The following was originally published on LinkedIn in March 2025.
By Dean Dalili, MD, MHCM, DeepScribe Chief Medical Officer
The transition to a value-based care payment environment in healthcare – in which quality outcomes and reduced costs are rewarded – has been going on for a long time. Longer than you might think.
At a healthcare policy level, the U.S. has flirted with value-based incentives since the 1960s, with the creation of Medicare, Medicaid, and the first Patient-Centered Medical Home (PCMH) model in 1967. These were the initial steps toward Health Maintenance Organizations (HMOs) and real efforts to keep healthcare costs down.
With the Affordable Care Act and popularity of Medicare Advantage, the last decade or so has seen a significant increase in the number of value-based programs, with as many as 50 now available. With that growth has come the practical application of technology as a pivotal ally; not just a tool, but an enabler of value-based care success.
Ambient AI, even at this early stage in its potential, is helping provider groups with patient information access and, of course, the documentation process. I’ll delve into the role of AI in value-based care, and technology's role in general, later in this article.
The promise of better outcomes
It’s commonly understood that a close alignment between healthcare outcomes and provider reimbursement can deliver better results. Yet, despite the growth of value-based programs, the healthcare industry's shift to linking payment to quality measures has been slow and challenging to implement.
A 2022 MGMA report states that value-based care collections accounted for 15% of revenue for non-surgical specialties, 6% for surgical specialties, and 7% within primary care. We all know the pace of healthcare change can be glacial, but value-based care adoption is difficult – especially compared to the “old” way, aka fee for service.
The traditional fee-for-service model is straightforward – a physician or group is compensated for each medical service delivered, regardless of outcome. Value-based care reimbursement is far more complex, with some VBC scenarios making provider groups responsible for every healthcare expense related to caring for and maintaining a patient’s health over a relatively long period of time. That includes medication fees, durable medical equipment costs, hospitalizations, home health, physician fees, and on and on.
The results and progress of value-based care - and when it works
Overall, the results from value-based and alternative payment models have been varied, especially for provider groups that have not taken on risk fully. They tend not to see a sizable impact in cost savings and quality improvement.
But for institutions with very high or full participation in risk-based models, we do see a difference – Intermountain, Geisinger, and our partners at Ochsner Health are notable examples. (For more on Ocshner Health, here’s an excellent video from Eric Bricker, MD breaking down their successful approach to controlling costs and improving care. Highly recommended.)
How are they doing it? Organizations that are most successful at value-based care build their operation specifically around a value-based model. They reimagine workflows to ensure they’re delivering very high-quality, patient-centered care.
They consider patient interventions that aren’t necessarily medical. Actions may be focused on social determinants of health (SDOH): resources to access or transportation needs or food security. These are the kind of items that, if not tended to, can lead to high utilization and unnecessarily high spend for chronically ill patients.
The beauty of a value-based care model is that it makes this type of outreach possible – and affordable – because providers can mobilize an entire budget for a person’s care, taking their total costs into account. To move the needle in terms of quality and value, the model encourages a holistic look at the patient, and a total approach to their care.
Bringing today’s technology into the process
As mentioned, we’re now seeing an increase in provider groups that are only partially managing risk. They straddle both business models: a fee-for-service system demanding high efficiency and productivity plus value-based care attention for a subset of patients. Without tailored technology, this juggling act is very difficult to sustain.
Managing any patient population with a focus on outcomes requires technology solutions and tools intentionally built around value-based care workflows. The very structure of care delivery and reimbursement changes; so too must the work within it.
One such option is working with an enablement provider like our partners at Pearl Health. They make it easier for primary care organizations to participate in value-based care as an alternate revenue stream. Pearl offers tools, insights, and services for population health management, and takes on risk for a subset of a group’s patient population.
This empowers the primary care group to benefit from revenue earned within a value-based setting like ACO Reach without converting their entire practice into a completely different payment model.
The role of AI and ambient documentation
One of the key facets of value-based care management is risk assessment and adjustment – appropriately identifying chronic conditions that are expected to impact a patient’s cost of care longitudinally. Doing this well, without massive manual work, demands specific, accurate documentation. The right nomenclature within the patient chart can have a significant impact on value-based program performance.
At its core, DeepScribe helps as a time-saving ambient documentation technology. By listening to the clinical conversation and producing a structured, billable note, ambient documentation saves the clinician time and helps lighten the work burden.
It’s something our system does remarkably well, and we are transparent about our results. Clinicians who use ambient AI convert their time savings into seeing more patients, or going home early and having a better quality of life.
But there’s another functionality we’ve created – and that’s in the risk adjustment realm.
Specifically, DeepScribe has developed intelligence around HCCs, or Hierarchical Condition Categories. When HCC codes are properly captured in the clinical note, the note accurately reflects the patient’s complete illness burden, a must in value-based care; in turn, that translates into a full reimbursement for that patient’s care over the next year. Without a comprehensive list of HCCs for each chronic care patient, the documentation of illness and need are incomplete.
DeepScribe works from previous problem lists for each patient to identify historical HCC codes that need recapturing. That information is shared with the clinician before the patient visit, prompting the provider to address the breadth of medical conditions with the patient. That conversation is then captured in the AI-generated documentation.
This is a step beyond documentation functionality - it’s ambient intelligence at the point of care to help providers more easily comply with a standard value-based requirement.
Our system can also access suspect codes from sources such as HIEs (Health Information Exchanges). In these instances, the HCC list is even more thorough, reflecting codes that need to be recaptured plus newer ones that will adjust the patient’s RAF (Risk Adjustment Factor score) with finer accuracy.
One system for greater efficiency, leaner workflow
When I see a patient, my source of truth is the EHR – to check their history, order labs, change medication doses, whatever their care requires. But if I need to access data about a patient who falls under a value-based model, I would ordinarily have to check a database on a separate system to find suspect codes or identify care gaps.
By “nudging” clinicians with HCC information, DeepScribe offers providers the ability to stay within the tool they’re already using at that moment to document care. When a patient comes in who is part of a value-based contract, the DeepScribe system pulls the historical HCC codes into the workflow. There’s no need for the clinician to look in another system.
Then, as the encounter progresses, the ambient AI analyzes the content of the conversation to validate that the clinician has addressed the conditions within those diagnosis codes against Medicare’s MEAT criteria – without prompting the doctor or interrupting with questions. Clinicians can adjust the visit conversation as needed.
Summary: Giving clinicians the ambient intelligence to thrive in any reimbursement model
Doctors did not attend medical school to learn the finer points of ICD-10. Regardless, getting it right is really hard; precise, intelligent ambient documentation can do it, taking a tremendous weight off the clinicians’ responsibilities for the day.
But let’s zoom out for a moment. Documentation is not the only lever to pull to succeed in value-based care. Yes, AI technology can generate excellent documentation, but that’s not a satisfactory goal in and of itself. Ambient AI should be applied to its fullest potential on behalf of improving care for patients.
There are plenty of other levers to pull, with AI identifying and surfacing the right information for clinicians. Consider care gaps: Let’s say a patient needs a screening or lab study as part of their health maintenance. Just as AI can list HCCs needing attention, it can also make sure patient needs are surfaced at the point of care to ensure the clinician can take action.
How about preventing hospitalization, emergency department visits, or hospital readmission? This can be done by leveraging AI to take a much more holistic, proactive view of a patient. Ideally, this is done via the EHR, but also by accessing other data resources and medical records – all to provide that essential information at the point of care.
Here’s what’s unique about ambient AI as an engine for knowledge and change in healthcare: It’s a technology in the room with the doctor and patient. At DeepScribe, we find that to be a privilege. It’s distinct and very valuable. But, we want – and need – to amplify the value of that opportunity by bringing in additional data and making it as impactful as we can, filtering it and presenting to make it most actionable for the clinician and most relevant for the patient’s care.
That is when our technology will not only drive better documentation outcomes, but also better clinical outcomes – a powerful combination for value-based care success. We’ve taken the first steps, and we’re very early in the journey.
text