Medical cost trend is the highest annual projection in more than a decade at a median of 9% for 2026. Utilization remains elevated post-pandemic, compressing plan margins and squeezing employers, while specialty drug costs continue to climb. GLP-1s alone accounted for nearly a third of last year's growth in drug spending. Health plans have already narrowed networks, intensified utilization management and members have taken on about as much cost-sharing as they reasonably can.
Each party optimizes the part it controls, and the cost accumulates in between. The party that bears the cost is rarely the party that controls it. A self-funded employer carries the claims risk but only sees the bill after care is delivered. The administrator is hired to adjudicate claims accurately, not to influence which providers a member sees. The network is assembled on negotiated discounts rather than demonstrated clinical performance. The provider, still paid predominantly fee-for-service, has every incentive to do more. No one owns the care journey end to end.
Addressing the cost means all parties must work from a shared reference, and clinical quality can be a powerful cost lever. The decisions that incur most of the dollars are clinical: whether a patient is referred to a high-performing specialist or a low-performing one, or whether a procedure is appropriate or one that should never have been booked. Choices made by providers influence roughly 80% of the healthcare dollar. Those decisions belong to individual providers, which is exactly why provider-level quality matters. The industry has spent decades accumulating the data to make that quality measurement, and it is time that measurement is used to inform a member’s choice in provider and drive better outcomes at lower costs.
Variation at the Provider Level
The variation at the provider level is larger than most cost models assume. Two providers treating the same condition in comparable patients can make very different decisions about it: whether to image, refer, prescribe physical therapy or operate. Across more than 230 million members, Embold Health's measurement methodology scores performance at the individual provider level. The data reveals surgical rates for the same condition can differ by more than a factor of 30 between the most conservative and most interventional providers. A single unwarranted operation can cost tens of thousands of dollars and expose the patient to complications and recovery time that follows a surgery they did not need.
Yet almost none of this is visible at the point of choosing a provider. A network that looks adequate on access and discount can contain the full spread of that variation. The measures that have been deployed, such as facility ratings and plan-level scores, lack insight about the individual provider and the member choosing a surgeon might not be equipped to tell the difference in quality and what that means for them.
The Measurement Gap
The data exists; what has lagged is where and how it gets applied. Historically, it has served for accountability and payment: public reporting, value-based purchasing and readmission penalties. All of it looks backward at care already delivered, operating at the level of the plan, the hospital, because that is where attribution and risk adjustment have been most defensible. That orientation made the system more honest about its results, but it also left the individual provider largely unmeasured and put the information everywhere except in front of the member choosing their provider.
The shift now underway is to move that measurement upstream to the start of the care journey, so it informs care decisions. Retrospective scoring doesn't go away; it remains necessary and will keep evaluating care after the fact. What changes is that the same quality signal is also used at the beginning, to guide members toward high-performing providers rather than explaining the outcome after the claim is paid.
Quality as the Foundation
Activating quality at the point of decision places real demands on the system. Provider directories, navigation and network design were built around measures other than provider performance, such as geography, contract status and cost tier. If clinical excellence is going to drive where members actually go for care, it has to be what the systems are built on: how providers surface in search, how a navigator guides a member and how a network or tier is constructed in the first place. A rating that arrives after those decisions are designed cannot redirect them.
If quality measurements are going to drive member provider decisions, the industry must raise the bar on the measurement itself. The data has to evaluate the whole care journey rather than a single visit, run on complete multi-payer claims and be risk-adjusted for patient complexity, so it reflects how a provider actually practices instead of a fraction of their panel. It also must weigh appropriateness alongside outcomes, since a procedure that never should have happened distorts both cost and risk no matter how well it is executed.
Starting From Quality
This only works if all parties agree to begin from the same place, and that place is quality. The employer funds toward measured quality rather than the deepest discount. The administrator routes members on it. The network is built and tiered around it. Providers are measured by it and engaged on it. Health plans hold the claims data that makes a single view of quality possible. It works best, and most effectively, when all parties are measuring from the same, shared definition of quality.
When working from the same place, the efforts are compounded. To the member it looks unremarkable: they reach the right doctor the first time. The savings sit in members receiving the right outcome the first time, avoiding the second surgery, the unnecessary complication or the referral that should never have been made.
The data is in hand, provider-level measurement has matured and holds up under scrutiny, and the parties are already connected to one another. What is missing is the agreement to start from quality and build it into the front of the care journey, the directories, the navigation and the networks, rather than reading it off the back end.
