Analytics has the potential to harness the growing availability of data and propel the development of cutting-edge models that improve the quality and efficiency of medical and health care. This workshop will focus on how different sources of healthcare data, including electronic medical records and clinical trial results, can be leveraged to fundamentally change modern organizations by improving not only healthcare operations but also patient outcomes. In addition, topics related to decision making for macro-scale healthcare policies will be discussed.
This workshop will include a poster session. In order to propose a poster, you must first register for the workshop, and then submit a poster proposal using the form that will become available on this page after you register. The registration form should not be used to propose a poster. The deadline for proposing a poster is March 20, 2023.
Optimal Patient Selection into Care Management Programs
Speaker: Dan Adelman (University of Chicago)
Care Management Programs (CMPs) coordinate the care for patients with complex care needs and older frail adults, who usually represent the top of healthcare spending. Although CMPs have appeared as credible avenues for reducing healthcare utilization, empirical evidence showed mixed results. Using patient-level data we evaluate the causal impact of the CMP of a major academic medical center, and we find no impact on five healthcare utilization measures. In the light of these negative results, one wonders how can CMPs be improved. To address this question, we use Markov Decision Processes (MDPs) and Dynamic Programming to model the task of optimally allocating treatment amongst patients while fulfilling some capacity constraints. The complexity of such a problem may be very high because healthcare populations may be large enough that gathering information of the current status of each patient and tracking the evolution of their covariates is untenable. To address this challenge we develop the so-called measurized theory, which allows to model MDPs that optimize the distribution of treated and untreated patients instead of dealing with identifed patients. This abstraction transforms a complicated problem into an intuitive formulation and sets the stage for delivering clinically implementable solutions in the future.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
Health AI will Fail without Data Sharing
Speaker: Leo Anthony Celi (Massachusetts Institute of Technology (MIT))
The application of artificial intelligence in healthcare requires a team science approach. A diverse set of expertise, perspectives and lived experiences are required to understand the various ways bias lurks in the data – from bias introduced by sampling selection (who made it to the database, who didn’t, and what’s the impact on downstream models), variation in the frequency of measurement that is not explained by the disease or patient phenotype (aka “shortcut” features in medical images), technology that performs differently across patient subgroups (e.g. pulse oximetry, wearable sensors optimized around fit individuals), etc. Data bias is the roadblock to realizing the promise of machine learning. Algorithmic bias is not just about evaluating model performance across patient subgroups post hoc. The goal is to ascertain that the model does not learn from features that should not affect decision making. Offering chemotherapy should not depend on whether a patient is on Medicaid or has a private insurance, predicting job performance should not be informed by the gender of the applicant, optimizing treatment for sepsis should be not be confounded by the use of infrared sensing technology. This is much easier said than done because of the discovery that computers can easily learn sensitive attributes that the human eye does not see. Using real world data to evaluate the models makes this extremely challenging. Excellent model accuracy means existing outcome disparities are fully encoded in the algorithms.
11:45-12:30 CDT
Journey of Integrating Advanced Analytic Models into Clinical and Operational Decision Making within Integrated HealthCare Delivery Network
Speaker: Barry Stein (Hartford HealthCare)
12:30-14:00 CDT
Lunch
14:00-14:45 CDT
Decision-Aware Learning for Global Health Supply Chains
Speaker: Hamsa Bastani (University of Pennsylvania)
The combination of machine learning (for prediction) and optimization (for decision-making) is increasingly used in practice. However, a key challenge is the need to align the loss function used to train the machine learning model with the decision loss associated with the downstream optimization problem. Traditional solutions have limited flexibility in the model architecture and/or scale poorly to large datasets. We propose a principled decision-aware learning algorithm that uses a novel Taylor expansion of the optimal decision loss to derive the machine learning loss. Importantly, our approach only requires a simple re-weighting of the training data, allowing it to flexibly and scalably be incorporated into complex modern data science pipelines, yet producing sizable efficiency gains. We apply our framework to optimize the distribution of essential medicines in collaboration with policymakers at the Sierra Leone National Medical Supplies Agency; highly uncertain demand and limited budgets currently result in excessive unmet demand. We leverage random forests with meta-learning to learn complex cross-correlations across facilities, and apply our decision-aware learning approach to align the prediction loss with the objective of minimizing unmet demand. Out-of-sample results demonstrate that our end-to-end approach significantly reduces unmet demand across 1000+ health facilities throughout Sierra Leone. Joint work with O. Bastani, T.-H. Chung and V. Rostami.
14:45-15:30 CDT
Stochastic Modeling to Personalize Disease Screening Decisions
Speaker: Oguzhan Alagoz (University of Wisconsin, Madison)
This talk describes the use of partially observable Markov decision processes (POMDPs) for personalizing cancer screening decisions. POMDP models can be used to address several controversial open research questions in cancer screening, such as when to start and stop screening and how often to screening. We demonstrate the development and application of a POMDP-based personalized cancer screening policy using breast cancer as an example. In addition, we briefly describe how nonadherence to the screening recommendations, limited screening resources, and existence of chronic conditions could be addressed using the POMDP modeling framework. Finally, we describe successful POMDP applications in other cancers including colorectal and lung cancer screening.
15:30-16:00 CDT
Networking Session
Tuesday, April 4, 2023
9:00-9:45 CDT
Waiting Online versus In-person in Outpatient Clinics: An Empirical Study on Visit Incompletion
Speaker: Carri Chan (Columbia Business School)
The use of telemedicine has increased rapidly over the last few years. To better manage telemedicine visits and effectively integrate them with in-person visits, we need to better understand patient behaviors under the two modalities of visits. Utilizing data from two large outpatient clinics, we take an empirical approach to study service incompletion for in-person versus telemedicine appointments. We focus on estimating the causal effect of physician availability on service incompletion; physician availability is likely to affect the likelihood a patient leaves without being seen behavior but only for patients who show up for the appointment. That is, physician availability should not impact the likelihood of patient no-show. We introduce a multivariate probit model with instrumental variables to handle estimation challenges due to endogeneity, sample selection bias, and measurement error. Our estimation results show that intra-day delay increases the telemedicine service incompletion rate by 7.40%, but does not have a significant effect on the in-person service incompletion rate. This suggests that telemedicine patients may leave without being seen when delayed, while in-person patients are not sensitive to intra-day delay. We conduct counterfactual experiments to optimize the intra-day sequencing rule when having both telemedicine and in-person patients. Our analysis indicates that not correctly differentiating the types of incompletions due to intra-day delays from no-show behavior can lead to highly suboptimal patient sequencing decisions.
9:45-10:30 CDT
Got (optimal) milk?
Speaker: Timothy Chan (University of Toronto)
Human donor milk is considered important nutrition for millions of infants that are born preterm each year. Donor milk is collected, processed, and distributed by milk banks. The macronutrient content of donor milk is directly linked to infant brain development and can vary substantially across donations, which is why multiple donations are typically pooled together to create a final product. Approximately half of all milk banks in North America do not have the resources to measure the macronutrient content of donor milk, which means pooling is done heuristically. We propose a data-driven framework combining machine learning and optimization, to predict macronutrient content of donations and then optimally combine them in pools, respectively. In collaboration with our partner milk bank, we collect a data set of milk to train our predictive models. We rigorously simulate milk bank practices to fine-tune our optimization models and evaluate operational scenarios such as changes in donation habits during the COVID-19 pandemic. Finally, we conduct a year-long trial implementation, where we observe the current nurse-led pooling practices followed by our intervention. Pools created by our approach meet clinical macronutrient targets between 31% to 76% more often than the baseline, while taking 67% less recipe creation time.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
TBA
Speaker: Andrew Schaefer (Rice University)
11:45-12:30 CDT
Digital Footprints: How Smartphone Location Data Can Track Health Behaviors
Speaker: Elisa Long (University of California, Los Angeles (UCLA))
Smartphone geolocation data are increasingly used by social science researchers to study human decision-making and health behaviors. I will summarize three of my recent studies related to nursing home staff networks and COVID; prescription stimulant (eg, Adderall) use in adolescents during periods of academic stress; and utilization of reproductive health clinics offering abortion services. I will discuss the benefits and challenges of working with individual device-level smartphone data, including implications for data privacy.
12:30-14:00 CDT
Lunch
14:00-14:45 CDT
TBA
Speaker: Justin Boutilier (University of Wisconsin, Madison)
15:00-16:00 CDT
Poster Session
Wednesday, April 5, 2023
9:00-9:45 CDT
A Granular Approach to Optimal and Fair Patient Placement in Hospital Emergency Departments
Speaker: Georgia Perakis (Massachusetts Institute of Technology (MIT))
Prolonged emergency department (ED) length of stay (LOS) is associated with detrimental effects on patient care and quality, including increased mortality, increased risk of hospital-acquired infections, and disrupted patient flow. There is also evidence that certain groups of patients experience longer LOS based on their gender or race, especially with regard to the part of LOS that is attributable to waiting to be seen by a clinician. This work tackles the patient prioritization and placement aspects of ED operations with the goal of improving throughput and wait time in a fair, equitable way. We present a novel Mixed Integer Linear Programming (MILP) predictive-prescriptive formulation that incorporates a breakdown of predicted patient ED LOS into actionable pieces and allows for a more granular model of ED operations. We show how to incorporate considerations for fairness and reformulate the MILP formulation into a compact and computationally tractable formulation that can be solved efficiently in real time. To deal with uncertainty, we propose a sampling-based solution, and provide provable guarantees regarding its convergence, stability and sample complexity. The proposed solution increases the throughput of the ED by more than 50% and decreases the average wait time by at least 75% compared to current hospital practice. In addition, the method is near optimal in terms of throughput, and produces high-quality solutions in terms of average wait time compared to a clairvoyant oracle. Finally and importantly, our proposed approach demonstrates desirable properties when it comes to fairness in patient prioritization, illustrating a path for addressing hidden biases in patient ED wait times and hospital operations as a whole. This study was conducted in collaboration with a large urban US academic medical center. Data from more than 40,000 patient visits were used to shape and evaluate the predictive-prescriptive models. The proposed method will be used by the hospital to improve patient flow and quality of care as well as to support more fair and consistent bed allocation decisions. (This is joint work with M. Canellas, D. Pachamanova, O. Skali Lami, A. Tsiourvas)
9:45-10:30 CDT
Helping the Captive Audience: Advance Notice of Diagnostic Service for Hospital Inpatients
Speaker: Nan Liu (Boston College)
Inpatients are often treated as the “captive audience” on-demand for hospital diagnostic service. They are standing by all the time, but notified only when service capacity is available. This arrangement causes significant chaos and inefficiencies in hospital operations. We propose an innovative scheduling approach called “advance notice” to manage hospital diagnostic practice. Advance notice is a brand-new scheduling paradigm in between the classic allocation scheduling and advance scheduling. Patients are placed in acommon queue waiting to be called for service, and they will be provided both a fixed preparation time and a guaranteed service time window in advance (neither a last-minute notice nor an exact service time in the future). The advance notice policy enjoys the benefit of allocation scheduling (giving the provider flexibility in using her capacity) and that of advance scheduling (reducing patient online waiting). It calls for two decisions: who to serve now and who to send advance notices to. We formulate a Markov Decision Process model to optimize these decisions dynamically. Via a novel variable transformation, we reveal the hidden antimultimodular structure of the problem and show how optimal decisions change with the system state. These structural and sensitivity results allow us to develop efficient solution approaches. Our numerical study, populated by real data from a large academic medical center in the United States, demonstrates significant improvement in operational efficiency by switching from current practice to adopting our proposed advance notice policy.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
Leapfrogging for Last-mile Delivery in Health Care
Speaker: Hummy Song (University of Pennsylvania)
Radical technological innovations may allow leapfrogging over traditional solutions to improve access to quality medical care, especially in hard-to-reach areas. Using data from Rwandan public hospitals, we examine the impact of using drones for the delivery of blood products on inventory management and health outcomes. We find that adopting drone delivery leads to a 62% reduction in on-hand inventory of blood products, 42% reduction in their wastage, and 88% decrease in inpatient mortality from postpartum hemorrhage (PPH). Hospitals that experienced road infrastructure improvements prior to adopting drone delivery see a quarter of the decline in PPH mortality compared to facilities that only adopted drone delivery, suggesting a leapfrogging effect.
11:45-12:30 CDT
TBA
Speaker: Bob Batt (University of Wisconsin-Madison)
12:30-14:00 CDT
Lunch
14:00-14:45 CDT
Data-Driven Hospital Admission Control: A Contextual Learning Approach
Speaker: Cong Shi (University of Michigan)
Hospitals are typically uncertain about the readmission impact of a care unit placement decision for a patient. The placement decision is challenging due to the wide variety of patient characteristics, uncertain needs of patients, and the limited number of beds in critical and intermediate care units. We develop an optimization-learning algorithm, called the Data-driven Admission Control (DAC) algorithm, under the presence of limited reusable hospital beds and delayed bandit feedback. The algorithm is designed to adaptively learn the readmission risk of patients and choose the best care unit placement for a patient based on the observed contextual information. The objective is to minimize patient readmissions while capturing the trade-off between the benefit of better health outcomes versus the opportunity cost of reserving high-demand beds for potentially more complex patients arriving in the future. We prove that our proposed online optimization-learning algorithm admits a sub-linear Bayesian regret bound. We also investigate and assess the effectiveness of our methodology using hospital system data. Our empirical results suggest that implementing our approach provides promising results compared to different benchmark policies and improves the current policy of our partner hospital.
14:45-15:30 CDT
TBA
Speaker: Andrew Li (Carnegie Mellon University)
15:30-16:00 CDT
Networking Session
Thursday, April 6, 2023
9:00-9:45 CDT
The burden of evidence for operations research in hospitals
Speaker: Martin Copenhaver (Massachusetts General Hospital and Harvard Medical School)
9:45-10:30 CDT
Does Machine Learning Improve Operational Efficiency? Evidence from the Design of an Emergency Department Vertical Processing Unit
Speaker: Agni Orfanoudaki (University of Oxford)
Addressing hospital emergency department (ED) overcrowding is a critical challenge for many healthcare systems worldwide. To address this challenge, many hospitals have been experimenting with innovative patient flow designs. A promising new design is to separate patients who can be served vertically (e.g., on a regular chair as opposed to horizontally on an ED bed) and route them to a different area termed the Vertical Processing unit, also known as the Rapid Medical Assessment (RMA) unit. While this can potentially increase operational efficiency by addressing the problem of bed availability, it can degrade performance if patients are not correctly routed through the system. Successful implementation of this design, thus, significantly depends on understanding which patients should be routed to the RMA unit. To assist our partner hospital, we developed a machine learning model trained on large-scale data capable of providing a personalized risk score for each arriving patient on whether they will eventually need an ED bed. We then feed these risk scores to an analytical model of patient flow to characterize the optimal protocol for utilizing the RMA unit. We find that the optimal protocol depends not only on the predicted risk scores but also on the machine learning model’s accuracy and ED characteristics. Finally, we use simulation analyses to compare the performance of our recommended RMA-based design with more traditional ED flow approaches such as a “fast track” or a “physician in triage” based design. Our results suggest that following the RMA design under our recommended protocol can bring several advantages to EDs. It outperforms traditional patient flow designs due to the dynamic and efficient use of ED resources, especially in settings with a higher prevalence of high-acuity patient cases. Overall, this work provides a roadmap to healthcare systems that seek to implement data-driven patient flow systems to improve ED operations.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
Policy Optimization for Personalized Interventions in Behavioral Health
Speaker: Jackie Baek (New York University)
Behavioral health interventions, delivered through digital platforms, have the potential to significantly improve health outcomes, through education, motivation, reminders, and outreach. We study the problem of optimizing personalized interventions for patients to maximize some long-term outcome, in a setting where interventions are costly and capacity-constrained. This paper provides a model-free approach to solving this problem. We find that generic model-free approaches from the reinforcement learning literature are too data intensive for healthcare applications, while simpler bandit approaches make progress at the expense of ignoring long-term patient dynamics. We present a new algorithm we dub DecompPI that approximates one step of policy iteration. Implementing DecompPI simply consists of a prediction task from offline data, alleviating the need for online experimentation. Theoretically, we show that under a natural set of structural assumptions on patient dynamics, DecompPI surprisingly recovers at least 1/2 of the improvement possible between a naive baseline policy and the optimal policy. At the same time, DecompPI is both robust to estimation errors and interpretable. Through an empirical case study on a mobile health platform for improving treatment adherence for tuberculosis, we find that DecompPI can provide the same efficacy as the status quo with approximately half the capacity of interventions. (Joint work with V.F. Farias, J.J. Boutilier, J.O. Jonasson, E. Yoeli)
IMSI is committed to making all of our programs and events inclusive and accessible. Contact to request accommodations.
In order to register for this workshop, you must have an IMSI account and be logged in. Please use one of the buttons below to login or create an account.