Continual learning has emerged as a powerful approach to address the challenges of non-stationary environments, allowing machine learning models to adapt to new data while retaining the previously acquired knowledge. In time-sensitive healthcare applications, where entities such as physicians, hospital rooms, and medications exhibit continuous changes over time, continual learning holds great promise, yet its application remains relatively unexplored. This paper aims to bridge this gap by proposing a novel framework, i.e., Continually-Adaptive Representation Learning, designed to adapt representations in response to changing data distributions in evolving healthcare applications. Specifically, the proposed approach develops a continual learning strategy wherein the context information (e.g., interactions) of healthcare entities is exploited to continually identify and retrain the representations of those entities whose context evolved over time. Moreover, different from existing approaches, the proposed approach leverages the valuable patient information present in clinical notes to generate accurate and robust healthcare embeddings. Notably, the proposed continually-adaptive representations are have practical benefits in low-resource clinical settings where it is difficult to training machine learning models from scratch to accommodate the newly available data streams. Experimental evaluations on real-world healthcare datasets demonstrate the effectiveness of our approach in time-sensitive healthcare applications such as Clostridioides difficile (C.diff) Infection (CDI) incidence prediction task and medical intensive care unit transfer prediction task.

Healthcare associated infections (HAIs) impose a substantial burden, both on patients and on the healthcare system. Designing effective strategies by using interventions such as vaccination, isolation, cleaning, mobility modification, etc., to reduce HAI spread is an important computational challenge. Spectral approaches are quite useful for modeling and solving problems of reducing disease spread over contact networks, but they have not been used for disease-spread models and contact networks that are specific for HAIs. Our main contribution in this paper is to close this gap. We make 3 specific contributions. (i) We present the first epidemic threshold results on temporal bipartite networks, i.e., a time-varying sequence of bipartite people-location network, for the Susceptible-Infected-Susceptible (SIS) model. (ii) We leverage our epidemic threshold result to pose the HAI mitigation problem as minimizing the spectral radius of the system matrix, while removing few nodes or edges. We present a scalable combinatorial algorithm that provides approximation guarantees. (iii) Through extensive experiments on actual healthcare contact networks derived from operations data from the University of Iowa Hospitals and Clinics, Carilion Clinic, and several other healthcare facilities, we show that our algorithm consistently outperforms a number of baselines (random, degree, top-k, eigen centrality) both in terms of reducing the spectral radius of the system matrix and in terms of reducing infections.

Recurrent neural networks (RNN) are widely used for handling sequence data. However, their black-box nature makes it difficult for users to interpret the decision-making process. We propose a new method to construct deterministic finite automata to explain RNN. In an automaton, states are abstracted from hidden states produced by the RNN, and the transitions represent input symbols. Thus, users can follow the paths of transitions, called patterns, to understand how a prediction is produced. Existing methods for extracting automata partition the hidden state space at the beginning of the extraction, which often leads to solutions that are either inaccurate or too large in size to comprehend. Unlike previous methods, our approach allows the automata states to be formed adaptively during the extraction. Instead of defining patterns on pre-determined clusters, our proposed model, AdaAX, identifies small sets of hidden states determined by patterns with finer granularity in data. Then these small sets are gradually merged to form states, allowing users to trade fidelity for lower complexity. Experiments show that our automata can achieve higher fidelity while being significantly smaller in size than baseline methods on synthetic and complex real datasets.

This paper presents a high-fidelity agent-based simulation of the spread of methicillin-resistant Staphylococcus aureus (MRSA), a serious hospital acquired infection, within the dialysis unit at the University of Iowa Hospitals and Clinics (UIHC). The simulation is based on ten days of fine-grained healthcare worker (HCW) movement and interaction data collected from a sensor mote instrumentation of the dialysis unit by our research group in the fall of 2013. The simulation layers a detailed model of MRSA pathogen transfer, die-off, shedding, and infection on top of agent interactions obtained from data. The specific question this paper focuses on is whether there are simple, inexpensive architectural or process changes one can make in the dialysis unit to reduce the spread of MRSA? We evaluate two architectural changes of the nurses' station: (i) splitting the central nurses' station into two smaller distinct nurses' stations, and (ii) doubling the surface area of the nursing station. The first architectural change is modeled as a graph partitioning problem on a HCW contact network obtained from our HCW movement data. Somewhat counter-intuitively, our results suggest that the first architectural modification and the resulting reduction in HCW-HCW contacts has little to no effect on the spread of MRSA and may in fact lead to an increase in MRSA infection counts in some cases. In contrast, the second modification leads to a substantial reduction - between 12% and 22% for simulations with different parameters - in the number of patients infected by MRSA. These results suggest that the dynamics of an environmentally mediated infection such as MRSA may be quite different from that of infections whose spread is not substantially affected by the environment (e.g., respiratory infections or influenza).

We present a new distributed algorithm for computing the minimum description length (MDL) in learning Bayesian inference networks from data. Our learning algorithm exploits both properties of the MDL-based score metric and a distributed, asynchronous, adaptive search technique called nagging. Nagging is intrinsically fault-tolerant, has dynamic load balancing features, and scales well. We demonstrate the viability, effectiveness, and scalability of our approach empirically with several experiments using networked machines. More specifically, we show that our distributed algorithm can provide optimal solutions for larger problems as well as good solutions for Bayesian networks of up to 150 variables.