September 27, 2022

The superior efficacy of machine learning based AI tools

By: Shawn Sefton, MBA, BSN, RN
Share

With all the factors involved in optimizing OR utilization, assigning OR slots is a complex proposition. When you consider that the average cost to run a fully equipped OR is now at approximately $1,000 per hour in staff and equipment costs, it is more important than ever for hospitals to keep as many ORs in use as possible. As hospitals work to clear elective surgery backlogs from the COVID-19 pandemic and get back to their normal business, they need solutions that can help them optimize the scheduling process, accommodate changes and cancellations, and increase OR utilization.

 

Block scheduling and the OR

The concept of block scheduling got its start in education as a way to organize the academic day into chunks of time to enable flexibility for instruction. Today, block scheduling is a buzzword for many industries seeking ways to improve productivity by scheduling a day into blocks for types of tasks or projects.

In recent years, block scheduling has become common practice in perioperative suites at many hospitals. This system essentially serves as a mutual agreement between surgeons and hospitals, in which surgeons commit to providing the hospital’s community with access to their surgical procedures, and the hospital ensures that the OR is resourced appropriately with staff, instrumentation, technology, equipment, and supplies.

Block scheduling has proven to be a superior approach to the traditional first come, first served open scheduling approach, particularly for addressing case variability across multiple surgeons and service lines. When a block schedule is managed well, it improves the way OR time—the OR’s most valuable resource—is managed. It can help hospitals improve utilization of OR capacity, maximize efficiency, and realize cost savings. For surgeons and specialties with high case volumes, it can ensure that they have predictable and reliable operating times. It also enables OR leaders to rearrange blocks of time when they need to adjust for usage and accommodate for business imperatives.

 

Getting block scheduling right

Accurate scheduling of case times for surgical procedures is important because it is the key to optimum utilization of OR capacity and resources. When an OR schedule is accurate and finalized on a timely basis, the OR and OR support leaders are better able to optimize planning and resourcing.

Yet, getting block scheduling right has been a daunting challenge for hospital staff for many reasons. First, they need to meet the often-conflicting priorities of various OR stakeholders. Surgeons demand OR access, while anesthesiologists are challenged with the need to balance multiple locations in and outside the OR. And OR staff appreciate planning for resource availability at the time of the procedure and do not appreciate unanticipated overtime work.

When procedures are not booked accurately, cases that run longer than the booked time create delays for subsequent cases, impact room availability, and disrupt staff and resources. Cases that finish in less time than originally scheduled result in valuable OR time that is unused. And, of course, where there are significant delays, a domino effect results, delaying all the cases that follow on the schedule. The results are often drops in patient satisfaction, frustrated staff, and wasted resources—with substantial revenue not realized and unnecessary costs incurred.

 

Improving surgical case duration prediction

One of the biggest challenges for OR schedulers is estimating the amount of time a specific surgical procedure will take. Typically, case duration estimates have been calculated in one of two ways: by using an estimate provided by the surgeon or by relying on historical averages for individual surgeons and procedures from electronic health record (EHR) scheduling systems.

The problem is that neither method is working well enough to meet the needs of today’s OR managers. Surgeon estimates are often unreliable. In some cases, surgeons may purposely overestimate case duration to prevent releasing OR time to another case or to account for anticipated complications. In other cases, they may underestimate a case duration to fit their allocated OR time, to account for potential cancellations, or they did not account for steps that do not involve them directly.

Relying on perioperative data from the EHR can also lead to inaccurate predictions. Today, all perioperative EHRs have the capability to predict the average case length for a surgical procedure by taking a rolling average and suggesting a case duration and automatically populating the OR schedule with a case length that can be overwritten by OR leaders. Yet, the fact that there is so much variability among cases and the factors involved in each case makes it difficult to use EHR data to pinpoint an average duration for a specific case.

 

A better way

While block time allocation and management provide benefits that support efforts to optimize OR operations, the need to create, find, and fill available OR time within the existing schedule—and pinpoint opportunities to reallocate existing block time—continues to challenge perioperative leaders.

What is needed is a model or methodology that provides a more accurate estimate of case duration that takes into consideration both actual versus scheduled time. Scheduled time for a procedure refers to the actual time allotted for the case, while actual time refers to the actual turnaround time for the case, including from when the patient entered to when they left the OR. The rule of thumb is that the acceptable difference between scheduled and actual case time should be less than 15 minutes at most.

The traditional methods for predicting surgery case duration are often inaccurate because they rarely take into consideration other important factors relating to the patient, case, staffing availability, or resources. Using the traditional method, OR managers and clinical staff need to review every case manually to pinpoint any that are scheduled incorrectly. This requires more time for the entire process of case scheduling and review and preparing OR resources.

Over the years, researchers have searched for new ways to improve the accuracy of predictive data, including regression and other statistical methods. Machine learning (ML) has more recently demonstrated greater accuracy in predicting case time duration.

A recent study comparing case duration predictions for colorectal surgery patients using existing moving average-based estimates with ML-based predictive models that considered additional patient, surgeon, and surgery-related factors at two hospital sites found that the ML-based models improved prediction of scheduled durations across multiple metrics over the moving average-based models.

Another study found that ML delivers significant improvements in predictive accuracy. In 2019, researchers at the University of Washington School of Medicine collected historical EHR data about patient, surgery type, staff, and scheduling information for 45,000 procedures to develop surgeon-specific ML models, which resulted in improvements over surgeon estimates of as much as 50%.

 

Benefits of ML-based AI

Artificial intelligence (AI), with its ability to make the large volumes of data in an EHR actionable, is uniquely suited for use by OR schedulers. One way AI can be valuable is in matching available time slots with the surgeon most likely to make use of the time. AI can be programmed to work within the hospital’s scheduling policies (eg, keeping OR time within a service line or length of procedure) and rank and prioritize the best possible matches for specific time slots.

Conversely, AI can be used to identify “hidden” OR time. By combining historical data about surgeons with AI and predictive analysis, administrators can pinpoint future time blocks or segments that a particular blockholder may not be likely to use and, employing a simple mobile workflow, prompt them to release that time so that it can be used by other surgeons.

The benefits of AI’s capabilities are the best of both worlds: Surgeons are presented with OR slots that match their needs and preferences, while the hospital benefits from improved OR utilization and increased case volume.

 

AI tools eliminate guesswork

Clearly, there is a need for a way to improve OR scheduling accuracy with case length predictions that factor in more than historical averages. Using a data-driven approach, a more accurate case length estimate can be achieved by considering historical averages as well as specific information, case details, and staffing availability.

Using data specific to the patient, the surgical team, and the case, ML-based AI is able to look beyond historical case length averages to accurately predict case length and improve the scheduling accuracy of an OR department.

When considering the implementation of a solution that leverages data science to predict more accurate case lengths, one should seek a tool that provides the ability to schedule cases more accurately and minimize time-consuming back-and-forth communication among staff, identify scheduling inaccuracies before surgery begins, and reduce the day-of-surgery impact of inaccurate scheduling.

To meet these objectives, the solution should enable surgical schedulers to provide a default requested case length specific to the patient and case details when entering a case booking request. It should enable OR schedulers to identify potential scheduling accuracy issues at the time they are reviewing case booking requests. It should also enable OR leaders and OR schedulers to identify any scheduling issues days prior to the day of surgery so they can proactively make adjustments to the schedule.

When the OR scheduler is provided with the details that generated the case duration prediction—such as procedures and modifiers, the surgical team, patient-specific data, and past surgeries—they have the ability to override the prediction and request any adjustment in the case length that may be necessary. When there is inadequate data to generate a prediction, the tool should default to historical averages.

A solution like this enables OR schedulers to estimate case times more accurately and to reduce the impact of inaccurate scheduling on resources. It also helps them be proactive by identifying scheduling inaccuracies before surgery begins.

A case prediction solution that leverages the power of AI not only enables schedulers to make adjustments before the day of surgery, it reduces the impact on stakeholders and resources. It also reduces the need for manual review of individual cases for potential issues.

Improving surgery case duration prediction can also give surgeons renewed confidence in the accuracy of the OR schedule, so they are able to schedule cases more accurately without the need for multiple contacts among surgeons, their clinics, and OR schedulers. And they experience fewer delays caused by inaccurate scheduling.

Now, more than ever, as hospitals ramp up to schedule surgeries that may have been put off during the pandemic, it will be critical that they integrate ML-based solutions into their block schedules and their daily clinical operations. The ability to accurately predict surgical or procedural case lengths and schedule cases accordingly enables OR leaders to optimize utilization of existing resources while also creating additional capacity and additional primetime cases, without the need for new capital investment.

Improved accuracy also reduces costs from unplanned overtime caused by cases that run longer than anticipated. This all results in increased patient and staff satisfaction.

 

—Shawn Sefton, MBA, BSN, RN, is a 35-year nursing executive and vice president of clinical operations and chief nursing officer at Hospital IQ.

 

References

Here’s a Better Way to Schedule Surgeries. Based on the research of Bandi C. and Gupta D. Kellogg Insight. September 5, 2018.

Macario A. Is It Possible to Predict How Long a Surgery Will Last? Medscape. July 14, 2010.

The START methodology to work up and manage surgery scheduling inaccuracy. Association of Anesthesia Clinical Directors. December 12, 2016.

Join our community

Learn More
Video Spotlight
Live chat by BoldChat