be influenced by predominantly clinical factors (
27,
28) or both clinical and nonclinical factors (
29,
30). A recent study with ED clinicians reported that clinical severity and social factors, such as homelessness or patient unwillingness to engage with outpatient services, influenced disposition planning to an inpatient psychiatric facility among patients boarding in the ED for suicide risk (
29).
Our findings support well-documented evidence that racial-ethnic disparities in mental health care exist the United States (
31,
32). African Americans and Latinos have either lower or equivalent rates of mental disorders than whites but are less likely to receive needed mental health services, treatment, or medications (
32). Past research has documented that blacks are overrepresented in inpatient psychiatric units (
33). The results of our study contradict this evidence. In the southern United States, disparities in access to mental health services and treatments have persisted over time for African Americans (
34,
35), which could explain this discrepancy. It could also explain why our results differed from those reported by Olfson and colleagues (
10). That study found differences between Hispanics and whites, but not between African Americans and whites, in the likelihood of receiving an emergency mental health assessment (
10). Because our data contained a small sample of Latinos, we were unable to examine differences in the outcomes specifically for that group. Further studies with larger sample sizes are needed for such an investigation.
Patients with Medicare, private, or other insurance were more likely than patients without insurance to be discharged to an inpatient psychiatric facility. This result was expected, given that shortages of inpatient psychiatric beds, coupled with the high cost of acute psychiatric hospitalization, have made it difficult even for insured persons to access inpatient mental health services (
36,
37). Follow-up contacts are one strategy for reducing suicide risk. A review of 11 studies conducted in ED and inpatient settings reported that postdischarge follow-up contacts may protect against suicidal behaviors (
38). Three of the reviewed studies found that postcard contacts were associated with a significant reduction in repeat deliberate drug overdose at 12 months (
39–
41). Follow-up contacts are less expensive than inpatient psychiatric care and may be used to improve access to treatment for patients who lack health coverage.
Patients ages 45–54 years were less likely to be discharged to an inpatient psychiatric facility than those ages 18–24 years. This is consistent with evidence that younger age is predictive of psychiatric hospitalization (
42). Although national suicide rates are highest among middle-aged adults (
1), young people have a high prevalence of suicide risk factors, such as risk-taking behaviors (
43) and mental disorders (
44). They are also at higher risk than older adults for nonsuicidal self-harm (
45), which is among the strongest predictors of future suicide attempts (
46). Any of these factors may influence disposition planning to an inpatient psychiatric facility for young adults and could account for the observed differences in disposition across age groups.
Our data show that patients hospitalized for an overdose of a drug with high abuse or misuse liability were more likely to receive a mental health assessment than those admitted for an overdose from opioids or sedative-hypnotics. Respiratory depression is the main feature of overdoses from opioids or a combination of opioids and sedative-hypnotics. Patients who overdose on these drugs generally do not present with psychological symptoms. In contrast, patients who overdose on drugs with high abuse or misuse liability, such as cocaine or methamphetamine, often exhibit altered mental status, such as agitation or acute psychosis (
47,
48). The differences in clinical features of patients by the drug or drugs ingested before hospitalization might explain our finding. In light of connections between opioid use and suicide (
49), all patients admitted to the hospital for an overdose involving opioids or sedative-hypnotics should receive a mental health assessment.
There are several limitations of this analysis. First, we could not determine whether deliberate drug overdoses were completed with the intention to attempt suicide, nonsuicidal self-harm, or neither. Second, all-payer data are limited and contain no information about certain clinical or nonclinical factors, such as clinical severity or patient preference for treatment location, that may influence study outcomes. Third, although administrative databases can be used with a strong degree of accuracy to reliably identify drugs involved in overdoses (
50–
52), clinical staff are subject to misclassifying drugs when comprehensive drug screening is not available (
53). Fourth, by selecting the study sample solely on the basis of primary diagnosis field, we may have underestimated cases of drug overdose. Moreover, we used E-codes to determine intentionality of overdose, which may have resulted in further underestimation of deliberate drug overdoses as a result of varying rates of E-code completeness across all-payer claims databases (
54). At the same time, our approach, coupled with all-payer data, strengthened the validity of our findings. Fifth, by excluding dispositions other than to home or an inpatient psychiatric facility, we may have missed patients who received a mental health assessment before hospital discharge. Finally, consistent with previous studies, we measured recent mental health services by using a 60-day review period (
10,
11). Yet, we acknowledge the limitation that longer-term mental health and medical history could influence study outcomes.