H. Zhang, Amy X. Lu, Mohamed Abdalla, Matthew B. A. McDermott, M. Ghassemi
In this work, we examine the extent to which embeddings may encode marginalized populations differently, and how this may lead to a perpetuation of biases and worsened performance on clinical tasks. We pretrain deep embedding models (BERT) on medical notes from the MIMIC-III hospital dataset, and quantify potential disparities using two approaches. First, we identify dangerous latent relationships that are captured by the contextual word embeddings using a fill-in-the-blank method with text from real clinical notes and a log probability bias score quantification. Second, we evaluate performance gaps across different definitions of fairness on over 50 downstream clinical prediction tasks that include detection of acute and chronic conditions. We find that classifiers trained from BERT representations exhibit statistically significant differences in performance, often favoring the majority group with regards to gender, language, ethnicity, and insurance status. Finally, we explore shortcomings of using adversarial debiasing to obfuscate subgroup information in contextual word embeddings, and recommend best practices for such deep embedding models in clinical settings.
{"title":"Hurtful words: quantifying biases in clinical contextual word embeddings","authors":"H. Zhang, Amy X. Lu, Mohamed Abdalla, Matthew B. A. McDermott, M. Ghassemi","doi":"10.1145/3368555.3384448","DOIUrl":"https://doi.org/10.1145/3368555.3384448","url":null,"abstract":"In this work, we examine the extent to which embeddings may encode marginalized populations differently, and how this may lead to a perpetuation of biases and worsened performance on clinical tasks. We pretrain deep embedding models (BERT) on medical notes from the MIMIC-III hospital dataset, and quantify potential disparities using two approaches. First, we identify dangerous latent relationships that are captured by the contextual word embeddings using a fill-in-the-blank method with text from real clinical notes and a log probability bias score quantification. Second, we evaluate performance gaps across different definitions of fairness on over 50 downstream clinical prediction tasks that include detection of acute and chronic conditions. We find that classifiers trained from BERT representations exhibit statistically significant differences in performance, often favoring the majority group with regards to gender, language, ethnicity, and insurance status. Finally, we explore shortcomings of using adversarial debiasing to obfuscate subgroup information in contextual word embeddings, and recommend best practices for such deep embedding models in clinical settings.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73808405","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Paidamoyo Chapfuwa, Chunyuan Li, Nikhil Mehta, L. Carin, Ricardo Henao
Conventional survival analysis approaches estimate risk scores or individualized time-to-event distributions conditioned on covariates. In practice, there is often great population-level phenotypic heterogeneity, resulting from (unknown) subpopulations with diverse risk profiles or survival distributions. As a result, there is an unmet need in survival analysis for identifying subpopulations with distinct risk profiles, while jointly accounting for accurate individualized time-to-event predictions. An approach that addresses this need is likely to improve the characterization of individual outcomes by leveraging regularities in subpopulations, thus accounting for population-level heterogeneity. In this paper, we propose a Bayesian nonparametrics approach that represents observations (subjects) in a clustered latent space, and encourages accurate time-to-event predictions and clusters (subpopulations) with distinct risk profiles. Experiments on real-world datasets show consistent improvements in predictive performance and interpretability relative to existing state-of-the-art survival analysis models.
{"title":"Survival cluster analysis","authors":"Paidamoyo Chapfuwa, Chunyuan Li, Nikhil Mehta, L. Carin, Ricardo Henao","doi":"10.1145/3368555.3384465","DOIUrl":"https://doi.org/10.1145/3368555.3384465","url":null,"abstract":"Conventional survival analysis approaches estimate risk scores or individualized time-to-event distributions conditioned on covariates. In practice, there is often great population-level phenotypic heterogeneity, resulting from (unknown) subpopulations with diverse risk profiles or survival distributions. As a result, there is an unmet need in survival analysis for identifying subpopulations with distinct risk profiles, while jointly accounting for accurate individualized time-to-event predictions. An approach that addresses this need is likely to improve the characterization of individual outcomes by leveraging regularities in subpopulations, thus accounting for population-level heterogeneity. In this paper, we propose a Bayesian nonparametrics approach that represents observations (subjects) in a clustered latent space, and encourages accurate time-to-event predictions and clusters (subpopulations) with distinct risk profiles. Experiments on real-world datasets show consistent improvements in predictive performance and interpretability relative to existing state-of-the-art survival analysis models.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-02-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83583192","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In several crucial applications, domain knowledge is encoded by a system of ordinary differential equations (ODE), often stemming from underlying physical and biological processes. A motivating example is intensive care unit patients: the dynamics of vital physiological functions, such as the cardiovascular system with its associated variables (heart rate, cardiac contractility and output and vascular resistance) can be approximately described by a known system of ODEs. Typically, some of the ODE variables are directly observed (heart rate and blood pressure for example) while some are unobserved (cardiac contractility, output and vascular resistance), and in addition many other variables are observed but not modeled by the ODE, for example body temperature. Importantly, the unobserved ODE variables are "known-unknowns": We know they exist and their functional dynamics, but cannot measure them directly, nor do we know the function tying them to all observed measurements. As is often the case in medicine, and specifically the cardiovascular system, estimating these known-unknowns is highly valuable and they serve as targets for therapeutic manipulations. Under this scenario we wish to learn the parameters of the ODE generating each observed time-series, and extrapolate the future of the ODE variables and the observations. We address this task with a variational autoencoder incorporating the known ODE function, called GOKU-net1 for Generative ODE modeling with Known Unknowns. We first validate our method on videos of single and double pendulums with unknown length or mass; we then apply it to a model of the cardiovascular system. We show that modeling the known-unknowns allows us to successfully discover clinically meaningful unobserved system parameters, leads to much better extrapolation, and enables learning using much smaller training sets.
{"title":"Generative ODE modeling with known unknowns","authors":"Ori Linial, D. Eytan, Uri Shalit","doi":"10.1145/3450439.3451866","DOIUrl":"https://doi.org/10.1145/3450439.3451866","url":null,"abstract":"In several crucial applications, domain knowledge is encoded by a system of ordinary differential equations (ODE), often stemming from underlying physical and biological processes. A motivating example is intensive care unit patients: the dynamics of vital physiological functions, such as the cardiovascular system with its associated variables (heart rate, cardiac contractility and output and vascular resistance) can be approximately described by a known system of ODEs. Typically, some of the ODE variables are directly observed (heart rate and blood pressure for example) while some are unobserved (cardiac contractility, output and vascular resistance), and in addition many other variables are observed but not modeled by the ODE, for example body temperature. Importantly, the unobserved ODE variables are \"known-unknowns\": We know they exist and their functional dynamics, but cannot measure them directly, nor do we know the function tying them to all observed measurements. As is often the case in medicine, and specifically the cardiovascular system, estimating these known-unknowns is highly valuable and they serve as targets for therapeutic manipulations. Under this scenario we wish to learn the parameters of the ODE generating each observed time-series, and extrapolate the future of the ODE variables and the observations. We address this task with a variational autoencoder incorporating the known ODE function, called GOKU-net1 for Generative ODE modeling with Known Unknowns. We first validate our method on videos of single and double pendulums with unknown length or mass; we then apply it to a model of the cardiovascular system. We show that modeling the known-unknowns allows us to successfully discover clinically meaningful unobserved system parameters, leads to much better extrapolation, and enables learning using much smaller training sets.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-02-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89526901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
N. Ravindra, Arijit Sehanobish, Jenna L. Pappalardo, D. Hafler, D. V. Dijk
Single-cell RNA sequencing (scRNA-seq) has revolutionized bio-logical discovery, providing an unbiased picture of cellular heterogeneity in tissues. While scRNA-seq has been used extensively to provide insight into health and disease, it has not been used for disease prediction or diagnostics. Graph Attention Networks have proven to be versatile for a wide range of tasks by learning from both original features and graph structures. Here we present a graph attention model for predicting disease state from single-cell data on a large dataset of Multiple Sclerosis (MS) patients. MS is a disease of the central nervous system that is difficult to diagnose. We train our model on single-cell data obtained from blood and cerebrospinal fluid (CSF) for a cohort of seven MS patients and six healthy adults (HA), resulting in 66,667 individual cells. We achieve 92% accuracy in predicting MS, outperforming other state-of-the-art methods such as a graph convolutional network, random forest, and multi-layer perceptron. Further, we use the learned graph attention model to get insight into the features (cell types and genes) that are important for this prediction. The graph attention model also allow us to infer a new feature space for the cells that emphasizes the difference between the two conditions. Finally we use the attention weights to learn a new low-dimensional embedding which we visualize with PHATE and UMAP. To the best of our knowledge, this is the first effort to use graph attention, and deep learning in general, to predict disease state from single-cell data. We envision applying this method to single-cell data for other diseases.
{"title":"Disease state prediction from single-cell data using graph attention networks","authors":"N. Ravindra, Arijit Sehanobish, Jenna L. Pappalardo, D. Hafler, D. V. Dijk","doi":"10.1145/3368555.3384449","DOIUrl":"https://doi.org/10.1145/3368555.3384449","url":null,"abstract":"Single-cell RNA sequencing (scRNA-seq) has revolutionized bio-logical discovery, providing an unbiased picture of cellular heterogeneity in tissues. While scRNA-seq has been used extensively to provide insight into health and disease, it has not been used for disease prediction or diagnostics. Graph Attention Networks have proven to be versatile for a wide range of tasks by learning from both original features and graph structures. Here we present a graph attention model for predicting disease state from single-cell data on a large dataset of Multiple Sclerosis (MS) patients. MS is a disease of the central nervous system that is difficult to diagnose. We train our model on single-cell data obtained from blood and cerebrospinal fluid (CSF) for a cohort of seven MS patients and six healthy adults (HA), resulting in 66,667 individual cells. We achieve 92% accuracy in predicting MS, outperforming other state-of-the-art methods such as a graph convolutional network, random forest, and multi-layer perceptron. Further, we use the learned graph attention model to get insight into the features (cell types and genes) that are important for this prediction. The graph attention model also allow us to infer a new feature space for the cells that emphasizes the difference between the two conditions. Finally we use the attention weights to learn a new low-dimensional embedding which we visualize with PHATE and UMAP. To the best of our knowledge, this is the first effort to use graph attention, and deep learning in general, to predict disease state from single-cell data. We envision applying this method to single-cell data for other diseases.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79660898","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Joseph D. Janizek, G. Erion, A. DeGrave, Su-In Lee
While deep learning has shown promise in the domain of disease classification from medical images, models based on state-of-the-art convolutional neural network architectures often exhibit performance loss due to dataset shift. Models trained using data from one hospital system achieve high predictive performance when tested on data from the same hospital, but perform significantly worse when they are tested in different hospital systems. Furthermore, even within a given hospital system, deep learning models have been shown to depend on hospital- and patient-level confounders rather than meaningful pathology to make classifications. In order for these models to be safely deployed, we would like to ensure that they do not use confounding variables to make their classification, and that they will work well even when tested on images from hospitals that were not included in the training data. We attempt to address this problem in the context of pneumonia classification from chest radiographs. We propose an approach based on adversarial optimization, which allows us to learn more robust models that do not depend on confounders. Specifically, we demonstrate improved out-of-hospital generalization performance of a pneumonia classifier by training a model that is invariant to the view position of chest radiographs (anterior-posterior vs. posterior-anterior). Our approach leads to better predictive performance on external hospital data than both a standard baseline and previously proposed methods to handle confounding, and also suggests a method for identifying models that may rely on confounders.
{"title":"An adversarial approach for the robust classification of pneumonia from chest radiographs","authors":"Joseph D. Janizek, G. Erion, A. DeGrave, Su-In Lee","doi":"10.1145/3368555.3384458","DOIUrl":"https://doi.org/10.1145/3368555.3384458","url":null,"abstract":"While deep learning has shown promise in the domain of disease classification from medical images, models based on state-of-the-art convolutional neural network architectures often exhibit performance loss due to dataset shift. Models trained using data from one hospital system achieve high predictive performance when tested on data from the same hospital, but perform significantly worse when they are tested in different hospital systems. Furthermore, even within a given hospital system, deep learning models have been shown to depend on hospital- and patient-level confounders rather than meaningful pathology to make classifications. In order for these models to be safely deployed, we would like to ensure that they do not use confounding variables to make their classification, and that they will work well even when tested on images from hospitals that were not included in the training data. We attempt to address this problem in the context of pneumonia classification from chest radiographs. We propose an approach based on adversarial optimization, which allows us to learn more robust models that do not depend on confounders. Specifically, we demonstrate improved out-of-hospital generalization performance of a pneumonia classifier by training a model that is invariant to the view position of chest radiographs (anterior-posterior vs. posterior-anterior). Our approach leads to better predictive performance on external hospital data than both a standard baseline and previously proposed methods to handle confounding, and also suggests a method for identifying models that may rely on confounders.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81387282","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Electronic Health Records (EHR) are high-dimensional data with implicit connections among thousands of medical concepts. These connections, for instance, the co-occurrence of diseases and lab-disease correlations can be informative when only a subset of these variables is documented by the clinician. A feasible approach to improving the representation learning of EHR data is to associate relevant medical concepts and utilize these connections. Existing medical ontologies can be the reference for EHR structures, but they place numerous constraints on the data source. Recent progress on graph neural networks (GNN) enables end-to-end learning of topological structures for non-grid or non-sequential data. However, there are problems to be addressed on how to learn the medical graph adaptively and how to understand the effect of medical graph on representation learning. In this paper, we propose a variationally regularized encoder-decoder graph network that achieves more robustness in graph structure learning by regularizing node representations. Our model outperforms the existing graph and non-graph based methods in various EHR predictive tasks based on both public data and real-world clinical data. Besides the improvements in empirical experiment performances, we provide an interpretation of the effect of variational regularization compared to standard graph neural network, using singular value analysis.
{"title":"Variationally regularized graph-based representation learning for electronic health records","authors":"Weicheng Zhu, N. Razavian","doi":"10.1145/3450439.3451855","DOIUrl":"https://doi.org/10.1145/3450439.3451855","url":null,"abstract":"Electronic Health Records (EHR) are high-dimensional data with implicit connections among thousands of medical concepts. These connections, for instance, the co-occurrence of diseases and lab-disease correlations can be informative when only a subset of these variables is documented by the clinician. A feasible approach to improving the representation learning of EHR data is to associate relevant medical concepts and utilize these connections. Existing medical ontologies can be the reference for EHR structures, but they place numerous constraints on the data source. Recent progress on graph neural networks (GNN) enables end-to-end learning of topological structures for non-grid or non-sequential data. However, there are problems to be addressed on how to learn the medical graph adaptively and how to understand the effect of medical graph on representation learning. In this paper, we propose a variationally regularized encoder-decoder graph network that achieves more robustness in graph structure learning by regularizing node representations. Our model outperforms the existing graph and non-graph based methods in various EHR predictive tasks based on both public data and real-world clinical data. Besides the improvements in empirical experiment performances, we provide an interpretation of the effect of variational regularization compared to standard graph neural network, using singular value analysis.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76835562","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
While machine learning is rapidly being developed and deployed in health settings such as influenza prediction, there are critical challenges in using data from one environment to predict in another due to variability in features. Even within disease labels there can be differences (e.g. "fever" may mean something different reported in a doctor's office versus in an online app). Moreover, models are often built on passive, observational data which contain different distributions of population subgroups (e.g. men or women). Thus, there are two forms of instability between environments in this observational transport problem. We first harness substantive knowledge from health research to conceptualize the underlying causal structure of this problem in a health outcome prediction task. Based on sources of stability in the model and the task, we posit that we can combine environment and population information in a novel population-aware hierarchical Bayesian domain adaptation framework that harnesses multiple invariant components through population attributes when needed. We study the conditions under which invariant learning fails, leading to reliance on the environment-specific attributes. Experimental results for an influenza prediction task on four datasets gathered from different contexts show the model can improve prediction in the case of largely unlabelled target data from a new environment and different constituent population, by harnessing both environment and population invariant information. This work represents a novel, principled way to address a critical challenge by blending domain (health) knowledge and algorithmic innovation. The proposed approach will have significant impact in many social settings wherein who the data comes from and how it was generated, matters.
{"title":"Population-aware hierarchical bayesian domain adaptation via multi-component invariant learning","authors":"V. Mhasawade, N. Rehman, R. Chunara","doi":"10.1145/3368555.3384451","DOIUrl":"https://doi.org/10.1145/3368555.3384451","url":null,"abstract":"While machine learning is rapidly being developed and deployed in health settings such as influenza prediction, there are critical challenges in using data from one environment to predict in another due to variability in features. Even within disease labels there can be differences (e.g. \"fever\" may mean something different reported in a doctor's office versus in an online app). Moreover, models are often built on passive, observational data which contain different distributions of population subgroups (e.g. men or women). Thus, there are two forms of instability between environments in this observational transport problem. We first harness substantive knowledge from health research to conceptualize the underlying causal structure of this problem in a health outcome prediction task. Based on sources of stability in the model and the task, we posit that we can combine environment and population information in a novel population-aware hierarchical Bayesian domain adaptation framework that harnesses multiple invariant components through population attributes when needed. We study the conditions under which invariant learning fails, leading to reliance on the environment-specific attributes. Experimental results for an influenza prediction task on four datasets gathered from different contexts show the model can improve prediction in the case of largely unlabelled target data from a new environment and different constituent population, by harnessing both environment and population invariant information. This work represents a novel, principled way to address a critical challenge by blending domain (health) knowledge and algorithmic innovation. The proposed approach will have significant impact in many social settings wherein who the data comes from and how it was generated, matters.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73829768","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Shirly Wang, Matthew B. A. McDermott, Geeticka Chauhan, Michael C. Hughes, Tristan Naumann, M. Ghassemi
Machine learning for healthcare researchers face challenges to progress and reproducibility due to a lack of standardized processing frameworks for public datasets. We present MIMIC-Extract, an open source pipeline for transforming the raw electronic health record (EHR) data of critical care patients from the publicly-available MIMIC-III database into data structures that are directly usable in common time-series prediction pipelines. MIMIC-Extract addresses three challenges in making complex EHR data accessible to the broader machine learning community. First, MIMIC-Extract transforms raw vital sign and laboratory measurements into usable hourly time series, performing essential steps such as unit conversion, outlier handling, and aggregation of semantically similar features to reduce missingness and improve robustness. Second, MIMIC-Extract extracts and makes prediction of clinically-relevant targets possible, including outcomes such as mortality and length-of-stay as well as comprehensive hourly intervention signals for ventilators, vasopressors, and fluid therapies. Finally, the pipeline emphasizes reproducibility and extensibility to future research questions. We demonstrate the pipeline's effectiveness by developing several benchmark tasks for outcome and intervention forecasting and assessing the performance of competitive models.
{"title":"MIMIC-Extract: a data extraction, preprocessing, and representation pipeline for MIMIC-III","authors":"Shirly Wang, Matthew B. A. McDermott, Geeticka Chauhan, Michael C. Hughes, Tristan Naumann, M. Ghassemi","doi":"10.1145/3368555.3384469","DOIUrl":"https://doi.org/10.1145/3368555.3384469","url":null,"abstract":"Machine learning for healthcare researchers face challenges to progress and reproducibility due to a lack of standardized processing frameworks for public datasets. We present MIMIC-Extract, an open source pipeline for transforming the raw electronic health record (EHR) data of critical care patients from the publicly-available MIMIC-III database into data structures that are directly usable in common time-series prediction pipelines. MIMIC-Extract addresses three challenges in making complex EHR data accessible to the broader machine learning community. First, MIMIC-Extract transforms raw vital sign and laboratory measurements into usable hourly time series, performing essential steps such as unit conversion, outlier handling, and aggregation of semantically similar features to reduce missingness and improve robustness. Second, MIMIC-Extract extracts and makes prediction of clinically-relevant targets possible, including outcomes such as mortality and length-of-stay as well as comprehensive hourly intervention signals for ventilators, vasopressors, and fluid therapies. Finally, the pipeline emphasizes reproducibility and extensibility to future research questions. We demonstrate the pipeline's effectiveness by developing several benchmark tasks for outcome and intervention forecasting and assessing the performance of competitive models.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79677379","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Michaela Hardt, A. Rajkomar, Gerardo Flores, Andrew M. Dai, M. Howell, Greg S. Corrado, Claire Cui, Moritz Hardt
Much work aims to explain a model's prediction on a static input. We consider explanations in a temporal setting where a stateful dynamical model produces a sequence of risk estimates given an input at each time step. When the estimated risk increases, the goal of the explanation is to attribute the increase to a few relevant inputs from the past. While our formal setup and techniques are general, we carry out an in-depth case study in a clinical setting. The goal here is to alert a clinician when a patient's risk of deterioration rises. The clinician then has to decide whether to intervene and adjust the treatment. Given a potentially long sequence of new events since she last saw the patient, a concise explanation helps her to quickly triage the alert. We develop methods to lift static attribution techniques to the dynamical setting, where we identify and address challenges specific to dynamics. We then experimentally assess the utility of different explanations of clinical alerts through expert evaluation.
{"title":"Explaining an increase in predicted risk for clinical alerts","authors":"Michaela Hardt, A. Rajkomar, Gerardo Flores, Andrew M. Dai, M. Howell, Greg S. Corrado, Claire Cui, Moritz Hardt","doi":"10.1145/3368555.3384460","DOIUrl":"https://doi.org/10.1145/3368555.3384460","url":null,"abstract":"Much work aims to explain a model's prediction on a static input. We consider explanations in a temporal setting where a stateful dynamical model produces a sequence of risk estimates given an input at each time step. When the estimated risk increases, the goal of the explanation is to attribute the increase to a few relevant inputs from the past. While our formal setup and techniques are general, we carry out an in-depth case study in a clinical setting. The goal here is to alert a clinician when a patient's risk of deterioration rises. The clinician then has to decide whether to intervene and adjust the treatment. Given a potentially long sequence of new events since she last saw the patient, a concise explanation helps her to quickly triage the alert. We develop methods to lift static attribution techniques to the dynamical setting, where we identify and address challenges specific to dynamics. We then experimentally assess the utility of different explanations of clinical alerts through expert evaluation.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72762737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Michael W. Dusenberry, Dustin Tran, E. Choi, Jonas Kemp, Jeremy Nixon, Ghassen Jerfel, K. Heller, Andrew M. Dai
In medicine, both ethical and monetary costs of incorrect predictions can be significant, and the complexity of the problems often necessitates increasingly complex models. Recent work has shown that changing just the random seed is enough for otherwise well-tuned deep neural networks to vary in their individual predicted probabilities. In light of this, we investigate the role of model uncertainty methods in the medical domain. Using RNN ensembles and various Bayesian RNNs, we show that population-level metrics, such as AUC-PR, AUC-ROC, log-likelihood, and calibration error, do not capture model uncertainty. Meanwhile, the presence of significant variability in patient-specific predictions and optimal decisions motivates the need for capturing model uncertainty. Understanding the uncertainty for individual patients is an area with clear clinical impact, such as determining when a model decision is likely to be brittle. We further show that RNNs with only Bayesian embeddings can be a more efficient way to capture model uncertainty compared to ensembles, and we analyze how model uncertainty is impacted across individual input features and patient subgroups.
{"title":"Analyzing the role of model uncertainty for electronic health records","authors":"Michael W. Dusenberry, Dustin Tran, E. Choi, Jonas Kemp, Jeremy Nixon, Ghassen Jerfel, K. Heller, Andrew M. Dai","doi":"10.1145/3368555.3384457","DOIUrl":"https://doi.org/10.1145/3368555.3384457","url":null,"abstract":"In medicine, both ethical and monetary costs of incorrect predictions can be significant, and the complexity of the problems often necessitates increasingly complex models. Recent work has shown that changing just the random seed is enough for otherwise well-tuned deep neural networks to vary in their individual predicted probabilities. In light of this, we investigate the role of model uncertainty methods in the medical domain. Using RNN ensembles and various Bayesian RNNs, we show that population-level metrics, such as AUC-PR, AUC-ROC, log-likelihood, and calibration error, do not capture model uncertainty. Meanwhile, the presence of significant variability in patient-specific predictions and optimal decisions motivates the need for capturing model uncertainty. Understanding the uncertainty for individual patients is an area with clear clinical impact, such as determining when a model decision is likely to be brittle. We further show that RNNs with only Bayesian embeddings can be a more efficient way to capture model uncertainty compared to ensembles, and we analyze how model uncertainty is impacted across individual input features and patient subgroups.","PeriodicalId":87342,"journal":{"name":"Proceedings of the ACM Conference on Health, Inference, and Learning","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81730399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：