How Can Hospitals Reduce the Burden of Quality Measure Reporting?

The quality reporting burden on U.S. hospitals has compounded steadily over the last decade. CMS, Leapfrog, Joint Commission, and the specialty registries each add to the abstraction queue, and the cumulative effect is now unsustainable at most quality departments.

What does the quality reporting burden actually look like?

A modern U.S. hospital reports to a portfolio of programs that no single document fully enumerates. On the federal side, the CMS Hospital Inpatient Quality Reporting (IQR) Program covers chart-abstracted process-of-care measures (SEP-1 sepsis, perinatal care, elective delivery), a growing set of electronic clinical quality measures (eCQMs), structural measures (patient safety, age-friendly hospital), claims-based outcomes (readmissions, mortality, complications), the HCAHPS patient experience survey, and hybrid measures that combine electronic data with claims data. CMS Hospital Outpatient Quality Reporting (OQR) and the Promoting Interoperability Program add further measures. The Joint Commission ORYX performance measurement program contributes its own core measure sets, with overlap and divergence from CMS at specific data elements.

On the registry side, the American Heart Association's Get With The Guidelines registries cover stroke, heart failure, and resuscitation, each with its own specification manual. Specialty society registries (the ACC NCDR for cardiovascular, the STS National Database for thoracic surgery, the ACS NSQIP for general surgery, and analogous registries in oncology, orthopedics, and others) layer additional abstraction requirements scoped to the procedures and clinical populations each registry covers. State hospital association registries add state-level requirements that vary by jurisdiction. Payer-driven quality programs (commercial bonus tiers, Medicare Advantage quality programs) add commercial-side reporting obligations on top of the federal stack. Leapfrog adds its own.

The empirical scale of the resulting burden is documented in the peer-reviewed literature. Casalino et al., 2016, published in Health Affairs, surveyed primary care, cardiology, orthopedic, and multispecialty practices across the United States and found that physician practices spend an average of 785 hours per physician per year dealing with the reporting of quality measures. Practices reported 15.1 hours per physician per week on quality reporting work: 2.6 hours by the physician personally, and 12.5 hours by clinical and administrative staff. The largest single component, 12.5 hours per physician per week, was data entry into medical records solely for quality reporting. The annual aggregate cost across the four specialties surveyed was $15.4 billion. A subsequent 2019 estimate by Shrank and colleagues, published in JAMA, raised the figure to $17.6 billion for physician quality reporting alone. The hospital-side burden, which covers a larger surface area, is structurally larger.

At the staffing level, the American Hospital Association's 2025 Cost of Caring report notes that the average U.S. hospital now employs roughly 64 administrative and billing staff, approximately 6.5% of total hospital employment. Quality reporting is a substantial slice of that figure. The credentialed nursing staff who perform chart abstraction, often with additional credentials such as CPHQ, sit in a parallel structure with its own headcount that has compounded steadily as the measure portfolio has grown.

Why does the burden keep growing?

The structural drivers are consistent across programs and persistent over time. New measures are added more often than old measures are retired. CMS layers new measures onto existing programs as part of annual rulemaking, including new electronic clinical quality measures, new structural measures, new patient-reported outcome measures, and new hybrid measures. The Joint Commission updates ORYX core measure sets on its own cadence. Specialty societies add new registry data elements as the underlying clinical evidence base evolves. Payers introduce new quality program tiers as part of value-based contracting. The aggregate work expands, year over year. The 2016 Casalino survey found that 81% of practices reported spending significantly more time on quality measure reporting than they had three years earlier; a decade later, no published estimate suggests that trend has reversed.

CMS's electronic clinical quality measure (eCQM) ramp is a concrete and well-documented example. Per the agency's published Hospital IQR Program Guides, the number of mandatory eCQM submissions a hospital must report is set to roughly quadruple between 2024 and 2028. The increase is intentional: CMS is shifting more of its quality measurement away from chart-abstracted measures toward eCQMs and hybrid measures that draw partially on structured electronic data. The transition is meant to reduce burden over the long term but does not, in the near term, replace the existing chart-abstracted obligations. It runs in parallel.

Beyond the federal ramp, two further dynamics push the burden upward. The first is documentation depth per measure. As measures become more clinically specific, the documentation required to support them gets longer and more granular. A single CMS SEP-1 sepsis case now generates the equivalent of roughly 177 single-spaced pages of clinical documentation, growing across the past decade as sepsis protocols themselves have become more detailed. The second is specification update cadence. CMS, The Joint Commission, and the specialty registries all publish specification updates on their own schedules. A hospital's quality department, or any vendor it uses, must track those updates and apply them to ongoing abstraction work without dropping accuracy on the prior version. The work of tracking updates is itself a non-trivial program management discipline.

The Casalino survey finding that only 27% of practices believed current quality measures correlated with quality care is a separate critique that has been raised consistently in the literature since at least 2016. Whether the burden is justified by the value of the measurement is a policy question. For the operational question this article addresses, the burden is the input regardless of the policy answer.

Where is meaningful operational relief actually possible?

Four categories of relief are sometimes proposed. The first is measure standardization and harmonization: getting CMS, payers, and registries to align on a smaller set of shared measures with shared specifications. The case for it is strong and the policy momentum is real, but the actual progress over the past decade has been incremental. The second is EHR direct collection through eCQMs, which moves measure data from manual chart abstraction into structured electronic data the EHR is already capturing. The transition is real but partial: eCQMs only cover the data the EHR is configured to capture correctly in structured form, and the most clinically important data elements (the timing of clinical events, the documented reasoning behind treatment decisions, the chronology of how care actually unfolded) remain in the unstructured narrative. The third is outsourcing to the registry itself, available from some specialty registries as a paid service. It transfers work but does not reduce the underlying volume.

The fourth category, and the one that has changed materially in the past three years, is reasoning-based automated abstraction: AI systems that read the full medical record, reason against the current measure specification, and return every required data element with a citation to the source documentation. The category is distinct from the older rule-based extractors that have been on the market for two decades. Rule-based extractors pattern-match structured fields and return blanks on the hard fields buried in clinical narrative; abstractors then do the hard fields manually. Reasoning-based abstraction reads the unstructured narrative the way a credentialed abstractor reads it, applies the specification the way a credentialed abstractor applies it, and produces the structured output with full audit traceability. The architectural distinction is treated at length in the companion article on how automated clinical quality abstraction works.

What reasoning-based automated abstraction actually relieves is specific. It removes per-chart data entry, which is the dominant time component of the Casalino estimate (12.5 of the 15.1 hours per physician per week). It removes the rework cycle that follows registry rejections of incomplete or specification-violating submissions, which compounds the original abstraction time by weeks per quarter. It tracks specification updates as they are published, applying them to ongoing abstraction work without requiring manual model retraining or per-abstractor retraining. And it prevents the abstractor backlog that forces shortcuts in submission, which is the leading cause of the rework cycle in the first place. What it does not relieve, deliberately, is the work that requires clinical and quality judgment: strategic measure selection, cross-functional improvement initiatives, complex case adjudication, and denial and appeal logic on individual cases. The clinical quality function does not disappear. The repetitive data-entry component of the function does.

The staffing implication, on the ground, is usually redeployment rather than headcount reduction. Most hospitals deploying reasoning-based abstraction report that credentialed nursing staff move from data entry to data governance: sampling and methodology audits to verify continued accuracy, working the small subset of cases the system flags as low-confidence, owning the relationship with the measure specification itself, and engaging directly with the clinical workflow improvement initiatives that the measures are designed to drive. The clinical nursing skill set is too scarce and too valuable to reduce by attrition when it can instead be redirected toward work that demands clinical judgment. The cost benefit, in practical financial terms, is the prevention of staffing growth that would otherwise be required to absorb the eCQM ramp, the measure additions, and the documentation depth increase. It is also the improvement in retention that follows from removing the per-chart data-entry component of a credentialed clinician's job, which has been documented in the burnout literature for over a decade.

The strict precondition on any of this is that the automated system has to be accurate enough at production scale to make routine per-chart human review unnecessary. A system whose accuracy is below the federal validation threshold (the CMS Hospital IQR Validation Program's 75% upper-bound confidence interval) is not a relief mechanism; it is an additional layer of review work bolted on top of the existing burden. A system whose accuracy substantially exceeds credentialed human inter-rater reliability, audited longitudinally and adjudicated independently, allows the workflow shift from data entry to data governance to actually happen in production. The accuracy methodology that supports that shift is treated at length on the clinical quality product page and in the companion accuracy methodology research.

About the author

Allen Fienberg, PhD

Chief Strategy Officer, AccuCode AI

Dr. Fienberg co-founded Intra-Cellular Therapies in 2002, leading business development and investor relations as the company grew from startup to a multi-billion-dollar NASDAQ-listed enterprise (ITCI) and developed CAPLYTA, the FDA-approved treatment for schizophrenia and bipolar depression. He holds a Ph.D. in Human Genetics from Yale and conducted postdoctoral research at The Rockefeller University under Nobel laureate Paul Greengard. At AccuCode he leads strategy and healthcare partnerships, with particular attention to the patterns that recur whenever a federally-regulated industry has to choose between absorbing rising compliance burden as labor cost or restructuring it through technology investment that lifts the burden off credentialed staff.

View full bio →