The Hidden Variable in Your Clinical Trial: AI Tools Already Embedded in 71% of U.S. Hospitals

Artificial intelligence clinical decision support tools have quietly infiltrated the majority of American hospitals, creating an unaddressed challenge for clinical trial integrity. According to data from the Office of the National Coordinator for Health Information Technology, 71% of U.S. hospitals were running predictive AI integrated into electronic health records by 2024. Yet a large-scale randomized trial in Kenya has exposed a troubling disconnect: these tools can improve how clinicians make decisions without changing what happens to patients—a finding with profound implications for trial sponsors whose studies run in AI-augmented environments.

The Evidence-Adoption Gap Widens

The Kenyan trial tested "AI Consult," a generative AI clinical support tool integrated into EHR systems across 16 primary care clinics serving more than 9,600 patients. The results revealed a paradox that should concern every clinical trial operator: the AI tool demonstrably improved clinician decision-making quality, yet produced no significant change in short-term patient outcomes.

This disconnect matters because AI decision support systems are now embedded in the clinical workflows where trials collect data, creating potential confounders that current monitoring plans don't address. The problem is particularly acute in oncology and cardiology, where AI adoption rates are highest and EHR integrations most complex. Decentralized and hybrid trials face additional exposure through AI-assisted telehealth platforms, electronic patient-reported outcome tools, and symptom checkers that shape data before it reaches electronic data capture systems.

Regulatory Guidance Lags Behind Deployment Reality

The FDA has attempted to keep pace with AI's clinical integration. The agency published its AI/ML-based Software as a Medical Device Action Plan in January 2021, followed by final guidance in December 2024 titled "Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions," updated as recently as August 2025. Aidoc received FDA 510(k) clearance for its CARE AI foundation model, a comprehensive triage solution operating across radiology and clinical decision workflows.

Yet five years after the initial action plan, post-market surveillance and adverse event reporting for AI tools remain open action items. Current FDA guidance addresses AI/ML device lifecycle management but lacks operational specificity for documenting AI modifications during active trials. This creates methodological challenges for regulatory affairs teams preparing submissions with real-world evidence from AI-augmented environments, particularly when facing advisory committee review without pre-registered study designs that account for AI as an effect modifier.

What Trial Operators Must Do Now

Clinical trial sponsors and contract research organizations face unaccounted confounders and data integrity risks from AI tools embedded in site EHR systems. ELIQUENT Life Sciences has documented data flow integrity challenges including AI tools used for protocol deviation assessment, data cleaning, and risk signal identification in active trials.

Sponsors must implement new site feasibility protocols documenting which AI tools are active in clinical workflows, whether they interact with EDC data fields, and what change control processes exist for mid-trial updates. Without these safeguards, trials risk protocol deviations and regulatory challenges during submission reviews.

Health systems and academic medical centers with high AI adoption rates are creating evidentiary gaps in clinical research conducted within their facilities. Sites must establish disclosure and change control standards for AI decision support tools to maintain research integrity and sponsor relationships. For AI vendors, the evidence-adoption gap creates both market opportunity and regulatory risk—while FDA clearances enable rapid deployment, the disconnect between process improvement and outcome changes may trigger heightened scrutiny for real-world evidence claims.

Sources

1. AI Clinical Decision Support Is Outrunning Its Own Evidence Base: What Trial Operators Need to Know Now — Clinical Trial Vanguard