Data science and AI continue to make healthcare headlines because of their potential to change everything from how drugs are developed to revolutionizing information workflows.
One area where we anticipate opportunities for significant impact is in the field of operations research. While the general public may not have heard as much about this field, it could be equally transformational.
Operations research is the science of using data and mathematics to optimize how organizations function, which can bring major improvements in quality and efficiency. In the case of a cancer center, this includes everything from scheduling procedures to implementing new technologies and design for hospitals of the future.
At Houston-based The University of Texas MD Anderson Cancer Center, advancing our high-stakes mission to end cancer means it is vital to constantly evaluate every step of the cancer care process so that we can evolve as treatments and populations evolve. Operations research ensures that our scientists’ discoveries are making it to patients in need as quickly and efficiently as possible.
Here are just a few examples of tools we are developing that utilize operations research.
Predicting inpatient length of stay and unplanned readmissions
A major theme of operations research is better leveraging data to predict the future. Two areas with tremendous opportunity are predicting the length of stay for inpatients and the likelihood of unplanned readmission following discharge.
We are working with Ryan Huey, MD, of the department of gastrointestinal medical oncology, to develop tools to predict costly unplanned readmission of chemotherapy patients. Our model estimates risk and provides oncologists with decision support specific to each patient, allowing them to weigh and consider whether a patient would benefit from extended recovery or have more frequent follow up to minimize the risk of return to the emergency room. The model could also provide oncologists with information to adapt treatment plans in a manner that further reduces the chance of side effects and readmission.
Our operations research team is developing algorithms to predict a patient’s length of stay. Using nearly a decade of data from MD Anderson, we worked with hospital administration and nurse managers to understand challenges in inpatient bed management and discharge to develop a tool that is 99% accurate in predicting the number of nights stay required for safe recovery and discharge. Exciting aspects of the model include the incorporation of patient-specific information in the prediction and the ability to update the prediction as a patient proceeds in their care pathway, adapting to information as it arises in the course of therapy and recovery. This approach truly demonstrates the transformational potential of operations research in healthcare not only by helping to guide optimal treatment and recovery pathways but also catching “outlier” cases with unexpectedly long recovery times and giving physicians opportunity to intervene earlier for a smoother recovery.
Surgical scheduling: Optimizing efficiency and improving workforce wellness
Over the past several decades, operations research has fueled dramatic shifts in a number of industries. For example, in air travel, increased efficiency wasn’t accomplished by building more runways but by optimizing schedules and workflow using data to predict the timing of every flight.
Yet surgical scheduling has remained a mostly manual process for more than 50 years, and the downstream effects of even one procedure going longer than expected can be taxing. Not only are such inefficiencies frustrating to patients, but they result in longer hours and shifts for staff. Improving the robustness, reliability and efficiency of surgical scheduling can therefore result in fewer delayed surgeries and major improvements in patients and employee experience.
An essential first step is a more accurate estimate of surgical case duration, which within the state of the art can be prone to error and often does not account for important patient-specific factors such as age, weight, frailty and extent of disease.
Working closely with a team led by Mark Clemens, MD, associate vice president of perioperative services, alongside advanced practice providers, we uncovered numerous factors leading to scheduling inaccuracy. This allowed us to develop an algorithm to more accurately predict case duration, including “outlier” cases for which surgeries are likely to take longer than normally expected.
As a next step, we are developing algorithms for surgical schedule optimization analogous to the “block scheduling” methods that have revolutionized other industries, optimizing not only for increased operating room utilization but also factors that directly relate to the quality of daily schedules and improved patient and staff satisfaction.
Capacity planning: Using data science to plan ahead
As operations data scientists, our work can often yield incremental improvements on the margins (which can add up), though it can be rare that we get to suggest major shifts in how things are done. One such case arises when we bring operations research insight to the planning and development of new facilities.
Traditional thinking may simply take what you’re currently doing and scale it up a bit. After all, if you’re just planning to do more of the same, then planning is easy. However, in the rapidly changing landscape of healthcare, we need to be ready for change and draw on deep insights from data to think strategically for the future.
A shift toward outpatient-focused care is clear, which presents a change in the landscape as we plan new healthcare facilities. For major questions such as the number of operating rooms and recovery beds, “more of the same” will not suffice.
Working closely with Robert Ghafar, our vice president of procedural and therapeutic operations, we again brought nearly a decade of data to bear on such questions and reached dramatically different projections than suggested by traditional thinking. Such approaches are helping to shape numerous questions, from which technologies are best integrated in each care setting to the overall footprint and expansion of our services in years ahead.
Enabling effective implementation
Operations research can be an invaluable asset for hospitals and health systems. Teams aiming to implement these data-driven solutions should keep in mind a few best practices. Through our joint initiative with Houston-based Rice University, the newly formed Center for Operations Research in Cancer, we look forward to continuing our approach utilizing data to better inform operational decisions.
While healthcare has been a relatively late adopter of operations research solutions, we’re now faced with a compelling opportunity to answer our industry’s most pressing needs. Success in transforming operations for clinical benefit relies on an organization’s ability to collect and use data to gain insight on important questions, attract and retain talent in these important domains of applied mathematics and engineering, and effectively implement and manage data science solutions in operations.
