Ask anyone who's worked in health plan quality operations about January through May, and they'll describe a specific kind of chaos. Chart chasers making thousands of calls to provider offices. Faxes coming in by the pallet. Abstractors retyping clinical information into HEDIS abstraction tools. Spreadsheets tracking which charts have been requested, received, reviewed, and loaded. Deadline pressure rising as the compliance audit date approaches.
The function exists because HEDIS hybrid measures require clinical documentation beyond what claims data captures. The plan needs to prove the mammogram happened, the blood pressure was controlled, the diabetes care was provided. Claims say payment was made; they don't always say what was clinically documented. So every year, the plan retrieves a sample of charts, abstracts the relevant clinical elements, and submits the results.
This is one of the great hidden operational drags in health plans. A single large plan may spend tens of millions of dollars on chart chase and abstraction annually. The work is labor-intensive, error-prone, and creates significant provider abrasion. And for most of the last 20 years, the alternatives were theoretical.
That's actually changing now. The combination of mature interoperability APIs, AI document processing, and the industry shift toward digital quality measures is making most of the manual chart chase function either automatable or obsolete. The plans that move aggressively are eliminating a substantial cost center while improving data quality. The plans that don't will still be running the manual sprint in 2030.
What the manual process actually costs
The direct costs are substantial but only part of the picture:
| Cost category | Typical scale | Hidden cost |
|---|---|---|
| Chart chase labor | $5-15M annually for mid-size plans | Provider abrasion, declining response rates |
| Abstraction labor | $3-8M annually | Error rates, rework on audit findings |
| Technology licensing | $500K-2M | Integration complexity, workflow fragility |
| Vendor services | $2-10M for hybrid approach | Quality variability, accountability gaps |
| Provider office burden | Borne by providers | Damages network relationships, affects contracts |
| Audit preparation | $1-3M | Diverts clinical staff time |
| Missed measure performance | Variable, can be millions | Stars rating impact, quality bonus erosion |
The FHIR dimension
The regulatory push toward FHIR-based data access — CMS interoperability rules, provider API requirements, state-level mandates — is finally producing real infrastructure for programmatic clinical data access. For HEDIS and Stars purposes, this changes the game.
- Patient Access APIs (CMS-9115-F) require payers to provide member access to their clinical data. The infrastructure built for this also supports internal use for quality measurement.
- Provider Access APIs extend similar access to providers, which in turn creates bidirectional clinical data flows.
- Payer-to-Payer APIs (CMS-0057-F) require payers to share clinical data on member transitions. The clinical data history that accumulates has quality measurement utility.
- Bulk FHIR allows efficient extraction of clinical data for populations rather than individual-patient queries. This specifically matters for measurement use cases.
The plans building production FHIR capability for regulatory compliance reasons are discovering that the same infrastructure eliminates significant chart chase work. Clinical elements needed for HEDIS — vitals, labs, diagnoses, procedures, medications — are retrievable through API for the portion of the provider network that supports FHIR, and that portion is growing.
AI document processing for the non-FHIR portion
Not every provider supports FHIR yet. For the non-FHIR portion of the network, clinical records still arrive as PDFs, faxes, and unstructured documents. This is where AI document processing has matured enough to replace manual abstraction.
- OCR quality on medical documents. Modern OCR handles handwriting, stamps, multi-column layouts, and poor-quality faxes at accuracy rates that make automated extraction viable. Edge cases still need human review, but the human review volume is much lower.
- Clinical NLP. Extracting clinical concepts from notes — the diagnosis, the procedure, the finding — has reached accuracy levels suitable for HEDIS abstraction in most measure types.
- Measure-specific logic. HEDIS specifications are specific — a mammogram has to meet specific criteria to count. AI abstraction embeds the measure logic, not just text extraction.
- Confidence scoring. High-confidence extractions can be accepted automatically; low-confidence cases route to human abstractors. This dramatically reduces the abstractor workload.
- Audit trail. Every extraction is traceable to the source document with the specific text that supported the decision, which is essential for audit defense.
The operational transition
The transition from manual to automated chart chase is operationally complex. Plans can't flip a switch — they have a compliance requirement that has to be met every year, and the existing process has to work until the replacement is proven.
- Phase 1: Instrument the current process. Before transforming anything, instrument. How long does each step take? Where are the bottlenecks? What are the error rates? Which providers are responsive? This baseline is essential and rarely exists in clear form.
- Phase 2: Deploy FHIR retrieval for capable providers. Start with the providers that support FHIR. Build the retrieval, map clinical elements to measure requirements, validate against manual processes, and prove the automated path produces equivalent or better results.
- Phase 3: Deploy AI abstraction for incoming documents. For documents that still arrive via chart chase, run them through AI abstraction. Compare results to manual abstractors. Tune until accuracy meets audit standards. Transition abstractors to QA and exception handling roles.
- Phase 4: Shrink the chart chase footprint. With automated paths working for the majority of the work, the residual manual chase should shrink significantly — focused on the most challenging providers and edge cases.
- Phase 5: Prepare for dQM transition. Digital quality measures use different data structures and workflows. Plans on the automated track are well-positioned; plans still running manual chase will struggle with the transition.
The hybrid vs. pure-digital question
The industry is transitioning from traditional hybrid HEDIS measures (which require chart review for a sample) to digital quality measures (which use direct clinical data access). The transition is multi-year but directional.
Clinical data infrastructure becomes central to quality measurement, not a supplement to claims
FHIR capability moves from "nice to have" to "required operational capability"
Chart chase and manual abstraction become declining functions
Data quality at the source (provider EHRs) becomes more visible and consequential
Provider data quality coaching becomes a plan quality activity
The plans investing in FHIR and clinical data integration now are preparing for the operational model dQMs require
The plans relying on mature chart chase operations are preparing for a transition they haven't started.
The provider burden dimension
Chart chase is a significant burden on provider offices. A single provider may receive hundreds of record requests during chase season, each requiring staff time to pull, copy or fax, and track. For small practices, this can be overwhelming. For large health systems, it's a significant administrative cost that's typically absorbed rather than billed.
Providers have legitimate complaints about chart chase. The same chart is requested repeatedly across plans. Requests arrive with unclear authority and vague timelines. Responses aren't acknowledged. The same records are requested year after year.
Automation that shifts to FHIR-based retrieval eliminates most of this burden — the provider system answers the query without staff involvement. This matters for provider relationships, particularly in value-based arrangements where the plan wants the provider aligned on quality goals rather than antagonized by chase operations.
The measurement quality improvement
Automated chart chase isn't just cheaper — it can be more accurate. Manual abstraction has known error rates, particularly under time pressure at year-end. Chart chasers sometimes collect the wrong document. Abstractors sometimes miss elements. Timing pressure compounds errors.
AI-based extraction, properly deployed, can be more consistent. Every document is processed against the same measure logic. Every element is extracted with the same criteria. Audit trails are complete. The human review that happens is focused on genuinely ambiguous cases where judgment adds value, not on repetitive extraction of obvious elements.
This improvement in data quality matters for more than cost. Stars ratings depend on measure performance. Measure performance depends partly on how accurately care is captured. Plans leaving measurable care uncaptured through flawed chart chase are leaving Stars performance on the table.
The transformation of HEDIS and Stars operations isn't a marginal cost-out — it's a structural shift in how health plans capture and validate clinical information. Plans that invest now eliminate a significant cost center, improve measure performance, reduce provider abrasion, and prepare for the digital quality measure future. For leadership teams assessing where HEDIS operations, clinical data infrastructure, and quality measurement capabilities sit within the broader health plan operating model, the Health Insurance Capability Model maps the capabilities — clinical data access, document AI, measure logic, audit readiness, provider integration — that determine whether quality operations remain a year-end scramble or become a year-round data function.