AI Model Outperforms Clinicians in HCM Registry Data Extraction, With Alexander Blood, MD

The artificial intelligence (AI) model RECTIFIER (RAG-Enabled Clinical Trial Infrastructure for Inclusion Exclusion Review) outperformed physician abstractors in sensitivity in hypertrophic cardiomyopathy (HCM) registry data extraction.1

These data were presented at the American College of Cardiology (ACC) Scientific Sessions 2026, in New Orleans, Louisiana, by Alexander Blood, MD, associate director of the Massachusetts General Brigham Accelerator of Clinical Transformation, instructor in medicine at Harvard Medical School, and co-founder and CEO of AIwithCare.

“The challenge with data extraction is that sometimes it’s in a scanned document and sometimes it’s in somebody’s notes somewhere and was never reported,” Blood told HCPLive in an exclusive interview. “The advantage of using AI in these situations is that AI can look at everything – it doesn’t get fatigued, it doesn’t get bored, it knows what it’s looking for, and it can be more comprehensive because of that.”

RECTIFIER was first developed at Mass General Brigham’s Accelerator for Clinical Transformation. A proof-of-concept study was published in June of 2024, followed by a randomized-controlled blinded trial in February of 2025. These publications have led to RECTIFIER’s expansion across >20 active and onboarding use cases in clinical operations and research in cardiology, oncology, gastroenterology, neurology, pathology, and psychiatry.2

The program utilizes generative AI to screen electronic health records for any information that can determine a patient’s eligibility for a given clinical trial, including diagnoses, current or past medications, or key health indicators. RECTIFIER analyzes unstructured data from notes and reports, which are generally inaccessible without a significantly involved manual review process from both coordinators and clinical staff.2

In the present study, Blood and colleagues adopted an HCM registry containing 423 questions, each of which reflected guideline-directed screening and management. RECTIFIER and a team of 3 physician reviewers were tasked with abstracting data from 44 patients with confirmed HCM. The program’s outputs were compared against consensus from the abstractors, with any discrepancies managed by a blinded cardiologist.1

Ultimately, 18,612 individual questions were answered by both RECTIFIER and the team of abstractors. Mean agreement between the 3 physicians was 97.1 +/- 1.3% - in 1001 adjudicated disagreements, the cardiologist agreed with 544 (54.3%) of RECTIFIER designations and 457 (44.7%) of physician abstractor consensuses. Due to this discrepancy, RECTIFIER achieved higher overall accuracy and sensitivity compared to physician abstractors (P <.05 each), while specificity was largely similar (P = .33).1

Blood also discussed the team’s plans for RECTIFIER moving forward. He acknowledged the existing data to support its broader implementation, but also noted that further clinical trials are necessary to determine the full scope of its capabilities and the degree to which it exceeds physician abstractors’ abilities.

“As an academic clinician, I very much believe this needs to be validated and studied as rigorously as possible to show that we’re doing high-quality evidence generation and that we’re not introducing any sort of biases into the research,” Blood said. “I think this type of research is essential to make sure that we’re putting out a high-quality product.”

Editors’ Note: Blood reports disclosures with AIwithCare, Astra Zeneca, Eli Lilly, Merck & Co., Novo Nordisk, Porter Health, and others.

References

1. Frost C, Bunn D, Lu J, et al. Automated Registry Completion Using a Retrieval Augmented Generation-Enabled Large Language Model: a Comparative Study in Hypertrophic Cardiomyopathy. Abstract presented at the American College of Cardiology Scientific Sessions 2026, New Orleans, LA. March 28-30, 2026.

2. Jaslow R. Mass General Brigham Announces New AI Company to Accelerate Clinical Trial Screening and Patient Recruitment. Mass General Brigham. December 12, 2025. Accessed April 14, 2026. https://www.massgeneralbrigham.org/en/about/newsroom/press-releases/aiwithcare-mass-general-brigham-spinout-new-company