Suicide After Deliberate Self-Harm in Adolescents and Young Adults

A self-harm cohort was extracted from 2001 to 2007 national (45 states, not including AZ, DE, NV, OR, or RI) Medicaid Analytic Extract data, which includes claims records for Medicaid-financed services and medications. Dates and cause of death information were derived from linkage to the National Death Index, which is the most complete resource available for tracing mortality in national samples.²⁶ US population death rates in 2001–2007 were derived from the National Center for Health Statistics.²⁵ This project was approved by the Rutgers University Institutional Review Board.

The cohort was restricted to individuals 12 to 24 years of age with clinical diagnoses of deliberate self-harm (International Classification of Diseases, Ninth Revision, Clinical Modification codes E950–E958). Patients were also required to be eligible for Medicaid services during the 180 days preceding their index self-harm event.

The first eligible instance of self-harm was selected, and no patient contributed more than 1 observation. The cohort was followed forward from their index self-harm until the end of a 365-day follow-up period, date of death, or end of available data, whichever came first. For self-harms treated in outpatient or emergency department settings, the index date was the date of self-harm treatment. For self-harms treated in inpatient settings, the index date was the hospital discharge date.²⁷ 

Following the Centers for Disease Control and Prevention, suicide was defined as International Classification of Diseases, 10th Revision codes X60–X84, Y87.0, or U03 as the immediate cause of death.²⁸ The primary outcomes of interest were repeat, nonfatal self-harm and suicide deaths. Suicide deaths coinciding with the index self-harm were not included in the analyses. Secondary outcomes included all causes of death and death due to accidents (V01–X59, Y85–Y86).

On the basis of Medicaid eligibility data, cohort members were classified by age at time of the index self-harm as adolescents (12–17 years) or young adults (18–24 years), sex, and race and/or ethnicity: white, non-Hispanic; African American, non-Hispanic; Hispanic; American Indian or Alaskan native; and “other” including Asian Americans, Native Hawaiians or other Pacific Islanders, and more than 1 race. Cohort members were also classified by geographic region (west, midwest, south, northeast), and Medicaid eligibility status (disability, foster care, low income, or other).

Variables representing clinical diagnoses on or within 180 days before the index self-harm were defined for ADHD, disruptive behavior disorders, depressive disorders, bipolar disorder, anxiety disorders, schizophrenia and related psychoses, substance use disorders, personality disorders, and a residual group of other mental disorders (Supplemental Table 6). At least 1 inpatient or 2 outpatient diagnoses defined each mental disorder group.²⁹ In analysis of these diagnostic groups, comparators were self-harm patients without the disorder of interest. A separate variable defined the number of mental health diagnostic groups (0, 1, 2, 3+). Separate variables defined any inpatient and outpatient mental health care and any outpatient care during the 180 days preceding the self-harm event.

Place of service codes were used to classify the index self-harm treatment setting hierarchically as (1) inpatient, (2) emergency department, or (3) outpatient for those receiving care in >1 setting. Index self-harm method was classified as violent, nonviolent, and other methods with external cause of injury codes (E-codes). Nonviolent methods were subclassified as cutting, poisoning, or other methods. Violent methods included firearms and other violent methods (drowning, hanging and suffocation, jumping from heights, motor vehicle or aircraft crashing, fire and scalding, and extreme cold) (Supplemental Table 6).^30,31 

Adolescents and young adults who were treated for nonfatal self-harm were compared with respect to demographic and clinical characteristics. For each demographic and clinical variable stratum, we determined the number of patients, person years of follow-up, number with ≥1 repeat self-harm events, and first repeat self-harm rates per 1000 person-years. Corresponding stratified rates of suicide risk per 100 000 person-years of follow-up were determined. For self-harm patients who survived the initial event, Cox proportional hazard models, which make no assumptions about the baseline hazard, estimated unadjusted hazard ratios (HRs) of repeat self-harm during follow-up with each stratification variable as the independent variable of interest. Adjusted Cox models controlled for age, sex, race and/or ethnicity, geographic region, and Medicaid eligibility status. Separate models were fit for adolescents and young adults. Models also included an interaction term (stratification variable by age group) to assess differences in repeat self-harm risk across the 2 age groups.

For comparison of suicide rates of the self-harm cohort with US population suicide rates, we determined the matched age group, sex, and race and/or ethnicity in the US population annual mean suicide rates for 2001–2007. All statistical analyses were performed with SAS 9.4 (SAS Institute, Inc, Cary, NC).

A total of 32 437 self-harm events were observed. Forty-two (0.13%) were fatal (suicide), including 20 adolescents and 22 young adults, and excluded from the prospective cohort (n = 32 395).

Most patients with nonfatal initial self-harm were white, adolescents, female subjects, and their events involved poisoning (Table 1). Depression and substance use disorders were the most commonly diagnosed mental disorders within the 180 days preceding self-harm. Approximately half of the patients with self-harm were treated in emergency departments, whereas 22.5% involved inpatient admission. As compared with adolescent self-harm patients, young adult self-harm patients were more likely to have been diagnosed with substance use disorders, personality disorders, or schizophrenia and related psychoses and less likely to have been diagnosed with ADHD, other disruptive behavioral disorders, or depression in the 180 days before self-harm. Approximately half of each age group received any outpatient mental health care during this period.

Approximately 17.1% (5545 of 32 395) of the self-harm patients had repeated nonfatal self-harm during the follow-up year for an unadjusted repeat self-harm rate of 222.7 per 1000 person-years (Table 2). The cumulative probability of first self-harm distributions were similar for adolescents and young adults, with both having a steep rise in the first few days after the initial event and then rising more gradually (Fig 1).

The unadjusted hazards of repeat nonfatal self-harm were higher for young adults than adolescents, female than male patients, and white than African American or Hispanic patients, although the age group difference was not evident in the adjusted model. Recent clinical diagnoses of each specific mental disorder other than ADHD were also associated with significantly increased hazards of repeated self-harm. The hazards also increased with increasing number of mental disorders. The risk of repeat self-harm was significantly higher for patients treated for personality or anxiety disorders than depressive disorders. Self-harm events that involved poisoning or that were treated in inpatient settings were also associated with increased risk of repeated self-harm.

There were few differences between adolescents and young adults in risk factors of repeated self-harm. However, female sex, Hispanic ethnicity, and treatment of the initial self-harm in an inpatient setting were more strongly associated with repeat self-harm for adolescents than for young adults (Table 3). By contrast, a recent diagnosis of disruptive behavior disorders or schizophrenia and the number of mental disorder diagnoses were more strongly associated with repeat self-harm for young adults than adolescents.

The 1-year suicide standardized mortality rate ratio (SMR) for the study cohort was 26.7 (95% confidence interval [CI]: 19.9–35.1) compared with the matched age, sex, or race and/or ethnicity in the US general population (Table 4). The corresponding suicide SMR was significantly higher for adolescents (46.0, 95% CI: 29.9–67.9) than for young adults (19.2, 95% CI: 12.7–28.0). The cumulative probability of suicide was similar for the 2 age groups during the first 111 days, after which the risk for young adults exceeded the risk for adolescents (Fig 2).

Unadjusted hazards of suicide after nonfatal self-harm were significantly higher for male subjects than female subjects, non-Hispanic white than African American patients, and patients whose initial self-harm involved violent rather than nonviolent methods (Table 5). After controlling for age and sex, the odds of suicide were more than 5 times higher for American Indian and Alaskan native self-harm patients than for white non-Hispanic self-harm patients.

The adjusted hazards of suicide after self-harm events with violent methods were significantly larger for adolescents (28.18, 95% CI: 9.98–80.36) than for young adults (7.44, 95% CI: 2.90–19.08) (interaction P value = .04). In a post hoc analysis, American Indian and Alaskan native self-harm patients were significantly more likely than their white non-Hispanic counterparts to have used violent methods during their initial self-harm event (7.36% vs 3.86%, P < .0001). Male self-harm patients were also more likely than female self-harm patients to use violent methods (7.88% vs 5.32%, P < .0001).

After self-harm, the 1-year SMR for all-cause mortality was significantly higher for young adults (10.3, 95% CI: 8.1–12.8) than for adolescents (4.1, 95% CI: 2.9–5.7) (Table 4). The standardized risk of accidental death was significantly elevated in the combined self-harm cohort in relation to matched controls (3.3, 95% CI: 2.2–4.8) and among the young adult cohort (5.0, 95% CI: 3.0–7.8) but not among the adolescent cohort (1.8, 95% CI: 0.8–3.6).

After nonfatal self-harm, adolescents and young adults in the Medicaid program were at 26.7 times higher risk of suicide than that of the demographically matched US general population. The standardized mortality rate ratio of suicide after self-harm was significantly higher for adolescents than young adults. Suicide risk was especially prominent for male subjects, individuals who were American Indians or Alaskan natives, and those whose initial self-harm involved violent methods, such as firearms or hanging. The link between violent initial self-harm method and subsequent suicide was stronger for adolescents than young adults. In the year after self-harm, young adults, but not adolescents, were also at significantly increased risk of dying of accidents.⁸ 

Consistent with past research, nonfatal self-harm was strongly associated with suicide during the first follow-up year.^7,8 In a UK self-harm cohort, suicide risk was 35 times higher than the population risk for male subjects and 75 times higher for female subjects,⁶ whereas in a Swedish self-harm cohort the corresponding suicide risks were 50.2 times higher for male subjects and 7.5 times higher for female subjects.⁷ One reason that our estimates of relative suicide risk tend to be lower than these estimates may be because our results included self-harm presenting to all settings, whereas the United Kingdom and Swedish studies involved only patients treated in hospitals who were presumably at higher risk.

Adolescents and young adults had roughly similar rates of suicide after self-harm. The diagnostic composition of the 2 age groups, including larger proportions of ADHD and disruptive behavior disorders among adolescents and substance use and personality disorders among young adults, likely reflects differences in ages of onset and clinical diagnostic practices. Because the general population risk of suicide is lower in adolescents than young adults, however, the standardized risk of suicide after self-harm was significantly higher for adolescents than young adults. It is possible that neurodevelopmental vulnerabilities during adolescence linked to the onset of mood and substance use disorders, risk-taking behaviors, or heightened susceptibility to negative social cues contribute to the relatively high suicide risk during adolescence following self-harm.^32,33 

Self-harm presented differently in adolescents and young adults. Although depression and anxiety diagnoses were common in both age groups, adolescents with self-harm were far more likely to have been recently diagnosed with ADHD and other disruptive behaviors, whereas young adults more commonly had substance use and personality disorder diagnoses. Similar age-related patterns have been reported among young people who received diagnostic assessments after suicide attempts.^34,–37 

The risk of repeated self-harm during the first follow-up year was over 100 times greater than the risk of suicide. High recurrence of self-harm underscores persistence of self-injurious behaviors in this age group. Consistent with previous research,³⁸ several clinical diagnoses including personality, anxiety, mood, substance use, schizophrenia, and disruptive behavior disorders were related to an elevated risk of repeated self-harm. The wide distribution of risk across common clinical psychiatric diagnoses complicates prediction of repeated attempts. Also in line with previous research,⁸ patients admitted for inpatient care were significantly more likely than outpatients to have repeated self-harm.

Suicide risk after self-harm was ∼4 times greater for male subjects than female subjects. Some potential explanations for the preponderance of suicide among male subjects include sex-related differences in mental health treatment–seeking behavior,³⁹ self-harm methods,⁴⁰ and pubertal development.⁴¹ Because the risk for repeated self-harm was greater for female subjects than male subjects, higher lethality of male than female self-harm events likely contributes to the preponderance of suicides by male subjects after self-harm.

Risks of suicide after self-harm were ∼5 times higher for white than African American young people, suggesting that self-harm poses a particular threat to white adolescents and young adults. By contrast, risk of repeat self-harm was only modestly higher for white than African American individuals. In nationally representative US samples, no significant differences have been found in the lifetime prevalence of suicide attempts of African American and white young adults.^42,43 However, as compared with their African American counterparts, white young adults may have a significantly higher prevalence of suicidal ideation⁴³ as well as mood, anxiety, and substance use disorders.⁴⁴ Greater exposure to these and other suicide risk factors may contribute to the relatively high suicide risk of white young people after self-harm.

Suicide risk after self-harm was particularly elevated for American Indian and Alaska Native young people, a group with increased suicide rates.⁴⁵ These results underscore the need for suicide prevention initiatives in American Indian and Alaska Native communities. In 1 recent report, a multitiered suicide prevention program among the Apache of Arizona, which included efforts to increase identification, coping skills, and resilience of at-risk young people, coincided with a 23% decrease in suicide among 15- to 24-year-olds.⁴⁶ 

In accordance with previous research with adults, violent self-harm methods posed a high risk for subsequent suicide.^8,27,47,48 Acquiring the capability to use more violent or lethal methods, although less common than lower lethality methods, may increase suicide risk by weakening a fear of death.⁴⁹ Suicide risk was especially great after initial nonfatal self-harm involving firearms. For high-risk young people with access to firearms, distributing trigger locks and urging family members to store firearms away from the patient’s home⁵⁰ can be a lifesaving precautionary measure. The association between violent methods and subsequent suicide was significantly stronger for adolescents than young adults, although this interaction was based on a small number of suicide deaths and therefore requires replication.

This study has several limitations. First, claims records do not distinguish self-harm injuries with suicidal from nonsuicidal intent.⁵¹ Variation in suicidal intent may mediate risks across clinical groups. For example, high suicidal intent might partially mediate the association between high lethality self-harm methods and suicide risk, whereas persistently low or no intent (nonsuicidal self-injury) might mediate the association between personality disorders and repeated self-harm. Second, we have no means of validating mental disorder diagnoses in Medicaid claims records. Third, information was not available concerning several factors thought to increase suicide risk, such as hopelessness, problem-solving deficits, stressful or traumatic life events, and access to lethal means. Because community mental treatment may be more intensive and effective after severe or violent self-harm events, it may have reduced risks of suicide and repeated self-harm events. Fourth, misdiagnosis and perceptions of stigma may have introduced errors in coding self-harm, and low autopsy rates may have resulted in misclassification of cause of death.⁵² Fifth, claims records do not capture self-harm that does not result in medical care. Sixth, the data reflect suicide mortality patterns in 2001–2007, which may have changed in recent years. Finally, although there are concerns over the validity⁵³ and completeness⁵⁴ of E-codes to measure deliberate self-harm, a high concordance exists between self-harm E-codes and medical record documentation of intentional injury with suicidal intent.⁵⁵ 

Adolescents and young adults who survive self-harm are at high risk for repeated self-harm and suicide over the following year. The risk of suicide was especially marked after nonfatal self-harm events that involved violent methods. Clinical priority should be given to ensuring the safety of young people after self-harm, which may include treating underlying psychiatric disorders, restricting access to lethal means, fortifying psychosocial supports, and close monitoring for emerging suicidal symptoms.

ADHD

attention-deficit/hyperactivity disorder

CI

confidence interval

E-code

external cause of injury code

HR

hazard ratio

SMR

standardized mortality rate ratio