HEDIS & CMS Star Improvement Cycles
John Frias Morales, DrBA, MS
DIRECTOR OF ANALYTICS | Value-Based Care, Healthcare Data Governance, Analytic Excellence | Lead the Charge in Optimizing Value-Based Care through Data-Driven Strategies and Innovations
A mindset of HEDIS & CMS Star compliance is holding back payers from achieving 90th percentile quality measure scores and attaining 5-Star Ratings. Tech feature releases that keep the lights on for quality measure platforms rarely address the core issue of significantly improving outcomes. For instance, increasing glycemic control from the national Medicaid average of 48% to the gold standard 90th percentile of 60% requires a targeted care delivery approach focused on outcomes and result (Medisolv, 2024; NCQA, 2025, Integrated Healthcare Association, 2024). ?How would you change provider workflows to improve Star ratings if presented with an uncontrolled diabetic who is readmitted for DKA?
By incorporating provider-centric tech components, payers can transform their quality measure platforms from mere compliance repositories into powerful tools that drive meaningful, sustained improvements in patient care and achieve their strategic goals. This necessitates a shift in focus towards actionable insights. HEDIS and CMS quality measure platforms must evolve to:
Quality Measure Cycle (graphic above)
In contrast to a purely technology-centric approach focused on HEDIS & CMS Star data compliance, a robust quality measurement cycle drives meaningful outcomes. By meticulously analyzing provider touchpoints across the care continuum (risk subgroups, well visits, controlling diagnoses, assessments, screenings, immunizations, vaccinations, testing, adherence, surveys, transitions, and follow-ups), we identify opportunities for improvement in CMS Star ratings and HEDIS scores. For example, this cycle can proactively address uncontrolled diabetic readmissions by pinpointing gaps in care and facilitating timely interventions.
Leveraging a sophisticated technology infrastructure (membership, claims, encounters, surveys, EHRs, digital communication platforms, ADT, and payer-to-payer data exchange) and standardized data models (FHIR-eCQM, HQMF, QDM, CQL, USCDI), we capture, analyze, and report on a comprehensive set of quality measures. This data-driven approach enables us to: 1) audit and certify data integrity; 2) benchmark performance against industry standards; 3) segment populations by demographics and acuity; and 4) facilitate data submission to key stakeholders (CMS, state agencies, payers, ACOs, HIEs). Furthermore, we actively engage providers through prospective record reviews, identifying patients who require interventions to address care gaps (e.g., missed well visits, overdue screenings).
Advanced analytics, including NLP and LLMs, are employed to optimize chart abstraction, identify impactful interventions, and personalize outreach strategies for complex patients. This proactive, data-driven approach fosters a culture of continuous improvement, ultimately enhancing patient outcomes and driving value-based care.
COTS Vendors Vs Open-Source In-House
A majority of commercial and public payers rely on 91 NCQA-certified Software-as-a-Service vendors (e.g., Inovalon, Cotivity, IQVIA, Cerner, Epic, IBM Watson, Cognizant) for their end-to-end HEDIS reporting needs (NCQA, 2021). These vendors facilitate data exchange by connecting to provider systems through interoperability standards like FHIR-eCQMs, enabling the transfer of low-grain EHR data sets. Payers typically process an average of 200 events per member, subsequently loading this data into their longitudinal patient records (e.g., data lakes, EDW, or ODS repositories). To ensure data quality, these repositories undergo rigorous validation with over 1,000 integrity checks to scrub and harmonize the data to standardized formats. Notably, the Inovalon One platform enjoys widespread adoption, serving all of the top 25 commercial insurance plans and top 25 pharmaceutical companies (Inovalon Holdings Inc, 2020).
Alternatively, the NCQA offers a suite of open-source tools, including the CQL Execution Engine, CQL Exec FHIR, and the Digital Starter Pack, to facilitate the processing of digital quality measures. These tools empower organizations to automate quality measure calculations, with a notable adoption rate: 94% of commercial payers, 75% of Medicaid plans, and 50% of Medicare plans currently utilize Electronic Clinical Data Submission (ECDS) for HEDIS reporting (NCQA, 2025). This approach leverages diverse data sources, including claims (33%), EHRs (25%), case management (1%), and registries (41%) (NCQA, 2024). While promising, the open-source approach may have limitations. It may not inherently include features such as complex integration mapping, rigorous data integrity checks, robust data repository management, comprehensive benchmarking capabilities, or advanced predictive analytics, which are often integral components of commercial vendor offerings.
Year-Round Chart Review
NLP, LLMs, and AI algorithms possess the capability to analyze unstructured text data within electronic health records (EHRs), such as physician notes, laboratory reports, and imaging results. This advanced technology enables proactive identification of care gaps by analyzing patient data in real-time. For example, these algorithms can identify patient subgroups at risk of missing critical care, such as those who have not received annual checkups, immunizations, or necessary screenings. This proactive identification allows for timely interventions to address these gaps in care.
Furthermore, these algorithms can effectively review and enhance the accuracy of medical documentation. By analyzing imperfect or incomplete records, they can guide abstractors towards the necessary information to accurately capture quality measures. For instance, Astrata (2022) demonstrated the significant impact of NLP solutions, achieving a 7x increase in chart review efficiency.
Manual Chart Review (MCR) for surgical data extraction presents inherent challenges, including time-consuming processes, human error, and inconsistencies. Dagli et al. (2024) developed an algorithm that automates this process by employing rule-based NLP to identify and classify relevant information within operative notes. This automated approach has significant implications for HEDIS reporting, particularly for surgical quality measures, such as complication rates following specific procedures. By accurately and efficiently extracting data from operative notes, this technology enhances the accuracy and reliability of HEDIS reporting, ultimately improving the overall quality of care.
Targeted Abstraction
Year-round prospective chart abstraction, utilizing a combination of Natural Language Processing (NLP), Large Language Models (LLMs), and Artificial Intelligence (AI), involves the continuous and ongoing process of extracting relevant medical information from patient records. These sophisticated algorithms efficiently condense complex medical information into concise medical codes, enabling potential adjustments to care plans and performance measures.
LLMs have demonstrated significant potential in automating this critical process. For example, the LLM system developed at UC San Diego analyzes patient data within the standardized Fast Healthcare Interoperability Resources (FHIR) format (Boussina et al., 2024). The LLM then extracts relevant information from the patient's electronic health record (EHR) to determine compliance with established quality measures. This automation significantly streamlines the process, reducing the time and resource demands associated with traditional manual chart review, a historically labor-intensive and costly aspect of HEDIS reporting.
Furthermore, as demonstrated by Schuemie et al. (2025), LLMs can be effectively applied to automate the identification of patients who meet specific HEDIS criteria. For instance, these algorithms can accurately identify patients with diabetes who have received the required annual eye exam.
Impactability
Impactability or propensity to succeed refers to the likelihood that a patient will benefit from specific interventions aimed at reducing readmissions and improving health outcomes. Impactability is defined as “the identification of patients most likely to respond to care, based on quantitative and qualitative factors” (Health Economics Unit, 2022). This concept helps healthcare providers prioritize resources and tailor interventions. To learn more about Impactability, see Advances in Transitional Care
Identity-driven engagement & personalization Platforms
A new class of identity-driven engagement platforms is dedicated to personalizing provider experiences with plans/payers. Numerous companies have developed intelligent engagement platforms that enhance provider experiences with payers. These tools consolidate data from portals, call centers, mobile apps, provider credentialing & contracting systems, surveys, and various other sources. For instance, the Praia Health Identity & Engagement Platform streamlines service delivery by crafting personalized profiles (Gildehaus, Allred, Naidoo, and Poduturri, 2021). These profiles amalgamate preferences, service experiences, and other data sources to establish a learning loop regarding stakeholders. The simplified model of the Praia Health playbook is depicted below. For more information, see Psychographic Profiling
AI-Driven Value
A strong correlation exists between higher CMS Star Ratings and increased Medicare Advantage enrollment. Reid et al. (2013) found that each star increase is associated with a significant 9.5 percentage point jump in enrollment likelihood, underscoring the critical role of quality and member experience in driving plan selection and retention.
Leveraging AI technologies can significantly enhance quality improvement initiatives. Basu et al. (2023) demonstrated the potential of Large Language Models (LLMs) to personalize member engagement strategies by measuring the effectiveness of interventions along different pathways. Moreover, AI has proven valuable in areas such as social service matching and targeting early interventions, including closing care gaps, improving medication adherence, and providing social support.
Research by Baum et al. (2024) further highlights the impact of such interventions. Their study observed an 11.8 absolute percentage point increase in HEDIS scores, coupled with a 22.9% reduction in all-cause acute events, a 48.3% decrease in ambulatory care-sensitive hospitalizations, and a 20.4% reduction in ambulatory care-sensitive ED visits among patients receiving these interventions.
领英推荐
Definitions
Fast Healthcare Interoperability Resources Electronic Clinical Quality Measure FHIR-eCQM are focused on measuring and reporting on healthcare quality.
Health Quality Measure Format or HQMF is a XML technical standard for representing eCQMs in a machine-readable format, and a source of metadata and specification detail. ?
Quality Data Model or QDM is a data model used to represent clinical data for quality measurement. ?
Clinical Quality Language or CQL is a language for expressing clinical logic in a human-readable way and is queryable. ?
US Core Data for Interoperability or USCDI is a set of standardized data elements that should be routinely available for electronic exchange.
Electronic Clinical Data Systems or ECDS is a method for collecting and reporting structured electronic clinical data for HEDIS quality measures using EHR clinical sources.
Summary
This article highlights that current IT systems often focus on data visualization and exchange, but lack the tools to significantly improve patient care. The article suggests that by incorporating provider-centric IT components, payers can transform their quality measure platforms into tools that drive meaningful, sustained improvements. This includes pinpointing providers for improvement, enabling year-round prospective record review, supporting targeted abstraction with NLP and LLM, and providing personalized engagement recommendations. The article emphasizes the potential of AI technologies like LLMs to personalize member engagement, improve care coordination, and enhance quality improvement initiatives. Finally, the article discusses how NLP, LLMs, and AI can be used for targeted abstraction, year-round chart review, and automated data extraction, ultimately leading to more accurate and efficient HEDIS reporting and better patient outcomes.
Sources
Astrata (2022). Scaling up year-round HEDIS with NLP insights: case study and ROI analysis. Downloaded from https://www.astrata.co/case-studies/scaling-up-year-round-hedis-with-nlp-insights/
Basu, S., Schillinger, D., Patel, S., and Rigdon, J. (2024). Simulating A/B testing versus SMART designs for LLM-driven patient engagement to close prevention care gaps. Digital Medicine, 7, 322. Downloaded from file:///C:/Users/jmorales3/Downloads/s41746-024-01330-2.pdf
Baum, A., Batniji, R., Ratcliffe, H., DeGosztonyi, M., and Basu, S. (2024). Supporting rising-risk Medicaid patients through early intervention. NEJM Catalyst, 5, 11. Downloaded from https://catalyst.nejm.org/doi/full/10.1056/CAT.24.0060
Bousina, A., Krishnamoorthy, R., Quintero, K., Joshi, S., Wardi, G., Pour, H., Hilbert, N., and Nemati, S. (2024). Large language models for more efficient reporting of hospital quality measures. NEJM AI, 1, 11. Downloaded from https://ai.nejm.org/doi/full/10.1056/AIcs2400420
Dagli, M., Ghenbot, Y., Ahmad, H., Chauhan, D., Turlip, R., Wang, P., Welch, W., Ozturk, A., and Yoon, J. (2024). Development and validation of a novel AI framework using NLP with LLM integration for relevant clinical data extraction through automated chart review. Nature Scientific Reports, 14, 26783. Downloaded from https://www.nature.com/articles/s41598-024-77535-y
Gildehaus, C., Allred, D., Naidoo, E., and Poduturri, A. (2021). Powering the innovation flywheel in the digital era. Boston Consulting Group. Downloaded from https://www.bcg.com/publications/2021/driving-business-impact-with-the-innovation-flywheel-approach
Health Economics Unit (2022). Building on expert knowledge: research and next steps on impactability in health and care. NHS Midlands and Lancashire. Downloaded from https://healtheconomicsunit.nhs.uk/wp-content/uploads/2024/05/Impactibility-report.pdf
Inovalon Holdings Inc (2020). SEC form 10-K annual report. Downloaded from https://www.sec.gov/Archives/edgar/data/1619954/000161995421000042/inov-20201231.htm
Integrated Health Association (2024). Align measure perform APM incentive design technical documentation. Downloaded from https://www.iha.org/wp-content/uploads/2025/01/AMP-Incentive-design-technical-documentation_final_Dec2024.pdf
Medisolv (2024). 2024 star ratings calculation. Downloaded from https://blog.medisolv.com/articles/2024-star-ratings-calculation
NCQA (2021). HEDIS and AMP vendor certification status. Downloaded from https://wpcdn.ncqa.org/www-prod/wp-content/uploads/2021/10/MY2021-Measure-Certification-Vendor-List.pdf
NCQA (2024). Special report: results for measures leveraging ECDS for HEDIS. Downloaded from https://wpcdn.ncqa.org/www-prod/wp-content/uploads/Special-Report-Nov-2024-Results-for-Measures-Leveraging-Electronic-Clinical-Data-for-HEDIS.pdf
NCQA (2025). Helping states move towards a digital quality system. Downloaded from https://www.ncqa.org/helping-states-move-towards-a-digital-quality-system/
NCQA (2025). Hemoglobin A1c control for patients with diabetes (HBD). Downloaded from https://www.ncqa.org/hedis/measures/hemoglobin-a1c-control-for-patients-with-diabetes/
NCQA (2025). Plan all-cause readmissions (PCR). Downloaded from https://www.ncqa.org/hedis/measures/plan-all-cause-readmissions/
Reid, R., Deb, P., Howell, B., and Shrank, W. (2013). Association between Medicare Advantage plan star ratings and enrollment. JAMA, 309, 3, 267-274. Downloaded from https://pubmed.ncbi.nlm.nih.gov/23321765/
Schuemie, M., Ostropolets, A., Zhuk, A., Korsik, U., Seo, S., Suchard, M., Hripcsak, G., and Ryan, P. (2025). Standardized patient profile review using large language models for case adjudication in observational research. Digital Medicine, 8, 18. Downloaded from https://www.nature.com/articles/s41746-025-01433-4