Overbooked and Overlooked: Machine Learning and Racial Bias in Medical Appointment Scheduling

50 Pages Posted: 23 Oct 2019 Last revised: 11 Sep 2020

See all articles by Michele Samorani

Michele Samorani

Santa Clara University - Information Systems and Analytics

Shannon Harris

Virginia Commonwealth University (VCU)

Linda Goler Blount

Black Women’s Health Imperative

Haibing Lu

Santa Clara University - Information Systems and Analytics

Michael A. Santoro

Santa Clara University

Date Written: August 31, 2020

Abstract

Problem definition: Overbooking is commonly employed by outpatient clinics to counteract no-shows. State-of-the-art appointment scheduling systems are composed of a machine learning component, which predicts the individual patients’ no-show probability, and an optimization component, which uses these predictions to schedule appointments. The goal is to minimize the schedule cost, computed as a weighted sum of patients’ waiting time and the provider’s overtime and idle time.

Academic/Practical Relevance: Despite its widespread use, we show that the objective of minimizing schedule cost may cause the patients at higher risk of no-show to experience longer waits at the clinic than the other patients. This may translate into undesirable racial disparities, as the patients’ no-show risk is typically correlated with their race.

Methodology: We analytically study racial disparity in this context. Then, we propose new objective functions that minimize both schedule cost and racial disparity, and that can be readily adopted by researchers and practitioners. We develop a “race-aware” objective, which instead of minimizing the waiting times of all patients, minimizes the waiting times of the racial group expected to wait the longest. We also develop “race-unaware” methodologies that do not consider race explicitly. We validate our findings both on simulated and real-world data.

Results: Motivated by the real-world case of a large specialty clinic whose black patients have a higher no-show probability than non-black patients, we demonstrate that state-of-the-art scheduling systems cause black patients to wait about 30% longer than non-black patients. Our race-aware methodology achieves both goals of eliminating racial disparity and obtaining a similar schedule cost as that obtained by the state-of-the-art scheduling method, whereas the race-unaware methodologies fail to obtain both efficiency and fairness.

Managerial Implications: Our work uncovers that the traditional objective of minimizing schedule cost may lead to unintended racial disparities. Both efficiency and fairness can be achieved by adopting a race-aware objective.

Keywords: Appointment Scheduling, Machine Learning, Algorithmic Bias, Socio-economic Bias, Racial Bias

Suggested Citation

Samorani, Michele and Harris, Shannon and Blount, Linda Goler and Lu, Haibing and Santoro, Michael A., Overbooked and Overlooked: Machine Learning and Racial Bias in Medical Appointment Scheduling (August 31, 2020). Available at SSRN: https://ssrn.com/abstract=3467047 or http://dx.doi.org/10.2139/ssrn.3467047

Michele Samorani (Contact Author)

Santa Clara University - Information Systems and Analytics ( email )

500, El Camino Real
Santa Clara, CA 95053-0382
United States

Shannon Harris

Virginia Commonwealth University (VCU) ( email )

1015 Floyd Avenue
Richmond, VA 23284
United States

Linda Goler Blount

Black Women’s Health Imperative ( email )

700 Pennsylvania Ave, SE
Ste. 2059
Washington, DC 2003
United States

Haibing Lu

Santa Clara University - Information Systems and Analytics ( email )

500, El Camino Real
Santa Clara, CA 95053-0382
United States

Michael A. Santoro

Santa Clara University ( email )

Leavey School of Business
500 El Camino Real
Santa Clara, CA 95050
United States
408-5516001 (Phone)

HOME PAGE: http://www.michaelAsantoro.com

Here is the Coronavirus
related research on SSRN

Paper statistics

Downloads
374
Abstract Views
3,010
rank
91,360
PlumX Metrics