This article is based on a presentation at the 2022 Healthcare Analytics Summit (HAS 2022) given by Julie Watson, Senior Vice President and Chief Medical Officer at INTEGRIS Health, and Jason Jones, Chief Analytics and Data Science Officer at Health Catalyst, titled, “How to Use AI to Improve Leadership Decisions and Motivate Change.”
Healthcare organizations have primarily used augmented intelligence (AI) for point-of-care decisions or predictive models, such as identifying which patient is more likely to be readmitted. Moving forward, the industry is starting to explore how AI-assisted decision making can support additional use cases and organizational analysis at the executive level.
For example, recent experience shows that AI can help leaders select incentive metrics for executive compensation that would motivate desired behavioral changes and generate measurable, reproducible results. Healthcare leaders are beginning to leverage AI to distill massive data sets into actionable insights, drive operational decisions, and set executive compensation KPIs that are meaningful, relevant, and fair across hospitals of different sizes and resources.
To implement AI-assisted decision making at the executive level, organizations first need to identify which areas would benefit from AI assistance. In a collaboration between INTEGRIS Health and Health Catalyst, leaders exploring how augmented intelligence (AI) could assist in setting executive compensation plans first listed the tasks that needed to be performed and then sorted them into three categories: 1) tasks that humans should always complete, 2) tasks that computers can do better than humans, 3) and tasks best accomplished by human efforts combined with AI.
The “obviously human” tasks included understanding organizational aims, creating meaning from data points, and evaluating how cultural and environmental influences impact decision-making. The “obviously computer” tasks included data storage, lengthy calculations, and data retrieval. For human- and AI-combined tasks, the collaboration explored how 1) AI could be incorporated into the selection of improvement focus areas and 2) how AI-powered insights could assist leaders in establishing equitable goals across the organization.
After identifying tasks, healthcare leaders can leverage AI to identify areas that would optimize patient care across the organization. Industry leaders have developed new AI-driven tools, such as benchmarking and statistical modeling functions, to help executives identify actionable insights from raw data.
Benchmarking current performance against peer organizations helps leaders identify areas where they already perform in the top quartile and areas where they can most improve. An organization already functioning in the top five percent of a particular area might want to focus on other areas needing improvement.
In addition to benchmarking, new statistical modeling using functions such as Observed/Expected (O/E) ratios and forest plots (graphical displays of estimated results from several case studies addressing the same question) have been developed to help executives identify how the organization as a whole is functioning versus individual hospitals. For example, should the organization focus on sepsis mortality, or is there a single hospital that has more of an opportunity to improve in that area?
Once improvement areas are identified, organizations need to focus on developing ways of making goals attainable, fair, and equitable across the organization. They also need a means of measuring, in a reproducible way, whether improvements beyond random fluctuation occur. INTEGRIS Health has hospitals of varying sizes and resources. They required a way to establish statistical confidence limits (to factor out randomness) and equalize measures for hospital size and current resources. In their collaboration, Health Catalyst proposed solving these problems by incorporating “boost factors,” such as hospital averages, to make the goals equitable for small and large hospitals.
“The most demotivating thing you can do is give people an unattainable goal,” said Jason Jones, Chief Analytics and Data Science Officer at Health Catalyst. “An organization focused on admissions and readmissions, for example, would need a way to achieve equity among hospitals of different sizes.”
To benefit from these AI-driven statistical models, leaders need to be trained to use them. In the collaboration mentioned above, INTEGRIS Health initiated a cascading training sequence in which the head of technical training first mastered the tools and then taught the organization’s executive leadership how to use them. They, in turn, trained vice presidents who trained managers. This training approach meant that people who had just learned how to use the tools taught others how useful they were. INTEGRIS Health discovered that this cascading training model increased adoption and motivated change at every level.
AI-assisted decision making can support executives tasked with incentivizing change throughout healthcare organizations. Through AI, leaders can identify organizational improvement opportunities that may otherwise remain obfuscated. AI-driven analyses can also assist in equalizing improvement aims across hospitals of varying sizes and resource allocation, resulting in improvement aims that are equitable and achievable.
And while statistical analysis concepts such as O/E may be abstract, Julie Watson, Senior Vice President and Chief Medical Officer at INTEGRIS Health, shared an additional learning from her team’s collaboration with Health Catalyst: “Saving lives motivates doctors. When sharing O/E reports and goals with your teams, focus on how many patients teams would save if they met these goals. Targets become meaningful when you talk about the number of lives impacted by an improvement.”