Researchers found that exposure to emotional abuse during the years around puberty showed the strongest associations with adult internalising mental disorders, such as depression and anxiety, while the windows of vulnerability differed between women and men. The study, published in Child and Adolescent Psychiatry and Mental Health, used machine-learning methods to rank which combinations of maltreatment subtype and age of exposure carried the most information about adult mental health diagnoses.
A professional opinion research institute recruited participants through randomly selected households, yielding 2,515 respondents and a response rate of 42.6 percent. The final analytical sample comprised 2,466 adults aged 18 and older: 1,278 women, or 51.8 percent, with a mean age of 50.3 years, and 1,188 men, or 48.2 percent, with a mean age of 51.1 years.
Child maltreatment was assessed with the retrospective version of the ISPCAN Child Abuse Screening Tools, a questionnaire that screens for neglect, physical abuse, emotional abuse and sexual abuse before the age of 18. For each described act, participants could indicate at what age or ages the exposure had taken place, allowing the researchers to build binary variables for every combination of maltreatment subtype and year of childhood. The questionnaire also captured multiplicity, meaning the number of different subtypes experienced, as well as the total duration in years and the frequency of exposure.
Mental health outcomes were based on participants’ reports of diagnoses they had received from a physician or psychotherapist, covering conditions such as depression, anxiety, burn-out, post-traumatic stress disorder, personality disorders, eating disorders, attention deficit disorder, schizophrenia, and alcohol or drug addiction. Following the Hierarchical Taxonomy of Psychopathology model, these were grouped into internalising mental disorders, including depression, anxiety, burn-out, post-traumatic stress and eating disorders, and externalising mental disorders, including personality disorders, attention deficit disorder, schizophrenia and addiction disorders.
To analyse the data, the team applied conditioned random forest regression, a machine-learning technique that builds many decision trees from different subsets of the data and estimates the importance of each predictor by randomly permuting it and measuring how much the model fit degrades. A higher and statistically significant variable importance score means a variable is more predictive of the outcome than the others in the model. To validate these scores, the procedure was repeated 5,000 times with reshuffled variables to simulate chance. Post-hoc SHapley Additive exPlanations, or SHAP, analyses were then used to examine the direction and strength of each timing variable, with positive values indicating an increased associated risk. All analyses were stratified for sex, and timing variables below the age of three were excluded because of limited recall of early childhood.
The results pointed to a critical window around (pre)puberty for emotional abuse. For internalising disorders in adulthood, timing of emotional abuse was the most strongly associated factor: for women, exposure between ages 10 and 16 showed variable importance values ranging from 0.255 to 1.014, while for men ages 13 and 14 stood out, with values of 0.349 and 0.504. For externalising disorders, men’s outcomes were most closely linked to emotional abuse between ages 10 and 16, with variable importance ranging from 0.174 to 0.968, whereas women’s externalising disorders were most associated with sexual abuse between ages 13 and 15, with values from 0.136 to 0.331, and neglect between ages 11 and 17, with values from 0.098 to 0.331.
Global indicators of maltreatment also proved important. Multiplicity, or exposure to multiple subtypes, was important for both internalising and externalising disorders in women and men, reaching a variable importance of 2.365 for internalising disorders in women and 2.126 for externalising disorders in men. Duration of exposure was likewise important for both outcomes in both sexes, especially for externalising disorders in men, where it reached 1.686. In competitive hypothesis testing, multiplicity was generally more important than timing of specific subtypes, with the notable exception of emotional abuse at age 18 in women for externalising disorders. For women, emotional abuse at ages 14 and 15 was more important than duration for internalising disorders, and for men, emotional abuse at ages 13 and 14 was more important than frequency for internalising disorders.
The descriptive data also revealed sex differences in reported diagnoses. Women reported significantly more internalising disorders than men, 216 women or 17.1 percent compared with 118 men or 10.1 percent, while men reported significantly more externalising disorders, 100 men or 8.5 percent compared with 56 women or 4.4 percent. The SHAP analyses confirmed the timing findings, showing a concentration of positive values for emotional abuse in pre-puberty in relation to internalising disorders, over a longer window of years for women than for men, and for emotional abuse in puberty in relation to externalising disorders in men.
The authors are careful about what these findings do and do not mean. The analysis does not constitute a validated predictive model and does not permit causal inference; permutation importance reflects the strength of association after accounting for correlated predictors, and the approach is described as exploratory and hypothesis-generating, helping to narrow a large and correlated variable space and highlight developmental periods worth examining in future predictive or causal studies. The retrospective design is a further limitation, since retrospective and prospective reports of maltreatment show limited congruence, and remembering the specific age of events may be problematic, although autobiographical memory may allow an approximate age to be recalled. Results for sexual abuse at ages 4 to 10 rest on small numbers and should be interpreted with caution.
Even so, the study underscores that both the type and the timing of child maltreatment may shape adult mental health differently in women and men, and that multiplicity and duration of exposure matter alongside specific windows of vulnerability. The authors suggest that assessing the timing of maltreatment subtypes could contribute to earlier diagnosis of mental disorders and to prevention strategies aimed at reducing lifetime mental health risks, while noting that child maltreatment remains a global challenge despite the protections of the United Nations Convention on the Rights of the Child.
Subject of Research: Timing of child maltreatment exposure and its association with adult internalising and externalising mental disorders in women and men
Article Title: Critical windows: timing of child maltreatment exposure and adult mental health
Article References: Otten, D., Schalinski, I., Fegert, J. M., Jud, A., Brähler, E., & Clemens, V. (2026). Critical windows: timing of child maltreatment exposure and adult mental health. Child and Adolescent Psychiatry and Mental Health, 20(1), Article 139. https://doi.org/10.1186/s13034-026-01188-0
Image Credits: AI Generated
DOI: 10.1186/s13034-026-01188-0
Keywords: child maltreatment, emotional abuse, internalising disorders, externalising disorders, puberty, random forest, SHAP, retrospective survey, sex differences, mental health, neglect, sexual abuse
News Source: Glenn Wilkins. (October 29, 2026). Timing of child maltreatment shapes adult mental health risks differently in women and men. Scienmag.
