Main Content

Data Analytics in Healthcare - 2018

Our first Round Table was held on May 14th, and focused on papers related to data analytics in healthcare. Aside from members of the Rotman community, many clinicians and administrators from surrounding healthcare organizations attended to learn more about the latest research in this area.

Agenda


8:00 am

Coffee and continental breakfast

8:30 am

Opening remarks (Opher Baron)

8:40 am

Session I, 3 Research presentations (20 m each, with 5 m for questions)

Andre Cire:
Dynamic Scheduling of Home Care Patients to Medical Providers

Avni Shah:
Surcharges Plus Unhealthy Labels Reduce Demand for Unhealthy Menu Items

Opher Baron:
Redesigning the Emergency Department (ED): Lessons from ED at Southlake RHC

9:55 am

Break

10:15 am

Session II, 3 Research presentations (20 m each, with 5 m for questions)

Olga Bountali:
On the impact of treatment protocol restrictions for the uninsured suffering from a chronic disease: The case of compassionate dialysis

Gonzalo Romero:
Data Analytics on the Operations of a Radiology Workflow Platform: Statistical Evidence of Cherry-Picking and its Impact on Service Level

Mihkel Tombak:
Competition between For-Profit and Nonprofit Healthcare Providers and Quality

11:30 am

Facilitated discussion on research and applications in healthcare analytics. We will touch upon research support available from school and the university (Data lab and the Centre for Health Sector Strategy). Opher Baron, Rosemary Hannam, and Jay Cao.  

12:15 pm

Lunch

Abstracts and titles


Session I, 8:40-9:55

Title: Dynamic Scheduling of Home Care Patients to Medical Providers

By Andre Cire. Jointly w.: Adam Diamant

Abstract: Home care aims at providing personalized medical care and social support to patients within their own home. It allows patients to avoid unnecessary hospitalization and either prevents or postpones institutionalization. Since 2014, it has been the fastest-growing US industry attending to more than 14 million patients per year. In this work we propose a dynamic scheduling framework to assist in the assignment of patients to home care practitioners (or HPs). An HP attends to the individual for the entirety of their care (continuity of care requirement) and must travel to their homes in order to serve them. We formulate the assignment of patients to HPs within a home care agency as a discrete-time Markov decision process (MDP). We consider the amount of service each HP provides per period, the expected number of remaining visits a patient will need with an HP, and the total time an HP spends in-transit serving their patient panel. Due to the curse of dimensionality and the complex underlying combinatorial structure of the problem, we propose a one-step policy improvement heuristic that builds upon the agencies existing assignment strategy. Specifically, we apply machine-learning techniques to learn different probabilistic policies from historical data, and formulate the one-step improvement problem as an exponentially-sized mathematical programming model. Such a model can be solved using L-shaped approaches that simultaneously provides upper and lower bounds at each iteration. We derive new relaxations to speed-up the convergence of our method and show sufficient conditions under which this relaxation be solved efficiently. Several extensions account for patients that return for service, multiple HP assignments per patient, and patients who need periodic service are also provided. We test the quality of our solution methodology with data from a Canadian home health care provider to assess the service improvement as compared to their existing policies.

Title: Surcharges Plus Unhealthy Labels Reduce Demand for Unhealthy Menu Items

By: Avni Shah. Jointly w.: James R. Bettman, Peter A. Ubel, Punam Anand Keller & Julie A. Edell

Abstract: Three laboratory experiments and a field experiment in a restaurant demonstrate that neither a price surcharge nor an unhealthy label are enough on their own to curtail the demand for unhealthy food.  However, when combined as an unhealthy label surcharge, they reduce demand for unhealthy food.  We also show that the unhealthy label is equally effective for women as the unhealthy label surcharge but backfires for men, who order more unhealthy food when there is an unhealthy label alone.  We demonstrate that an unhealthy surcharge, which highlights both the financial disincentive and potential health costs, can significantly drive healthier consumption choices.  From a policy and government perspective, if the goal is to reduce demand for unhealthy food, increasing the transparency of the health rationale for any financial disincentive is a necessity in order to effectively lower unhealthy food consumption.

Title: Redesigning the Emergency Department (ED): Lessons from ED at Southlake RHC

By Dmitry Krass. Jointly w.: Opher Baron, Marko Duic, and Tianshu Lu

Abstract: The emergency department (ED) of Southlake Regional Health Centre (Southlake RHC) started a waiting time management project in June, 2011. Since then, the time to physician initial analysis (TPIA) has dropped significantly. In this report, we document the key changes and empirically investigate their impact on four key performance measures: 90th percentiles of TPIA, and the lengths of stay of admitted patients (AdmLOS), non-admitted acute patients (AcuLOS), and non-admitted non-acute patients (NonAcuLOS). We use daily scale data on volume, resource capacity, and seasonal fluctuation. Based on our statistical analysis, we find that TPIA is decreased by nearly two hours; AdmLOS is decreased by more than five hours,; AcuLOS drops one hour when TPIA effect is taken into account; NonAcuLOS drops 15 minutes when TPIA effect is taken into account. Moreover, the total number of patient treated in the ED increased by 23.9% since the project started and until June, 2016. Our findings support the implementation of similar projects to improve the waiting time management in other EDs.

Session II, 10:15-11:30

Title: On the impact of treatment protocol restrictions for the uninsured suffering from a chronic disease: The case of compassionate dialysis

By Olga Bountali

Abstract: Due to protocols restrictions, uninsured patients with a chronic condition are not immediately eligible for systematic treatment. In the case of End Stage Renal Disease, protocol admits such patients for dialysis only if, during a screening process in the Emergency Room, their condition is deemed as life threatening (a.k.a. compassionate dialysis). Motivated by observations at a county hospital, we combine a stylized queueing model and simulation techniques to evaluate the impact of protocol restrictions on system operation and patient welfare, and investigate policies for systemic improvement.

Title: Data Analytics on the Operations of a Radiology Workflow Platform: Statistical Evidence of Cherry-Picking and its Impact on Service Level

By Gonzalo Romero. Jointly w.: T. C. Y. Chan, N. Howard, and S. Lagzi

Abstract: Most hospitals in the U.S. follow an unstructured process for assigning images to radiologists, where each doctor has ample freedom to select the next image to work on out of a common pool. This can result in long processing times and sub-optimal use of the specialists’ time. Motivated by our ongoing collaboration, we analyze a large and unique dataset from the operations of a radiology workflow platform. In particular, we explore whether the point system that Medicare uses to measure the complexity of each type of image -and ultimately for compensating hospitals and radiologist- is aligned with the amount of work required to process them. We find that there exist some misalignment, which opens the door for radiologists to process images that give a higher bang-per-buck first. Although this issue has been discussed in the literature, to the best of our knowledge we provide the first statistical evidence of cherry-picking in the processing of radiological images. Specifically, we show that images with a higher bang-per-buck are processed stochastically faster. Work in progress will address whether this phenomenon has a measurable impact on the time images spent on the platform before being processed, and hence on the service level provided by the platform to its customers.

Tittle: Competition between For-Profit and Nonprofit Healthcare Providers and Quality

By Mihkel Tombak. Jointly w.: Rune Stenbacka

Abstract: We develop a model including many features of healthcare systems: a limited number of approved treatments of certain qualities, insurance schemes reimbursing costs of a standard service, and nonprofit organizations competing with for-profit providers. All the equilibria exhibit quality differentiation and the nonprofit captures a higher market share. Nonprofits (for-profits) supply the standard service when the production cost increase induced by a quality upgrade is sufficiently high (low). When the nonprofit provides the standard quality all patients are served. In contrast, in a for-profit duopoly the standard quality provider charges a price premium, implying that a market segment is excluded.

Research Roundtable: Data Analytics in Healthcare

Surcharges Plus Unhealthy Labels Reduce Demand for Unhealthy Menu Items

Avni Shah, Assistant Professor of Marketing, University of Toronto Scarborough and Research Fellow, Behavioural Economics in Action at Rotman

Redesigning the Emergency Department (ED): Lessons from ED at Southlake RHC

Opher Baron, Professor of Operations Management, Rotman School of Management

»