Cohorts: Using More Meaningful Data to Identify Unwarranted Clinical Variation
Written by Megan Bultema, Ph.D., Chief Data Scientist
Healthcare systems across the country rely on data to help solve one of the largest challenges they face — how to improve the value of the care they deliver to patients by reducing cost and improving quality.
Data, however, is not a magic wand, and traditional ways of looking at data can result in analyses with a lot of “noise”. Successfully using data starts by making it meaningful and credible to the audience.
A New Approach to Identifying and Reducing Unwarranted Clinical Variation
Unwarranted clinical variation, unnecessary differences in clinical approach and delivery, often results in the misuse- and even waste- of expensive healthcare resources. This can have a negative impact on patient outcomes as well as their out-of-pocket costs.
In 2012, Utah-based Intermountain Healthcare set out to reduce unwarranted clinical variation — the largest driver of unnecessary cost for its surgical cases. A team of operating room nurses, surgeons, analysts, and data engineers devised a revolutionary approach to grouping data in a more clinically meaningful way leveraging cohort study design.
Surgical Cohort Studies
Cohort studies represent a useful statistical technique to gather retrospective and prospective evidence on populations in which random control trials are not plausible. Traditionally cohorts, or groups of individuals that share defining characteristics, provide a statistical structure to determine the correlation in inputs (e.g., surgical technique, supplies) to outcomes. In order to identify unwarranted variation in both costs and outcomes, cohorts containing comparable patients and surgical procedures are formed through extensive analysis of discrete and non-discrete data . Provided with clinically relevant surgical cohorts, clinicians can investigate clinical variation amongst peers and identify opportunities to reduce costs and improve outcomes.
Traditional cost and outcome analytics, which were based on basic procedure codes, were too broad and failed to give real insight into the reasons for clinical variation. With procedures (such as ICD-10) or procedure groupers (such as MS-DRG) that include expected surgical and patient variation, the outcomes and costs cannot be analyzed in a meaningful way. Unlike previous coding driven analytics, the Empiric cohort concept accounts for a number of variables, including patient type, surgical technique, equipment and supplies used, secondary procedures performed during the case, and insights the surgeon captures in the text of the operative report.
As a simple example, Total Knee Arthroplasty, which can be performed as a revision of a previous knee operation or as an initial knee operation, would receive the same billing procedure coding. However, when performed as a revision or conversion of a previous surgery, the surgical approach for this type of case will likely require more time, different supplies, and have different expected patient outcomes. It would be nonsensical to advise a surgeon performing a revision or conversion to use the same supplies or expect them to have the same outcomes as the surgeon performing an initial Total Knee Arthroplasty.
The same is true for patient presentation during surgery. A patient whose appendix has ruptured and needs a laparoscopic appendectomy is very different than a patient whose appendix has not ruptured. The approach, follow up care, and expected complications for each of these procedures is very different. By creating new categories of appendectomies that separated patients into “ruptured” and “unruptured” cohorts, the team was able to glean more meaningful insights from the data.
Empiric’s unique cohort groupings make data not only more insightful, but more actionable as well. With the creation of clinically meaningful cohorts, hospital and service line leadership are able to identify and analyze the precise clinical variation drivers within a given procedure. Using cohorts, surgical encounters can be meaningfully compared across surgeons and locations to quickly and easily find best practices and areas of opportunity.
At Intermountain Healthcare, surgeons, and their clinical care teams recognized the value of analyzing and acting upon the cohort-based data. Soon, a system-wide culture change began to take hold. Surgeons shared best practices with each other and collaborated on ways to standardize those practices across the health system.
Automating the Process
Empiric Health built on the hard work of the Intermountain team, introducing new Artificial Intelligence (AI) and Natural Language Processing (NLP) techniques to automate cohort assignment and facilitate data comparison. This approach significantly reduced the labor hours to build clinically comparable cohorts. This technology has led to the development of new cohorts and now serves as a key component of Empiric Health’s approach.
A Proven Concept
As the culture change at Intermountain grew, so did the savings. Today, there are more than 360 cohorts for surgical procedures that have helped the health system save more than $90 million. As the analytics became more meaningful, physician engagement with the data grew, and physician service line meetings saw a 150 percent increase in attendance.
Healthcare providers want to offer their patients the highest quality, and most economical care possible. Much of the value in healthcare, however, is determined by many small decisions that clinical care teams make each and every day. To make more evidence-based decisions, clinicians need data. Data that is clinically-relevant, accurate, and actionable– data that allows them to draw true insights.
Empiric Health is partnering with health systems to look at data in a new and innovative way. By grouping information in clinically meaningful cohorts, Empiric has removed the “noise” and significantly shortened the feedback loop to allow clinicians to learn from their own data as well as that of their peers.
To learn more about Empiric Health and how our proprietary data cohorts are actually bending the cost curve in healthcare, visit www.empirichealth.com.