Last year, Nordic’s CEO Jim Costanzo wrote about “healthier data, healthier people.” He noted that our electronic health records (EHRs) are missing essential information about what’s going on between infrequent visits to the clinic or emergency department, and he wrote about how this lack of data negatively affects patients. But it’s not just too little information that is a concern. Wearables and home monitoring tools can generate a mountain of readings. How can a clinician properly find the signal while ignoring the noise of every-five-minute blood glucose readings and daily weights? The implication is clear: we’re missing many of the data points we need to make good decisions while we’re simultaneously drowning in data. What’s a clinician to do?
There’s a joke among medical students rotating through general surgery. As you might imagine, students who are just beginning their training don’t get to do much of the “fun stuff” during surgery. I recall doing a lot of retractor holding and suctioning. Depending on the attending, sometimes we were tasked with cutting sutures after they’ve been tied. The punchline was the student asking: “Should I cut the suture too high or too low?” Naturally, it wasn’t really a joke because it seemed that no matter what we did, it wasn’t right. Our current dilemma with patient data involves a similar catch-22.
We surely are missing essential pieces of information from the EHR today. As Jim wrote in his blog post, “patients visit health professionals only periodically, which leads to point-in-time snapshots of their health. There are no data about the person between encounters, which is problematic: most of what accounts for patient health can be tied back to things other than acute care, such as behavior, genetics, socio-economic status, and environmental factors.” While acute care (aka inpatient or ED visits) is a data-rich environment, what happens outside the hospital is more of a mystery. When a patient is seen at a clinic visit, we might record smoking status, weight, and current medications. This information isn’t updated until the next visit, if ever. Yet still, physicians are expected to manage the chronic care of large populations of patients with this sparse data set.
If too little information is a problem, one might expect that it would be difficult to have too much. But clinicians learned about a decade ago that too many data points can be just as troublesome to navigate during patient care. When I was a resident physician a million years ago, we went to the medical records department with hopes and prayers to see if we could find an old chart on a newly-admitted patient. Many questions could be quickly answered by thumbing through the pages to see what diagnoses the patient might have previously had and which treatments were successful. About 10–15 years ago, basic information began to be exchanged between EHRs of different health systems.
Clinicians can now access an abundance of information from all over the country, if not the world. While this was a huge improvement over no data, sometimes the knowledge that we needed was hidden in the nooks and crannies of the chart. It seems like we went directly from famine to feast – and not a “good feast” either. If physicians are looking for a specific lab result or medication name, having virtually unlimited access to past patient history is sublime. But when we’re not exactly sure what we’re looking for, the feeling of drowning in note after note can be a bit much.
What can we do about our data problem? To help improve our ability to effectively manage population health, we need more data and more robust data. Refill information for prescription medications is not readily available in most EHRs; it should be. We know that we can identify early signs of certain behavioral health issues by monitoring social media accounts (given patient permission, of course) yet this is infrequently done. It should be easier for patients to message their primary care physicians and get timely responses from their care team, not necessarily the doctor themselves.
To help us with an abundance of data (some might even say big data), we need to apply tried-and-true methods of having machines use algorithms to identify patterns or outliers that we humans can then validate. Daily weight or glucometer data can be assessed using AI techniques to look for trends that need to be addressed. As these tools become more sophisticated, the ability for a team of clinicians to meaningfully care for a large group of chronically ill patients becomes a reality.
Using technology to solve problems (partially created by technology) is not a new idea…as much as I’d like to claim it. We’ve identified the problems, which is a great initial step. As we create and iterate on solutions, we’ll see incremental improvement in the data problem leading to better outcomes for all of us, or rather: healthier data, healthier people!