The evolution of analytics in OR scheduling

Feb. 20, 2017

Modern OR scheduling tools have become commodities that compete mostly on the price of installation and ongoing service. Most modern vendors offer products with block reservations, interactive scheduling, resource requirement mapping, day-of room assignments, work-in prioritization, and more. These vendors also offer some form of reporting. However, if it is not a core focus of the product, the functionality and features may have very limited customization and/or drilldown capability.

A robust OR scheduling tool will go beyond just reporting to include predictive analytics that allow managers to simulate and understand the impact of different scheduling policies. For example, how will adding more orthopedic cases to Fridays affect inpatient bed demand over the weekend?

But unless hospitals leverage the electronic data they have with modern reporting and analysis tools, smaller hospitals are probably better off scheduling on whiteboards — and larger hospitals … on larger whiteboards. The cost of installing and maintaining an OR scheduling system that doesn’t deliver analytics is like buying a car without wheels. The return on investment from electronic scheduling cannot be fully realized unless paired with interactive tools based on modern management science used by every other industry.

Three reasons why major OR scheduling systems offering report capabilities fail to deliver actionable, data-driven insights:

  1. The reports are canned and non-interactive: A good report asks more questions than it answers — and a good data analytic platform keeps answering those questions until the answers catch up to the questions and the lines cross. Reports that are canned may be interesting for a few minutes but they can’t convey the nuance and details necessary for hospital leaders to make decisions. Reports must be customizable to each hospital’s unique culture and policies. Furthermore, they must offer interactive, real-time drilldown capabilities in order to address the questions that come up during OR meetings. When a department or service has poor utilization, managers want to know, but more importantly they want to understand why. Was it a few bad actors? A few bad weeks? A specific day of the week? Good analytic platforms answer those questions in real-time, with everyone seated at the table, instead of a week later in another meeting.
  2. The methodology is opaque: Healthcare professionals need crisp, clear explanations of the metrics presented to them. For example, when reporting on volumes does this refer to procedures, surgeries, visits, or patients? Are block hours prime or non-prime? Are holidays included or excluded? OR meetings that focus on reverse-engineering these questions are not a productive use of anyone’s time. We’ve all seen too many meetings get derailed because these basic questions are not answered with confidence.
  3. The reports lack predictive analytics: To fully leverage the data in hand, hospitals need the ability to run what-if simulations showing the effect of various policy decisions on operations. This is the capstone of a robust data-driven operation and it is what every other modern industry expects from their data operation teams. Hospitals should receive the same benefit.

Looking towards the horizon

OR scheduling has progressed from paper & pencil, to spreadsheets, to the custom-built software that is predominant within the industry today. Many vendors now support scheduling via mobile devices and tablets, with voice control probably not far behind.

For today’s implementations, interactive reports and predictive analytics is fast becoming table stakes, yet major vendors are far behind in delivering this capability. Progressive hospitals must therefore rely on additional third-party solutions.

The next wave of tools will move beyond interactive reporting, beyond predictive analytics, and into the realm of recommendations. This next generation of analytic tools will integrate data analysis with hospital policy to produce actionable recommendations. A report says that volume is low on Fridays. Predictive analytics show what will happen to work-in wait times if more surgeries are scheduled for Friday. A recommendation engine will suggest the optimal number of additional surgeries to schedule to balance opportunity for work-ins with utilization of OR resources.

Users will still want to validate recommendations using predictive analytics and reporting but a computationally generated recommendation will enlighten busy users in a way that might not have happened otherwise. A recommendation engine based on AI or machine learning won’t replace managers but will work as a partner to leverage soft skills and knowledge of unique hospital culture.