Association Between Social Determinants of Health and Deliberate Self-Harm Among Youths With Psychiatric Diagnoses

Research has consistently recognized the critical role of social determinants in health concerns and disparities. In Healthy People 2030 (1), the U.S. Department of Health and Human Services defines social determinants of health (SDoHs) as “the conditions in the environments where people are born, live, learn, work, play, worship, and age that affect a wide range of health, functioning, and quality-of-life outcomes and risks.” Approximately 80%–90% of U.S. health outcomes are influenced by a combination of SDoHs and health behaviors (2, 3). Healthy People 2030 (1) underscores the importance of these conditions by focusing on improvement of SDoHs as one of its overarching goals to promote health and well-being. One important health outcome influenced by social determinants is self-harm, defined here as encompassing both nonsuicidal self-injury and suicide attempts (4). Self-harm is a key risk factor for youth suicide and is a major public health concern (5). Between 7% and 31% of youths may have exhibited self-injurious behaviors (6–9), and 2019 national surveillance data indicate that 8.9% of U.S. high school students reported at least one suicide attempt during the previous 12 months (10).

Consistent with the social-ecological model (11), which incorporates individual, relationship, community, and societal factors, evidence suggests that adverse SDoHs (e.g., economic instability, inadequate health care and education, family and other interpersonal relationship problems, and lack of social connectedness) are associated with increased risk for self-harm and suicide among youths. A study examining the association between parental socioeconomic status and youth self-harm (12) found that youths whose parents reported low levels of income had about 1.5 times greater risk for self-harm compared with youths whose parents reported higher income levels. Another study (13) reported that youths ages 10–19 years who were living in disadvantaged neighborhoods were twice as likely to report suicidal thoughts and four times as likely to have a suicide attempt as control-group youths. Poor parental relations and family discord including childhood maltreatment have also been linked to increased suicide risk among youths. In a study of youths ages 9–10 (14), high family conflict was significantly associated with suicidal ideation, even after the analyses controlled for demographic and psychosocial variables. Gomez et al. (15) also examined whether child maltreatment significantly increased suicide risk among adolescents ages 13–18 and found that youths who had experienced physical or sexual abuse were 5.8 times and 4.2 times more likely to report suicide attempts, respectively, than those without these adverse experiences.

Knowledge of the association between SDoHs and youth self-harm and suicidal behavior has been limited, because of the lack of standardized, population-level data linking SDoHs to these outcomes. Capturing SDoH information in medical records may assist in identifying individuals in health care settings affected by adverse SDoHs and in targeting interventions to youths who have experienced these adverse events to prevent self-harm. Such identification may be especially critical for youths at risk for suicide, who frequently have contact with health care professionals before a suicide attempt or death. Approximately 80% of youths who die by suicide are seen by their primary care clinicians during the year before their death, whereas only 20% see a mental health provider (16, 17). A 2019 survey by the American Academy of Pediatrics (18) found that >75% of pediatricians reported having had a patient who had attempted or died by suicide, and 48% said that this situation had occurred during the past year. To address this research gap, the primary aims of this study were to quantify the prevalence of adverse SDoHs captured in medical claims data and to examine the association between such determinants and self-harm among Medicaid-enrolled youths with primary psychiatric diagnoses. A better understanding of SDoHs and suicide risk in pediatric primary care and other health care settings could inform targeted suicide prevention strategies to reduce youth suicide rates.

Methods

Study Design and Cohort

We used a retrospective longitudinal cohort design to examine the association between adverse SDoHs and nonfatal self-harm. The study population consisted of all youths ages 10–17 years in Ohio with at least one claim with a primary psychiatric diagnosis (ICD-10 codes F00–F99) between April 1, 2016, and December 31, 2018, and who had been continuously enrolled in Medicaid during the 180-day period prior to the index claim. Data from youths whose records were missing data on sex (N=4) or area of residence (N=35) were excluded, leaving a final sample of N=244,958. Youths were followed up until the first occurrence of self-harm, age 25, the end of Medicaid enrollment, death, or December 31, 2018 (the end of the study), whichever occurred first. Follow-up time ranged from 1 to 1,004 days (mean±SD=600.5±330.4 days). The Ohio State University Institutional Review Board approved all procedures, and we followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines (19).

Data Sources

We extracted data from Ohio Medicaid claims and eligibility data. Medicaid claims data include information such as service dates, CPT or Healthcare Common Procedure Coding System procedure codes, and up to 16 ICD-10 diagnosis codes for paid inpatient and outpatient service claims. Monthly Medicaid enrollment status and youth demographic information (e.g., age, sex, and race-ethnicity) were obtained from Medicaid eligibility files.

Measures

Dependent variable.

The primary outcome of interest was time (measured in days) to first nonfatal self-harm event during the follow-up period (see the online supplement to this article for ICD-10 codes) (20). Our operationalization of self-harm included both nonsuicidal self-injury and suicide attempts (4).

SDoHs.

The primary exposures of interest were adverse SDoHs, identified by using ICD-10 codes and classified into 14 categories (see the online supplement for the specific ICD-10 codes). These categories, modified from previous research (21) to better capture social determinants of interest among youth populations, included abuse and neglect, child welfare placement, educational problems, employment or financial problems, exposure to violence, housing instability, legal issues, disappearance or death of a family member, family disruption by separation or divorce, family alcohol or drug use, parent-child conflict, other family problems (e.g., inadequate parent supervision, parental overprotection, excessive parental pressure, or needing to care for a dependent relative), social and environmental problems (e.g., discord with neighbors or landlord, problems related to living in a residential institution, acculturation difficulty, or social exclusion and rejection), and nonspecific psychosocial needs or psychosocial issues not directly defined by the ICD-10 code used (e.g., other specified problems related to psychosocial circumstances or problems related to unspecified psychosocial circumstances). Each category was treated as a time-varying variable. Youths were considered as experiencing a given SDoH for the 365 days after a claim with a diagnosis code in that category.

Covariates.

Covariates included demographic and clinical characteristics. Demographic variables included age on the index date, sex (male or female), race-ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other, including American Indian/Alaskan Native, Asian American/Native Hawaiian/other Pacific Islander, more than one race, and unknown), area of residence (metropolitan or nonmetropolitan), and Medicaid eligibility status (disability, foster care, poverty, or other). Comorbid conditions, both psychiatric and general medical conditions, were identified from medical claims during the 180 days preceding the index claim with a primary psychiatric diagnosis or during the follow-up period and were treated as time-varying covariates. Evidence of a comorbid psychiatric condition (i.e., attention-deficit hyperactivity disorder [ADHD], anxiety disorders, conduct or oppositional defiant disorders, bipolar disorders, major depression, intellectual disability, personality disorders, schizophrenia, and substance use disorders) was based on the presence of at least one inpatient claim or at least two outpatient claims with a diagnosis code for that condition (see online supplement for ICD-10 codes). We identified complex chronic conditions (e.g., cancer or cystic fibrosis) on the basis of the presence of at least one relevant diagnosis code, by using a system previously used for administrative data (22, 23), and we identified noncomplex chronic conditions (e.g., asthma or obesity) on the basis of at least one claim with a relevant diagnosis code, as identified through the chronic condition indicator of the Healthcare Cost and Utilization Project (22, 24). Psychiatric conditions were not considered chronic conditions. Prior self-harm and suicidal ideation, as defined in previous research (25, 26), included any claim with a relevant diagnosis code during the 180 days preceding or including the index primary psychiatric diagnosis claim. Any inpatient, emergency room, or outpatient mental health care visits during the 180 days before the index claim were included as measures of previous health care utilization.

Statistical Analysis

Descriptive statistics were used to examine the distribution of demographic and clinical characteristics, service history, and adverse SDoH prevalence. A Cox proportional hazards analysis was used to examine the association between each SDoH category and time to first nonfatal self-harm event (27). Unadjusted and adjusted hazard ratios (HRs) and 99% CIs were calculated for each SDoH category. To determine the independent contribution of adverse SDoHs to self-harm risk, we developed two Cox proportional hazards models. Model 1 included the demographic and clinical covariates, and model 2 added the adverse SDoHs. A likelihood ratio test was used to determine whether model 2 had significantly improved fit compared with model 1. All analyses were performed in SAS, version 9.4, and R, version 4.0.3 (28, 29).

Results

Most individuals in the final sample (N=244,958) were male (54.4%) and non-Hispanic White (64.6%), with a mean±SD age of 12.9±2.4 years (Table 1). Most youths were eligible for Medicaid because of poverty (86.9%) and lived in a metropolitan area (78.8%). Approximately two-thirds (64.8%, N=158,713) had a chronic medical condition. Common psychiatric diagnoses included ADHD (39.7%), depression (26.2%), and anxiety disorders (23.9%).

Prevalence of Adverse SDoHs

Adverse SDoHs were documented for 51,796 youths (21.1%) in the period between the index date and the earliest of age 25, death, end of Medicaid enrollment, or December 31, 2018. The most prevalent adverse determinant category was abuse and neglect (13.3%), followed by other family problems (3.1%), educational problems (2.7%), parent-child conflict (2.2%), and child welfare placement (2.2%) (Table 2). Only 153 youths (0.1%) had a diagnosis code related to legal issues during follow-up, and none of these youths had legal issues at the time of the first self-harm event. Because of the small sample size, we did not investigate the association between legal issues and self-harm.

Associations Between Adverse SDoHs and Nonfatal Self-Harm

Self-harm occurred among 3,262 youths (1.3%) during the follow-up period. The mean follow-up time before a first self-harm event was 352.5±274.3 days (range 1–1,000 days). At the time of a first self-harm event, one-third of the youths (33%) had experienced documented abuse and neglect, 6% had experienced parent-child conflict, and 6% had experienced other family problems. Of these youths, 5% had a child welfare placement documented within the past 365 days (Table 2).

Table 3 shows the HRs and 99% CIs estimated with the multivariable Cox proportional hazards models. In model 1, demographic characteristics associated with increased hazard for self-harm included older age (HR=1.07) and female sex (HR=2.23). Medicaid eligibility because of disability was associated with decreased hazard of self-harm (HR=0.77), compared with eligibility because of poverty. The following psychiatric disorders were associated with significantly increased hazard of self-harm: anxiety (HR=1.44), bipolar disorder (HR=1.92), conduct disorder or oppositional defiant disorder (HR=1.63), depression (HR=3.91), personality disorder (HR=1.60), schizophrenia (HR=1.75), and substance use disorder (HR=1.37). Compared with no chronic general medical conditions, presence of a noncomplex chronic general medical condition (HR=1.30) or a complex chronic general medical condition (HR=1.30) was also associated with increased hazard of self-harm, as was a history of suicidal ideation (HR=1.81) and self-harm (HR=2.02). Previous inpatient mental health care (HR=0.76) and outpatient mental health care (HR=0.77) were associated with a significantly decreased hazard of self-harm.

In model 2, inclusion of adverse SDoHs significantly improved model fit (χ²=335.6, df=13, p<0.001) compared with model 1, which included only demographic and clinical characteristics. After the analyses controlled for demographic and clinical characteristics, abuse and neglect (HR=1.90), child welfare placement (HR=1.32), parent-child conflict (HR=1.52), other family problems (HR=1.25), and nonspecific psychosocial needs (HR=1.41) were associated with significantly increased hazard of self-harm.

Discussion

Our study’s primary purpose was to characterize the association between adverse SDoHs and risk for self-harm among youths with a primary psychiatric diagnosis. We built on the previous literature by examining a broad range of adverse SDoHs and by using data applicable to real-world clinical settings, thereby allowing for the broad application of our results to daily clinical practice settings, including primary care. In this statewide, population-based sample of Medicaid-enrolled youths with a primary psychiatric diagnosis, 21.1% had at least one documented adverse SDoH, and 1.3% had at least one self-harm event during the follow-up period. Several adverse SDoHs (including abuse and neglect, child welfare placement, parent-child conflict, other family problems, and nonspecific psychosocial needs) were associated with increased risk for self-harm.

Our findings suggest that adverse SDoHs provide vital information beyond demographic and clinical factors in understanding risk for self-harm. Identification of these factors in health care settings could improve youth suicide prevention strategies by helping to integrate medical and social service needs. In the United States, four main screening tools have been designed for identification of SDoHs in primary care: the Protocol for Responding to and Assessing Patients’ Assets, Risks, and Experiences (PRAPARE) tool by the National Association of Community Health Centers, the Accountable Health Communities Screening Tool, the Health Leads Screening Tool, and Health Begins Upstream Risks Screening Tool (30). Future research should investigate how to most effectively increase documentation of SDoHs in medical records, possibly through use of these screening tools and through quality improvement studies (e.g., plan-do-study-act cycles) coupled with educational initiatives or standardized protocols.

The category of abuse and neglect was strongly associated with increased risk for self-harm. After the analyses controlled for demographic and clinical factors, the risk for self-harm among youths with documented abuse and neglect within the past 365 days was found to be nearly two times greater than that of youths without such histories. Previous research (4, 31) has identified abuse and neglect as key risk factors for self-harm and suicide among adolescents. The association between abuse and neglect and self-harm has been reported for both males and females and across racial-ethnic groups (31). Identification of youths who have experienced abuse and neglect—possibly through use of SDoH ICD-10 codes in the medical records—will help ensure that these youths at high risk for self-harm receive targeted screening and interventions. For example, youths who self-report abuse or neglect on a screening tool embedded within an electronic health record, or youths whose health care provider enters an ICD-10 code for abuse or neglect, could receive an automatic referral to community or counseling services that could help support these youths and help prevent them from engaging in self-harm.

Several other factors related to family issues were also associated with increased self-harm risk, even after analyses controlled for demographic and clinical factors. Family issues, including poor parent-child attachment, family support and cohesion, parental psychiatric symptoms, and frequent arguing with adult authority figures, have been identified as key risk factors for self-harm and suicide among youths (31). We found that parent-child conflict and other family problems were significantly associated with increased self-harm risk. Consistent with previous research, child welfare placement (likely a proxy of poor family functioning, history of child abuse or neglect, and other stressors) was also associated with elevated risk for self-harm. In a nationally representative sample of U.S. adolescents, risk for suicide attempt in the past 12 months was about four times greater among those with a history of foster care involvement (32). Addressing underlying family and related social issues may help reduce self-harm risk. For example, multisystemic family-based therapy, a family-focused intervention that targets the multiple systems encompassing individual youths and their support networks, has shown promising evidence of success in reducing attempted suicide rates, compared with psychiatric hospitalization for adolescents with psychiatric emergencies (33). Identification of issues related to child welfare placement or other family-related issues may increase the chance that youths experiencing such problems receive appropriate interventions to prevent negative health outcomes, such as self-harm.

Strengths of this study included its large, diverse, and population-based sample of Medicaid-enrolled youths with a primary psychiatric diagnosis and its design as a longitudinal observational study rather than as a cross-sectional analysis. To our knowledge, this study was one of the first examinations of the association between adverse SDoHs, as reported by ICD-10 codes, and nonfatal self-harm among youths. However, this study had several limitations. First, our study included data from only one state’s Medicaid-enrolled population of youths with a primary psychiatric condition; therefore, our findings may not be generalizable to other Medicaid programs, populations without psychiatric conditions, or privately insured or uninsured populations. Medicaid populations with psychiatric conditions have higher rates of general medical comorbid conditions, compared with the general population (34). Second, suicidal intent of self-harm injuries was not distinguished in the claims data and nonsuicidal self-injury and suicide attempts could therefore not be differentiated. Third, the diagnoses obtained from claims data were not validated or obtained through standardized methods. Fourth, adverse SDoHs are likely underreported in claims records (35), and the degree of underreporting may differ by clinical setting and diagnosis. This limitation underscores the importance of increasing standardized documentation of SDoHs in medical records. Finally, self-harm that does not result in medical care is not captured in medical claims.

Conclusions

Adverse SDoHs were significantly associated with nonfatal self-harm among youths with a primary psychiatric diagnosis, even after the analyses controlled for demographic and clinical characteristics. Our findings underscore the importance of addressing medical and social factors in efforts to prevent self-harm. These determinants are particularly important among Medicaid-enrolled populations, who disproportionately experience suicide risk factors (e.g., mental illness) and adverse SDoHs. Use of ICD-10 codes to identify clinical and social factors related to self-harm risk in medical records could help to identify youths for appropriate prevention efforts and may allow health care providers to bill for efforts to address adverse SDoHs in general clinical settings (36). Although reporting of information related to SDoHs in electronic medical records has been championed (36–38), documentation of SDoHs through ICD codes remains underused, despite the potential benefits of the codes’ standardized definitions to population-level studies and interventions to track patients’ social needs. Future research should focus on effective methods to increase the recording of adverse SDoHs in medical records and to translate such information into meaningful interventions.