arXiv subjects

David Scheinker

Publications and source records attributed to David Scheinker.

13 recordsLinked to original sources

arXiv2026

An ambiguity taxonomy for evaluating large language model performance on clinical registry abstraction: a multi-site prospective study

Objective: To evaluate large language model (LLM) performance on unprocessed electronic medical record (EMR) data for clinical registry abstraction. Methods: We evaluated LLM performance answering registry questions for the American College of Cardiology National Cardiovascular Data Registry (ACC NCDR). In a pilot study at an academic medical center, the model identified candidate data sources for each registry question and experienced abstractors used these results to define question-specific document sets. In a validation study at a second center with a second ACC NCDR registry, the LLM answered questions using the question-specific document sets. Before reviewing any output, two abstractors independently established the ground truth and assigned each question to one of six categories, ordered by the ambiguity and clinical reasoning required to resolve it: Medication/Event Flag, Binary Clinical Presence, Administrative, Quantitative Laboratory/Physiologic, Clinical Interpretation, and Event Timing. Results: The analytical sample comprised 9,430 abstractor answers reconciled to 4,715 consensus answers (501 pilot; 4,214 validation). In the pilot, candidate data sources per question averaged between 14.6 (SD 13.9) for demographics and 89.2 (SD 56.1) for history and risk factors. In validation, human inter-rater agreement was approximately 98\% while 87\% of LLM answers exactly matched consensus, 2\% partially, and 9\% did not. Mean question-level accuracy was 91.5\% (SD 13.4\%) across 157 questions with at least 20 answers, and declined as ambiguity increased, from 96\% for Medication/Event Flag to 62\% for Event Timing questions. Conclusions: LLMs answering clinical registry questions on unprocessed EMR data achieved far lower accuracy than human abstractors. LLM accuracy fell steadily as ambiguity and the level of required clinical reasoning increased.

cs.CL↗

arXiv2025

Optimal Control for Remote Patient Monitoring with Multidimensional Health States

Selecting the right monitoring level in Remote Patient Monitoring (RPM) systems for e-healthcare is crucial for balancing patient outcomes, various resources, and patient's quality of life. A prior work has used one-dimensional health representations, but patient health is inherently multidimensional and typically consists of many measurable physiological factors. In this paper, we introduce a multidimensional health state model within the RPM framework and use dynamic programming to study optimal monitoring strategies. Our analysis reveals that the optimal control is characterized by switching curves (for two-dimensional health states) or switching hyper-surfaces (in general): patients switch to intensive monitoring when health measurements cross a specific multidimensional surface. We further study how the optimal switching curve varies for different medical conditions and model parameters. This finding of the optimal control structure provides actionable insights for clinicians and aids in resource planning. The tunable modeling framework enhances the applicability and effectiveness of RPM services across various medical conditions.

eess.SY↗

arXiv2024

Learning Explainable Treatment Policies with Clinician-Informed Representations: A Practical Approach

Digital health interventions (DHIs) and remote patient monitoring (RPM) have shown great potential in improving chronic disease management through personalized care. However, barriers like limited efficacy and workload concerns hinder adoption of existing DHIs; while limited sample sizes and lack of interpretability limit the effectiveness and adoption of purely black-box algorithmic DHIs. In this paper, we address these challenges by developing a pipeline for learning explainable treatment policies for RPM-enabled DHIs. We apply our approach in the real-world setting of RPM using a DHI to improve glycemic control of youth with type 1 diabetes. Our main contribution is to reveal the importance of clinical domain knowledge in developing state and action representations for effective, efficient, and interpretable targeting policies. We observe that policies learned from clinician-informed representations are significantly more efficacious and efficient than policies learned from black-box representations. This work emphasizes the importance of collaboration between ML researchers and clinicians for developing effective DHIs in the real world.

cs.LG↗

arXiv2024

Tiered Service Architecture for Remote Patient Monitoring

We develop a remote patient monitoring (RPM) service architecture, which has two tiers of monitoring: ordinary and intensive. The patient's health state improves or worsens in each time period according to certain probabilities, which depend on the monitoring tier. The patient incurs a "loss of quality of life" cost or an "invasiveness" cost, which is higher under intensive monitoring than under ordinary. On the other hand, their health improves faster under intensive monitoring than under ordinary. In each period, the service decides which monitoring tier to use, based on the health of the patient. We investigate the optimal policy for making that choice by formulating the problem using dynamic programming. We first provide analytic conditions for selecting ordinary vs intensive monitoring in the asymptotic regime where the number of health states is large. In the general case, we investigate the optimal policy numerically. We observe a threshold behavior, that is, when the patient's health drops below a certain threshold the service switches them to intensive monitoring, while ordinary monitoring is used during adequately good health states of the patient. The modeling and analysis provides a general framework for managing RPM services for various health conditions with medically/clinically defined system parameters.

eess.SY↗

arXiv2022

The Design and Implementation of a Broadly Applicable Algorithm for Optimizing Intra-Day Surgical Scheduling

Surgical scheduling optimization is an active area of research. However, few algorithms to optimize surgical scheduling are implemented and see sustained use. An algorithm is more likely to be implemented, if it allows for surgeon autonomy, i.e., requires only limited scheduling centralization, and functions in the limited technical infrastructure of widely used electronic medical records (EMRs). In order for an algorithm to see sustained use, it must be compatible with changes to hospital capacity, patient volumes, and scheduling practices. To meet these objectives, we developed the BEDS (better elective day of surgery) algorithm, a greedy heuristic for smoothing unit-specific surgical admissions across days. We implemented BEDS in the EMR of a large pediatric academic medical center. The use of BEDS was associated with a reduction in the variability in the number of admissions. BEDS is freely available as a dashboard in Tableau, a commercial software used by numerous hospitals. BEDS is readily implementable with the limited tools available to most hospitals, does not require reductions to surgeon autonomy or centralized scheduling, and is compatible with changes to hospital capacity or patient volumes. We present a general algorithmic framework from which BEDS is derived based on a particular choice of objectives and constraints. We argue that algorithms generated by this framework retain many of the desirable characteristics of BEDS while being compatible with a wide range of objectives and constraints.

cs.AI↗

arXiv2022

Surgical Scheduling via Optimization and Machine Learning with Long-Tailed Data

Using data from cardiovascular surgery patients with long and highly variable post-surgical lengths of stay (LOS), we develop a modeling framework to reduce recovery unit congestion. We estimate the LOS and its probability distribution using machine learning models, schedule procedures on a rolling basis using a variety of optimization models, and estimate performance with simulation. The machine learning models achieved only modest LOS prediction accuracy, despite access to a very rich set of patient characteristics. Compared to the current paper-based system used in the hospital, most optimization models failed to reduce congestion without increasing wait times for surgery. A conservative stochastic optimization with sufficient sampling to capture the long tail of the LOS distribution outperformed the current manual process and other stochastic and robust optimization approaches. These results highlight the perils of using oversimplified distributional models of LOS for scheduling procedures and the importance of using optimization methods well-suited to dealing with long-tailed behavior.

cs.LG↗

arXiv2020

The Development and Deployment of a Model for Hospital-level COVID-19 Associated Patient Demand Intervals from Consistent Estimators (DICE)

Hospitals commonly project demand for their services by combining their historical share of regional demand with forecasts of total regional demand. Hospital-specific forecasts of demand that provide prediction intervals, rather than point estimates, may facilitate better managerial decisions, especially when demand overage and underage are associated with high, asymmetric costs. Regional forecasts of patient demand are commonly available as a Poisson random variable, e.g., for the number of people requiring hospitalization due to an epidemic such as COVID-19. However, even in this common setting, no probabilistic, consistent, computationally tractable forecast is available for the fraction of patients in a region that a particular institution should expect. We introduce such a forecast, DICE (Demand Intervals from Consistent Estimators). We describe its development and deployment at an academic medical center in California during the `second wave' of COVID-19 in the Unite States. We show that DICE is consistent under mild assumptions and suitable for use with perfect, biased, unbiased regional forecasts. We evaluate its performance on empirical data from a large academic medical center as well as on synthetic data.

stat.AP↗

arXiv2018

Extremum Seeking Optimal Controls of Unknown Systems

We present a method for finding optimal controllers for unknown, time-varying, dynamic systems which can be re-initialized from a given initial condition repeatedly, in which the performance measure is available for sampling with noise, but analytically unknown. Such systems are present throughout industry where processes must be repeated many times, such as a voltage source which is repeatedly turned on for a fraction of a second from zero initial conditions and then turned off again, whose output must track a specific trajectory, while the system's components are slowly drifting with time due to temperature variations. For systems with convex cost functions we prove that our algorithm will produce controllers that approach the minimal cost, e.g., the cost minimizing LQR optimal controller that could have been designed analytically had the system and objective function been known. We demonstrate the algorithm's effectiveness with simulation studies of noisy and time-varying systems.

math.OC↗

arXiv2016

Extremum Seeking for Stabilization of Systems Not Affine in Control

In [22] a form of extremum seeking for control (ESC) was developed for the stabilization of uncertain nonlinear systems. In ESC the extremum seeker itself controls the systems through feedback rather than fine tuning a controller. The ESC results, and other related results, apply only to systems affine in control. However, in most physical systems the control effort enters the system's dynamics through a nonlinear function, such as an input with deadline and saturation. In this work, we utilize our previous results on ESC to develop stabilizing controllers for systems of practical interest that are non-affine in control.

math.DS↗

arXiv2016

Predicting Pediatric Surgical Durations

Effective management of operating room resources relies on accurate predictions of surgical case durations. This prediction problem is known to be particularly difficult in pediatric hospitals due to the extreme variation in pediatric patient populations. We propose a novel metric for measuring accuracy of predictions which captures key issues relevant to hospital operations. With this metric in mind we propose several tree-based prediction models. Some are automated (they do not require input from surgeons) while others are semi-automated (they do require input from surgeons). We see that many of our automated methods generally outperform currently used algorithms and even achieve the same performance as surgeons. Our semi-automated methods can outperform surgeons by a significant margin. We gain insights into the predictive value of different features and suggest avenues of future work.

stat.AP↗

arXiv2013

Hilbert function spaces and the Nevanlinna-Pick problem on the polydisc II

In \cite{ds_hfs}, a geometric procedure for constructing a Nevanlinna-Pick problem on $\D^n$ with a specified set of uniqueness was established. In this sequel we conjecture a necessary and a sufficient condition for a Nevanlinna-Pick problem on $\DT$ to have a unique solution. We use the results of \cite{ds_hfs} and Bezout's theorem to establish three special cases of this conjecture.

math.CV↗

arXiv2011

Hilbert function spaces and the Nevanlinna-Pick problem on the polydisc

In the paper `Distinguished Varieties,' Agler and McCarthy used Hilbert function spaces to study the uniqueness properties of the Nevanlinna-Pick problem on the bidisc. In this work we give a geometric procedure for constructing a Nevanlinna-Pick problems on D^n with a specified set of uniqueness. On the way to establishing this procedure, we prove a result about Hilbert function spaces and partially answer a question posed by Agler and McCarthy.

math.FA↗

arXiv2010

A uniqueness theorem for bounded analytic functions on the polydisc

For each n,N>0 we construct a set of points x_1,...,x_M in D^n with the following property: if f is a rational inner function on D^n of degree strictly less than N and g is an analytic function mapping D^n to D that satisfies g(x_i)=f(x_i) for each i=1,...,M, then g=f on D^n. In terms of the Pick problem on D^n, our result implies that for any rational inner f of degree less than N, the Pick problem with data x_1,...,x_M and f(x_1),...,f(x_M) has a unique solution.

math.CV↗