Analytics for Improved Healthcare

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.

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.

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.

Organ transplantation programs in the United States have seen increased scrutiny of outcomes in the past twenty years.  Under regulations by the Organ Procurement Transplantation Network (OPTN) and Centers for Medicare and Medicaid (CMS), the United States has seen a rise in risk-averse patient selection among transplant programs, resulting in decreased transplantation volume for some programs.  However, there is debate in the clinical literature over whether this observed response is rational.  In this work, we develop a chance-constrained mixed-integer program to model the perspective of a transplant program that seeks to simultaneously maximize transplant volume and control the risk of OPTN/CMS penalization.  Using our model, we demonstrate that under realistic conditions, it may be rational for a transplant program to curtail its transplant volume in order to avoid penalization.  Moreover, we demonstrate that this incentive does not disappear even if regulators use accurate risk adjustment for high-risk patients.  Our findings provide the first rigorous theoretical evidence that OPTN/CMS regulations create incentives for programs to reject certain medically-suitable patients.

Tuberculosis (TB) is a global health priority and ending the TB pandemic is part of the United Nations Sustainable Development Goals. Lack of patient adherence to treatment protocols is the main barrier to reducing the global disease burden of tuberculosis. In this talk, we will study the operational design of a treatment adherence support platform that requires patients to verify their treatment adherence on a daily basis. To do this, we partner with Keheala, a TB treatment adherence support provider in Kenya and use data from a completed randomized controlled trial. First, we investigate who should be enrolled on the platform by evaluating who benefits from treatment adherence support. We use a causal forest framework to estimate heterogenous treatment effects and demonstrate that differentiated care can improve program efficiency and reduce inequity in treatment outcomes. Second, we use an empirical framework to quantify the impact of costly human support sponsor outreach on subsequent patient verification behavior. We then develop a rolling-horizon machine learning framework to generate dynamic risk predictions for patients enrolled on the platform. Our analysis establishes that patient verification can be increased by personal sponsor outreach and that patient behavior data can be used to identify at-risk patients for targeted outreach. This work represents an important step towards shifting the current TB treatment adherence paradigm from observational (i.e., monitoring and collecting data on patient adherence) to actionable (i.e., combining behavioral data with analytics to improve patient adherence in a personalized manner).

A key step in every emergency department visit is the assignment of the patient to a physician. We compare the operational performance of two emergency department patient-physician assignment system. One system is a Physician-choice system in which the physician chooses what patient to serve next. The second system is a Nurse-assignment system in which a nurse assigns the patient to an attending physician (i.e., senior physician). We test for the presence of “free riding” in the physician choice system in which physicians are able to reduce their total work load and to avoid difficult or undesirable patients. We also test for the presence of “bed blocking” or “foot dragging” in the nurse assignment system which physicians have an incentive to keep patients in treatment longer than necessary so as to avoid being assigned a new patient. Key measures of interest include wait time, doctor “pickup” time, and treatment time. We also examine whether the assignment system affects care intensity and quality measures such as charges, relative value units, orders, and revisits.

An accurate blood test for early-stage cancer (a “liquid biopsy”) is arguably the most important open problem in oncology, and the race to a solution is tantalizingly close to the finish. In this talk, we will discuss the state of this race as of 2023, and how optimization will play a role in reaching the finish line. In particular, we address a set of active learning problems that occur in the development of liquid biopsies via the lens of active sequential hypothesis testing (ASHT). This is a classic problem in which a learner seeks to identify the “true” hypothesis from among a known set of hypotheses, given a set of actions. Motivated by applications in which the number of hypotheses or actions is massive, we propose efficient (greedy, in fact) algorithms and provide the first approximation guarantees for ASHT, under two types of adaptivity. Both of our guarantees are independent of the number of actions and logarithmic in the number of hypotheses. We numerically evaluate the performance of our algorithms using both synthetic and real-world DNA mutation data, demonstrating that our algorithms outperform previously proposed heuristic policies by large margins. This is joint work with Kyra Gan, Su Jia, and Sridhar Tayur.

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.

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)

Healthcare systems are limited resource environments where scarce capacity is often reserved for the most critical/urgent patients. However, there has been a growing interest in the use of proactive care when less urgent patients may become urgent while waiting. On one hand, providing care for patients when they are less urgent could mean that fewer resources are needed to return them to a healthy, stable state. On the other hand, utilizing limited capacity for patients who may never need that level of care in the future takes the capacity away from the other more urgent patients who need it now. To understand this tension, we propose a queueing model with two customer classes: moderate and urgent, and customers can transition from the moderate class to the urgent class while waiting. We characterize how moderate and urgent customers should be prioritized for service when proactive service for moderate customers is an option. Our focus is on the transient setting, where a demand surge pushes the system away from its normal state of operation. The goal is to cope with the surge in a cost-effective way. In this context, we demonstrate how optimal control theory can be applied to derive structural insights into the optimal scheduling policy.  This is based on joint work with Carri Chan and Yue Hu.