Each year in the United States, roughly two-thirds of a million patients present to emergency departments for the treatment of deliberate self-harm (
1). Because patients who present after a deliberate self-harm event, which may involve varying degrees of suicidal intent, are as a group at exceedingly high risk of repeat deliberate self-harm (
2,
3) and suicide (
4,
5), their clinical management is of considerable public health importance. A substantial proportion of publicly (62.7%) and privately (46.9%) insured adult emergency department self-harm patients are discharged to the community directly from the emergency department (
6).
Patients who present to emergency departments after deliberate self-harm events may benefit from mental health evaluations. Three observational studies conducted in the United Kingdom (
7–
9) have compared the short-term risk of repeat self-harm patients who did and did not receive psychosocial assessments. Among a cohort with deliberate self-poisoning, 10% who received a psychosocial assessment and 18% who did not receive one poisoned themselves again within 12 weeks (
7). Among 246 deliberate self-harm patients discharged directly from the emergency department, 37.5% of the nonassessed patients and 18.2% of assessed patients engaged in subsequent self-harm over the following year (
8). A six-hospital observational study, however, found no overall significant association between psychosocial assessment and self-harm repetition (
9).
National clinical practice guidelines in the United Kingdom recommend providing all emergency self-harm patients with a psychosocial assessment, including a thorough evaluation of the social, psychological, and motivational factors specific to the self-harm event and an assessment of mental health and social risks and needs (
10). Although comparable national guidelines do not currently exist in the United States, the recently released
National Strategy for Suicide Prevention (
11) recommends standardized emergency department protocols for patients at high risk of suicide, continuity of care of patients treated for suicide risk in emergency departments, and collaborations between emergency departments and other health care professionals to provide appropriate alternatives to hospital admission and promote rapid follow-up.
In the United States, there is substantial variation in the availability and quality of mental health services in emergency departments. In a statewide survey in California, 34.0% of emergency departments reported having on-call access to a psychiatrist and only 12.8% reported having access to a psychologist (
12). Despite the near-universal occurrence of mental disorders among individuals who self-harm (
13,
14), a recent national study (
15) found that only about one-half (47.5%) of adult Medicaid self-harm patients who are discharged to the community receive a mental disorder diagnosis or a psychological assessment in the emergency department. The naturally occurring variation in mental health service availability in hospital emergency departments provides an opportunity to compare the outcomes of self-harm patients who are and are not recognized in the emergency department as having a mental disorder.
Method
Data Source
The source of data was the 2005 Medicaid Analytic Extract files from all 50 states and the District of Columbia, which were obtained from the Centers for Medicare and Medicaid Services. Medicaid is a state and federal entitlement program that pays for medical care of approximately 49 million individuals with low incomes and limited resources. Within federal guidelines, each state establishes eligibility standards. Although national Medicaid data include large numbers of vulnerable individuals with detailed longitudinal information on a full continuum of care, data are collected from clinicians and health care organizations rather than patients and coded for administrative rather than clinical or research purposes. As a result, raw claims data must be organized into clinically meaningful diagnoses and treatment variables.
The study was reviewed and determined to be exempt from human subjects review by the New York State Psychiatric Institute Institutional Review Board.
Sample Selection
Data on individuals 21–64 years of age were examined through Medicaid medical claims for a 60-day period before each emergency department visit and a 30-day period after emergency department discharge. Emergency visits were selected for intentional self-injury codes (ICD-9-CM codes E950–E958) in any position on the claim, excluding late effects of deliberate self-harm (code E959). Visits were included only if the patient was continuously eligible for Medicaid services for the 60 days before emergency department admission and the 30 days after emergency department discharge. There was no requirement for mental health contact before the index emergency department visit. The earliest index emergency department visit was March 3, 2005, and the latest was December 1, 2005. Recording external-cause-of-injury codes (E-codes) is mandatory in about half of the states (
17). Because our analysis focused on patients who were discharged from the emergency department to the community, we excluded patients who were discharged to an inpatient hospital or who resided in another institutional residential setting. These other settings included prisons and correctional facilities, assisted living facilities, group homes, nursing and custodial care facilities, hospices, and residential care facilities.
Dependent Variables
The two dependent variables were repeat visits for deliberate self-harm and admission for inpatient mental health treatment during the 30 days following the emergency visit. Repeat deliberate self-harm, which may or may not have been fatal, was defined as any Medicaid-reimbursed health service billed for intentional self-injury (codes E950–E958) in any position on the claim during the 30 days following emergency department discharge. Inpatient psychiatric treatment was defined as admission for inpatient treatment during the 30 days following emergency department discharge in which the first listed inpatient discharge diagnosis was a mental disorder (codes 290–319).
Independent Variables
Mental disorder recognition in the emergency department was an independent variable of interest. It was defined by the presence of a diagnosis of a mental disorder (codes 290–319) in any position on the claim during the emergency department visit (
15).
Other independent variables included patient age at emergency department discharge (21–34, 35–44, 45–64 years), sex, race/ethnicity, and Medicaid eligibility (poverty related or disability related). On the basis of claims during the 60-day period preceding the index emergency department visit, visits were also classified with respect to the presence of one or more outpatient visits, inpatient episodes, emergency department visits, or any other health care contact in which a mental disorder (codes 290–319) was diagnosed during the preceding 60 days. Emergency department treatment episodes were further classified by recently diagnosed mental disorders, defined by one or more claims with a diagnosis of a depressive disorder (codes 296.2, 296.3, 298.0, 300.4, 311), bipolar disorder (codes 296.0, 296.4, 296.5, 296.6, 296.7, 296.8), an anxiety disorder (codes 300.0, 300.2, 300.3, 293.84, 300.83, 309.81), an adjustment disorder (codes 308.3, 309.0, 309.1, 309.2, 309.4, 309.9), schizophrenia and related disorders (codes 295, 297, 298, 299), a substance use diorder (codes 291, 292, 303, 304, 305), a personality disorder (code 301), or other mental disorders (codes 290–319 not otherwise classified) during the 60-day period preceding the index emergency visit. A diagnosis of borderline personality disorder (code 301.83) was treated as a distinct subgroup of personality disorder.
The index emergency department visit for deliberate self-harm was classified on the basis of E-codes by method: violent methods (firearm, drowning, suffocation, fall, fire, electrocution, extreme cold, and motor vehicle), nonviolent (cutting and poisoning), and a residual group of unspecified or poorly specified methods (
18).
Analytic Plan
The percentages of emergency self-harm visits that were followed by a repeat deliberate self-harm visit within 30 days were determined overall and stratified by patient-level characteristics. Similar sets of analyses were performed with inpatient psychiatric care within 30 days of the emergency visit as the dependent variable. For each independent variable in both sets of analyses, unadjusted risk ratios were calculated with log-binomial regressions using the SAS GENMOD procedure (SAS Institute, Cary, N.C.). In adjusted models, each independent variable of interest was first forced into each model, and all covariates were then stepped in with specified entry (p<0.05) and retention (p<0.05) criteria. The number of retained control variables varied across the models depending on whether they improved model fit, with a theoretical range from one to 26.
The percentages of initial self-harm visits with and without mental disorder recognition in the emergency department that were followed by repeat deliberate self-harm visits or a psychiatric hospital admission were also stratified by each of the recent mental disorder diagnosis groups. To help balance the diagnostic groups that did and did not include mental disorder recognition in the emergency department, a logistic regression model was fitted that estimated the probability of mental disorder recognition given each of the demographic and clinical covariates and extracted the predicted probabilities for each observation as a decile propensity score. In the multivariate models that were stratified by recent diagnosis of various mental disorders, the emergency department mental disorder recognition variable and the propensity score variable were first forced into each model, and all covariates were then stepped in with the same entry (p<0.05) and retention (p<0.05) criteria.
A total of 5,567 self-harm emergency visits from 4,866 individuals met all eligibility criteria. Because the self-harm visits are nonindependent, generalized estimating equations were used to adjust the confidence intervals to accommodate clustering of observations within individual subjects.
Discussion
In the emergency care of adult Medicaid self-harm patients, recognition of a mental disorder is related to a lower short-term risk of repeat self-harm. Recognizing mental disorders in the emergency department had a particularly strong protective association with subsequent self-harm visits among self-harm patients who had not recently received mental health care. Significant protective correlations were also observed among deliberate self-harm patients who had recently been treated for depression, bipolar, or substance use disorder. These results support programmatic reforms to expand access to mental health evaluations for patients who are treated in general hospital emergency departments after an episode of deliberate self-harm.
Our findings extend earlier U.K. observations on the protective effects of psychosocial assessments for patients after an episode of deliberate self-harm (
7,
8) to the U.S. national population of publicly insured low-income and disabled adults. In contrast to the U.K. research, which involved psychosocial assessments as defined by the Royal College of Psychiatrists, we focused on mental disorder diagnoses in the emergency department claims record. A mental disorder diagnosis does not necessarily imply that a thorough mental health assessment occurred. More detailed practice-based research is needed to assess which specific aspects of emergency department mental health evaluation, management, and referral activities are associated with reducing the risk of repeat deliberate self-harm. Randomized controlled trials of well-defined interventions will then be required to test these emergency mental health interventions against routine care.
We found that only about one-third of the deliberate self-harm emergency department patients in our sample had not recently received mental health care. In this group, recognition of mental disorders appeared to have an especially strong protective effect on the risk of repeat self-harm. Recognition of mental disorders may be particularly important for discharged self-harm patients who lack a usual source of outpatient mental health care. Emergency self-harm patients who do not have existing relationships with outpatient mental health clinicians are at high risk for not receiving follow-up outpatient mental health care (
15). For these individuals, mental health care delivered in the emergency department may offer opportunities to initiate mental health services that can help prevent future crises (
16).
Patients who had recently received a diagnosis of a personality disorder, especially borderline personality disorder, were the highest risk group for repeat deliberate self-harm. Recurrent suicidal behavior is one of the defining characteristics of borderline personality disorder. Among emergency department patients who present after a suicide attempt, borderline personality disorder is strongly associated with a history of multiple suicide attempts (
19). Because of the challenging clinical nature of deliberate self-harm in individuals with personality disorders, it is perhaps not surprising that among patients with a recent diagnosis of borderline personality disorder, recognition of a mental disorder in the emergency department was not associated with a lower risk of repeat self-harm events. Far more intensive psychosocial interventions, including a 52-week course of dialectical behavioral therapy (
20) and cognitive-behavioral therapy (
21), have achieved sustained reductions of deliberate self-harm in adults with borderline personality disorder. In contrast to a recent diagnosis of borderline personality disorder, a recent self-harm visit was not significantly related to risk of a repeat self-harm visit shortly after the index visit.
Hispanic patients with self-harm were significantly less likely than their white counterparts to receive services for repeat deliberate self-harm during the follow-up period. Without more detailed clinical information, it is not possible to determine the sources of this ethnic difference in service use. Cultural attitudes and beliefs, including moral objections to suicide, greater responsibility to the family, and survival and coping beliefs (
22), may help to protect some Hispanic patients from repeat deliberate self-harm (
23). Because Hispanic individuals who attempt suicide may be less likely than non-Hispanics to seek treatment after a suicide attempt (
24), however, it is also possible that ethnic differences in help seeking contribute to the relatively low proportion of Hispanic self-harm patients who received care for repeat deliberate self-harm. In this regard, Hispanic self-harm patients have been found to be less likely than non-Hispanic self-harm patients to receive any follow-up mental health care after a self-harm event (
15).
There was a smaller, though significant, protective association between recognition of a mental disorder in the emergency department and short-term risk of psychiatric hospital admission. In one of the previously mentioned U.K. studies, a similar but statistically nonsignificant trend was reported toward lower psychiatric hospital admission rates among self-harm patients who received emergency mental health assessments (
8). In the present study, a significant protective correlation with psychiatric admission was limited to patients who had not received mental health care in the 60 days preceding the index the self-harm visit. It is possible that the protective effects of recognizing mental health problems during emergency evaluations are partially mediated by referrals for outpatient mental health care, which have a greater impact on outcome among patients who do not have established connections with outpatient mental health services.
Our findings provide empirical support for calls to increase access to mental health services in the emergency department management of deliberate self-harm. Some potential strategies for improving access to mental health evaluations in emergency departments include training emergency staff in assessment and acute management of self-harm (
25), developing liaison services with mental health specialists (
26), and integrating mental health teams into the emergency department service (
27).
This study had several limitations. First, because of the naturalistic design, causal inferences cannot be established between mental disorder recognition and the outcomes. Second, data were not available concerning many factors that may influence the quality of the emergency mental health assessments. For example, no measures were available concerning emergency department staffing, use of standardized suicide risk assessment tools (
28), referral activities, or whether patients left before the completion of the emergency department evaluation (
29). Third, concerns have been expressed about the validity (
30) and completeness (
31) of E-codes to measure deliberate self-harm, and some self-harm events are not brought to medical attention (
32), although high correlations have been reported between deliberate self-harm E-codes and medical record confirmation of attempted suicide (
33,
34). Fourth, diagnoses were based on clinician judgment. Without standardized psychiatric assessments, the validity of the diagnostic categories remains unknown, and patients who did not receive health care during the 60 days preceding the emergency visit had no opportunity to receive a recent mental disorder diagnosis. Fifth, as previously mentioned, the mere presence of a mental disorder diagnosis in the billing record does not necessarily denote a thorough mental health assessment or provision of substantial mental health care (
35). Self-harm patients with recognized mental disorders may present with more prominent psychopathology than patients who were not so recognized in the emergency department. Such a bias, if it were related to higher risk of adverse outcomes, would diminish the observed protective effects of mental disorder recognition on the adverse outcomes. Finally, the analyses were limited to Medicaid recipients and may not generalize to privately insured patient populations (
6).
Deliberate self-harm carries a high risk for future deliberate self-harm and suicide, especially in the near term (
3–
5). Moreover, approximately 20%−25% of adults who complete suicide have visited an emergency department for deliberate self-harm in the preceding year (
4,
36,
37). For these reasons, even incremental declines in suicide risk following a deliberate self-harm event might substantially diminish the overall number of suicides. Immediately after a deliberate self-harm event, emergency mental health evaluations may provide opportunities to evaluate aggression, impulsivity, hopelessness, and other symptoms that bear on suicide risk (
38) and to assess the need for ongoing mental health care. In this context, mental health evaluations that result in the diagnosis of a mental disorder may help reduce the short-term risk of repeat self-harm and psychiatric hospital admission.