Assessment of Suicidal Behaviors Among Individuals With Autism Spectrum Disorder in Denmark

Key Points

Question  Do people with autism spectrum disorder have higher rates of suicide attempt and suicide compared with people without autism spectrum disorder?

Findings  In this nationwide retrospective cohort study that included 6 559 266 persons aged 10 years or older living in Denmark during the period from 1995 to 2016, individuals with a diagnosed autism spectrum disorder had more than 3-fold higher rates of suicide attempt and suicide compared with all other persons after adjusting for sex, age, and time period.

Meaning  In this Danish cohort, diagnosis of an autism spectrum disorder was associated with suicide attempt and suicide.

Importance  There is limited evidence supporting an association of autism spectrum disorder (ASD) with suicidality and the risk factors for suicide attempt and suicide among people with ASD. Existing research highlights the need for national cohort studies.

Objectives  To analyze whether people with ASD have higher rates of suicide attempt and suicide compared with people without ASD using national register data, identify potential risk factors for suicide attempt and suicide among those with ASD, and examine associations with comorbid disorders.

Design, Setting, and Participants  In this cohort study, nationwide register data from January 1, 1995, to December 31, 2016, were gathered on 6 559 266 individuals in Denmark aged 10 years or older. Statistical analysis was performed from November 20, 2018, to November 21, 2020.

Main Outcomes and Measures  Rates of suicide attempt and suicide among persons with ASD were compared with rates among persons without ASD, using Poisson regression models to calculate incidence rate ratios adjusted for sex, age, and time period.

Results  Of the total study population of 6 559 266 individuals, 35 020 individuals (25 718 male [73.4%]; mean [SD] age at diagnosis, 13.4 [9.3] years) received a diagnosis of ASD. A total of 64 109 incidents of suicide attempts (587 [0.9%] among individuals with ASD) and 14 197 suicides (53 [0.4%] among individuals with ASD) were recorded. Persons with ASD had a more than 3-fold higher rate of suicide attempt (adjusted incidence rate ratio [aIRR], 3.19; 95% CI, 2.93-3.46) and suicide (aIRR, 3.75; 95% CI, 2.85-4.92) than those without ASD. For individuals with ASD, the aIRR for suicide attempt among female individuals was 4.41-fold (95% CI, 3.74-5.19) higher compared with male individuals; for individuals without ASD, the aIRR for female individuals was 1.41-fold (95% CI, 1.39-1.43) higher compared with male individuals. Higher rates of suicide attempt were noted across all age groups for those with ASD. Persons with a diagnosis of ASD only had an aIRR of 1.33 (95% CI, 0.99-1.78) for suicide attempt, whereas those with other comorbid disorders had an aIRR of 9.27 (95% CI, 8.51-10.10) for suicide attempt compared with those without any psychiatric disorders. A total of 542 of 587 individuals with ASD (92.3%) who attempted suicide had at least 1 other comorbid condition and 48 of 53 individuals with ASD (90.6%) who died by suicide had at least 1 other comorbid condition.

Conclusions and Relevance  This nationwide retrospective cohort study found a higher rate of suicide attempt and suicide among persons with ASD. Psychiatric comorbidity was found to be a major risk factor, with more than 90% of those with ASD who attempted or died by suicide having another comorbid condition. Several risk factors are different from the risk factors in the general population, which suggests the need for tailored suicide prevention strategies.

Autism spectrum disorder (ASD) comprises a set of chronic neurodevelopmental disorders with a wide range of symptoms and levels of severity.^1,2 Globally, the prevalence of ASD has been estimated to be 1% to 1.5%,³ with a cumulative incidence up to 2.8% in recent birth cohorts in Denmark.⁴ Although the onset of ASD is generally in childhood, it may be recognized and diagnosed later in life.² The number of children with a diagnosis of ASD has increased during recent decades,^1,2,5,6 and professionals have debated whether this increase is due to changes in diagnostic criteria, increased clinical or parental awareness, or increased prevalence of etiologic factors.^2,4-7

Lack of social integration, unemployment, and psychiatric disorders have been found to be associated with ASD in adults⁸; the same factors are associated with suicidal behavior,^9,10 which would suggest a potential link between ASD and suicidal behavior. Nevertheless, little evidence from large-scale studies exists regarding an association between ASD and suicidality.^11-15 A recent population-based case-cohort study from Sweden showed an increased risk of suicide and suicide attempt among those with ASD, especially among those without intellectual disability.¹⁶ It has yet to be determined what factors are associated with suicidal behavior in people with ASD and whether they differ from the factors associated with suicidal behavior in the population without ASD. To make evidence-based decisions and inform the design of intervention studies, there is a need for large-scale national cohort studies on the risk of suicide attempts and suicides among persons with ASD. Therefore, the aims of this retrospective cohort study were to analyze whether people with a diagnosis of ASD had higher rates of suicide attempts and suicides compared with people without ASD, identify risk factors for suicide attempt and suicide among those with ASD, and examine associations with psychiatric comorbid disorders.

A cohort design was applied to nationwide register data. The unique personal identification number assigned to each individual living in Denmark facilitated an individual-level data linkage of the Danish Civil Registration System¹⁷ with the Psychiatric Central Research Register (PCRR),¹⁸ the National Hospital Register,¹⁹ the Cause of Death Registry (CDR),²⁰ the Populations Education Register, and the Income Statistics Register.²¹ From the latter 2 registers, data on educational level and socioeconomic status were gathered. The Civil Registration System contains information on sociodemographic characteristics, the PCRR and National Hospital Register contain information on morbidity, and the CDR contains information on mortality. This study, which used an anonymized, register-based data set, was approved by the Danish Data Protection Agency and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Because this study is a national register–based study, all information is deidentified and is allowed to be linked for research purposes without need for consent.

All individuals living in Denmark from January 1, 1995, through December 31, 2016, were included in the cohort study. Given that suicide attempts and deaths are rare events among children, the inclusion age was set to 10 years or older.

The primary outcome was suicide attempt, while death by suicide was examined separately as an exploratory outcome. Suicide attempt was recorded when an individual presented at either a psychiatric or somatic hospital or emergency department with one of the following International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes: X60 to X84 or Y87.0, or when the reason for contact was listed as being self-harm. Death by suicide was identified in the CDR with the same ICD-10 codes or when the manner of death was listed as suicide.²¹

Information on ASD and other psychiatric diagnoses was obtained from the PCRR. This register contains data on inpatient admissions since 1969, while information on emergency department and outpatient contacts were available since 1995. During this period, diagnoses were recorded according to the International Classification of Diseases, Eighth Revision (ICD-8) and ICD-10. People who received the following codes for main diagnoses or subdiagnoses were considered as having ASD: ICD-8 codes 299.00 to 299.01 and ICD-10 codes F84.0 to F84.1 and F84.5 to F84.9.⁷ Dichotomization into low-functioning and high-functioning ASD was based on co-recorded ICD codes for intellectual disability (ICD-8 codes 310-315 and ICD-10 codes F70-F73 and F78-F79). We screened all supplementary diagnoses to capture comorbidity. Every person who had ever received a diagnosis of a mental disorder (see eTable 1 in the Supplement for complete list) was considered exposed from the date of diagnosis. In addition, parental psychiatric disorders and parental suicidal behavior (no parent vs ≥1 parent) were included.

Several socidemographic factors were included as following: sex (male or female), age group (10-19, 20-29, 30-39, and ≥40 years), marital and cohabitational status (married, registered partnership, or cohabiting vs not), educational level (basic, vocational, high school, university degree, or unknown or missing), and socioeconomic status (employed, unemployed, disabled pensioner or retired, children or students, or unknown or missing). Age at ASD diagnosis was divided into 3 groups: children (<13 years), adolescent or young person (13-24 years), and adult (≥25 years). Presence of chronic physical disorders (0 [none] vs ≥1 [any]) was measured using the Charlson Comorbidity Index.²²

Follow-up began on January 1, 1995, and individuals reaching 10 years of age or migrating into the country were included on the date of those events. Persons were followed up until December 31, 2016, unless they migrated out of the country or died, in which case they were censored at the date of the respective event. In total, 233 493 persons emigrated and 1 148 407 died by causes other than suicide during the period from 1995 to 2016. Once the outcome of suicide attempt was observed, the individual was censored (ie, the follow-up ended). In addition, we did not consider the exposure (ie, ASD) if the person was reported with a first diagnosis of ASD and a suicide attempt on the same day; the suicide attempt was still counted but just as a suicide attempt among an unexposed individual.

Statistical analysis was performed from November 20, 2018, to November 21, 2020. Poisson regression models using person-years as offset were conducted using the PROC GENMOD procedure in SAS version 9.4 (SAS Institute Inc).²³ The obtained estimates presented the suicide incidence rate, for instance, among people with a diagnosis of ASD compared with those with no disorders, expressed as incidence rate ratios. Covariates were added one at a time, while the model fit was evaluated by comparing the log likelihood of the models. We evaluated the model fit by comparing the log likelihood values of different models²⁴ and checked that the model converged. We assessed for interactions between ASD and specific exposures. Multivariable models were adjusted for sex (male or female), age (10-19, 20-29, 30-39, or ≥40 years), and time period (1995-1999, 2000-2004, 2005-2009, or 2010-2016); the adjusted incidence rate ratios (aIRRs) with 95% CIs are presented. Considering that marital and cohabitation status, educational level, socioeconomic status, and other psychiatric disorders may be on the causal pathway between the ASD and suicidal behavior, they were not included as confounding factors. However, this issue was addressed in sensitivity analyses in which the main models for suicide and suicide attempt were adjusted for educational level, socioeconomic status, and marital and cohabitation status. Associations with other psychiatric disorders were examined in separate models. Additional sensitivity analyses were conducted in the study population, which was restricted to those born 1955 or later (ie, those who were ≤40 years in 1995). All variables were time varying and updated on either exact date of change or by calendar year (eg, change in the employment status).

Of the total study population of 6 559 266 persons aged 10 years or older living in Denmark during the period from 1995 to 2016, 35 020 individuals (25 718 male [73.4%]; mean [SD] age at diagnosis, 13.4 [9.3] years) received a diagnosis of ASD. eTable 2 in the Supplement provides a more detailed description of the people who received a diagnosis of ASD. A total of 64 109 people had at least 1 suicide attempt recorded, of whom 587 had ASD (0.9%) (Table 1). During the follow-up, 14 197 died by suicide; 53 of these individuals had ASD (0.4%).

The incidence rates of suicide attempt were 266.8 per 100 000 person-years among those with ASD and 63.4 per 100 000 person-years among those without ASD (Table 1). When adjusting for sex, age, and period, we found that persons with ASD had a higher rate of suicide attempt (aIRR, 3.19; 95% CI, 2.93-3.46) compared with those without ASD.

Analysis by sex showed that male individuals with ASD had a 1.93-fold (95% CI, 1.71-2.18) higher incidence rate of suicide attempt compared with male individuals without ASD (Table 1). For individuals with ASD, the aIRR for female individuals was 4.41-fold (95% CI, 3.74-5.19) higher compared with male individuals; for individuals without ASD, the aIRR for female individuals was 1.41-fold (95% CI, 1.39-1.43) higher compared with male individuals. With the use of persons aged 10 to 19 years as a reference group, a different pattern for those with ASD than for the general population was revealed; persons with ASD did not show differences between age groups. However, for those without ASD, the aIRR decreased with age. The rate ratio for suicide attempt increased with the age at first diagnosis.

People with ASD who were unemployed had a 2.24-fold (95% CI, 1.52-3.30) higher incidence rate of suicide attempt compared with those who were employed; for individuals without ASD, those who were unemployed had a 5.89-fold (95% CI, 5.83-6.13) higher incidence rate (Table 1). The difference between those with and those without ASD was even wider for disabled pensioners and retired individuals; they had a 1.62-fold (95% CI, 1.11-2.36) higher rate compared with employed individuals among those with ASD and an 8.59-fold (95% CI, 8.37-8.83) higher rate compared with employed individuals among those without ASD. However, compared with the respective socioeconomic groups, the rate of suicide attempt was highest for employed individuals: 3.89-fold (95% CI, 2.81-5.40) higher for those with ASD compared with those without ASD. For people with ASD, a higher rate of suicide attempt was noted across all educational categories compared with those without ASD. More specifically, for those without ASD, the rate of suicide attempt decreased with educational level, but among those with ASD, individuals with vocational and university education did not have a lower rate of suicide attempt compared with individuals with a basic education. Individuals with ASD who were in a relationship (married, civil union, or cohabiting) had a 3.61-fold (95% CI, 3.16-4.13) higher rate compared with those without ASD. For individuals with ASD, the aIRR for those not in a relationship was 1.56-fold (95% CI, 1.31-1.84) higher compared with those in a relationship; for individuals without ASD, the aIRR for those not in a relationship was 2.34-fold (95% CI, 2.30-2.37) higher compared with those in a relationship. Physical comorbidities were associated with a 2.13-fold (95% CI, 2.08-2.17) higher rate of suicide attempt for those without ASD vs no change for those with ASD (aIRR, 1.21; 95% CI, 0.94-1.55).

A total of 72.5% of those with ASD (25 401 of 35 020) had received a diagnosis of other psychiatric disorders (eTable 2 in the Supplement). The aIRR for individuals with a diagnosis only of ASD was 1.33 (95% CI, 0.99-1.78), while those with other comorbid disorders had an aIRR of 9.27 (95% CI, 8.51-10.10) compared with individuals without any psychiatric disorders (Table 2). The aIRR was higher for persons with psychiatric disorders other than ASD (aIRR, 21.00; 95% CI, 20.67-21.34). The most prevalent type of psychiatric comorbidity among those with ASD was attention-deficit/hyperactivity disorder (ADHD); 11 456 [32.7%]), followed by those with anxiety, dissociative, stress-related, and somatoform disorders (9646 [27.5%]) and affective disorders (5770 [16.5%]) (eTable 2 in the Supplement). Although higher suicide attempt rates were noted for most of the examined comorbid psychiatric disorders, no difference was noted for ADHD and intellectual disability. Adjusted incidence rate ratios were highest for relatively rare comorbidities, including posttraumatic stress disorder, substance use disorders, and borderline personality disorder (Table 2).

People with a parent who had a psychiatric disorder had a 2.26-fold (95% CI, 2.21-2.30) higher rate of suicide attempt if they did not have ASD and a 1.28-fold (95% CI, 1.07-1.52) higher rate if they did have ASD. Relatively similar trends were measured for parental suicidal behavior (Table 1).

With respect to suicide, persons with ASD had a 3.75-fold (95% CI, 2.85-4.92) higher rate compared with those without ASD after adjusting for sex, age, and period (Table 3). Male individuals with ASD had an aIRR of 3.48 (95% CI, 2.57-4.74), and female individuals with ASD had an aIRR of 2.63 (95% CI, 1.46-4.76), compared with those without ASD. For individuals with ASD, there was no difference by sex in suicide rate (aIRR, 0.75; 95% CI, 0.39-1.46); for those without ASD, female individuals had a 0.36-fold (95% CI, 0.35-0.37) lower suicide rate compared with male individuals. The aIRR increased gradually with age for those without ASD; however, this pattern was not confirmed for those with ASD. Although most results showed relatively similar trends to suicide attempts, they were less pronounced and included some differences. For example, in people without ASD, those who were not married or cohabiting had a 2.81-fold (95% CI, 2.72-2.91) higher rate compared with those who were married or cohabiting; people with ASD who were married or cohabiting had no difference in the suicide rate than those who were not married or cohabiting (0.77; 95% CI, 0.42-1.38). With regard to psychiatric comorbidities, similarly, 48 of 53 individuals (90.6%) had at least 1 comorbid condition; the most prevalent type of psychiatric comorbidity among those with ASD was affective disorders (27 of 53 [50.9%]), followed by anxiety, dissociative, stress-related, and somatoform disorders (25 of 53 [47.2%]) and schizophrenia spectrum disorders (24 of 53 [45.3%]). The IRRs for suicide were highest for the comorbidity with depression and substance use disorders (Table 4).

A significant association remained between ASD suicide attempt (aIRR, 1.97; 95% CI, 1.81-2.14) and suicide (aIRR, 2.10; 95% CI, 1.81-2.14) when additionally adjusting for educational level, socioeconomic status, and marital and cohabitational status. Also, when we restricted the analyses to those born after 1955, persons with ASD were found to have an aIRR of 3.18 (95% CI, 2.92-3.45) for suicide attempt and an aIRR of 3.83 (95% CI, 2.91-5.05) for suicide compared with those without ASD.

This registry-based national cohort study showed that persons with a diagnosis of ASD have more than 3-fold higher rates of suicide attempt and suicide compared with those without ASD, after adjusting for sex, age, and time period. Factors that have been identified as protective against suicide attempt in the general population, such as older age and higher educational level, were not found to have this association among individuals with ASD, and some factors, such as being married or cohabiting and employed, were associated with being less protective among those with ASD. Most factors associated with suicide in the general population were not associated with suicide among those with ASD (eg, male sex or not being married or cohabiting). Psychiatric comorbidity was found to be a major risk factor, with more than 90% of those with ASD who attempted or died by suicide having another comorbid condition (with anxiety and affective disorders being most common). These factors are crucial for assessing suicide risk by practitioners working with people with ASD.

To our knowledge, this is the first nationwide cohort study to examine the association of ASD with suicide attempt as well as with suicide. A recent prospective birth cohort study from the United Kingdom²⁵ showed an association of the impairment of social communication with suicidal thoughts, suicidal plans, and self-harm in adolescence, but it was unable to show an association between ASD and suicidality (ideation and self-harm) owing to limited statistical power. A nested case-control study from Taiwan²⁶ found that ASD was significantly associated with suicide attempts among adolescents (12-17 years) and young adults (18-29 years) during the follow-up period, showing a similar risk for both age groups. It has been suggested that adolescents and young adults with ASD have a higher risk of suicide attempts²⁷; nevertheless, most studies have included younger age groups. Although the general population showed decreasing suicide attempt rates with increasing age, there was no notable change between age groups for those with ASD. The suicide rate was higher among those with ASD who were aged 20 years or older; nevertheless, suicide risk is increasing with age in the general population of Denmark.

A possible causal mechanism linking ASD to suicidality, particularly in adults, may be a combination of social isolation and poor access to health care.²⁸ Although it is possible that the inability to establish and retain social and intimate relationships is associated with suicide attempts among adult women with ASD, they might also receive a diagnosis and treatment later in the course of the disorder by being able to camouflage their autistic traits.^29,30 This possibility might explain the higher rates of suicidal behavior among women in our study, which is supported by findings from Swedish linkage studies in which higher risk of suicidal behavior was noted for female individuals with ASD compared with their controls than for male individuals.^16,31

The highest rates of suicide attempt were found for people with ASD who were unemployed; however, the difference with the group without ASD was highest for those who were employed, with employment having a lesser protective association for those with ASD. Some authors have suggested a mitigating association of employment for those with ASD.¹⁴ We can only speculate whether this finding might be partly associated with higher levels of peer victimization and other types of workplace bullying, as shown in other studies.¹⁴ Another possibility is that those who are employed in low-paying jobs are presumably experiencing stress owing to poor finances and structural inequality.

Increased risk of suicidal behavior among those with psychiatric disorders has been long established for the general population,¹⁰ and psychiatric comorbidity has also been shown to be associated with increased risk in studies for individuals with ASD.^2,8 Nevertheless, the association of psychiatric comorbidity with suicidal behaviors of individuals with ASD is still understudied, to our knowledge.¹⁶ A recent population-based case-cohort study from Sweden found that the risks of suicide attempt and suicide among those with ASD without comorbidities remained significantly higher compared with matched controls after adjusting for comorbidities.¹⁶ Results from our cohort study showed that individuals with ASD without comorbidities did not have higher rates of suicide attempt or suicide. However, our study might not have had sufficient power or the odds ratios may have overestimated the incidence rate ratios. Our results indicate that psychiatric comorbidity is a major risk factor for suicide attempt and suicide among people with ASD; more than 90% of those with ASD who attempted or died by suicide had another comorbid condition. Anxiety and affective disorders were the most common comorbidities, followed by schizophrenia spectrum disorders and ADHD. Attention-deficit/hyperactivity disorder, the most common comorbid condition among those with ASD, did not increase the rate of suicide attempt compared with individuals who had ASD without comorbid conditions. It has been suggested that suicidality is more common among high-functioning individuals with ASD,²⁹ which has been shown in some studies measuring functioning by the presence of intellectual disability.^16,31 This suggestion is in line with our findings of higher rates of suicide attempt and suicide among those with ASD only (high-functioning group). This result seems to be associated with the finding that, in contrast to the general population, the risk of suicide attempt among people with ASD increased with educational level and was highest among those with a university degree (although a university degree was relatively rare in people with ASD). Higher levels of cognitive functioning and education may imply a wider exposure to different risk factors,²⁷ but also there is the realization that limited social and problem-solving skills may increase self-imposed pressure to cope with and alter expectations of success.

Our results point toward important implications for clinicians working with people with ASD and those working with suicidal patients, highlighting a need for tailored suicide prevention.²⁷ Early intervention to improve social skills in children with ASD is likely to lower risks of suicidal behavior later in life. Nevertheless, it is essential to expand support and services for adults with ASD, especially those with psychiatric comorbidity, considering the higher risk of suicide attempt throughout the life span.^28,32 The high rates of suicide attempt and suicide among female individuals with ASD suggest a need to improve diagnostic tools to avoid delays in required treatment. Further work to identify the best tools to measure suicidality among those with ASD is needed.³³

This study has some strengths, including the longitudinal nationwide register data with few missing values.¹⁷ The analyses were adjusted for period effects to avoid an increase in the number of cases over time, which might be associated with different diagnostic criteria. Furthermore, the inclusion criterion of a hospital-based ASD diagnosis, which was determined by a child psychiatrist in a mental health setting using standardized diagnostic tools, was likely to have improved the validity of the measure. In addition to time period, our analyses were adjusted for variations associated with differences with respect to sex and age. Still, the sensitivity analyses supported that significant associations remained when adjusting for potential confounders or restricting the sample to younger adults.

Some limitations should also be noted. Although the diagnosis of childhood autism in the PCRR has been evaluated as valid,³² there is less information about possible underreporting of ASD in the registers, and it is possible that we have missed some cases (for instance, if some persons received a diagnosis only in primary care). Suicide attempts are underrecorded in the Danish hospital registries; therefore, our estimates may be considered conservative.³⁴

Our findings show a higher rate of suicide attempt and suicide associated with persons with ASD in a nationwide cohort study in Denmark. Psychiatric comorbidity was found to be a major risk factor, with more than 90% of those with ASD who attempted or died by suicide having another comorbid condition. A number of risk factors for suicidality among individuals with ASD are different from risk factors in the general population.

Accepted for Publication: November 23, 2020.

Published: January 12, 2021. doi:10.1001/jamanetworkopen.2020.33565

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Kõlves K et al. JAMA Network Open.

Corresponding Author: Kairi Kõlves, PhD, Australian Institute for Suicide Research and Prevention, Griffith University, Mt Gravatt Campus, Mt Gravatt, QLD 4122, Australia (k.kolves@griffith.edu.au).

Author Contributions: Dr Erlangsen had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Kõlves, Nordentoft, Erlangsen.

Acquisition, analysis, or interpretation of data: Kõlves, Fitzgerald, Wood, Erlangsen.

Drafting of the manuscript: Kõlves.

Critical revision of the manuscript for important intellectual content: Fitzgerald, Nordentoft, Wood, Erlangsen.

Statistical analysis: Kõlves, Fitzgerald, Nordentoft, Erlangsen.

Obtained funding: Kõlves, Nordentoft, Erlangsen.

Administrative, technical, or material support: Kõlves.

Supervision: Nordentoft.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the Danish Health Foundation and by the Griffith University Research Fellowship.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.