JMIR Publications

ABSTRACT

Background:

Clinical practice guidelines are central to evidence-based mental health care, yet their implementation remains inconsistent across clinical settings. Digital health technologies and AI-enabled decision-support systems offer new opportunities to support guideline use, but their translation into routine mental health practice remains limited by clinical, organisational, technical, and governance barriers.

Objective:

This scoping review aimed to identify and synthesise barriers to implementing clinical guidelines in mental health services and to examine their implications for computable, interoperable, and governable digital decision support.

Methods:

A structured scoping review was conducted using systematic search, screening, and narrative synthesis. Studies addressing guideline implementation, digital health technologies, computable knowledge, interoperability, AI-enabled decision support, or governance in mental health were included. Barriers were extracted, coded, and synthesised across clinical, organisational, technical, digital, governance, and human-relational domains.

Results:

Forty-six studies were included. Barriers spanned clinical and organisational constraints, limitations in evidence and knowledge representation, interoperability and data challenges, AI and decision-support issues, governance and ethical concerns, and human-relational factors. These barriers were frequently interconnected, limiting the translation of evidence-based recommendations into consistent, computable, and accountable clinical workflows. The findings indicate that technology alone is insufficient to improve guideline implementation unless it is embedded within service workflows, interoperable data infrastructures, governance mechanisms, and clinician-facing accountability structures.

Conclusions:

Mental health guideline implementation requires more than the availability of clinical recommendations or digital tools. Current approaches lack an integrated pathway linking evidence, computable guideline logic, interoperable data, decision support, and governance. This review proposes the Guideline-as-Guardrail model, in which computable guidelines function as executable constraints for digital and AI-enabled decision support while preserving clinical judgement, patient values, shared decision-making, and professional accountability.