Beyond Behavioral Signs: Toward a Multidomain Risk Architecture for Early Detection of Autism Spectrum Disorder

Background: Autism spectrum disorder (ASD) prevalence has increased markedly over recent decades, yet primary tools for early detection remain focused almost exclusively on observable behavioral signs in children aged 18 months and older. Instruments such as the M-CHAT-R/F, ADOS-2, and ADI-R, while clinically valuable, operate within a narrow temporal and conceptual window that systematically excludes a wide range of prenatal, perinatal, medical, environmental, and psychosocial signals now supported by robust empirical evidence.Objective: This paper presents a theoretically grounded critique of current early detection frameworks for ASD and proposes a Multidomain Early Risk Architecture (MERA). MERA is a detection architecture rather than a causal model of aetiology: it organizes ten empirically supported domains of clinically accessible signals according to the developmental window in which each becomes observable, and specifies through a four-tier taxonomy what inference each class of signal can bear.Approach: A structured narrative synthesis of peer-reviewed literature published between 2006 and 2025 was conducted across PubMed, PsycINFO, Scopus, and Web of Science. Each retained source was mapped to one of 75 candidate risk factors and classified by study design and journal indexing quartile. The resulting evidence base comprises 152 factor–source pairings drawn from 148 distinct sources, organized into ten thematic domains and four evidential tiers.Findings: The retained literature documents ASD-relevant signals distributed across all ten domains and four developmental windows. Sixty-nine percent of pairings derive from quartile-1 journals and 38 from systematic reviews or meta-analyses. Half of the evidence base corresponds to factors established before any behavioural observation of the child is possible, and which no current screening instrument incorporates. Domains differ systematically in evidential status: some comprise antecedent exposures, others early prodromal markers, and others features of the diagnostic phenotype itself or associated conditions of contested directionality.Conclusion: A multidomain architecture that stratifies signals by developmental window and by evidential status offers a conceptual foundation for tiered developmental surveillance. The framework has implications for instrument development, primary care protocols, and public health policy—particularly in settings such as Chile where diagnostic delays remain substantial. Its domain structure and clinical utility require prospective validation, which the framework is explicitly designed to enable.