The fifteen-minute appointment has become one of the infamous fixtures of American healthcare. It isn't what physicians themselves chose but rather what the fee-for-service (FFS) model requires: panels of as many as 3,500 patients, with days crammed full of visits, a system where sheer volume equates to revenue. And yet, not so long ago, a 45-to-60-minute visit with your primary care physician was an unremarkable event. Look a little further back in history, and doctors carried their black medical bags with them, central figures in their respective communities who visited patients in their homes. Though commercialization and physician employment within health systems has diminished physician autonomy and contributed to widespread clinician burnout, recovering aspects of earlier models of care isn't mere nostalgia but a matter of changing infrastructure and incentives, both of which are shifting across healthcare faster than many realize.
Value-based care attempts to reestablish certain features of the older model of healthcare through direct re-alignment of that infrastructure and incentive. When physicians are compensated for keeping patients healthy rather than for volume, the super-structure of healthcare is permitted to reorganize around those goals, meaning smaller panels (closer to 400 patients, not 2,000), longer and more frequent visits–which sometimes are not even be permitted by FFS insurance models–more coordination across specialists, and more room for clinical programs or treatment options that fall outside of a billing code. For example, I've seen full risk provider organizations pay for a dog walker for a patient with high fall risk, buy a companion cat for a patient whose isolation was doing as much health damage as a diagnosis, and purchase a new fridge for a diabetic patient who only had access to a styrofoam cooler after their fridge broke. These examples aren't just gestures: they're what happens when the system finally pays you to care about the thing that actually drives the outcome: the patient’s health. Moreover, they represent an increased ability of the physician to choose how best to allocate resources based on ideal patient outcomes, acknowledge that true health is a matter beyond what is billable, and represent greatly increased physician autonomy in truly meeting a patient's needs.
Yet, for all its promise, VBC has primarily been available to organizations large enough to absorb its complex repayment model. Shared savings arrangements have historically required practices to cover their own operating costs for up to eighteen months before any payout arrives, leaving smaller and rural practices to finance the transition themselves with no guarantee of when and how much it will pay off, whereas fee-for-service has continued to pay on a rolling basis. Unsurprisingly, these practices stayed on the sidelines, including many of the practices serving the communities that stood to benefit most from a shift toward preventive, relationship-based care.
New policy guidance from CMS is able to address shortcomings in prior models by responding directly to how past policy played out in the healthcare ecosystem. The LEAD model, CMS’ successor to REACH, is a direct answer to the gap described above. Prospective payments mean practices get paid upfront rather than absorbing a year and a half of operating costs. Lower participation thresholds bring in organizations that couldn't previously meet ACO minimums. Perhaps most importantly, the change to the annual rebasing model means the goal posts benchmarked at the beginning stay in place. In other words, organizations that improve year over year are not penalized for it and experience an accumulating effect of shared savings rather than competing against their own prior success.
And yet, practicable access to the model as a hypothetical isn't the same as the ability to compete within it in reality. VBC rewards documentation quality, care coordination, and real patient outcomes, all categories that traditionally require dedicated teams of coders, care managers, and population health analysts. A five-physician practice doesn't possess a clinical support bench deep enough to succeed nor the reasonable means of obtaining one. What that means is that while CMS can make the door easier to open and more palatable to walk through, it can't make it so that a small practice gains the operational muscle of a large organization. That’s the second gap, and the one where technology–and AI particularly–matters, allowing even small practices to initiate an automated system of insight into their existing workflows.
Risk adjustment is a clear example of this. Reviewing a patient's chart for missed or under-documented conditions used to mean a coder or even a physician working through potentially hundreds of pages per patient, hunting for diagnoses that were documented or treated but never properly coded. The nature of that work is slow and expensive–nor is it the best dispensation of a physician’s time–and it’s easy to under-resource if you're not a large system. When I worked near a VA hospital, boxes of dot-matrix printed records would arrive regularly, chock full of labs, imaging records, and primary care notes that would take hours to read through. The experience, along with being exhausting, left me worrying at night, wondering whether I could have missed something key in the third hour of reading through a given patient’s information. Thankfully, now there are automated tools that can surface the most likely missed diagnoses directly into the EHR, at the point of care, for a physician to confirm or reject in seconds rather than a care team to chase down over weeks. Importantly, humans remain in the loop, but the tools themselves streamline those parts of the workflow least compatible with the realities of human time and attention.
The same situation extends to hospital admissions prevention and to population health tools that attempt to cohort patients automatically by risk so that the most at-risk patients get the fastest outreach. However, many of these tools rely on machine-learning driven analyses that ingest claims only, missing the bulk of the record–as much as 80 percent–that comes from unstructured documents such as primary care notes. Consider the example of a quadriplegic father whose daughter historically has picked up his medicine, but in a recent PCP visit it is noted that the daughter recently moved across the country: Mere claims would never catch this sudden gap in support. Or consider the example where a patient’s son leaves a EHR message for the primary care doctor that his mother got lost for 2 hrs on the way to the grocery store she’s been frequenting for 25 years. This calls for a new cognitive assessment that would never be suggested by the structured record.
The choice of vendor matters immensely here, as much as the choice to adopt the tool in the first place. There are several considerations a practice should look at before taking the step. Any AI tool should meet the “set-it-and-forget-it” criteria, meaning a tool that is going to be trusted by its clinical users because it consistently surfaces what is salient and actionable. It should also offer the flexibility—both of implementation and price–that a small operation will require. That means any AI tool that lives outside the EHR with a separate login, adding one more system to check, will create friction a small practice doesn't have staff to absorb (we all know just how much doctors love getting “yet another login”). Essential also is that the tool be LLM-native, since tools whose AI layer is enabled by machine-learning models will fail in their ability to draw on the majority-unstructured data where, as noted above, the majority of the insights come from. Lacking a sophisticated reasoning layer, non LLM-native tools may likewise generate an amount of noise likely to frustrate clinicians into abandonment and will lack the complex chain of reasoning needed for audit defense. Finally, the pricing structure should mirror the problem LEAD is solving: a per-patient model that scales with the practice, not a flat license that assumes the deep pockets a five-physician group doesn't have.
None of this makes VBC turnkey, at least not yet. Changing how American medicine is organized and paid for is not a single policy or product away from being fully solved. However, the combination of LEAD and other future policy models along with LLM-native tools are solving different halves of the same problem: one opens the door, the other gives small to medium sized practices the operational capacity to walk through that door and actually succeed once inside. Together, they're making something possible that wasn't a few years ago: a rural or independent practice delivering the kind of coordinated and preventive care that requires a health system’s resources while enabling relationship-oriented care the FFS model is likely to lack.