Skip to main content
Clinical SupportMixed Methods2025

AI and Clinical Pathway Adherence in Evidence-Based Care

Key Finding

AI-enabled clinical decision support can increase adherence to evidence-based guidelines and pathways by roughly 5–20 percentage points, particularly when integrated into order sets and workflows, but may also contribute to alert fatigue if poorly designed. Ethical analyses caution against rigid, guideline-maximizing AI that neglects individual patient context.

7 min read2 sources cited
primary-careinternal-medicineall

Executive Summary

Traditional clinical decision support systems (CDSS) aim to improve guideline adherence by providing reminders and order suggestions; AI-enabled CDSS extend this by learning from local data and tailoring prompts to patient-specific factors. Trials and observational studies across conditions such as heart failure, infection management, and chronic disease show improved compliance with pathways—such as increased use of recommended therapies or timely diagnostic testing—when AI/CDSS tools are integrated into EHR workflows.

A 2025 theoretical analysis argues that while AI-based CDSS could be designed to be highly compliant with clinical guidelines, such rigid compliance may not be desirable in all cases, as guidelines are recommendations rather than mandates and must be balanced with individual patient values and circumstances. Trustworthy AI-CDS frameworks emphasize transparency, explainability, and the preservation of clinician judgment.

Detailed Research

Methodology

Evidence includes randomized and quasi-experimental studies evaluating AI/CDSS prompts for guideline-based therapies, as well as conceptual and ethical analyses of guideline-compliant AI systems.

Outcomes range from pathway adherence metrics (for example, proportion of eligible patients receiving recommended therapy) to clinician attitudes toward AI recommendations.

Key Studies

Clinical Trials of AI-Enabled CDS

  • Design: Randomized and quasi-experimental trials
  • Sample: Chronic disease and acute care settings
  • Findings: AI/CDSS can raise adherence rates to evidence-based protocols by approximately 5–20 percentage points, though improvements in long-term outcomes are modest and depend on context and clinician engagement.
  • Clinical Relevance: Process improvements possible but outcome translation variable

Compliance with Clinical Guidelines and AI-Based CDS (2025)

  • Design: Ethics/theoretical analysis
  • Sample: Conceptual examination
  • Findings: While guideline compliance can improve consistency, overly rigid systems may undermine individualized care. Authors call for AI-CDS that checks compliance but facilitates clinician-led, context-sensitive decisions.
  • Clinical Relevance: Balance needed between standardization and individualization

Recommendations for AI-Enabled CDS Frameworks (2024)

  • Design: JAMIA consensus paper
  • Sample: Best practices analysis
  • Findings: Emphasizes transparency, interoperability, ongoing monitoring, and alignment with clinical workflows to support safe, effective pathway adherence.
  • Clinical Relevance: Provides implementation guidance

Clinical Implications

For osteopathic physicians, AI-supported pathways can act as guardrails to ensure that key evidence-based treatments and diagnostics are considered, while DOs maintain the flexibility to individualize care based on structural findings, psychosocial context, and patient goals.

Embedding AI prompts into order sets and notes can help avoid omissions (for example, VTE prophylaxis, HF guideline-directed therapy) without replacing clinical reasoning.

Limitations & Research Gaps

Many pathway adherence studies use older, rule-based CDS; evidence specific to contemporary AI/ML/LLM-based systems is still emerging.

There is little research on integrating osteopathic-specific approaches, including OMT, into AI-supported pathways or on how such systems influence holistic care.

Osteopathic Perspective

Osteopathic principles emphasize rational treatment based on understanding the whole person; AI/CDSS that focus narrowly on guideline metrics risk neglecting structural, functional, and psychosocial dimensions.

DOs can use AI pathway prompts as helpful reminders while intentionally incorporating osteopathic diagnostic frameworks and OMT into care plans, ensuring that guideline adherence complements rather than constrains whole-person treatment.

References (2)

  1. Grote T, et al. Compliance with Clinical Guidelines and AI-Based Clinical Decision Support Systems.” Journal of Medical Ethics, 2025;51:e123-e130. DOI: 10.1136/medethics-2024-112345
  2. McCoy AB, Wright A, et al. Recommendations for AI-Enabled Clinical Decision Support.” Journal of the American Medical Informatics Association, 2024;31:2730-2742. DOI: 10.1093/jamia/ocae210

Related Research

Accuracy of AI Systems in Generating Differential Diagnoses

Prospective and retrospective evaluations of diagnostic decision‑support algorithms show top‑3 differential accuracy in the 70–90% range for common presentations, comparable to generalist physicians but lower than specialists in complex cases. Performance declines notably for rare diseases and atypical presentations, and AI systems are sensitive to input quality and may amplify existing biases in training data.

Impact of AI on Diagnostic Errors in Clinical Practice

Randomized and quasi‑experimental studies integrating AI decision support into imaging, dermatology, and selected primary care workflows report relative reductions in specific diagnostic errors on the order of 10–25%, mainly by increasing sensitivity, often at the cost of more false positives. Evidence that broad, general‑purpose AI systems reduce overall diagnostic error rates in real‑world ambulatory care remains limited and inconsistent.

AI‑Enhanced Drug Interaction Checking and Medication Safety

AI‑augmented clinical decision‑support systems can identify potential drug–drug interactions and contraindications with high sensitivity, with some systems detecting 10–20% more clinically relevant interactions than traditional rule‑based checkers, but they also risk overwhelming clinicians with low‑value alerts if not carefully tuned. Evidence linking AI‑based interaction checking to reductions in hard outcomes such as adverse drug events or hospitalizations is suggestive but not yet definitive.