Noémie Elhadad

Griffin Adams, Alexander Fabbri, Faisal Ladhak et al.

Selecting the ``right'' amount of information to include in a summary is a difficult task. A good summary should be detailed and entity-centric without being overly dense and hard to follow. To better understand this tradeoff, we solicit increasingly dense GPT-4 summaries with what we refer to as a ``Chain of Density'' (CoD) prompt. Specifically, GPT-4 generates an initial entity-sparse summary before iteratively incorporating missing salient entities without increasing the length. Summaries generated by CoD are more abstractive, exhibit more fusion, and have less of a lead bias than GPT-4 summaries generated by a vanilla prompt. We conduct a human preference study on 100 CNN DailyMail articles and find that that humans prefer GPT-4 summaries that are more dense than those generated by a vanilla prompt and almost as dense as human written summaries. Qualitative analysis supports the notion that there exists a tradeoff between informativeness and readability. 500 annotated CoD summaries, as well as an extra 5,000 unannotated summaries, are freely available on HuggingFace (https://huggingface.co/datasets/griffin/chain_of_density).

Griffin Adams, Han-Chin Shing, Qing Sun et al.

In real-world scenarios with naturally occurring datasets, reference summaries are noisy and may contain information that cannot be inferred from the source text. On large news corpora, removing low quality samples has been shown to reduce model hallucinations. Yet, for smaller, and/or noisier corpora, filtering is detrimental to performance. To improve reference quality while retaining all data, we propose a new approach: to selectively re-write unsupported reference sentences to better reflect source data. We automatically generate a synthetic dataset of positive and negative revisions by corrupting supported sentences and learn to revise reference sentences with contrastive learning. The intensity of revisions is treated as a controllable attribute so that, at inference, diverse candidates can be over-generated-then-rescored to balance faithfulness and abstraction. To test our methods, we extract noisy references from publicly available MIMIC-III discharge summaries for the task of hospital-course summarization, and vary the data on which models are trained. According to metrics and human evaluation, models trained on revised clinical references are much more faithful, informative, and fluent than models trained on original or filtered data.

Griffin Adams, Jason Zucker, Noémie Elhadad

Long-form clinical summarization of hospital admissions has real-world significance because of its potential to help both clinicians and patients. The faithfulness of summaries is critical to their safe usage in clinical settings. To better understand the limitations of abstractive systems, as well as the suitability of existing evaluation metrics, we benchmark faithfulness metrics against fine-grained human annotations for model-generated summaries of a patient's Brief Hospital Course. We create a corpus of patient hospital admissions and summaries for a cohort of HIV patients, each with complex medical histories. Annotators are presented with summaries and source notes, and asked to categorize manually highlighted summary elements (clinical entities like conditions and medications as well as actions like "following up") into one of three categories: ``Incorrect,'' ``Missing,'' and ``Not in Notes.'' We meta-evaluate a broad set of proposed faithfulness metrics and, across metrics, explore the importance of domain adaptation (e.g. the impact of in-domain pre-training and metric fine-tuning), the use of source-summary alignments, and the effects of distilling a single metric from an ensemble of pre-existing metrics. Off-the-shelf metrics with no exposure to clinical text correlate well yet overly rely on summary extractiveness. As a practical guide to long-form clinical narrative summarization, we find that most metrics correlate best to human judgments when provided with one summary sentence at a time and a minimal set of relevant source context.

Mert Ketenci, Shreyas Bhave, Noémie Elhadad et al.

Survival analysis is a widely-used technique for analyzing time-to-event data in the presence of censoring. In recent years, numerous survival analysis methods have emerged which scale to large datasets and relax traditional assumptions such as proportional hazards. These models, while being performant, are very sensitive to model hyperparameters including: (1) number of bins and bin size for discrete models and (2) number of cluster assignments for mixture-based models. Each of these choices requires extensive tuning by practitioners to achieve optimal performance. In addition, we demonstrate in empirical studies that: (1) optimal bin size may drastically differ based on the metric of interest (e.g., concordance vs brier score), and (2) mixture models may suffer from mode collapse and numerical instability. We propose a survival analysis approach which eliminates the need to tune hyperparameters such as mixture assignments and bin sizes, reducing the burden on practitioners. We show that the proposed approach matches or outperforms baselines on several real-world datasets.

Mert Ketenci, Adler Perotte, Noémie Elhadad et al.

We introduce a novel stochastic variational inference method for Gaussian process ($\mathcal{GP}$) regression, by deriving a posterior over a learnable set of coresets: i.e., over pseudo-input/output, weighted pairs. Unlike former free-form variational families for stochastic inference, our coreset-based variational $\mathcal{GP}$ (CVGP) is defined in terms of the $\mathcal{GP}$ prior and the (weighted) data likelihood. This formulation naturally incorporates inductive biases of the prior, and ensures its kernel and likelihood dependencies are shared with the posterior. We derive a variational lower-bound on the log-marginal likelihood by marginalizing over the latent $\mathcal{GP}$ coreset variables, and show that CVGP's lower-bound is amenable to stochastic optimization. CVGP reduces the dimensionality of the variational parameter search space to linear $\mathcal{O}(M)$ complexity, while ensuring numerical stability at $\mathcal{O}(M^3)$ time complexity and $\mathcal{O}(M^2)$ space complexity.

Tony Y. Sun, Shreyas Bhave, Jaan Altosaar et al.

Disease identification is a core, routine activity in observational health research. Cohorts impact downstream analyses, such as how a condition is characterized, how patient risk is defined, and what treatments are studied. It is thus critical to ensure that selected cohorts are representative of all patients, independently of their demographics or social determinants of health. While there are multiple potential sources of bias when constructing phenotype definitions which may affect their fairness, it is not standard in the field of phenotyping to consider the impact of different definitions across subgroups of patients. In this paper, we propose a set of best practices to assess the fairness of phenotype definitions. We leverage established fairness metrics commonly used in predictive models and relate them to commonly used epidemiological cohort description metrics. We describe an empirical study for Crohn's disease and diabetes type 2, each with multiple phenotype definitions taken from the literature across two sets of patient subgroups (gender and race). We show that the different phenotype definitions exhibit widely varying and disparate performance according to the different fairness metrics and subgroups. We hope that the proposed best practices can help in constructing fair and inclusive phenotype definitions.

Georgianna "Blue" Lin, Rencong Jiang, Noémie Elhadad et al.

Although artificial intelligence (AI) agents are increasingly proposed to support potentially longitudinal health tasks, such as symptom management, behavior change, and patient support, most current implementations fall short of facilitating user intent and fostering accountability. This contrasts with prior work on supporting longitudinal needs, where follow-up, coherent reasoning, and sustained alignment with individuals' goals are critical for both effectiveness and safety. In this paper, we draw on established clinical and personal health informatics frameworks to define what it would mean to orchestrate longitudinal health interactions with AI agents. We propose a multi-layer framework and corresponding agent architecture that operationalizes adaptation, coherence, continuity, and agency across repeated interactions. Through representative use cases, we demonstrate how longitudinal agents can maintain meaningful engagement, adapt to evolving goals, and support safe, personalized decision-making over time. Our findings underscore both the promise and the complexity of designing systems capable of supporting health trajectories beyond isolated interactions, and we offer guidance for future research and development in multi-session, user-centered health AI.

Griffin Adams, Jason Zucker, Noémie Elhadad

Clinician must write a lengthy summary each time a patient is discharged from the hospital. This task is time-consuming due to the sheer number of unique clinical concepts covered in the admission. Identifying and covering salient entities is vital for the summary to be clinically useful. We fine-tune open-source LLMs (Mistral-7B-Instruct and Zephyr-7B-beta) on the task and find that they generate incomplete and unfaithful summaries. To increase entity coverage, we train a smaller, encoder-only model to predict salient entities, which are treated as content-plans to guide the LLM. To encourage the LLM to focus on specific mentions in the source notes, we propose SPEER: Sentence-level Planning via Embedded Entity Retrieval. Specifically, we mark each salient entity span with special "{ }" boundary tags and instruct the LLM to retrieve marked spans before generating each sentence. Sentence-level planning acts as a form of state tracking in that the model is explicitly recording the entities it uses. We fine-tune Mistral and Zephyr variants on a large-scale, diverse dataset of ~167k in-patient hospital admissions and evaluate on 3 datasets. SPEER shows gains in both coverage and faithfulness metrics over non-guided and guided baselines.

Robin Linzmayer, Georgianna Lin, Di Coneybeare et al.

We introduce AcuityBench, a benchmark for evaluating whether language models identify the appropriate urgency of care from user medical presentations. Existing health benchmarks emphasize medical question answering, broad health interactions, or narrow workflow-specific triage tasks, but they do not offer a unified evaluation of acuity identification across these settings. AcuityBench addresses this gap by harmonizing five public datasets spanning user conversations, online forum posts, clinical vignettes, and patient portal messages under a shared four-level acuity framework ranging from home monitoring to immediate emergency care. The benchmark contains 914 cases, including 697 consensus cases for standard accuracy evaluation and 217 physician-confirmed ambiguous cases for uncertainty-aware evaluation. It supports two complementary task formats: explicit four-way classification in a QA setting, and free-form conversational responses evaluated with a rubric-based judge anchored to the same framework. Across 12 frontier proprietary and open-weight models, we find substantial variation in clear-case acuity accuracy and error direction. Comparing task formats reveals a systematic tradeoff: conversational responses reduce over-triage but increase under-triage relative to QA, especially in higher-acuity cases. In ambiguous cases, no model closely matches the distribution of physician judgments, and model predictions are more concentrated than expert clinical uncertainty. We also compare expert and model adjudication on a subset of maximally ambiguous cases, using those cases to examine the role of clinical uncertainty in label disagreement. Together, these results position acuity identification as a distinct safety-critical capability and show that AcuityBench enables systematic comparison and stress-testing of how well models guide users to the right level of care in real-world health use.

Griffin Adams, Alexander R. Fabbri, Faisal Ladhak et al.

Two-step approaches, in which summary candidates are generated-then-reranked to return a single summary, can improve ROUGE scores over the standard single-step approach. Yet, standard decoding methods (i.e., beam search, nucleus sampling, and diverse beam search) produce candidates with redundant, and often low quality, content. In this paper, we design a novel method to generate candidates for re-ranking that addresses these issues. We ground each candidate abstract on its own unique content plan and generate distinct plan-guided abstracts using a model's top beam. More concretely, a standard language model (a BART LM) auto-regressively generates elemental discourse unit (EDU) content plans with an extractive copy mechanism. The top K beams from the content plan generator are then used to guide a separate LM, which produces a single abstractive candidate for each distinct plan. We apply an existing re-ranker (BRIO) to abstractive candidates generated from our method, as well as baseline decoding methods. We show large relevance improvements over previously published methods on widely used single document news article corpora, with ROUGE-2 F1 gains of 0.88, 2.01, and 0.38 on CNN / Dailymail, NYT, and Xsum, respectively. A human evaluation on CNN / DM validates these results. Similarly, on 1k samples from CNN / DM, we show that prompting GPT-3 to follow EDU plans outperforms sampling-based methods by 1.05 ROUGE-2 F1 points. Code to generate and realize plans is available at https://github.com/griff4692/edu-sum.

Griffin Adams, Bichlien H Nguyen, Jake Smith et al.

Summarization models often generate text that is poorly calibrated to quality metrics because they are trained to maximize the likelihood of a single reference (MLE). To address this, recent work has added a calibration step, which exposes a model to its own ranked outputs to improve relevance or, in a separate line of work, contrasts positive and negative sets to improve faithfulness. While effective, much of this work has focused on how to generate and optimize these sets. Less is known about why one setup is more effective than another. In this work, we uncover the underlying characteristics of effective sets. For each training instance, we form a large, diverse pool of candidates and systematically vary the subsets used for calibration fine-tuning. Each selection strategy targets distinct aspects of the sets, such as lexical diversity or the size of the gap between positive and negatives. On three diverse scientific long-form summarization datasets (spanning biomedical, clinical, and chemical domains), we find, among others, that faithfulness calibration is optimal when the negative sets are extractive and more likely to be generated, whereas for relevance calibration, the metric margin between candidates should be maximized and surprise--the disagreement between model and metric defined candidate rankings--minimized. Code to create, select, and optimize calibration sets is available at https://github.com/griff4692/calibrating-summaries

Chao Pang, Xinzhuo Jiang, Nishanth Parameshwar Pavinkurve et al.

Synthetic Electronic Health Records (EHR) have emerged as a pivotal tool in advancing healthcare applications and machine learning models, particularly for researchers without direct access to healthcare data. Although existing methods, like rule-based approaches and generative adversarial networks (GANs), generate synthetic data that resembles real-world EHR data, these methods often use a tabular format, disregarding temporal dependencies in patient histories and limiting data replication. Recently, there has been a growing interest in leveraging Generative Pre-trained Transformers (GPT) for EHR data. This enables applications like disease progression analysis, population estimation, counterfactual reasoning, and synthetic data generation. In this work, we focus on synthetic data generation and demonstrate the capability of training a GPT model using a particular patient representation derived from CEHR-BERT, enabling us to generate patient sequences that can be seamlessly converted to the Observational Medical Outcomes Partnership (OMOP) data format.

Aparajita Kashyap, Sara Matijevic, Noémie Elhadad et al.

When training machine learning (ML) models for potential deployment in a healthcare setting, it is essential to ensure that they do not replicate or exacerbate existing healthcare biases. Although many definitions of fairness exist, we focus on path-specific causal fairness, which allows us to better consider the social and medical contexts in which biases occur (e.g., direct discrimination by a clinician or model versus bias due to differential access to the healthcare system) and to characterize how these biases may appear in learned models. In this work, we map the structural fairness model to the observational healthcare setting and create a generalizable pipeline for training causally fair models. The pipeline explicitly considers specific healthcare context and disparities to define a target "fair" model. Our work fills two major gaps: first, we expand on characterizations of the "fairness-accuracy" tradeoff by detangling direct and indirect sources of bias and jointly presenting these fairness considerations alongside considerations of accuracy in the context of broadly known biases. Second, we demonstrate how a foundation model trained without fairness constraints on observational health data can be leveraged to generate causally fair downstream predictions in tasks with known social and medical disparities. This work presents a model-agnostic pipeline for training causally fair machine learning models that address both direct and indirect forms of healthcare bias.

Xuhai Xu, Haoyu Hu, Haoran Zhang et al.

Artificial intelligence (AI) is increasingly permeating healthcare, from physician assistants to consumer applications. Since AI algorithm's opacity challenges human interaction, explainable AI (XAI) addresses this by providing AI decision-making insight, but evidence suggests XAI can paradoxically induce over-reliance or bias. We present results from two large-scale experiments (623 lay people; 153 primary care physicians, PCPs) combining a fairness-based diagnosis AI model and different XAI explanations to examine how XAI assistance, particularly multimodal large language models (LLMs), influences diagnostic performance. AI assistance balanced across skin tones improved accuracy and reduced diagnostic disparities. However, LLM explanations yielded divergent effects: lay users showed higher automation bias - accuracy boosted when AI was correct, reduced when AI erred - while experienced PCPs remained resilient, benefiting irrespective of AI accuracy. Presenting AI suggestions first also led to worse outcomes when the AI was incorrect for both groups. These findings highlight XAI's varying impact based on expertise and timing, underscoring LLMs as a "double-edged sword" in medical AI and informing future human-AI collaborative system design.

Mert Ketenci, Vincent Jeanselme, Harry Reyes Nieva et al.

Survival analysis is a fundamental tool for modeling time-to-event outcomes in healthcare. Recent advances have introduced flexible neural network approaches for improved predictive performance. However, most of these models do not provide interpretable insights into the association between exposures and the modeled outcomes, a critical requirement for decision-making in clinical practice. To address this limitation, we propose Additive Deep Hazard Analysis Mixtures (ADHAM), an interpretable additive survival model. ADHAM assumes a conditional latent structure that defines subgroups, each characterized by a combination of covariate-specific hazard functions. To select the number of subgroups, we introduce a post-training refinement that reduces the number of equivalent latent subgroups by merging similar groups. We perform comprehensive studies to demonstrate ADHAM's interpretability at the population, subgroup, and individual levels. Extensive experiments on real-world datasets show that ADHAM provides novel insights into the association between exposures and outcomes. Further, ADHAM remains on par with existing state-of-the-art survival baselines in terms of predictive performance, offering a scalable and interpretable approach to time-to-event prediction in healthcare.

Chao Pang, Jiheum Park, Xinzhuo Jiang et al.

Electronic Health Records (EHRs) provide a rich, longitudinal view of patient health and hold significant potential for advancing clinical decision support, risk prediction, and data-driven healthcare research. However, most artificial intelligence (AI) models for EHRs are designed for narrow, single-purpose tasks, limiting their generalizability and utility in real-world settings. Here, we present CEHR-XGPT, a general-purpose foundation model for EHR data that unifies three essential capabilities - feature representation, zero-shot prediction, and synthetic data generation - within a single architecture. To support temporal reasoning over clinical sequences, CEHR-XGPT incorporates a novel time-token-based learning framework that explicitly encodes patients' dynamic timelines into the model structure. CEHR-XGPT demonstrates strong performance across all three tasks and generalizes effectively to external datasets through vocabulary expansion and fine-tuning. Its versatility enables rapid model development, cohort discovery, and patient outcome forecasting without the need for task-specific retraining.

Jennifer L. Chen, Faisal Ladhak, Daniel Li et al.

Given the black-box nature and complexity of large transformer language models (LM), concerns about generalizability and robustness present ethical implications for domains such as hate speech (HS) detection. Using the content rich Social Bias Frames dataset, containing human-annotated stereotypes, intent, and targeted groups, we develop a three stage analysis to evaluate if LMs faithfully assess hate speech. First, we observe the need for modeling contextually grounded stereotype intents to capture implicit semantic meaning. Next, we design a new task, Stereotype Intent Entailment (SIE), which encourages a model to contextually understand stereotype presence. Finally, through ablation tests and user studies, we find a SIE objective improves content understanding, but challenges remain in modeling implicit intent.

Mert Ketenci, Iñigo Urteaga, Victor Alfonso Rodriguez et al.

We propose probabilistic Shapley inference (PSI), a novel probabilistic framework to model and infer sufficient statistics of feature attributions in flexible predictive models, via latent random variables whose mean recovers Shapley values. PSI enables efficient, scalable inference over input-to-output attributions, and their uncertainty, via a variational objective that jointly trains a predictive (regression or classification) model and its attribution distributions. To address the challenge of marginalizing over variable-length input feature subsets in Shapley value calculation, we introduce a masking-based neural network architecture, with a modular training and inference procedure. We evaluate PSI on synthetic and real-world datasets, showing that it achieves competitive predictive performance compared to strong baselines, while learning feature attribution distributions -- centered at Shapley values -- that reveal meaningful attribution uncertainty across data modalities.

Griffin Adams, Emily Alsentzer, Mert Ketenci et al.

Summarization of clinical narratives is a long-standing research problem. Here, we introduce the task of hospital-course summarization. Given the documentation authored throughout a patient's hospitalization, generate a paragraph that tells the story of the patient admission. We construct an English, text-to-text dataset of 109,000 hospitalizations (2M source notes) and their corresponding summary proxy: the clinician-authored "Brief Hospital Course" paragraph written as part of a discharge note. Exploratory analyses reveal that the BHC paragraphs are highly abstractive with some long extracted fragments; are concise yet comprehensive; differ in style and content organization from the source notes; exhibit minimal lexical cohesion; and represent silver-standard references. Our analysis identifies multiple implications for modeling this complex, multi-document summarization task.

Kathy Li, Iñigo Urteaga, Amanda Shea et al.

Mobile health (mHealth) apps such as menstrual trackers provide a rich source of self-tracked health observations that can be leveraged for health-relevant research. However, such data streams have questionable reliability since they hinge on user adherence to the app. Therefore, it is crucial for researchers to separate true behavior from self-tracking artifacts. By taking a machine learning approach to modeling self-tracked cycle lengths, we can both make more informed predictions and learn the underlying structure of the observed data. In this work, we propose and evaluate a hierarchical, generative model for predicting next cycle length based on previously-tracked cycle lengths that accounts explicitly for the possibility of users skipping tracking their period. Our model offers several advantages: 1) accounting explicitly for self-tracking artifacts yields better prediction accuracy as likelihood of skipping increases; 2) because it is a generative model, predictions can be updated online as a given cycle evolves, and we can gain interpretable insight into how these predictions change over time; and 3) its hierarchical nature enables modeling of an individual's cycle length history while incorporating population-level information. Our experiments using mHealth cycle length data encompassing over 186,000 menstruators with over 2 million natural menstrual cycles show that our method yields state-of-the-art performance against neural network-based and summary statistic-based baselines, while providing insights on disentangling menstrual patterns from self-tracking artifacts. This work can benefit users, mHealth app developers, and researchers in better understanding cycle patterns and user adherence.

Tony Y. Sun, Oliver J. Bear Don't Walk, Jennifer L. Chen et al.

Sex and gender-based healthcare disparities contribute to differences in health outcomes. We focus on time to diagnosis (TTD) by conducting two large-scale, complementary analyses among men and women across 29 phenotypes and 195K patients. We first find that women are consistently more likely to experience a longer TTD than men, even when presenting with the same conditions. We further explore how TTD disparities affect diagnostic performance between genders, both across and persistent to time, by evaluating gender-agnostic disease classifiers across increasing diagnostic information. In both fairness analyses, the diagnostic process favors men over women, contradicting the previous observation that women may demonstrate relevant symptoms earlier than men. These analyses suggest that TTD is an important yet complex aspect when studying gender disparities, and warrants further investigation.

Griffin Adams, Mert Ketenci, Shreyas Bhave et al.

We introduce Latent Meaning Cells, a deep latent variable model which learns contextualized representations of words by combining local lexical context and metadata. Metadata can refer to granular context, such as section type, or to more global context, such as unique document ids. Reliance on metadata for contextualized representation learning is apropos in the clinical domain where text is semi-structured and expresses high variation in topics. We evaluate the LMC model on the task of zero-shot clinical acronym expansion across three datasets. The LMC significantly outperforms a diverse set of baselines at a fraction of the pre-training cost and learns clinically coherent representations. We demonstrate that not only is metadata itself very helpful for the task, but that the LMC inference algorithm provides an additional large benefit.

Gal Levy-Fix, Jason Zucker, Konstantin Stojanovic et al.

Identifying a patient's key problems over time is a common task for providers at the point care, yet a complex and time-consuming activity given current electric health records. To enable a problem-oriented summarizer to identify a patient's comprehensive list of problems and their salience, we propose an unsupervised phenotyping approach that jointly learns a large number of phenotypes/problems across structured and unstructured data. To identify the appropriate granularity of the learned phenotypes, the model is trained on a target patient population of the same clinic. To enable the content organization of a problem-oriented summarizer, the model identifies phenotype relatedness as well. The model leverages a correlated-mixed membership approach with variational inference applied to heterogenous clinical data. In this paper, we focus our experiments on assessing the learned phenotypes and their relatedness as learned from a specific patient population. We ground our experiments in phenotyping patients from an HIV clinic in a large urban care institution (n=7,523), where patients have voluminous, longitudinal documentation, and where providers would benefit from summaries of these patient's medical histories, whether about their HIV or any comorbidities. We find that the learned phenotypes and their relatedness are clinically valid when assessed qualitatively by clinical experts, and that the model surpasses baseline in inferring phenotype-relatedness when comparing to existing expert-curated condition groupings.

Iñigo Urteaga, Tristan Bertin, Theresa M. Hardy et al.

We present an end-to-end statistical framework for personalized, accurate, and minimally invasive modeling of female reproductive hormonal patterns. Reconstructing and forecasting the evolution of hormonal dynamics is a challenging task, but a critical one to improve general understanding of the menstrual cycle and personalized detection of potential health issues. Our goal is to infer and forecast individual hormone daily levels over time, while accommodating pragmatic and minimally invasive measurement settings. To that end, our approach combines the power of probabilistic generative models (i.e., multi-task Gaussian processes) with the flexibility of neural networks (i.e., a dilated convolutional architecture) to learn complex temporal mappings. To attain accurate hormone level reconstruction with as little data as possible, we propose a sampling mechanism for optimal reconstruction accuracy with limited sampling budget. Our results show the validity of our proposed hormonal dynamic modeling framework, as it provides accurate predictive performance across different realistic sampling budgets and outperforms baselines methods.

Gal Levy-Fix, Gilad J. Kuperman, Noémie Elhadad

Deep learning, an area of machine learning, is set to revolutionize patient care. But it is not yet part of standard of care, especially when it comes to individual patient care. In fact, it is unclear to what extent data-driven techniques are being used to support clinical decision making (CDS). Heretofore, there has not been a review of ways in which research in machine learning and other types of data-driven techniques can contribute effectively to clinical care and the types of support they can bring to clinicians. In this paper, we consider ways in which two data driven domains - machine learning and data visualizations - can contribute to the next generation of clinical decision support systems. We review the literature regarding the ways heuristic knowledge, machine learning, and visualization are - and can be - applied to three types of CDS. There has been substantial research into the use of predictive modeling for alerts, however current CDS systems are not utilizing these methods. Approaches that leverage interactive visualizations and machine-learning inferences to organize and review patient data are gaining popularity but are still at the prototype stage and are not yet in use. CDS systems that could benefit from prescriptive machine learning (e.g., treatment recommendations for specific patients) have not yet been developed. We discuss potential reasons for the lack of deployment of data-driven methods in CDS and directions for future research.

Iñigo Urteaga, Mollie McKillop, Sharon Lipsky-Gorman et al.

We investigate the use of self-tracking data and unsupervised mixed-membership models to phenotype endometriosis. Endometriosis is a systemic, chronic condition of women in reproductive age and, at the same time, a highly enigmatic condition with no known biomarkers to monitor its progression and no established staging. We leverage data collected through a self-tracking app in an observational research study of over 2,800 women with endometriosis tracking their condition over a year and a half (456,900 observations overall). We extend a classical mixed-membership model to accommodate the idiosyncrasies of the data at hand (i.e., the multimodality of the tracked variables). Our experiments show that our approach identifies potential subtypes that are robust in terms of biases of self-tracked data (e.g., wide variations in tracking frequency amongst participants), as well as to variations in hyperparameters of the model. Jointly modeling a wide range of observations about participants (symptoms, quality of life, treatments) yields clinically meaningful subtypes that both validate what is already known about endometriosis and suggest new findings.

Alexandre Yahi, Rami Vanguri, Noémie Elhadad et al.

Generative Adversarial Networks (GANs) represent a promising class of generative networks that combine neural networks with game theory. From generating realistic images and videos to assisting musical creation, GANs are transforming many fields of arts and sciences. However, their application to healthcare has not been fully realized, more specifically in generating electronic health records (EHR) data. In this paper, we propose a framework for exploring the value of GANs in the context of continuous laboratory time series data. We devise an unsupervised evaluation method that measures the predictive power of synthetic laboratory test time series. Further, we show that when it comes to predicting the impact of drug exposure on laboratory test data, incorporating representation learning of the training cohorts prior to training GAN models is beneficial.

Iñigo Urteaga, David J. Albers, Marija Vlajic Wheeler et al.

In this paper, we introduce a novel task for machine learning in healthcare, namely personalized modeling of the female hormonal cycle. The motivation for this work is to model the hormonal cycle and predict its phases in time, both for healthy individuals and for those with disorders of the reproductive system. Because there are individual differences in the menstrual cycle, we are particularly interested in personalized models that can account for individual idiosyncracies, towards identifying phenotypes of menstrual cycles. As a first step, we consider the hormonal cycle as a set of observations through time. We use a previously validated mechanistic model to generate realistic hormonal patterns, and experiment with Gaussian process regression to estimate their values over time. Specifically, we are interested in the feasibility of predicting menstrual cycle phases under varying learning conditions: number of cycles used for training, hormonal measurement noise and sampling rates, and informed vs. agnostic sampling of hormonal measurements. Our results indicate that Gaussian processes can help model the female menstrual cycle. We discuss the implications of our experiments in the context of modeling the female menstrual cycle.

Rajesh Ranganath, Adler Perotte, Noémie Elhadad et al.

The electronic health record (EHR) provides an unprecedented opportunity to build actionable tools to support physicians at the point of care. In this paper, we investigate survival analysis in the context of EHR data. We introduce deep survival analysis, a hierarchical generative approach to survival analysis. It departs from previous approaches in two primary ways: (1) all observations, including covariates, are modeled jointly conditioned on a rich latent structure; and (2) the observations are aligned by their failure time, rather than by an arbitrary time zero as in traditional survival analysis. Further, it (3) scalably handles heterogeneous (continuous and discrete) data types that occur in the EHR. We validate deep survival analysis model by stratifying patients according to risk of developing coronary heart disease (CHD). Specifically, we study a dataset of 313,000 patients corresponding to 5.5 million months of observations. When compared to the clinically validated Framingham CHD risk score, deep survival analysis is significantly superior in stratifying patients according to their risk.