What care gap closure involves, and why it is slow today

A care gap is a recommended service a member is eligible for but has not received yet: a colorectal cancer screening, a retinal eye exam for a member with diabetes, a controlled blood pressure reading, a follow-up after an emergency visit. HEDIS, the quality measurement set maintained by NCQA, includes more than 90 measures across domains such as effectiveness of care and access to care. Closing a gap means the service gets done, the evidence gets captured electronically, and the measure flips from open to met.

The work is slow because it is fragmented across teams and tools. A typical month looks like this:

  • The chart chase. Coordinators request records from physician offices, wait, then read charts by hand to find evidence that a gap is already closed but was never reported. This is the most expensive labor in quality and it peaks in a five-month season at the start of each year.
  • Reconciling supplemental files. Open gaps arrive from claims, but also from managed care organization (MCO) gap files, lab feeds, and payer supplemental extracts that each use a different member ID and a different update cadence. Reconciling them is manual and never finished.
  • Outreach by phone and spreadsheet. A coordinator builds a call list in a spreadsheet, dials down it, leaves voicemails, and re-dials. The list is stale the moment a member is reached, and three teams often call the same member in the same week for three different gaps.

None of this is a measurement problem. The plan usually knows the gap exists. It is an execution problem: the distance between knowing and done is full of manual steps.

The cost compounds. Chart-chasing labor is expensive and seasonal, so it pulls coordinators away from members for months. Stale call lists waste dials on members who already closed the gap or already moved. And because each team sees only its slice, the same member is worked more than once while a different member with a higher-value open gap is never reached at all. The work is busy without being effective.

Two vendors published survey data this year that describes the same thing from the other side of the table. Reveleer's 2026 State of Technology in Value-Based Care report, produced with Mathematica and fielded by Harris Poll among 200 senior payer and provider decision-makers between March 2 and 13, 2026, found that 94% of providers still rely on manual value-based-care processes and 93% report vendor overpromising, with only 13% of payers and 29% of providers feeling well prepared for current CMS requirements. Arcadia's survey of 281 healthcare leaders, published June 16, 2026, found that only 14% say AI insights are fully integrated into the decision points that matter, and named embedding AI into workflows as the top barrier at 31%. Both companies sell analytics, and both are describing an execution problem.

How AI changes care gap closure

AI compresses that distance. The pattern that works is not a smarter report. It is an agent that runs the full loop and hands a human the cases that need judgment.

Ingest and prioritize gaps across every source

The system pulls open gaps from claims, EHR, lab, ADT, and payer supplemental files into one canonical record per member, resolved across the different IDs. Then it prioritizes. Not every gap is worth the same call. The model weights each gap by its measure weight, by how close the member's panel sits to a Star cut point, by member reachability, and by whether the gap can be closed in a visit the member already has scheduled. A high-value gap on a member with a visit next week ranks above a low-value gap on an unreachable member.

Route each gap to the right place

Some gaps need a coordinator with clinical judgment. Many do not. A reminder to schedule a screening, a refill nudge, a record request to a practice, these can be handled by an agent. The system maps each gap to either a coordinator worklist or an autonomous agent, so human time goes to the cases that need a human.

Reach, schedule, document, close

The action layer makes the contact across the channel each member actually responds to: email, SMS, or an outbound voice call. It books the appointment, confirms it, and after the service is done it captures the evidence as structured, reportable data and closes the gap. The same record that opened the gap records its closure, with the source document attached.

Two details make this work in practice. First, deduplication. Because every gap and every contact attempt lives on one canonical record per member, a member with three open gaps gets one coordinated touch, not three uncoordinated calls from three teams in the same week. That alone removes a large share of the wasted outreach that plagues siloed quality programs. Second, closed-loop documentation. The evidence that closes a gap is captured at the moment of the action and attached to the record, rather than reconstructed from a chart weeks later. When the measure is computed, the proof is already in place.

Keep a human where judgment is required

Autonomy is not the goal; throughput with accuracy is. The agent handles the volume work: reminders, record requests, refill nudges, scheduling. A coordinator reviews anything that touches a clinical decision, an unusual member situation, or a low-confidence match. The division of labor is the point. People do the work that needs a person, and the system absorbs the work that never should have needed one.

+41%
Care gap closure improvement across Pelica deployments
3x
Outreach capacity per coordinator, with no new headcount
200,000+
Patients managed live across one risk-bearing contract

The ECDS shift, and why year-round digital quality matters

The way HEDIS is reported is changing in a way that rewards exactly this kind of execution. NCQA is moving measures from the hybrid model, which permits a once-a-year chart sample, to Electronic Clinical Data Systems (ECDS) reporting, which requires continuous electronic evidence for the full eligible population. The remaining hybrid measures retire by reporting year 2029.

That ends the five-month chart-chase season as a viable strategy. You cannot sample-and-project your way to an ECDS score; you need evidence flowing for every member, all year. A system that ingests, closes, and documents gaps continuously is no longer a nice-to-have. It is the reporting substrate. We covered the data-engineering reality of that shift in our ECDS transition guide, where supplemental data volumes rise 35x to 75x per measure.

Retiring hybrid is only half the timeline, and most vendor pages stop there. NCQA's stated goal is fully digital HEDIS by measurement year 2030. A digital quality measure, or dQM, ships as a self-contained package: human-readable specifications plus computable specifications in HL7 FHIR and CQL, so every vendor runs the same logic instead of each coding a PDF spec its own way. Both dates matter for different reasons: 2029 changes where the evidence has to come from, and 2030 changes how the measure itself is defined and computed.

The measure set is also getting shorter. The CY2027 Medicare Advantage and Part D final rule, published April 6, 2026, cuts 11 measures out of the Star Ratings calculation starting with the 2027 measurement period. The Diabetes Care Eye Exam measure, the running example in this article, was proposed for removal and survived. A depression screening and follow-up measure applies from the 2027 measurement year and first shows up in the 2029 Star Ratings. Fewer measures means each remaining one carries more of the rating, so every gap on a survivor is worth more per close.

Analytics versus execution

Most quality tools are analytics. They show you open gaps, a target, and a glide path. That is useful, but the dashboard does not pick up the phone, book the visit, or write the documentation. Those are the steps that actually move a measure, and they are where the time goes.

Nearly every quality tool can show you the gap. The question is whether it can close one. Can it make the call, schedule the appointment, capture the evidence, and reflect the closure in the measure without a coordinator stitching four systems together by hand? Vendors who have built automated abstraction report large efficiency gains on the reading work alone. Reveleer, for example, states a 75% increase in abstractor efficiency, records collected up to 80% faster, outreach efficiency up to 50%, EVE support for more than 40 quality measures, and one firm saving 1,400 hours with 26% higher retrieval rates. Those are real numbers on real work. Abstraction is one slice of the loop, and the loop only pays when it runs end to end, from open gap to documented close.

This distinction also explains why "more dashboards" rarely improves a quality score. Each new analytics tool adds another view of the same gaps and another tab for a coordinator to check. None of them reduce the number of calls that have to be made or charts that have to be requested. The score moves when the work gets done faster and with fewer hands, and that requires an execution layer sitting on top of the data, not beside it.

Analytics tells you the gap exists. An execution layer closes it. The measure only moves on the second one.

Every vendor says it closes gaps now. Four ways to tell them apart

In 2026 the claim itself got crowded, which has made the buyer's job harder. League's Spring '26 release, announced March 3, 2026, put a "Care Gaps Agent Team" in the market that checks coverage and provider availability and books appointments directly. Linear Health sells care gap closure automation to community health centers and clinic networks, ingesting payer gap files, reaching patients by voice, SMS and email, and documenting back to the chart. Innovaccer published a six-platform HEDIS comparison in July 2026 built on the same question this article asks, and answers it with the argument that one unified data model across risk and quality is what closes gaps.

That argument is half right, and we build on the same premise: one canonical record per member, so a plan is not retrieving and paying for the same chart twice. What a unified model does not settle is who performs the next action. Four checks separate the claims, and they apply to us as much as to anyone named above:

  • The channel and who operates it. Provider office, member phone, community pharmacy, in-home visit. A list of channels describes where work can happen. Ask who dials.
  • Time from event to first contact. Ask for it in hours, not adjectives. ADT-triggered outreach latency is the version of this that care management buyers already ask about, and quality buyers should borrow it.
  • Where the evidence is written back. The result has to land in the system of record at the moment of the action, as ECDS-eligible structured data, rather than being reconstructed from a chart the following spring.
  • Shared queue or one per program. Cross-payer gap consolidation means one member appears once across every contract and every program, so quality, pharmacy, and care management are not each calling her in the same week.

FHIR-native architecture answers a different question than any of these. It describes how a measure is stored and computed, which matters a great deal by 2030. It says nothing about whether the member picked up the phone.

Sources and further reading