The Operationalization Gap. It's the chasm between innovation and everyday practice that is a challenge for all large organizations.
Hospital executives will likely not be surprised to learn that the so-called "O-Gap" is a term coined by one of their own – former Boston Children's hospital Chief Innovation Officer Naomi Fried. It clearly describes the struggle for healthcare organizations as they seek to take advantage of myriad new technologies to become what the Institute of Medicine has called "a learning health system."
How can hospitals and health systems meet that challenge? First, they need to find ways to evolve long-held clinician habits into new and improved best practices. Then they need to operationalize best practice with minimal disruption, with full clinician support, and without jeopardizing the overall quality of care.
It's a tall order. In the push to realize the benefits of process improvement without losing clinician buy-in or impacting quality, healthcare organizations need technology that can clearly pinpoint the value of process changes and deliver a roadmap to achieve better outcomes.
Machine learning—a term that describes the ability of algorithms to adjust and learn from data and to take or suggest actions—may be the answer they're searching for.
Learning with machines
The Institute of Medicine (IOM) has been promoting the concept of the learning health system for almost 10 years ago. IOM envisions it this way: "science, informatics, incentives, and culture are aligned for continuous improvement and innovation, with best practices seamlessly embedded in the delivery process and new knowledge captured as an integral by-product of the delivery experience" (The Learning Healthcare System: Workshop Summary (IOM Roundtable on Evidence-Based Medicine, http://www.nap.edu/catalog/11903.html).
As healthcare systems strive to realize IOM's vision for continuous improvement in care delivery, many are recognizing that they have outgrown their data management and reporting capacity. Those that have turned to new machine-learning approaches have found they can expand capacity and capabilities while reducing administrative burden on clinicians.
Here's an example of how one health system used machine-learning tools to improve care delivery for intestinal surgery:
Until recently, the health system's surgical services team used traditional methods of hospital data analysis to inform their creation of order sets, protocols, and provider and patient education materials spanning the pre-op, intraoperative and post-op phases of care.
Then they applied a "machine intelligence" platform that pairs machine learning algorithms with topological data analysis (TDA)—a mathematical process that uses shape as an organizing principal for understanding complex data. By giving visible form to their data, the health system was able to replicate and validate years of analytical insights in a matter of days. This fast, powerful analysis also identified new insights that could further improve the health system's protocols for intestinal surgery.
The TDA analysis first confirmed known best practices. It revealed that a small subset of the health system's hospitals and surgeons had begun to adopt newer methods included in the medical literature. Most exciting was the association of the perioperative use of a drug that blocks the effects of opioids on the gut, and shortens lengths of stay. The use of this drug, alvimopan, reduced the slowing effect on the gut (ileus) caused by opioids, thereby allowing patients to eat sooner post-operatively and enabling quicker discharges.
The machine intelligence analysis also validated a variety of other best practices that were part of the health system's surgical protocols—the use of certain opioid analgesics for pain control, early ambulation, early removal of foley catheters and early oral feedings post-op. All of these practices had high, statistically significant correlations with lower length of stay and cost of care. The analysis demonstrated that these best practices really were leading to higher-value care for patients.
Second, the analysis identified several correlations that might represent potential new best practices among patients undergoing colon surgery. The analysis looked at the group of patients with the best clinical and financial outcomes and identified similarities in their treatment (whether it was the use of a certain drug, timing of drug administration or the use of particular surgical supplies). The surgical services team is evaluating these findings and assessing how they might be used to further refine care protocols.
Finally, the analysis identified areas for improvement. Today the health system is planning to use machine intelligence to track adherence to best practices across hospitals and physician groups. The technology will provide the rapid-cycle analysis and reporting needed for effective feedback to surgeons and care teams. Importantly, machine intelligence will make best practices visible in routine care—one of the pillars of continuous improvement. The health system plans to expand the use of machine intelligence to analyze dozens of procedures and hundreds of surgeons on a routine basis. By consistently uncovering new best practices and operationalizing them, this health system will be better able to meet the challenges of today's value-based healthcare payment arrangements while improving patient safety, reducing costs and enhancing the patient and provider experience.
Machine-driven feedback loops
Machine intelligence enables the learning health system by providing a continuous feedback loop of what is working, and what is not, to improve patient care. Presented with clear analytic foundations for process changes, clinicians can appreciate the rationale for adjusted workflows. And not only does machine intelligence help validate existing best practices and discover new ones, but it also delivers the roadmap needed to monitor protocol adoption and give effective feedback to care teams.
Healthcare systems that adopt machine-learning tools now will be well-positioned to tackle the rapid pace of change in the years to come. The tools can help healthcare systems reap the rewards and avoid the penalties associated with upcoming payment reforms such as bundled Medicare payments for total joint replacement. Most importantly, healthcare organizations that have committed to becoming a learning health system will be best prepared to deliver the best possible care for their patients well into the future.
Francis X. Campion, MD, FACP, is Chief Medical Officer for Ayasdi, a developer of machine intelligent applications for health systems and payer organizations.
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.