A few years ago, a physician I worked with told me that her young daughter had started greeting her at the door with a question: “Mommy, how many charts do you have to finish today?” Her daughter had learned that the higher the number, the less of her mother she was going to get that night.
That story has stayed with me through more than two decades in clinical informatics, because it captures something the data has been telling us for years. Recent research from the American Medical Association found that primary care physicians spend an average of 36.2 minutes on the electronic health record for every patient visit, often longer than the visit itself. More than one in five physicians still spend over eight hours a week on the EHR outside of normal working hours, a figure that has not budged since 2022. Burnout has eased modestly from its pandemic peak but still affects 41.9% of U.S. physicians, with documentation cited as a leading driver.
We have spent the better part of a decade asking clinicians to feed the EHR, but we have not yet asked the EHR to give them anything back.
That, to me, is the single most important (and most overdue) pivot in healthcare IT today.
We collected the data. Now what?
For roughly fifteen years, the dominant project on every health system’s IT roadmap has been data capture: implement the EHR, get clinicians documenting in it, layer in clinical systems that don’t talk to the EHR, then build a data warehouse to pull it all back together. Each step was necessary. None of them, by themselves, made anyone’s clinical day better.
What we built, in the end, is a generation of systems that are very good at describing what already happened. We can pull a list of diabetic patients. We can show how many people came through the ED last week. We can build a dashboard for almost any metric a regulator asks about. These are descriptive analytics, and these have been the ceiling rather than the floor for most mid-sized health systems.
Meanwhile, the data sits idle. The signals we need to predict which diabetic patient is most likely to land in the ED this month and the staffing pattern that correlates with the worst patient experience scores are mostly buried inside systems running the same playbook from 2016.
The shift that needs to happen now is what we believe the EHR is for. It is not, fundamentally, a documentation system. It has the potential to be a clinical reasoning system that happens to capture documentation along the way.
The good news is this is a problem health IT leaders are finally positioned to solve, because the computational tools to do it have caught up to the data we’ve been collecting.
What separates the systems that will get there
The largest academic medical centers are already moving in this direction. They are running machine learning against their own data, building decision support that reflects their own patient population, and treating their data scientists as core clinical infrastructure. The mid-sized health systems and regional networks where most Americans get their care are still thinking that the next medical record upgrade will solve their analytics problem. It will not.
Three things separate the organizations that I think will close this gap from the ones that won’t.
- Treating data as a capital asset. Health systems have no trouble issuing bonds for new buildings and depreciating them over fifteen years. The same financial discipline rarely gets applied to data infrastructure, which is still funded out of operating expense and held to a six-month payback. That mismatch is the single biggest reason mid-sized systems remain stuck in descriptive analytics.
- Building models against local data. A predictive algorithm trained on the patient population of a downtown academic hospital will not perform the same way at a rural community hospital ninety miles away. Different demographics, different chronic disease profiles, different social drivers, different air quality, different weather. The plug-and-play promise on a vendor slide deck almost always erodes on contact with real local data. Health systems that succeed will be the ones that invest in tuning models to their own population, not the ones that buy the most generic capability.
- Closing the loop into the clinical workflow. An insight that lives on a separate dashboard, in a separate system, behind a separate login, will not change a clinician’s day. The decision support has to land where the work is already happening — inside the chart, at the moment of care, integrated with the existing scheduling and ordering tools.
A virtual scribe I once worked with caught a cancer diagnosis the physician had missed: the patient mentioned a symptom in passing, and the scribe, trained to listen, flagged it. That is what closing the loop looks like in human form. The technology to do it at scale, faithfully, and across an entire health system is finally within reach.
About Lori Runion
Lori Runion is a Director and Portfolio Leader at Resultant, specializing in healthcare data and analytics. She has more than two decades of experience in clinical informatics, EHR implementation, and provider workflow optimization. She has led data and analytics modernization initiatives across both private health systems and state public health agencies.
