Modeling the time-series of high-dimensional, longitudinal data is important for predicting patient disease progression. However, existing neural network based approaches that learn representations of patient state, while very flexible, are susceptible to overfitting. We propose a deep generative model that makes use of a novel attention-based neural architecture inspired by the physics of how treatments affect disease state. The result is a scalable and accurate model of high-dimensional patient biomarkers as they vary over time. Our proposed model yields significant improvements in generalization and, on real-world clinical data, provides interpretable insights into the dynamics of cancer progression.
Unsupervised learning seeks to uncover patterns in data. However, different kinds of noise may impede the discovery of useful substructure from real-world time-series data. In this work, we focus on mitigating the interference of left-censorship in the task of clustering. We provide conditions under which clusters and left-censorship may be identified; motivated by this result, we develop a deep generative, continuous-time model of time-series data that clusters while correcting for censorship time. We demonstrate accurate, stable, and interpretable results on synthetic data that outperform several benchmarks. To showcase the utility of our framework on real-world problems, we study how left-censorship can adversely affect the task of disease phenotyping, resulting in the often incorrect assumption that longitudinal patient data are aligned by disease stage. In reality, patients at the time of diagnosis are at different stages of the disease -- both late and early due to differences in when patients seek medical care and such discrepancy can confound unsupervised learning algorithms. On two clinical datasets, our model corrects for this form of censorship and recovers known clinical subtypes.
We prove that the alpha-expansion algorithm for MAP inference always returns a globally optimal assignment for Markov Random Fields with Potts pairwise potentials, with a catch: the returned assignment is only guaranteed to be optimal in a small perturbation of the original problem instance. In other words, all local minima with respect to expansion moves are global minima to slightly perturbed versions of the problem. On "real-world" instances, MAP assignments of small perturbations of the problem should be very similar to the MAP assignment(s) of the original problem instance. We design an algorithm that can certify whether this is the case in practice. On several MAP inference problem instances from computer vision, this algorithm certifies that MAP solutions to all of these perturbations are very close to solutions of the original instance. These results taken together give a cohesive explanation for the good performance of "graph cuts" algorithms in practice. Every local expansion minimum is a global minimum in a small perturbation of the problem, and all of these global minima are close to the original solution.
Treatment policies learned via reinforcement learning (RL) from observational health data are sensitive to subtle choices in study design. We highlight a simple approach, trajectory inspection, to bring clinicians into an iterative design process for model-based RL studies. We inspect trajectories where the model recommends unexpectedly aggressive treatments or believes its recommendations would lead to much more positive outcomes. Then, we examine clinical trajectories simulated with the learned model and policy alongside the actual hospital course to uncover possible modeling issues. To demonstrate that this approach yields insights, we apply it to recent work on RL for inpatient sepsis management. We find that a design choice around maximum trajectory length leads to a model bias towards discharge, that the RL policy preference for high vasopressor doses may be linked to small sample sizes, and that the model has a clinically implausible expectation of discharge without weaning off vasopressors.
Data cleaning is naturally framed as probabilistic inference in a generative model, combining a prior distribution over ground-truth databases with a likelihood that models the noisy channel by which the data are filtered, corrupted, and joined to yield incomplete, dirty, and denormalized datasets. Based on this view, we present PClean, a unified generative modeling architecture for cleaning and normalizing dirty data in diverse domains. Given an unclean dataset and a probabilistic program encoding relevant domain knowledge, PClean learns a structured representation of the data as a relational database of interrelated objects, and uses this latent structure to impute missing values, identify duplicates, detect errors, and propose corrections in the original data table. PClean makes three modeling and inference contributions: (i) a domain-general non-parametric generative model of relational data, for inferring latent objects and their network of latent connections; (ii) a domain-specific probabilistic programming language, for encoding domain knowledge specific to each dataset being cleaned; and (iii) a domain-general inference engine that adapts to each PClean program by constructing data-driven proposals used in sequential Monte Carlo and particle Gibbs. We show empirically that short (< 50-line) PClean programs deliver higher accuracy than state-of-the-art data cleaning systems based on machine learning and weighted logic; that PClean's inference algorithm is faster than generic particle Gibbs inference for probabilistic programs; and that PClean scales to large real-world datasets with millions of rows.
Clinical studies often require understanding elements of a patient's narrative that exist only in free text clinical notes. To transform notes into structured data for downstream use, these elements are commonly extracted and normalized to medical vocabularies. In this work, we audit the performance of and indicate areas of improvement for state-of-the-art systems. We find that high task accuracies for clinical entity normalization systems on the 2019 n2c2 Shared Task are misleading, and underlying performance is still brittle. Normalization accuracy is high for common concepts (95.3%), but much lower for concepts unseen in training data (69.3%). We demonstrate that current approaches are hindered in part by inconsistencies in medical vocabularies, limitations of existing labeling schemas, and narrow evaluation techniques. We reformulate the annotation framework for clinical entity extraction to factor in these issues to allow for robust end-to-end system benchmarking. We evaluate concordance of annotations from our new framework between two annotators and achieve a Jaccard similarity of 0.73 for entity recognition and an agreement of 0.83 for entity normalization. We propose a path forward to address the demonstrated need for the creation of a reference standard to spur method development in entity recognition and normalization.
We present a system that uses a learned autocompletion mechanism to facilitate rapid creation of semi-structured clinical documentation. We dynamically suggest relevant clinical concepts as a doctor drafts a note by leveraging features from both unstructured and structured medical data. By constraining our architecture to shallow neural networks, we are able to make these suggestions in real time. Furthermore, as our algorithm is used to write a note, we can automatically annotate the documentation with clean labels of clinical concepts drawn from medical vocabularies, making notes more structured and readable for physicians, patients, and future algorithms. To our knowledge, this system is the only machine learning-based documentation utility for clinical notes deployed in a live hospital setting, and it reduces keystroke burden of clinical concepts by 67% in real environments.
Healthcare providers are increasingly using learned methods to predict and understand long-term patient outcomes in order to make meaningful interventions. However, despite innovations in this area, deep learning models often struggle to match performance of shallow linear models in predicting these outcomes, making it difficult to leverage such techniques in practice. In this work, motivated by the task of clinical prediction from insurance claims, we present a new technique called reverse distillation which pretrains deep models by using high-performing linear models for initialization. We make use of the longitudinal structure of insurance claims datasets to develop Self Attention with Reverse Distillation, or SARD, an architecture that utilizes a combination of contextual embedding, temporal embedding and self-attention mechanisms and most critically is trained via reverse distillation. SARD outperforms state-of-the-art methods on multiple clinical prediction outcomes, with ablation studies revealing that reverse distillation is a primary driver of these improvements.