Much of medical practice is as much a mystery to doctors as it is to patients.
Human physiology is so complex, and the external variables so numerous, that we often have no sure knowledge of why one patient did well or another patient didn't. Every physician longs for some way to really know what will work for each patient.
While we have come a long way, even in just the past five years, there is still so much left to be learned. The one thing that can help us reach greater knowledge faster is data and analytics.
That's really the underlying value of electronic medical records: they represent a treasure trove of data waiting to be mined. With the right algorithms, we can use that data to find patterns that tell us what factors make a tangible difference in outcomes. It's the wisdom of the ages waiting to be read.
Perhaps the most valuable medical team member of the future will be a data scientist. These are the experts who understand how to tag and mine data and how to construct algorithms that find patterns accurately and can help us be more effective in delivering the best possible care every time.
For example, there is a great study from the University of Iowa Medical Center, in which gastroenterology surgeons are using real-time patient data in the operating room, combined with past data from gastric surgery patients, to predict who is at risk for developing a surgical site infection. This helps guide decisions in the OR as well as post-surgical care. While doctors know that a variety of modalities can reduce infection risk and promote healing, resources are not endless. By identifying patients who need high-level care, they can ensure that resources are targeted where they are needed most. The project has reduced surgical site infections by more than 50 percent in the gastroenterology patients whose care was guided by the analytics.
So Dr. Data (as New York Times writer Steve Lohr calls one data scientist) is saving lives, even without a medical degree.
How do we know the predictions are accurate?
But here's the catch: we've got to get the algorithms right. If we aren't careful, we can draw conclusions that aren't really there. To make giant leaps forward in understanding, we need a colleague on the case who really understands how to create algorithms that have practical value and accurate results.
Tom Hill, a colleague of mine at Dell, recently wrote a blog in which he noted the necessity of using a systematic, transparent approach to predictive analytics. "Harvesting big data carries with it the responsibility to do-the-right-thing with those data. Big or any data and predictive models in healthcare must be correct, access and tamper-proof (secure), must not discriminate, generally do-good, and not-do-any-harm."
He goes on to talk about the need for transparency in analytics, so that those using the results understand what data is being used and how it is being analyzed. As changes or improvements are made, they must be documented, so that the transparency lives on.
I think this is a critical point for physicians who will be using the algorithms in the future. If our patients' lives will depend on the quality of the analytics used to guide treatment decisions, we need to know that the algorithms are correct. We don't want a black box that dispenses treatment prescriptions; instead, we want to know how the results are created, so that we can trust the advice offered and help guide future improvements.
Adopting analytics in ways that don't risk lives
Dr. Hill's point about "not doing harm" is well taken. As healthcare organizations add analytics to patient care, projects like the one at the University of Iowa is a good place to start. It takes a body of existing knowledge about a large population of gastroenterology surgical patients and analyzes what factors were associated with certain outcomes. It then takes that analysis and compares it to a specific patient, providing insight into how that patient may do in post-surgical care.
The likelihood of a result that harms a patient is small. At worst, a patient might receive more care than is really necessary, or might not be recommended for care that would help. But that happens all the time without any analytics intervention, so the risk to patients is not increased by using the insights from the analytics. And the care team can monitor to see that, if the patient needs more extensive post-surgical care, that care can be ordered.
Other initial analytics projects in healthcare are looking at ways to predict surges in demand for care, based on environmental factors, and those projects also aren't likely to put patients in harm's way.
These kinds of project allow an organization to use analytics for practical improvements, while also learning how to use these new insights. As the organization's expertise grows, the complexity of the analytics projects will likely grow, too. But starting with a project of limited scope and low risk for patient harm is a smart idea.
It's also a way to help build trust. Physicians may be somewhat leery of trusting an analytics program to help them make treatment decisions, especially if a recommendation flies in the face of what that doctor's always done in the past. So institutions must be careful to build trust in analytics as they move forward. As physicians see the effectiveness of using these tools, they'll be more willing to engage in analytics themselves. So how, when and why you use analytics really matters. And making sure that you're working with a really good Dr. Data is important, because at least for the foreseeable future, medical practitioners will be working very closely with Dr. Data to make analytics a powerful force for good.
The views, opinions and positions expressed within these guest posts are those of the author alone and do not represent those of Becker's Hospital Review/Becker's Healthcare. The accuracy, completeness and validity of any statements made within this article are not guaranteed. We accept no liability for any errors, omissions or representations. The copyright of this content belongs to the author and any liability with regards to infringement of intellectual property rights remains with them.