Written Self-Reports Catch Suicidality in Autistic Girls That Interviews Miss

Suicidal thoughts in autistic girls may be going undetected because of how clinicians ask about them, according to a new cross-sectional study published in the Journal of Autism and Developmental Disorders. Researchers led by Mona J. Rousan and Blythe A. Corbett at Vanderbilt University Medical Center compared three ways of assessing suicidality in autistic and non-autistic girls aged 6 to 13, and found that a simple written questionnaire completed by the children themselves flagged suicidal thinking far more often than verbal interviews conducted with parents or by clinicians directly with the child.

The study included 213 girls, 112 of whom had a confirmed autism spectrum disorder diagnosis and 101 of whom were non-autistic. All participants were between 6 and 13 years old, with an average age of 9.60 years, and all had intelligence quotient scores of at least 55. Suicidality was assessed using three established measures: the Kiddie Schedule for Affective Disorders and Schizophrenia, a structured verbal interview administered to parents; the Children’s Depression Inventory, Second Edition, a written self-report questionnaire completed by the children; and the Columbia Suicide Severity Rating Scale, a verbal interview administered by licensed clinicians to the children themselves.

Across all informants and measures, autistic girls had roughly three times greater odds of endorsing suicidality than their non-autistic peers, with an odds ratio of 3.05 and a 95 percent confidence interval of 1.64 to 5.67, a statistically significant difference. This finding held regardless of which informant or measure was used, suggesting that elevated suicide risk in autistic girls is not an artifact of any single assessment method but a consistent pattern across parent reports, written child self-reports, and clinician-administered interviews.

The more striking result concerned the informants themselves. Written child self-report on the Children’s Depression Inventory yielded the highest rate of suicidality endorsement. Compared to that written format, parent report on the K-SADS was significantly less likely to capture suicidality, with an odds ratio of 0.51, and clinician-rated child self-report on the Columbia scale was even less likely, with an odds ratio of 0.39. In other words, when a girl wrote privately on a questionnaire that she thought about killing herself, that risk was substantially more likely to be recorded than when the same concern had to be voiced aloud to a parent or a clinician.

Notably, there was no significant difference between parent reports and clinician-rated child self-reports, which showed similar odds ratios of capturing risk. The researchers also tested whether the pattern of reporting differed between autistic and non-autistic girls, and found no significant interaction, meaning that the tendency for written self-report to outperform verbal formats was not specific to autistic traits. Agreement across the three sources was slight to fair, with a Fleiss’ kappa of 0.24, indicating substantial discrepancy in who endorses suicidality depending on who is asked and how.

The researchers propose several mechanisms that might explain these discrepancies. Autistic girls are more likely to engage in camouflaging, internalizing distress while appearing emotionally regulated on the outside, and this masking may persist in social interactions with parents and clinicians. Engagement with another person may trigger masking of suicidal thoughts, meaning internalizing symptoms may not become apparent until distress escalates to overt dysregulation severe enough for parental awareness. A written, private format, by contrast, may offer a less confrontational way to communicate feelings of distress before severity reaches that threshold.

Measurement language may also play a role. The written questionnaire asks children directly whether they think about killing themselves, using active and serious wording, while the clinician-administered Columbia scale focuses on ideation, asking whether the child has wished she were dead or had thoughts of killing herself. The parent interview asks about both ideation and actions. The authors note that these measures pull along a continuum consistent with the ideation-to-action framework, and that even within that framework, the written format still produced higher endorsement. The verbal interviews, though often built around closed yes-or-no questions, rely on interactive social exchange, which may itself suppress disclosure.

The study’s context matters for interpreting its scope. Data were drawn from a larger longitudinal study of puberty, psychosocial development, and physiological changes in girls with and without autism, and the analyses were cross-sectional. Participants were recruited through self-referral, community outreach, online registries, and clinic referral, and all resided within a 300-mile radius of Nashville, Tennessee. Of 263 initial participants, 50 were excluded due to insufficient data, refusal to complete suicidality measures, or limited cognitive capacity. The sample was predominantly White, and the authors acknowledge that findings may not generalize across the full autism spectrum, particularly to individuals with intellectual disability or limited expressive ability.

The measures themselves carry caveats. None of the assessment tools used have been formally validated specifically for autistic populations, although the Columbia scale has shown promise as a reliable tool in initial research with autistic early adolescents, and the depression inventory’s suicide item has previously identified risk in autistic youth. The researchers collapsed response categories across measures into a simple present-versus-absent coding of suicide risk to enable comparison, prioritizing detection over severity. They also note that the written questionnaire contains only a single item assessing suicidal thoughts, with no intermediate option between having no such thoughts and having definite intent to act, which may discourage some children from reporting uncertain or ambivalent feelings.

The implications for clinical practice are significant. Current guidance from the American Academy of Pediatrics recommends universal suicide risk screening for youth aged 12 and older, screening when clinically indicated for those aged 8 to 11, and screening only when warning signs are present for children under 8. Yet this study found evidence of suicidality in girls as young as 6, and prior research suggests autistic youth may experience the onset of suicidal ideation earlier than the general population. Because screening in middle childhood is uncommon without previously identified risk, early signs may be overlooked. The authors argue that multi-informant, autism-informed approaches that include both written and verbal response methods can reduce underestimated risk and facilitate timely intervention, and that autism-sensitive training in suicide risk assessment is urgently needed, particularly in fast-paced settings such as emergency rooms where symptoms may be missed or masked.

Subject of Research: Differences in how suicidality is reported by parents, children, and clinicians in autistic versus non-autistic girls

Article Title: Differences in Reporting Suicidality Among Autistic Girls: Comparing Parent-, Child Self-, and Clinician-Rated Child Self-Reports Between Distinct Measures

Article References: Rousan, M. J., McGonigle, T., Muscatello, R. A., Vandekar, S., & Corbett, B. A. (2026). Differences in Reporting Suicidality Among Autistic Girls: Comparing Parent-, Child Self-, and Clinician-Rated Child Self-Reports Between Distinct Measures. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07588-7

Image Credits: AI Generated

DOI: 10.1007/s10803-026-07588-7

Keywords: autism, suicidality, autistic girls, child self-report, CDI-2, C-SSRS, K-SADS, suicide risk assessment, informant discrepancies, middle childhood, camouflaging, screening