Yugang Jia

Jackson A. Killian, Manish Jain, Yugang Jia et al.

Restless multi-armed bandits (RMABs) are a popular framework for algorithmic decision making in sequential settings with limited resources. RMABs are increasingly being used for sensitive decisions such as in public health, treatment scheduling, anti-poaching, and -- the motivation for this work -- digital health. For such high stakes settings, decisions must both improve outcomes and prevent disparities between groups (e.g., ensure health equity). We study equitable objectives for RMABs (ERMABs) for the first time. We consider two equity-aligned objectives from the fairness literature, minimax reward and max Nash welfare. We develop efficient algorithms for solving each -- a water filling algorithm for the former, and a greedy algorithm with theoretically motivated nuance to balance disparate group sizes for the latter. Finally, we demonstrate across three simulation domains, including a new digital health model, that our approaches can be multiple times more equitable than the current state of the art without drastic sacrifices to utility. Our findings underscore our work's urgency as RMABs permeate into systems that impact human and wildlife outcomes. Code is available at https://github.com/google-research/socialgood/tree/equitable-rmab

Eric Yang, Jonathan Amar, Jong Ha Lee et al. · deepmind

Deploying Large Language Models (LLMs) for healthcare question answering requires robust methods to ensure accuracy and reliability. This work introduces Query-Based Retrieval Augmented Generation (QB-RAG), a framework for enhancing Retrieval-Augmented Generation (RAG) systems in healthcare question-answering by pre-aligning user queries with a database of curated, answerable questions derived from healthcare content. A key component of QB-RAG is an LLM-based filtering mechanism that ensures that only relevant and answerable questions are included in the database, enabling reliable reference query generation at scale. We provide theoretical motivation for QB-RAG, conduct a comparative analysis of existing retrieval enhancement techniques, and introduce a generalizable, comprehensive evaluation framework that assesses both the retrieval effectiveness and the quality of the generated response based on faithfulness, relevance, and adherence to the guideline. Our empirical evaluation on a healthcare data set demonstrates the superior performance of QB-RAG compared to existing retrieval methods, highlighting its practical value in building trustworthy digital health applications for health question-answering.

Pawel Renc, Yugang Jia, Anthony E. Samir et al.

Integrating modern machine learning and clinical decision-making has great promise for mitigating healthcare's increasing cost and complexity. We introduce the Enhanced Transformer for Health Outcome Simulation (ETHOS), a novel application of the transformer deep-learning architecture for analyzing high-dimensional, heterogeneous, and episodic health data. ETHOS is trained using Patient Health Timelines (PHTs)-detailed, tokenized records of health events-to predict future health trajectories, leveraging a zero-shot learning approach. ETHOS represents a significant advancement in foundation model development for healthcare analytics, eliminating the need for labeled data and model fine-tuning. Its ability to simulate various treatment pathways and consider patient-specific factors positions ETHOS as a tool for care optimization and addressing biases in healthcare delivery. Future developments will expand ETHOS' capabilities to incorporate a wider range of data types and data sources. Our work demonstrates a pathway toward accelerated AI development and deployment in healthcare.

Rory Sayres, Kejia Chen, Ayush Jain et al.

Patient-managed Personal Health Records (PHRs) promises to empower patients to better understand their health; but information in the record is complex, potentially hindering insights. In this study, we assess the potential of large language models (LLMs, Gemini 3.0 Flash) to provide helpful answers to user health queries, when provided clinical data from PHRs as context. A total of 2,257 user queries were drawn from 3 different distributions to represent patient questions: shorter web search queries, longer questions derived from templates of chatbot conversations, and questions patients asked to their healthcare team (patient calls). Queries were matched with de-identified PHRs (from a pool of 1,945). Gemini responses were generated (1) without PHR context; (2) with a basic summary of demographics, conditions, and medications; (3) with full, extensive clinical notes. For evaluation, we leveraged an existing rating framework (SHARP), and developed a new framework for specific error modes when interpreting PHRs. Evaluation was performed using autoraters for the full set, and with clinician ratings for a subset (n=95), with both sets of raters knowing the full PHR context. We see significant improvements in the helpfulness of answers to all question types with PHR data (p < 0.001, paired t-test). We also observe potential gains in safety, accuracy, relevance and personalization of answers. Our PHR evaluation framework further identifies gaps in LLM understanding of particular aspects of complex PHRs, such as temporal disorientation, and rare but meaningful confabulations. These results suggest potential for PHR data to help people with a wide range of user needs; and provide a framework for monitoring for gaps in LLM answers based on PHR context. This study motivates further work to assess and realize potential benefits to users from understanding their health records.

Xiao Xiang, David Restrepo, Hyewon Jeong et al.

Learning from electronic health records (EHRs) time series is challenging due to irregular sam- pling, heterogeneous missingness, and the resulting sparsity of observations. Prior self-supervised meth- ods either impute before learning, represent missingness through a dedicated input signal, or optimize solely for imputation, reducing their capacity to efficiently learn representations that support clinical downstream tasks. We propose the Augmented-Intrinsic Dual-Masked Autoencoder (AID-MAE), which learns directly from incomplete time series by applying an intrinsic missing mask to represent naturally missing values and an augmented mask that hides a subset of observed values for reconstruction during training. AID-MAE processes only the unmasked subset of tokens and consistently outperforms strong baselines, including XGBoost and DuETT, across multiple clinical tasks on two datasets. In addition, the learned embeddings naturally stratify patient cohorts in the representation space.

Rajna Fani, Rafi Al Attrach, David Restrepo et al.

Masked autoencoders (MAEs) are increasingly applied to electronic health records (EHR) for learning general-purpose representations that support diverse clinical tasks. However, existing approaches typically rely on uniform random masking, implicitly assuming all features are equally predictable. In reality, laboratory tests exhibit substantial heterogeneity in volatility: some biomarkers (e.g., sodium) remain stable, while others (e.g., lactate) fluctuate considerably and are more difficult to model. Clinically, volatile biomarkers often signal acute pathophysiology and require more sophisticated modeling to capture their complex temporal patterns. We propose a volatility-aware pretraining strategy, Coefficient of Variation Masking (CV-Masking), that adaptively adjusts masking probabilities according to the intrinsic variability of each feature. Combined with a value-only masking objective aligned with clinical workflows, CV-Masking yields systematic improvements over random and variance-based strategies. Experiments on a large panel of laboratory tests show that CV-Masking enhances reconstruction, improves downstream predictive performance, and accelerates convergence, producing more robust and clinically meaningful EHR representations.

Gyubok Lee, Elea Bach, Eric Yang et al.

The recent shift toward the Health Level Seven Fast Healthcare Interoperability Resources (HL7 FHIR) standard opens a new frontier for clinical AI, demanding LLM agents to navigate complex, resource-based data models instead of conventional structured health data. However, existing benchmarks have lagged behind this transition, lacking the realism needed to evaluate recent LLMs on interoperable clinical data. To bridge this gap, we introduce FHIR-AgentBench, a benchmark that grounds 2,931 real-world clinical questions in the HL7 FHIR standard. Using this benchmark, we systematically evaluate agentic frameworks, comparing different data retrieval strategies (direct FHIR API calls vs. specialized tools), interaction patterns (single-turn vs. multi-turn), and reasoning strategies (natural language vs. code generation). Our experiments highlight the practical challenges of retrieving data from intricate FHIR resources and the difficulty of reasoning over them, both of which critically affect question answering performance. We publicly release the FHIR-AgentBench dataset and evaluation suite (https://github.com/glee4810/FHIR-AgentBench) to promote reproducible research and the development of robust, reliable LLM agents for clinical applications.

Rafi Al Attrach, Rajna Fani, David Restrepo et al.

Tokenization strategies shape how models process electronic health records, yet fair comparisons of their effectiveness remain limited. We present a systematic evaluation of tokenization approaches for clinical time series modeling using transformer-based architectures, revealing task-dependent and sometimes counterintuitive findings about temporal and value feature importance. Through controlled ablations across four clinical prediction tasks on MIMIC-IV, we demonstrate that explicit time encodings provide no consistent statistically significant benefit for the evaluated downstream tasks. Value features show task-dependent importance, affecting mortality prediction but not readmission, suggesting code sequences alone can carry sufficient predictive signal. We further show that frozen pretrained code encoders dramatically outperform their trainable counterparts while requiring dramatically fewer parameters. Larger clinical encoders provide consistent improvements across tasks, benefiting from frozen embeddings that eliminate computational overhead. Our controlled evaluation enables fairer tokenization comparisons and demonstrates that simpler, parameter-efficient approaches can, in many cases, achieve strong performance, though the optimal tokenization strategy remains task-dependent.

Pawel Renc, Michal K. Grzeszczyk, Nassim Oufattole et al.

Hospitals struggle to predict critical outcomes. Traditional early warning systems, like NEWS and MEWS, rely on static variables and fixed thresholds, limiting their adaptability, accuracy, and personalization. We previously developed the Enhanced Transformer for Health Outcome Simulation (ETHOS), an AI model that tokenizes patient health timelines (PHTs) from EHRs and uses transformer-based architectures to predict future PHTs. ETHOS is a versatile framework for developing a wide range of applications. In this work, we develop the Adaptive Risk Estimation System (ARES) that leverages ETHOS to compute dynamic, personalized risk probabilities for clinician-defined critical events. ARES also features a personalized explainability module that highlights key clinical factors influencing risk estimates. We evaluated ARES using the MIMIC-IV v2.2 dataset together with its Emergency Department (ED) extension and benchmarked performance against both classical early warning systems and contemporary machine learning models. The entire dataset was tokenized resulting in 285,622 PHTs, comprising over 360 million tokens. ETHOS outperformed benchmark models in predicting hospital admissions, ICU admissions, and prolonged stays, achieving superior AUC scores. Its risk estimates were robust across demographic subgroups, with calibration curves confirming model reliability. The explainability module provided valuable insights into patient-specific risk factors. ARES, powered by ETHOS, advances predictive healthcare AI by delivering dynamic, real-time, personalized risk estimation with patient-specific explainability. Although our results are promising, the clinical impact remains uncertain. Demonstrating ARES's true utility in real-world settings will be the focus of our future work. We release the source code to facilitate future research.

Taedong Yun, Eric Yang, Mustafa Safdari et al. · berkeley

We present an end-to-end framework for generating synthetic users for evaluating interactive agents designed to encourage positive behavior changes, such as in health and lifestyle coaching. The synthetic users are grounded in health and lifestyle conditions, specifically sleep and diabetes management in this study, to ensure realistic interactions with the health coaching agent. Synthetic users are created in two stages: first, structured data are generated grounded in real-world health and lifestyle factors in addition to basic demographics and behavioral attributes; second, full profiles of the synthetic users are developed conditioned on the structured data. Interactions between synthetic users and the coaching agent are simulated using generative agent-based models such as Concordia, or directly by prompting a language model. Using two independently-developed agents for sleep and diabetes coaching as case studies, the validity of this framework is demonstrated by analyzing the coaching agent's understanding of the synthetic users' needs and challenges. Finally, through multiple blinded evaluations of user-coach interactions by human experts, we demonstrate that our synthetic users with health and behavioral attributes more accurately portray real human users with the same attributes, compared to generic synthetic users not grounded in such attributes. The proposed framework lays the foundation for efficient development of conversational agents through extensive, realistic, and grounded simulated interactions.

Eric Yang, Tomas Garcia, Hannah Williams et al. · deepmind

Effective management of cardiometabolic conditions requires sustained positive nutrition habits, often hindered by complex and individualized barriers. Direct human management is simply not scalable, while previous attempts aimed at automating nutrition coaching lack the personalization needed to address these diverse challenges. This paper introduces a novel LLM-powered agentic workflow designed to provide personalized nutrition coaching by directly targeting and mitigating patient-specific barriers. Grounded in behavioral science principles, the workflow leverages a comprehensive mapping of nutrition-related barriers to corresponding evidence-based strategies. A specialized LLM agent intentionally probes for and identifies the root cause of a patient's dietary struggles. Subsequently, a separate LLM agent delivers tailored tactics designed to overcome those specific barriers with patient context. We designed and validated our approach through a user study with individuals with cardiometabolic conditions, demonstrating the system's ability to accurately identify barriers and provide personalized guidance. Furthermore, we conducted a large-scale simulation study, grounding on real patient vignettes and expert-validated metrics, to evaluate the system's performance across a wide range of scenarios. Our findings demonstrate the potential of this LLM-powered agentic workflow to improve nutrition coaching by providing personalized, scalable, and behaviorally-informed interventions.

David Restrepo, Chenwei Wu, Yueran Jia et al.

Accurate imputation of missing laboratory values in electronic health records (EHRs) is critical to enable robust clinical predictions and reduce biases in AI systems in healthcare. Existing methods, such as XGBoost, softimpute, GAIN, Expectation Maximization (EM), and MICE, struggle to model the complex temporal and contextual dependencies in EHR data, particularly in underrepresented groups. In this work, we propose Lab-MAE, a novel transformer-based masked autoencoder framework that leverages self-supervised learning for the imputation of continuous sequential lab values. Lab-MAE introduces a structured encoding scheme that jointly models laboratory test values and their corresponding timestamps, enabling explicit capturing temporal dependencies. Empirical evaluation on the MIMIC-IV dataset demonstrates that Lab-MAE significantly outperforms state-of-the-art baselines such as XGBoost, softimpute, GAIN, EM, and MICE across multiple metrics, including root mean square error (RMSE), R-squared (R2), and Wasserstein distance (WD). Notably, Lab-MAE achieves equitable performance across demographic groups of patients, advancing fairness in clinical predictions. We further investigate the role of follow-up laboratory values as potential shortcut features, revealing Lab-MAE's robustness in scenarios where such data is unavailable. The findings suggest that our transformer-based architecture, adapted to the characteristics of EHR data, offers a foundation model for more accurate and fair clinical imputation. In addition, we measure and compare the carbon footprint of Lab-MAE with the a XGBoost model, highlighting its environmental requirements.

Eric Yang, Jong Ha Lee, Jonathan Amar et al.

Large Language Models (LLMs) trained for average correctness often exhibit mode collapse, producing narrow decision behaviors on tasks where multiple responses may be reasonable. This limitation is particularly problematic in ordinal decision settings such as clinical triage, where standard alignment removes the ability to trade off specificity and sensitivity (the ROC operating point) based on contextual constraints. We propose STEER (Steerable Tuning via Evolutionary Ensemble Refinement), a training-free framework that reintroduces this tunable control. STEER constructs a population of natural-language personas through an offline, constrained quality-diversity search that promotes behavioral coverage while enforcing minimum safety, reasoning, and stability thresholds. At inference time, STEER exposes a single, interpretable control parameter that maps a user-specified risk percentile to a selected persona, yielding a monotonic adjustment of decision conservativeness. On two clinical triage benchmarks, STEER achieves broader behavioral coverage compared to temperature-based sampling and static persona ensembles. Compared to a representative post-training method, STEER maintains substantially higher accuracy on unambiguous urgent cases while providing comparable control over ambiguous decisions. These results demonstrate STEER as a safety-preserving paradigm for risk control, capable of steering behavior without compromising domain competence.

Jonathan Amar, Edward Liu, Alessandra Breschi et al.

This paper introduces an innovative Electronic Health Record (EHR) foundation model that integrates Polygenic Risk Scores (PRS) as a foundational data modality, moving beyond traditional EHR-only approaches to build more holistic health profiles. Leveraging the extensive and diverse data from the All of Us (AoU) Research Program, this multimodal framework aims to learn complex relationships between clinical data and genetic predispositions. The methodology extends advancements in generative AI to the EHR foundation model space, enhancing predictive capabilities and interpretability. Evaluation on AoU data demonstrates the model's predictive value for the onset of various conditions, particularly Type 2 Diabetes (T2D), and illustrates the interplay between PRS and EHR data. The work also explores transfer learning for custom classification tasks, showcasing the architecture's versatility and efficiency. This approach is pivotal for unlocking new insights into disease prediction, proactive health management, risk stratification, and personalized treatment strategies, laying the groundwork for more personalized, equitable, and actionable real-world evidence generation in healthcare.

Sina Rashidian, Nan Li, Jonathan Amar et al.

Background: We present a Patient Simulator that leverages real world patient encounters which cover a broad range of conditions and symptoms to provide synthetic test subjects for development and testing of healthcare agentic models. The simulator provides a realistic approach to patient presentation and multi-turn conversation with a symptom-checking agent. Objectives: (1) To construct and instantiate a Patient Simulator to train and test an AI health agent, based on patient vignettes derived from real EHR data. (2) To test the validity and alignment of the simulated encounters provided by the Patient Simulator to expert human clinical providers. (3) To illustrate the evaluation framework of such an LLM system on the generated realistic, data-driven simulations -- yielding a preliminary assessment of our proposed system. Methods: We first constructed realistic clinical scenarios by deriving patient vignettes from real-world EHR encounters. These vignettes cover a variety of presenting symptoms and underlying conditions. We then evaluate the performance of the Patient Simulator as a simulacrum of a real patient encounter across over 500 different patient vignettes. We leveraged a separate AI agent to provide multi-turn questions to obtain a history of present illness. The resulting multiturn conversations were evaluated by two expert clinicians. Results: Clinicians scored the Patient Simulator as consistent with the patient vignettes in those same 97.7% of cases. The extracted case summary based on the conversation history was 99% relevant. Conclusions: We developed a methodology to incorporate vignettes derived from real healthcare patient data to build a simulation of patient responses to symptom checking agents. The performance and alignment of this Patient Simulator could be used to train and test a multi-turn conversational AI agent at scale.

Bhawesh Kumar, Jonathan Amar, Eric Yang et al.

Large Language Models (LLMs) have demonstrated their efficacy across a broad spectrum of tasks in healthcare applications. However, often LLMs need to be fine-tuned on task-specific expert annotated data to achieve optimal performance, which can be expensive and time consuming. In this study, we fine-tune PaLM-2 with parameter efficient fine-tuning (PEFT) using noisy labels obtained from gemini-pro 1.0 for the detection of Schedule-of-Event (SoE) tables, which specify care plan in clinical trial protocols. We introduce a filtering mechanism to select high-confidence labels for this table classification task, thereby reducing the noise in the auto-generated labels. We show that fine-tuned PaLM-2 with those labels achieves performance that exceeds the gemini-pro 1.0 and other LLMs. Furthermore, its performance is close to a PaLM-2 fine-tuned on labels obtained from non-expert annotators. Our results show that leveraging LLM-generated labels through powerful models like gemini-pro can potentially serve as a viable strategy for improving LLM performance through fine-tuning in specialized tasks, particularly in domains where expert annotations are scarce, expensive, or time-consuming to obtain.

Riyi Qiu, Yugang Jia, Mirsad Hadzikadic et al.

Deep learning models have exhibited superior performance in predictive tasks with the explosively increasing Electronic Health Records (EHR). However, due to the lack of transparency, behaviors of deep learning models are difficult to interpret. Without trustworthiness, deep learning models will not be able to assist in the real-world decision-making process of healthcare issues. We propose a deep learning model based on Bayesian Neural Networks (BNN) to predict uncertainty induced by data noise. The uncertainty is introduced to provide model predictions with an extra level of confidence. Our experiments verify that instances with high uncertainty are harmful to model performance. Moreover, by investigating the distributions of model prediction and uncertainty, we show that it is possible to identify a group of patients for timely intervention, such that decreasing data noise will benefit more on the prediction accuracy for these patients.