Professor Song-Hee Kim, Usc Marshall School of Business
Most data-driven tools do not ask people for their input. We study whether and how to incorporate the discretion/expertise/intuition of physicians into data-driven decision support tools for improved operational decision-making in hospitals, in an empirical setting of predicting the surgery duration.
We consider three families of models for predicting the surgery duration: models with physician input, statistical models, and models that combine the two. Using the operating room scheduling and usage data from an academic hospital collected over three years, we empirically evaluate and compare the performances of the different models. We find that physician input offers predictive power beyond that of statistical models in our empirical setting. The best performing model is the combined model which includes the physician input as a feature in a statistical model with other predictors. The corrected physician input model, which applies a simple correction to the physician input, performs comparably well (the mean squared error increases by five per cent), when corrections are applied at the surgeon-procedure level. Without the physician input, the mean squared error of the best performing model increases by 17 per cent. We also compare the performances of the operating-room schedules resulting from our models and show that the results carry over.
Our findings suggest that hospital managers should consider eliciting physician input and incorporating it into data-driven decision support tools. We also show how physician input can be best leveraged in the context of predicting the surgery duration. Understanding the benefits/costs of allowing expert input in decision-making has been attracting the interest of the OM community in various application areas. We contribute to this line of research by studying the effect of incorporating physician input on operating room use. For hospital managers, we show the potential value of incorporating physician input in data-driven decision support tools.
This is joint work with Rouba Ibrahim (University College London).
Song-Hee Kim is an Assistant Professor of Operations Management in the Marshall School of Business at the University of Southern California. Her research interests are on using empirical and statistical analysis to build models for decision-making. She has a particular interest in healthcare applications, and has been actively involved in collaborative research projects with several hospitals including the Keck Hospital of USC, Yale New Haven Hospital, Kaiser Permanente in Northern California and Samsung Medical Centre in South Korea. Her research has been published in Management Science, Operations Research, and Manufacturing & Service Operations Management. She has received several academic awards including the Best OM Paper in Management Science Award (winner), MSOM Best Paper Award (finalist), INFORMS Pierskalla Award (finalist), MSOM Student Paper Award (first place), and INFORMS Health Applications Society Student Paper Award (finalist). Song-Hee completed her PhD in Operations Research from Columbia University and earned her BS in Operations Research and Industrial Engineering from Cornell University. Prior to joining the Marshall School, she was a postdoctoral associate at the Yale School of Management.
This seminar is preceded by Dr Daniel Barnett presenting his paper, “Examining Health Workforce Perceptions of Organisational Expectations in Disasters: Leadership Considerations”.