Data geeks use regression analysis to shape treatment at Rogers10/14/17 05:00:pm
Let’s say Jack and John come to Rogers with the same condition, severity, age, gender, etc. After the same length of treatment, Jack makes significantly more progress than John. Since they are similar in so many ways, it would be valuable to determine what it was about Jack’s treatment that facilitated his greater improvement. However, getting to the bottom of knowing what treatment approaches produce the best results is no easy task.
That’s where the Rogers Clinical Effectiveness team comes in. By using what’s known as regression analysis, the team has the ability to compare two patients who are exactly the same with regard to all the variables being reviewed – except for one. By controlling for, or holding variables constant, they can make a change to the variable of interest to see what impact that has on the dependent variable.
Jessica Cook, Clinical Effectiveness coordinator, shares, “Regression analysis is superior to comparing how just two variables correlate over time, because it attempts to determine causation, while correlation cannot. Utilizing multiple regression allows us to better understand our patients and treatment programs as well as begin to understand what factors may predict outcomes.”
Dr. Jerry Halverson, psychiatrist and chief medical officer, says the outcomes tracked by Rogers already tell us that our evidence-based treatments overall work better than other psychiatric treatment providers, but this unique way of using cutting edge statistics goes a major step beyond. “This should lead to us learning why certain patients do better than others. That knowledge carries with it the promise of even better outcomes for all of our patients. In a best case scenario, we’ll be able to intervene earlier and more effectively if a treatment isn’t working for a specific individual,” he explains.
Early use of the technique points to better results when patients with major depressive disorder undergo transcranial magnetic stimulation (TMS) in addition to other treatment modalities. Specifically, patients have a greater reduction in their scores on the Quick Inventory of Depressive Symptomology (QIDS) assessment during treatment if they also had TMS treatment. The QIDS is a 16-item measure that covers nine diagnostic symptom domains used to characterize a major depressive episode. Higher scores indicate more depressive symptoms.
Dr. Tyler Rickers, a psychiatrist who specializes in treating resistant depression is excited about what this use of data analysis means for patients and their options for treatment. “For so many years we have had only one proven effective, non-medication, treatment for depression, ECT. While effective, ECT has significant risks and severe side-effects. Invasive techniques have been dubious and medications have many unintended effects, not to mention surprisingly low efficacy for some. Now we have a treatment that is well tolerated and very specific, and through regression analysis, we can demonstrate that it’s working well. Thanks to this data methodology, we’re able to provide the thing our patients perhaps need most-- hope.”
In addition, regression analysis methodology is being used to suggest areas to explore from an internal perspective. “We’re looking at how and why outcomes differ in similar services between different Rogers locations. For example, what factors in one location might produce different outcomes from a different location? Is it the patient population, the treatment structure, or differences in practice? These are some of the questions we are asking to better drive quality of care throughout the system. We’re in the very early stages of this project and any initial results will be used for internal quality and improvement purposes,” Jessica explains.
Regression analysis as a discipline isn’t necessary unique as there are an estimated 2.3 million individuals in the U.S. who do some form of it. What is unusual is the way the Clinical Effectiveness team is using regression analysis to understanding what treatments work best.
“Only a handful of large healthcare systems have committed resources to doing this work. We are lucky to work in a forward thinking organization that supports data science. Our patients will truly benefit from this kind of rigor,” comments Brian Kay, director, quality and clinical effectiveness.
Ultimately, regression analysis may be the key to clinical effectiveness. Dr. Halverson is thinking big: “This has the promise of being a really important method in ensuring we provide the right, best treatment and outcomes for every individual we treat,” he says.