The Collective “Blog” of Health IT- WEDI Podcast Episode 260
Beyond the Billing Code: Making Healthcare Data More Meaningful with Dr. Jay Anders, Medicomp
Healthcare generates enormous amounts of data, yet one of the most fundamental challenges in health information remains surprisingly difficult: accurately representing what is happening with the patient.
A billing code may tell a payer why a patient was seen, but it does not always tell the complete clinical story. As healthcare becomes increasingly digital and AI-driven, the distinction between a billing code and a clinical diagnosis—and the quality and specificity of the information captured in the first place—has become increasingly important.
On episode 260 of The Collective Voice of Health IT, WEDI’s Michael McNutt spoke with Dr. Jay Anders, Chief Medical Officer at MediComp Systems, about the challenges of translating clinical knowledge into meaningful digital information and how AI might help improve the process.
A billing code isn't necessarily the clinical picture
Anders, who moved from clinical medicine into healthcare technology to have an impact on a broader population of patients, explained that ICD-10 codes serve an important purpose: they provide insurers with information needed to process claims. But a billing code doesn't always capture the level of clinical detail necessary to understand a patient's actual condition.
Consider a patient with Charcot-Marie-Tooth disease. A general billing code may identify the disease, but it may not identify the particular variant affecting that patient. That distinction can become increasingly important as treatments become more targeted and dependent on factors such as a patient's genetic makeup.
The same issue can arise in other areas of medicine, including cancer treatment, where the specific characteristics of a patient's disease can determine which therapies may be appropriate.
The challenge, then, isn't that billing codes are inherently inadequate. It's that they were never designed to represent the entire clinical picture.
When clinical context gets lost
That distinction becomes even more significant when information moves between healthcare organizations.
Interoperability has made it possible to move information electronically between providers and organizations, but Anders noted that much of what gets transmitted can still be centered around billing codes rather than the more detailed clinical information behind them.
The result can be a loss of context.
A diagnosis that begins with a relatively nuanced clinical understanding can become increasingly simplified as information is documented, coded and exchanged. Michael compared the process to a game of telephone: as information moves from one person or organization to another, small inaccuracies can compound over time.
Once an inaccurate diagnosis or piece of information enters the medical record, correcting it can also be difficult. Anders noted that in systems where medical records are retained for long periods, an error can potentially follow a patient for years.
That makes the quality of information at the point of capture critically important.
Can AI solve the documentation problem?
AI offers an intriguing opportunity—but Anders cautioned against viewing it as a magic solution.
AI can help generate clinical language, summarize information and identify discrepancies between a clinician's documentation and other information in the patient's record. It can also flag potential inconsistencies before a record is finalized, giving clinicians an opportunity to review and correct them.
But the effectiveness of AI depends heavily on the information it is given.
If the underlying clinical information is incomplete, inaccurate or overly generalized, AI may simply process and reproduce those shortcomings. In other words, AI cannot reliably extract a clinical story that was never captured in the first place.
Anders also questioned whether some current implementations of AI are actually reducing clinician burden as much as expected. In some cases, work that once involved creating documentation is being replaced by reviewing and correcting AI-generated documentation—work that still requires clinical expertise.
The opportunity may therefore be less about asking AI to do everything and more about designing AI tools that provide the right assistance at the right point in the clinical workflow.
Keeping the clinician in the loop
One of Anders' key messages was the importance of using AI to assist clinicians at the point of care, rather than simply deploying broad AI capabilities across an electronic medical record system.
AI can identify a potential discrepancy, but a clinician needs to determine whether the information is actually wrong. AI can suggest language, but the clinician needs to ensure that the language accurately reflects the patient's condition.
That human review is particularly important when the information will ultimately influence decisions involving treatment, prior authorization, care coordination or other aspects of patient care.
Anders described the need for appropriate “guardrails” and medical knowledge applications around AI so that the technology is operating within a framework designed for healthcare rather than simply being applied broadly to healthcare data.
Start with the problem, then choose the technology
Perhaps the most important takeaway from the conversation was Anders' advice to be deterministic about AI implementation.
Rather than starting with the question, “Where can we use AI?”, healthcare organizations should first determine what problem they are trying to solve.
Is the goal to reduce documentation burden? Identify errors before they enter the medical record? Improve the specificity of clinical information? Support prior authorization? Improve care coordination?
Once the objective is clear, organizations can determine where AI augmentation can provide meaningful value.
That approach is particularly important as healthcare organizations look beyond simply processing claims toward creating a more complete representation of the patient.
From processing data to understanding the patient
Healthcare's next interoperability challenge may not simply be moving more data between organizations. It may be ensuring that the information being exchanged retains enough clinical meaning to actually help the next person who needs it.
A billing code has an important role in the healthcare system. But it is only one piece of the story.
The opportunity ahead is to capture and preserve the clinical context behind that code—and to use AI thoughtfully to help clinicians identify, validate and act on that information without removing them from the process.
As Anders emphasized throughout the conversation, the goal shouldn't be to apply AI everywhere. It should be to use the right technology, with the right guardrails, to solve the right problem—ultimately making healthcare data more accurate, meaningful and useful for the people who depend on it.
Listen to every episode of The Collective Voice of Health IT at www.wedi.org/category/podcasts