Clinical Registries
7 min read
Key Takeaways
Clinical registries like GWTG Stroke, GWTG Heart Failure, NSQIP, NCDR and NCDB turn patient records into benchmarks health systems use to improve care, but every one depends on manual chart abstraction.
A 2023 JAMA analysis found one academic medical center spent roughly 108,000 person-hours and $5 million a year on quality reporting, with manual chart review among the most resource-intensive measures.
AI-assisted abstraction shifts the work from finding evidence to actioning quality improvement initiatives. Users adopting AI are abstracting cases 65% faster while holding IRR-validated accuracy at 98% or higher.
Registries are also where Layer Health proves the reasoning that later supports quality measurement and clinical guidance beyond registry reporting.
What Clinical Registries Do
Clinical registries help health systems understand how care is being delivered, compare outcomes across sites, and identify where practice needs to change. They support:
Benchmarking
Accreditation
Quality improvement
Research
This gives leaders a view of performance that would otherwise remain buried across thousands of patient records.
The value of a registry depends on the speed, completeness and consistency of the data entering it. Yet registry programs still rely heavily on expert abstractors reading charts one at a time, finding the relevant evidence and applying detailed definitions to determine the right answer.
As registries have become more valuable, the manual process behind them has grown harder to sustain.
Clinical Registries Have Become Part of the Infrastructure for Better Care
Quality leaders increasingly use clinical registries as operating assets. Disease registry data helps health systems compare performance with national and peer benchmarks, identify variation across sites and service lines and evaluate whether improvement efforts are working.
Programs such as the following turn patient-level data into benchmarks health systems use to evaluate and improve care:
American Heart Association’s Get With The Guidelines Stroke (GWTG Stroke)
Get With The Guidelines Heart Failure (GWTG HF)
American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP)
American College of Cardiology’s National Cardiovascular Data Registry (NCDR)
National Cancer Database (NCDB)
The evidence for that impact is substantial. A 20-year analysis of GWTG-Stroke published in Stroke examined more than 7.8 million stroke cases from over 2,800 hospitals. Adherence to evidence-based care improved across nearly every measure the program tracks:
Anticoagulation for atrial fibrillation: 55.7% (2003) to 97.2% (2022)
Dysphagia screening: 53.8% to 83.5%
Door-to-needle times within 60 minutes: 19.0% to 75.3%
Surgical quality shows a similar pattern. Among hospitals participating in ACS NSQIP for at least three years, 69% reduced their mortality rate, 79% reduced complications and 71% reduced surgical site infections.
The stroke improvements were achieved alongside the American Heart Association's Target: Stroke initiative, a coordinated quality improvement campaign built on registry data. The registry made performance visible. Clinical and quality teams did the work that changed it.
Each registry serves a different clinical area, but all depend on reliable abstraction. Delayed, incomplete, inconsistent or inaccurate registry data limits a health system’s ability to understand performance and act on the findings.
What Makes Clinical Data Abstraction So Time Intensive
Clinical data abstraction transforms evidence from a patient’s medical record into structured data a registry can use. Trained clinical staff perform the work because it requires interpreting evidence in clinical context and applying detailed registry definitions. That evidence may be distributed across:
Physician notes
Nursing documentation
Procedure reports
Pathology results
Medication records
Flowsheets and scanned documents
What counts as thorough abstraction differs by registry:
GWTG-Stroke: reconciling several versions of a treatment timestamp
ACS NSQIP: reviewing documentation across the full surgical episode
NCDB: pathology, staging, and treatment information recorded across months of care
A 2023 JAMA analysis illustrates the broader burden. At the Johns Hopkins Hospital, one academic medical center, reporting 162 quality metrics required approximately 108,000 person-hours and $5 million in personnel costs in one year. Measures requiring manual chart review were among the most resource-intensive, consuming an estimated 691 person-hours per metric annually. Johns Hopkins Medicine has since centralized its clinical abstraction operations and, in 2025, began a multi-year collaboration with Layer Health to apply AI-guided abstraction across surgical and oncology registries.
The challenge is that these questions are clinically nuanced. Nurses are clicking through many notes by different providers, often with conflicting information.
David Sontag
,
CEO
Layer Health
Chart Abstraction Volume and Standardization Compound the Burden
Some records contain hundreds or thousands of documents, with evidence for a single variable appearing in several places. Growing case volume increases both search time and the need for quality review.
Maintaining standardization across abstractors, cases and facilities creates another challenge. Two experienced abstractors can review the same documentation and reach different conclusions when registry variables require clinical judgment. Changes to definitions and measure logic can introduce further variation, requiring teams to update internal guidance, retrain abstractors and monitor consistency over time.
Structured Data Still Requires Critical Interpretation Work
Interoperability and structured data exchange have improved access to clinical information, but registry abstraction still depends on understanding that information in context. An EHR may record one procedure-related time in a discrete field, while the registry requires a different milestone documented only in a narrative note.
Effective automation must connect evidence across both sources and apply the registry’s clinical definitions.
Limited Abstraction Capacity Becomes a Data Problem
Registry expertise develops through practice, so experienced abstractors take time to train and are hard to add quickly. When capacity is limited, quality leaders may:
Prioritize required registries
Abstract only a sample of eligible cases
Operate with backlogs that delay feedback to clinical teams
Manual abstraction also consumes expert hours that could be spent investigating performance, working with clinical teams and advancing quality improvement.
How AI-Guided Abstraction Changes the Operating Model
Layer Health’s AI-guided abstraction reasons across the full medical record, identifies evidence relevant to a registry variable and presents a proposed answer with its source for an expert abstractor to review.
The abstractor remains accountable for the final submission. Their time shifts toward reviewing the evidence surfaced by AI, resolving ambiguity and confirming the answer. Registry teams can validate performance on their own data, understand how each proposed answer was reached and maintain human oversight where clinical judgment is required.
Layer Health customers are abstracting cases 65% faster while maintaining 98%+ validated accuracy. That capacity allows them to:
Address backlogs
Absorb growing case volume
Increase registry coverage
Return expert capacity to the quality improvement work registry data is intended to support
By reducing administrative burdens and streamlining inefficiencies, we allow providers to focus on their ultimate priority—delivering exceptional patient care.
David Sontag
,
CEO
Layer Health
Clinical Registries Are a Proving Ground for Clinical AI
Clinical registries provide a rigorous, measurable environment for evaluating healthcare AI. Solutions must:
Produce clinically accurate answers
Show the supporting evidence
Maintain performance across large, inconsistent records and detailed registry specifications
Defined variables, expert review and measurable baselines for time and accuracy allow health systems to assess whether an AI-supported workflow improves productivity while preserving data quality.
Registry abstraction is one application of a broader capability. Layer Health's platform serves two connected functions:
Quality measurement establishes and validates clinical truth, an accurate and evidence-backed account of what happened in a patient's care.
Clinical guidance uses that foundation to help deliver the right care at the right time and place.
Registries are where quality measurement is proven. Establishing clinical truth reliably enough to satisfy a registry specification is what makes it possible to extend the same approach to quality measurement beyond registries, case finding and clinical guidance at the point of care.
AI-guided abstraction gives health systems a practical way to reduce registry burden, expand program capacity and put high-quality data to work sooner.
What Comes Next
Each registry concentrates the abstraction challenge in a different part of the workflow. For example:
GWTG Stroke requires reconstruction of time-critical events
GWTG Heart Failure depends on constantly changing definitions
ACS NSQIP spans an entire surgical episode
NCDB follows longitudinal cancer journeys
NCDR spans multiple procedure-specific registries, each with its own requirements for linking clinical records to follow-up outcomes
