illustration for Automating Fairness

Image: Chad Hagen

David Rea: Automating Fairness

Rea studies ways to embed fairness in operational systems for service industries, including health care and food distribution.

Story by

Lori Friedman

When David Rea was working on his Ph.D. in business administration at the University of Cincinnati, his focus was mainly statistical. Then, administrators of a local hospital system approached him with an operational challenge that had the potential to improve human well-being. Since then, much of his work has focused on solving problems that present both technical and human-related challenges.

“My field is traditionally very focused on efficiency, which makes sense considering that it comes out of manufacturing principles,” says Rea, now a faculty member in the Decision and Technology Analytics Department of the College of Business. “If a widget waits for a while, or gets lost in the system, as long as your overall efficiency is high, that’s OK. In my work, I’m asking: How do you take those same principles and apply them to situations where efficiency is not the only central objective?”

The local hospital system administrators were looking to manage the allocation of physicians’ clinical time across the different emergency medicine departments they operated throughout the city. While the hospital needed to ensure optimal service levels, it also wanted to accommodate physicians’ location preferences. With physician burnout an ongoing challenge affecting the health systems nationwide, administrators were looking to balance operational needs with a need to ensure their employees’ job satisfaction.

Could Rea help them identify allocations that would be both operationally efficient and be received well by the medical staff?

“You’re essentially trying to automate a fair process,” says Rea. “And the problem with that is that people have differing perceptions of what makes something ‘fair.’”

Illustration for automating fairness

Image: Chad Hagen

Rea and his colleagues developed a model that combines both predictive analytics, using data to identify the likelihood of a future outcome, and prescriptive analytics, using data to determine an optimal course of action. The goal was to balance the inherent trade-offs between two aspects of what is known as distributive justice, or the just allocation of resources: equity and equality. Equality, or sameness of outcomes among individuals, is a common focus of operational models, says Rea. Equity is based on an individual’s outcomes aligning with their inputs.

The model was a success. In an article called “Unequal but Fair: Incorporating Distributive Justice in Operational Alloca- tion Models,”published in the journal Production and Operations Management, Rea and colleagues demonstrate that it is possible to operationalize equality and equity in a way that provides objectively fair outcomes. Using the model, the time that physicians were sent to locations that they did not prefer was substantially reduced. Pre- and post-implementation surveys also revealed significant improvements in employee reported perceptions of fairness, transparency and overall satisfaction with the work-time-al- location process, reports Rea.

“We demonstrated that a system designed to provide objectively unequal allocations can improve employee satisfaction,” says Rea, who, in addition to his primary appointment in the College of Business, has a courtesy appointment with the College of Health.

The hospital administrators gained a solution for a tricky operational and human resource challenge, and they are still using the model today. For Rea, the Emergency Medicine Department case study led to a broadening of his research pursuits, which became to improve human well-being inside service systems.

“How can you design an algorithm to create ‘fair solutions’?” asks Rea. “There’s a human element to it. You have to account for people’s preferences or different definitions, different stakeholders. And that becomes both a really challenging technical problem and a challenge in terms of convincing people that what is going on inside an automated process is in their best interest.”

Devising An Operational Model to Include Multiple Stakeholders

Rea is now working on applying his operational modeling approach to a new context: a situation with more than one stakeholder group. He’s looking at the scheduling needs of another group of medical personnel: non-emergency inter-hospital transport personnel. Like physicians, burnout is a potential issue affecting transport crews, too, making equitable distribution of the workload an important goal. These crews move patients from one facility to another in non-emergency situations, such as from a nursing home to a specialty facility, like a dialysis center or laboratory.

illustration for automating fairness

Image: Chad Hagen

Rea received a real-world look at some of the challenges when he went for several ride-alongs. He observed that there were vastly differing levels of urgency depending on the individual patient, some of which required personnel with highly specialized skills.

He also noted a severe imbalance in workload.

“There are certain types of shifts and roles where the crew will be working nonstop for their full eight- to 12-hour shift. These are kind of the low acuity ones, transporting patients back and forth between a specialty facility and the main hospital,” says Rea.

“You don’t want to be sending ambulances all over the place, for many extra miles, just because it might slightly balance the workload,” says Rea. “I am looking at developing a model that makes allocation or dispatch decisions in a way that is fair to all stakeholders, trying to balance the needs of the workforce and patients, while also creating an efficient system.”

I am looking at developing a model that makes allocation or dispatch decisions in a way that is fair to all stakeholders, trying to balance the needs of the workforce and patients, while also creating an efficient system.

David Rea

Aligning Food Distribution to Community Need

Rea is now working on a long-term project focused on improving efficiency and equity in food distribution. Allocation is, once again, the theme.

“We are still looking at a resource,” says Rea. “Instead of physicians allocating their time to locations, we are looking at how food is being allocated across a region.”

Rea has found one major difference between working with a hospital system and working with regional nonprofit
food organizations: a severe lack of data. Building a data infrastructure is a key first step in the process and one that he expects will take a while.

To understand the local or regional need, explains Rea, nonprofits rely on data from the American

Community Survey, a federal program from the Census Bureau that annually collects data on a variety of topics. The data is aggregated state by state to provide a picture of each area’s need. According to Rea, it takes about two years before the states receive the data.

“Food distribution organizations are trying to use this data to decide where to allocate their resources,” says Rea, “but they’re doing it with this data that is not really reflective of the problem at hand.”

Rea hopes to combine the government data with information he gathers from regional food distribution organizations. This, he says, could provide a real-time picture of the region’s needs. The next step would be for the organizations to utilize the analysis to help make the most effective use of resources. Once the data infrastructure is built and the model to analyze it developed, Rea would then seek to make it scalable and replicable.

“There are some very clear justice aspects to how those food distribution decisions are being made,” says Rea.

Then, he says, there are these highly skilled crews that staff vehicles known as “mobile ICUs.” A nurse practitioner or a nurse is usually onboard to provide care to the patient. These crews are reserved for the much-less-frequent high-acuity cases. So those crews may not be doing as much work as the crews staffing the more day-to- day transportation.

“Ultimately, the question is: How do you take these scarce resources and allocate them in a way that is fair?”

Developing a way for cash-strapped nonprofit organizations to invest in data infrastructure presents an inherent challenge, says Rea. Organizations want to ensure that every dollar is going toward the mission of providing food to people who need it.

Adds Rea: “An investment in infrastructure could ensure that those dollars can be used more efficiently to get food to people and ensure the best use of scarce resources.”

David Rea’s research aspires to improve human well-being inside service systems. His work integrates predictive and prescriptive analytics methodologies, largely focusing on operational problems in the delivery of basic-needs services. Rea has a Ph.D. in business administration from the University of Cincinnati.

Story by

Lori Friedman

Related Stories

Lehigh's new Catastrophe Modeling and Resilience Center

Lehigh's First University Research Center to Focus on Catastrophe Modeling

The first of three new University Research Centers, The Center for Catastrophe Modeling and Resilience aims to explore how hurricanes, pandemics and other large-scale events impact communities.

money photo with stethescope

Nursing Home Industry May Obscure Some of its Profits, Lehigh Researcher Finds

Andrew Olenski, assistant professor of economics, publishes finding in a National Bureau of Economic Research working paper.

children raising hands in classroom

Research Promotes Equitable Reclassification for Multilingual Learners with Disabilities

Sara Kangas, associate professor of special education, presents policy brief to special education state agency leaders from across the U.S.