Our team dove immediately into analyzing data related to surgical services — supply cost, efficiency metrics, etc., but realized there were limitations to the data. Grouped under traditional billing and procedure codes, the data lacked specificity and often failed to return any meaningful insights related to clinical variation.
There was significant pushback from the surgeons regarding the validity of this data. Those who had variations in price and outcomes for their procedures often responded to the data in a similar fashion — “You’re not comparing me to the right peers. My patients are sicker and more complicated.”
In many cases, they were right. The data was simply too broad to account for variation in patient types, surgical approaches, and other meaningful differentiators. For example, every Anterior Cruciate Ligament (ACL) reconstruction surgery was treated the same, whether the procedure was an ACL repair or also included a repair of the Lateral Cruciate Ligament (LCL) or Medial Cruciate Ligament (MCL), which make the operation much more complicated. In short, the data was not comparing “apples to apples.”
Our team discovered they needed a more precise data set to identify why clinical variations were occurring. More accurate data would in turn help the team gain credibility with the surgeons, as they would realize the benefit of adopting the best practices identified by the research.
Our team devised a unique approach to analyzing cost and clinical data that presented results in a more meaningful way and made more sense to surgeons: cohorts.
The team worked closely with surgeons to develop hundreds of cohorts that aggregated comparable surgical cases based on more complete information, eliminating the variability. In cases of ACL surgeries, for instance, those procedures where both the ACL and LCL were repaired were compared only to each other, while ACL-only procedures were compared only to other ACL-only procedures.
Creating cohorts wasn’t simply a matter of looking at data in a new way. It took thousands of work hours from cross-functional work teams that included perioperative nurses, consultants, and data specialists. The effort required pouring through clinical notes and collaborating with surgeons over a period of years to develop cohort definitions that were more precise, specific, and meaningful than broader CPT, DRG, or ICD-defined groupings.
Moving forward called for more than cohort data alone – it required buy-in from surgeons of every specialty and cooperation among perioperative staff, directors, administrators, and supply chain leaders.
Though the cohort data proved to be precise and actionable, it required strong engagement from the surgeons, who ultimately made the decisions on how procedures were performed.
We met with surgeons regularly to showcase lessons from the cohort data, share what other surgeons were doing as a result of the data, and worked daily to open more internal communication lines between surgeons and the frontline staff.
Over the course of the project, a culture change was achieved across the entire health system, with surgeons engaging closely with the data and their colleagues to change the way they approached procedures.
With the cohorts and clinician engagement efforts seeing great success in reducing clinical variations and costs in surgery, Intermountain wanted to expand the effort across all surgical procedures. There was only one problem — labor capacity.
Each cohort is the result of hours of data review, validation and grouping. The volume of work needed to significantly expand the program would overwhelm the staff — and hiring a team large enough to handle the job would cut into much needed savings.
To solve the issue, Empiric Health engineers developed proprietary technology that could review reams of clinical information in a fraction of the time it would take humans to do the same. The result was an artificial intelligence-driven platform that featured natural language processing (NLP) technology with 98% accuracy to capture thousands of data points from a variety of sources, including clinician notes.
The machine learning technology allowed Empiric and Intermountain to expand the cohort program quickly across the entire system — and proved to be the tipping point for this new approach to healthcare expenditure reduction.
Today, thanks to the efforts of the Empiric team and the Intermountain clinical staff, there are more than 300 cohorts for surgical procedures — and the number keeps growing. The project far exceeded Intermountain’s target of $60 million saved, totaling $90 million saved over a period of just 4 years. In orthopedics, the average savings was $1500 per case. In urology, $600 per case and in pediatrics, $470 per case. Physician engagement also skyrocketed, with 80% of all neuro and spine surgeons attending meetings — up from just 30% prior to the project.
The results continue to roll in for Intermountain, as each year more cohorts are added, more physicians are engaged, and more money is saved. What began as a novel idea in 2012 has turned into a proven process through which Empiric Health is changing the way modern healthcare is delivered. As we continue to build new and even more meaningful cohorts and assess the variations within cohorts with an eye on episodic care, health systems will be able to realize even lower costs — and better outcomes.
*Source: Andi: J. Ryan, C. Lewis, B. Doster and S. Daily, “A Business Process Management Approach to Perioperative Supplies/Instrument Inventory and Workflow,” 2014 47th Hawaii International Conference on System Sciences(HICSS), Waikoloa, HI, USA USA, 2014, pp. 2868-2877. doi:10.1109/HICSS.2014.359
In orthopedics, the average savings was $1500 per case.
In urology, $600 per case and in pediatrics, $470 per case.
Average savings across all surgical cases was $350 per case.