Autism in Criminal-Justice and Forensic Settings: A Systematic Review and Meta-Analysis with an Exploratory Bidirectional Synthesis

Abstract

Background: Estimates of autism in criminal-justice and forensic populations vary widely, while studies involving autism-defined populations report heterogeneous justice outcomes. Broad pooling can obscure differences in recruitment, diagnostic ascertainment, legal context, and outcome definition. Methods: We conducted a systematic review and meta-analysis addressing two complementary questions: the proportion of autistic people in defined criminal-justice and forensic populations (Axis A) and criminal-justice outcomes within autism-defined populations (Axis B). Searches covered PubMed/MEDLINE, Lens.org, NCJRS, Scopus, Web of Science Core Collection, and ProQuest Central Premium, supplemented by citation searching. Compatible exact-count proportions were synthesized using binomial generalized linear mixed models with profile-likelihood confidence intervals; inverse-variance REML models with modified Hartung–Knapp inference were used for sensitivity analyses. Results: The reconciled register contained 300 candidate reports, 161 retained reports, and 37 selected numerical contributions from 30 reports representing 29 source-study units. Only three clinically compatible Axis A strata supported pooling. Estimated proportions of autism were 23.57% (95% CI 10.67–42.69) in specialist forensic intellectual-disability services, 0.89% (0.36–2.23) in general forensic responsibility evaluations, and 23.51% (20.10–27.08) in forensic youth referral services. Each synthesis contained two cohorts, and modified Hartung–Knapp intervals were substantially wider. General-prison administrative estimates ranged from 0.62% to 1.09%. Axis B findings differed by endpoint, population, sex, observation period, and adjustment set and were therefore not pooled. Conclusions: Autism frequency in justice and forensic settings is strongly setting- and ascertainment-dependent. The evidence does not support a single criminal-justice prevalence estimate or a unitary association between autism and offending. Clinically and legally defined populations should be synthesized separately, with explicit management of cohort overlap and diagnostic definitions.