Shift Effect in Emergency Departments
By Tianshu Lu, Opher Baron, Professor of Operations Management, Dmitry Krass, Professor of Operations Management and Statistics, Rotman School of Management, University of Toronto
Abstract
We study how a self-interested physician in an emergency department (ED) allocates her capacity between new patients and re-entrant patients, focusing on the time-dependent behavior pattern in a shift. Physicians are often paid according to the amount of service they provide, while EDs consider both average and variability of delay measures such as the time to physician initial assessment (TPIA). The variability is often measured by a specific (high) percentile of the delay. We characterize the physician’s optimal strategy in maximizing throughput; we show that this strategy has undesirable effect of on the TPIA in the ED. We analyze the physician's optimal strategy by using Markov decision process (MDP), and estimate its impact with a fluid queueing model. In both the MDP and the fluid model, we consider a tandem queueing system, composed of two stations: station 0 for new patients, and station 1 for re-entrant patients. This system incorporates two important features: abandonment of new patients, and capacity allocation between both stations. To our knowledge, this is the first paper investigating structural properties of queueing control MDP with abandonment in finite horizon. From the MDP we show that physician's optimal strategy is of two phases: in the first phase, she is more willing to serve new patients, while in the second phase, she is more willing to serve re-entrant patients. From the fluid model we show that in a busy ED, the p-th (p>50) percentile of TPIA increases when this two-phase strategy is applied. Our findings explain the behavior pattern observed in EDs and its impact on the system, shed light on the conflict of interest between ED physicians and managers, and provide guidelines on resolving this conflict.
Watch a video of the presentation →
Presentation slides →
Applying Discrete-Event Simulation to Produce Measurable Change.
By Arun Dixit, Talha Hussain, Millicent Brown, Andrea Ennis, Sandy Marangos, Dr. Ryan Margau, Dr. Bonnie O'Hayon, Mike Sharma, Ann Shook, Dr. Kuldeep Sidhu, Jennifer Zadravec, Jennifer Quaglietta, North York General Hospital
Abstract
Background: In early 2018, North York General Hospital (NYGH) employed specialized software to produce a Discrete Event Simulation (DES) which modeled the processes for ultrasound imaging services at the hospital. The model was built using advanced statistical analyses and methods and was refined with numerous iterative input sessions with leaders and subject matter experts. Implementing ideas to produce measurable change requires a rigorous understanding of how a system or process behaves. This requires understanding if a change implemented to one area may lead to unintended impacts to other elements of a system. Furthermore, significant costs may be associated with the implementation of unsuccessful changes to a system. The DES model was used to analyze multiple potential change ideas in a low-risk environment and determine if the changes would be expected to produce unintended impacts to other areas of the system. Following this analysis, a specific recommendation was tested with a live-trial. All changes implemented were resource neutral for the Medical Imaging (MI) department, and the total number of ultrasound rooms and technologists were kept stable. One Team Attendant was added to support patient transport. As a result, NYGH was able to see a reduction in wait times for patients visiting the Charlotte and Lewis Steinberg Emergency who required ultrasound imaging. Results: As a result of a live-trial, the exam completion turnaround times have decrease by over 15% when compared to the baseline period, with a reduction in the variation in exam completion turnaround times. Conclusion: DES can be an effective tool to evaluate potential changes in quality improvement projects, particularly for operational processes where large amounts of data are available for analysis. Combined with expert clinical knowledge, a DES serves as a low-cost and low-risk technique for understanding the impacts of changes in a system. As a result of the trial, NYGH is exploring opportunities to adapt the simulation model to improve the service for other areas of the hospital.
Watch a video of the presentation →
Presentation slides →
The Power of Hybrid Medical Research: Merging Analytic Modeling With Clinical and Operations Data
By Professor Abraham (Avi) Seidmann, The Xerox Professor of Computers & Information Systems, Electronic Commerce, and Operations Management, Simon Business School, University of Rochester.
Abstract
The rapid deployment of electronic medical records and various digital sensors such as EKG, MRI, and ultrasound, are presenting medical decision makers with a rapidly growing stock of Big Data. Medicine is a data intensive profession, and Big Data for healthcare has a huge upside potential. For instance, recent innovations in image analytics already support early detection, treatment planning and disease monitoring in oncology, cardiology and other areas. Solutions analyze radiology scans, pathology slides and other images to identify and quantify tumors, blood flow, strokes, body composition, gaps in care and several other issues. Yet, not all such applications of data intensive clinical systems have been successful and the bedside test is still a high hurdle to pass. The recent collapse of M.D. Anderson Cancer Center’s ambitious venture to use cognitive computing system to expedite clinical decision-making is a case in point.
analyze radiology scans, pathology slides and other images to identify and quantify tumors, blood flow, strokes, body composition, gaps in care and several other issues. Yet, not all such applications of data intensive clinical systems have been successful and the bedside test is still a high hurdle to pass. The recent collapse of M.D. Anderson Cancer Center’s ambitious venture to use cognitive computing system to expedite clinical decision-making is a case in point. of using (big) data for medical practice management applications, without having a complete understanding of the underlying mechanism. The first study will explain why the performances of clinical workflows depend not only on how various steps are carried out, but also on when certain clinical information items are collected along the workflow. Using our results from a long-term empirical study that looked at the implementation of a Radiology Information System (RIS) at a large regional network of radiology clinics, we reveal how clinics can achieve faster reports turnaround times ― even when it significantly increases the utilization of their bottleneck servers. (Journal of the American College of Radiology, 2009, MSOM, 2012). The second study I plan to mention, investigates the effects of information technology (IT)- enabled automation on staffing decisions in healthcare facilities. Integrating unique nursing home IT data sets from 2006 to 2012, we found that the licensed nurse staffing level decreases by 5.8% in high-end nursing homes but increases by 7.6% in low-end homes after the adoption of automation technology. Combing the above data with a mathematical model of a nursing home staffing, helps us explain that paradox. (Management Science, 2018). The final study I would plan on briefly touching on will discuss our empirical and analytical studies of Telemedicine, and some of the unexpected implications of these powerful technologies in chronic care delivery. (JAMA 2013, Management Science 2018). I plan to conclude the talk with six important implications for data-intensive medical research.
Watch a video of the presentation →
Presentation slides →