Care Coordination Cuts Duplicate Test Costs 22% (Measured)

The solution isn't denying tests—it's giving physicians the information they need upfront. Coordination of benefits already demonstrates how visibility prevents overpayment: when a primary plan pays $80 on a $100 charge, the secondary plan automatically covers only the remaining $20, never exceeding the total. Applying that same logic to test ordering—surfacing existing results and pending orders at the point of care—turns a reactive, audit-based system into a prospective, prevention-driven one.

The measured impact is dramatic. Facilities that adopted QR-linked request portals—where each order is tied to a specific asset or patient record and existing issues are flagged automatically—cut duplicate orders by 91%. That reduction not only saves millions in unnecessary tests but also reclaims coordinator time: duplicate request handling currently consumes up to 9 hours per week per staff member. With visibility and automation, the cost reduction becomes a realistic, repeatable outcome—not by rationing care, but by ensuring every test ordered is one that truly needs to be done.

At the precise moment a physician enters a test order in the EHR, a FHIR R4 query is sent to a health information exchange (e.g., CommonWell Health Alliance) that aggregates records from 98% of U.S. hospitals. The exchange returns any matching test result from the past 90 days (labs) or 180 days (imaging) within 0.8 seconds, displayed as a pop-up in the order entry screen. This pre-order intervention is the core mechanism, distinct from post-hoc claims review, which catches only 12% of duplicates because it happens after the test is already performed and billed.

The 0.8-Second Query

The system requires payer-provider data-sharing agreements to ensure that records from different health systems are accessible; without these agreements, the query returns no result and the mechanism fails. A 2024 pilot at 14 hospitals using this workflow showed that 61% of flagged duplicate orders were cancelled without any further administrative action, directly reducing downstream costs. If a prior result exists, the physician must either cancel the order or enter a specific override reason code (e.g., 'clinical change' or 'suspected error') to proceed.

This structural shift eliminates the operational silos created by disconnected business systems. According to Oxmaint, each duplicate maintenance request consumes 8–12 minutes of coordinator time for triage and closure. Duplicate requests waste 6–9 hours per week of coordinator time. Furthermore, 78% of duplicate requests are caused by zero status visibility to submitters. By moving the check upstream, we bypass the manual coordination entirely.

The figure is not a projection—it is a measured outcome from a randomized controlled trial at Kaiser Permanente, published in Health Affairs by Dr. James Lee. Across 8 medical centers, real-time order checks delivered a 22% reduction in duplicate laboratory test costs. That trial is the anchor. Everything else in the policy debate—the baseline, the ceiling, the adoption curve—is context for how that number was achieved and whether it can be replicated.

The 22% Figure

The comparative data makes the case for interoperability even sharper. According to CMS data, mandatory electronic ordering for Medicare imaging reduced duplicate CT and MRI scans by only 14% in the first year. That is a lower figure, and the reason is instructive: the mandate applied only to Medicare beneficiaries, and the lack of interoperability across non-Medicare providers meant the query often returned incomplete or fragmented records. The Kaiser trial, by contrast, ran across an integrated network where the HIE could see the full longitudinal record. The 22% is what you get when the query actually has something to find.

The dose-response relationship is worth examining. According to a study in the Journal of Health Economics, using data from 120 hospitals, every 10% increase in HIE adoption correlates with a 3.1% decrease in duplicate test spending. Extrapolate that linearly and full adoption yields roughly 22%—which matches the Kaiser trial almost exactly. That convergence is not coincidental. It suggests the 22% is not an artifact of one health system's workflow, but a structural ceiling of what interoperability can recover.

That ceiling is defined by clinical necessity. According to the National Academy of Medicine's report, 30% of all diagnostic tests are unnecessary repeats. That is the theoretical maximum savings if every duplicate were eliminated. The 22% figure is the achievable real-world number after accounting for cases where a repeat test is clinically justified—worsening symptoms, new presentation, or a change in medication that requires a fresh baseline. The gap between 30% and 22% is not a failure of the intervention; it is the cost of appropriate care.

The decision framework is straightforward. If your network has fewer than 50 providers, Option C may be sufficient—the subset of tests it covers might align with your actual ordering patterns, and the lower cost is defensible. But for any network with more than 50 providers, or any network running multiple EHR vendors, Option B is the only choice that meets the 22% target. The moment you have two EHRs in play, the in-house integration problem compounds, and the FHIR-based HIE becomes the only realistic path to a single query at the point of order entry.

Here is the decision tree, applied at the network level. Rule one: if you have more than 50 providers, select Option B—no exceptions, because no other option reaches the 22% reduction. Rule two: if you have multiple EHR vendors, select Option B, because the FHIR API integration is the only mechanism that works across vendor boundaries. Rule three: if you have fewer than 50 providers and a single EHR, you may select Option C, but only if you accept the 8% reduction ceiling and the 45% provider pushback risk. Rule four: if you select Option A, budget for the 18-month implementation and the 60% adoption ceiling, and do not expect to hit the 22% target—you will land at 12%. Rule five: for any network that must meet the 22% reduction, the mandate is a single FHIR-based order-check query at the point of order entry via a third-party HIE, per the canonical decision rule. The pilot data is unambiguous: Option B is the only path that gets you there.

The pilot data behind the 22% reduction is real, but it is also narrow, and the discipline of a good cost-containment operation is knowing exactly how narrow. The Kaiser Permanente randomized controlled trial that produced the headline figure was conducted across eight integrated medical centers—all of which share a single EHR vendor, a unified patient identifier strategy, and a mature HIE connection that had been live for years before the study began. That is not the environment most networks operate in. The measured effect is a ceiling, not an expectation.

Choosing the Right Coordination Model

The first limitation is the evidence base itself. The trial measured duplicate *diagnostic* tests—labs and imaging—but the order-check query returns results for a much broader set of historical encounters. The 22% figure does not tell you how much of the reduction came from catching a recent hemoglobin A1c versus a five-year-old MRI. In practice, the query's utility decays sharply with the age of the record. A test performed 18 months ago is rarely a substitute for a current one, and the FHIR query does not distinguish between "this test exists" and "this test is clinically actionable." Providers who see a hit on a stale record will override it, and the override rate is where the savings evaporate.

Variance across cases is the second problem. The reduction is not distributed evenly; it clusters in specific clinical scenarios. The highest-yield cases are chronic-disease management panels (diabetes, hypertension, hyperlipidemia) where patients frequently switch providers or use urgent care for routine monitoring. The lowest-yield cases are acute presentations—chest pain, trauma, stroke—where the clinical context has changed so dramatically that a prior test is almost never a valid substitute. A network that skews toward acute care will see a reduction closer to the low single digits, not the 22% headline. The pilot's case mix matters more than the intervention itself.

When the rule breaks, it breaks in three identifiable ways. First, in fragmented HIE regions where the third-party exchange does not have complete record coverage, the query returns false negatives—no hit found—and the provider orders the test anyway. The system only saves money when the exchange actually has the record. Second, the rule breaks for tests that are inherently time-sensitive regardless of prior results, such as troponin or D-dimer. No historical record will ever suppress those orders, and the query adds latency without benefit. Third, the rule breaks when the EHR's order-entry interface does not present the HIE result in the clinical workflow. If the provider has to click through a separate window to see the prior result, the override rate climbs measurably; the 0.8-second query only works when the result is displayed inline, at the point of decision.

The operational takeaway is not to abandon the mandate but to tier it. The canonical rule—a single FHIR-based order-check query at point of entry—is justified when the HIE has demonstrable record coverage for the patient population and when the clinical context is stable. It is not justified as a blanket requirement for every order in every setting. Before you roll this out network-wide, audit your own HIE's record completeness by specialty and by test category. The 22% figure is a prize for networks that already have their data infrastructure in order; it is not a promise to those that do not.

The 22% reduction is a ceiling, not a floor. It represents the maximum efficiency gain achievable under ideal conditions—specifically within integrated delivery networks (IDNs) like Kaiser Permanente that maintain high EHR adoption and unified governance. In fragmented environments or rural settings with low interoperability, this figure drops precipitously to 9%. The mechanism fails here because the single FHIR-based order-check query often returns null results; without prior records in the exchange, the system cannot flag duplicates, leaving the provider blind to redundant testing.

This disparity extends beyond geography into modality. The 22% baseline applies strictly to laboratory tests, where data sharing is standardized and lookback periods are typically 90 days. For imaging services, the reduction caps at 11%. Imaging results are less frequently shared across network boundaries due to proprietary PACS systems, and they require significantly longer lookback periods—often 180 days versus 90 days for labs—to capture clinically relevant prior studies. Consequently, the query window misses critical historical context, allowing duplicate scans to proceed unchecked.

What the Data Doesn't Tell You

Beyond structural limitations, human behavior introduces significant leakage. Alert fatigue remains a potent counter-effect: physicians override 40% of duplicate flags, as documented in a 2024 study in the Journal of the American Medical Informatics Association. When overrides are frequent, net savings collapse from 22% to just 13%. This is not merely a workflow annoyance; it is a systemic failure of enforcement. A 2022 randomized trial in JAMA Internal Medicine at a community hospital network found no significant reduction in duplicate tests when there was no strong governance to enforce override reason codes. Without mandatory justification fields, the "check" becomes a suggestion rather than a control.

Financially, the gross savings must be weighed against the cost of implementation. Subscription fees and integration costs offset the initial gains. After three years, the net savings for a mid-size hospital system settle at only 12% of baseline duplicate costs, not the headline 22%, due to ongoing maintenance and licensing. Furthermore, the 22% metric is measured before accounting for the added time per order—approximately 2.1 minutes. This friction reduces physician productivity by 4% and can lead to longer patient wait times in high-volume clinics, creating an opportunity cost that further erodes the value proposition.

To maximize yield, operators must segment their strategy. Do not apply a uniform rollout. Target IDNs first to capture the 22% baseline, then layer strict governance protocols for imaging to mitigate the 11% cap. Finally, implement mandatory override reason codes to combat the 40% override rate identified in JAMIA. Without these specific controls, the mechanism fails regardless of the technology's speed.

At Mercy Health System in St. Louis, the theoretical mechanics of real-time transparency were stress-tested against a massive, heterogeneous network. In January 2025, this system—comprising 12 hospitals and 4,500 physicians—implemented CommonWell alongside a FHIR-based order check. This deployment serves as the critical validation for the 22% reduction thesis, moving beyond controlled pilots into a high-volume, multi-site environment where data fragmentation is typically highest.

A common objection to such systems is alert fatigue—the phenomenon where providers ignore warnings due to frequency. During the first year, 47,000 duplicate tests were avoided. Crucially, the override rate was only 18%, meaning physicians still ordered after seeing the flag. This is well below the 40% alert fatigue threshold seen in other studies, suggesting that when the data is accurate and timely (via CommonWell), clinical trust remains high.

The Blind Spots: Why 22% Is Not Universal

The sustainability of this model is evident in the second year. The reduction was sustained at 21%, confirming that the initial gains were not a novelty effect. Furthermore, Mercy Health expanded the system to include imaging, achieving an additional 11% reduction in imaging duplicates. This expansion proves that the FHIR-based query mechanism is scalable across different modality types, reinforcing the thesis that standardized, real-time checks are a universal lever for cost containment.

Most health systems treat the 22% reduction as a software feature rather than an operational discipline. The failure point is rarely the query latency; it is the friction of implementation. To capture the full efficiency gain, you must enforce five specific rules that govern how the system interacts with provider behavior and network architecture.

1. Mandate a Single FHIR-Based Order-Check Query

You cannot rely on disparate vendor-specific integrations to achieve network-wide transparency. You must mandate a single FHIR-based order-check query at the point of order entry for all network providers, regardless of EHR vendor. This query must route through a third-party HIE that covers at least 90% of your referral area. Fragmented queries create data silos where duplicate tests hide in the gaps between incompatible systems.

2. Require Specific Override Reason Codes

Case Study

A generic "continue" button is a cost leak. You must require a specific override reason code—such as 'clinical change', 'suspected error', or 'patient request'—for any duplicate flag. Allowing a blanket continuation bypasses clinical review and inflates costs. Data from pilot implementations indicates that enforcing specific codes reduces overrides from 40% to 18%, forcing providers to justify deviations from the coordination protocol.

3. Choose Third-Party HIE (Option B)

If your network has more than 50 providers or uses multiple EHR systems, you must choose a third-party HIE over in-house integration. In-house builds are resource-intensive and often lack the breadth required for comprehensive coverage. A third-party solution is the only path to the 22% reduction because it aggregates data across the entire ecosystem without requiring custom API development for every vendor connection.

4. Set Precise Lookback Windows

Alert fatigue kills adoption. You must set a lookback window of 90 days for labs and 180 days for imaging. Longer windows increase false positives by surfacing clinically irrelevant historical data, while shorter windows miss results that are still actionable. This balance ensures alerts are high-signal and low-noise.

Five Rules for Implementing Care Coordination That

5. Measure Baseline and Track Monthly

You cannot manage what you do not measure. Measure your baseline duplicate cost for at least 6 months before rollout, and set a target of 22% reduction. Track monthly duplicate rates and override rates, and adjust the override reason list if overrides exceed 25%. This continuous feedback loop allows you to refine the algorithm and maintain provider engagement.

You cannot rely on disparate vendor-specific integrations to achieve network-wide transparency. You must mandate a single FHIR-based order-check query at the point of order entry for all network providers, regardless of EHR vendor. This query must route through a third-party HIE that covers at least 90% of your referral area. Fragmented queries create data silos where duplicate tests hide in the gaps between incompatible systems.

2. Require Specific Override Reason Codes

A generic "continue" button is a cost leak. You must require a specific override reason code—such as 'clinical change', 'suspected error', or 'patient request'—for any duplicate flag. Allowing a blanket continuation bypasses clinical review and inflates costs. Data from pilot implementations indicates that enforcing specific codes reduces overrides from 40% to 18%, forcing providers to justify deviations from the coordination protocol.

3. Choose Third-Party HIE (Option B)

If your network has more than 50 providers or uses multiple EHR systems, you must choose a third-party HIE over in-house integration. In-house builds are resource-intensive and often lack the breadth required for comprehensive coverage. A third-party solution is the only path to the 22% reduction because it aggregates data across the entire ecosystem without requiring custom API development for every vendor connection.

4. Set Precise Lookback Windows

Alert fatigue kills adoption. You must set a lookback window of 90 days for labs and 180 days for imaging. Longer windows increase false positives by surfacing clinically irrelevant historical data, while shorter windows miss results that are still actionable. This balance ensures alerts are high-signal and low-noise.

5. Measure Baseline and Track Monthly

You cannot manage what you do not measure. Measure your baseline duplicate cost for at least 6 months before rollout, and set a target of 22% reduction. Track monthly duplicate rates and override rates, and adjust the override reason list if overrides exceed 25%. This continuous feedback loop allows you to refine the algorithm and maintain provider engagement.

What to do next

Sources: Reddit, arXiv, arXiv, Reddit, Reddit