Active Research

With Harshita Kajaria-Montag
Last Update: April 2025
Working Paper: Upon Request 

Aim:  As a proxy for workload, most health systems carefully monitor the “nurse-to-patient ratio." We propose a fundamentally different tactic: using a painstakingly detailed workload tool, we observe how managers react to more accurate definitions of workload, including how they can better assign and balance work among their staff, as well as how it affects the nurses' turnover.
Methods: We use archival data both before and after the launch of the workload tool from over 60 unique nursing units operated by a large midwestern academic health system in the US. The data ranges from January 2024 to March 2025 and encompasses over 4000 nurses observed interacting with over 60,000 patient encounters.
Results: We find the introduction of the workload-assignment tool leads to a decrease in load imbalance for nurses, it reduces nurse attrition, and it improves patient care.

With Eric Xu, Eunae Yoo, and Hyoju Jeong
Last Update: December 2024
Working Paper: Upon Request 

Background: Providing education is inherently a service operation, an operation with a particular interest in quality outcomes. In today's learning environment, the quality of the service increasingly depends on the quality of the service channel between instructor and student. An internet connection has become the service channel that unlocks resources that can enhance student learning.

Aim: Though prior literature has studied how internet access affects student learning globally, the results remain mixed and context-dependent. We aim to fill the gap by answering: Does access to a reliable internet connection affect student learning outcomes across the US? A longitudinal, US-based study can leverage the diversity between states to connect internet access to student outcomes and so inform national policy and broader international discussions.
Conclusion: While broadband acts as a crucial enabler, disparities in the quality and mode of connectivity highlight persistent gaps in the digital divide. Our results illustrate the importance of infrastructure and policy in the education delivery process, in line with other literature on social impact operations.

With Danqi Luo
Last Update: December 2024
Working Paper: In Preparation

Background: Emergency departments (EDs) are under constant pressure to balance efficiency and quality of care, yet diagnostic test ordering patterns remain a critical and understudied factor in this equation. We know operational characteristics influence test ordering, where ordering tests early in a visit reduces congestion, or where the testing process can alter test-ordering likelihood. In recent years, the frequency of diagnostic test orders has increased substantially, leading to our primary research question: How do physicians structure their test orders over multiple patients, and what implications does this have for efficiency and patient outcomes?
Aim: One understudied aspect of test ordering is batching behavior, where physicians see multiple patients in succession before submitting orders collectively. While prior research has examined when and how frequently tests are ordered, little attention has been given to the ordering process itself — specifically, whether batching affects the quantity of tests ordered, alters resource utilization, or changes patient outcomes. This study addresses this gap by evaluating the impact of batching on test-ordering patterns, ED operational efficiency, and healthcare costs. Our findings provide insight into whether batching should be discouraged as a source of unnecessary test overuse or recognized as an efficiency-enhancing practice in emergency care.

Last Update: August 2023
Working Paper:  SSRN Link
Award: Decision Science Institute, Doctoral Showcase "Best Paper" (runner-up)

Methods:  Our empirical analysis combines observations from healthcare clinics across several US states, anonymized cellphone mobility data, COVID-19 severity measures, and stay-at-home orders. A Lasso-based procedure selects instruments and generates county-level measures of individual mobility, and a random forest forecast validates the insights of our descriptive approach for traffic prediction.
Conclusion: Combining observations from multiple sources allows us to evaluate how traffic to healthcare clinics changed with willingness to travel, stay-at-home orders, and other signals of environmental safety. During our study period, though patients exercised discretion in some negative ways (e.g., foregoing care), patients also shouldered some of the burden of their own wellness. In studying patient discretion, we characterize powerful predictors of healthcare traffic, we suggest how our findings might generalize post-COVID, and we encourage researchers to continue exploring the role of patient discretion in the co-production of wellness.