International Neuropsychiatric Disease Journal

Precision psychiatry proposes that the combination of genomic, neuroimaging, peripheral biological, digital and health-record data will permit treatment to be matched to individual patients rather than to diagnostic categories. Opioid use disorder in the United States has been advanced repeatedly as a priority application, on the grounds that the condition is common, heterogeneous in course, associated with high mortality, and treated with a small number of medicines whose comparative effectiveness varies across individuals. Despite more than a decade of investment, no multimodal stratification tool is in routine use in American opioid use disorder care. This critical narrative review examines why. Literature was identified through structured searching of biomedical and open-access scholarly sources, supplemented by citation tracking and examination of authoritative institutional documents, with a final search date of 2 July 2026. Evidence was appraised for design adequacy, validation status, representativeness and relevance to implementation rather than merely to discovery. Five interacting classes of barrier are identified. The measurement layer remains immature: individual predictors from genomics, neuroimaging and digital sensing produce effect sizes that are reliable in aggregate but weak at the level of the individual patient. The modelling layer is dominated by retrospective, internally validated models with unstable outcome definitions and few prospective evaluations of clinical utility. The delivery layer is characterised by a segregated, under-resourced treatment system in which first-generation care is still not reliably delivered, so that stratified care is being proposed for a system that has not achieved coverage. The economic and regulatory layer offers no established route to reimbursement or oversight for multimodal decision tools. The equity layer carries a demonstrable risk that ancestry-limited genomic reference data and biased administrative training data will reproduce existing disparities in access. Two conclusions follow with reasonable confidence: the principal obstacle is not the absence of biological signal but the absence of demonstrated incremental clinical utility within a functioning delivery system, and evaluation frameworks that measure implementation outcomes alongside predictive performance are required before wider adoption can be justified.