Machine Learning for Personalized Medicine: Predicting Primary Myocardial Infarction from Electronic Health Records

Weiss, Jeremy C.; Natarajan, Sriraam; Peissig, Peggy L.; McCarty, Catherine A.; Page, David

doi:10.1609/aimag.v33i4.2438

Cited by 70 publications

(43 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…ILP usage in the medical domain has focused on predicting patient outcomes [14,19–22]. Supervision for the prediction task comes from positive examples (POS—patients with a medical outcome) and negative examples (NEG—patients without the medical outcome), given some common exposure.…”

Section: Methodsmentioning

confidence: 99%

“…ILP has been used in medical studies ranging from predictive screening for breast cancer [19,20] to predicting adverse drug events [14,21,22] or adverse clinical outcomes [23–25]. Unlike rule induction and other propositional machine learning algorithms that assume each example is a feature vector or a record, ILP algorithms work directly on data distributed over different EHR tables.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Relational machine learning for electronic health record-driven phenotyping

Peissig

Costa

Caldwell

et al. 2014

Journal of Biomedical Informatics

Self Cite

View full text Add to dashboard Cite

Objective Electronic health records (EHR) offer medical and pharmacogenomics research unprecedented opportunities to identify and classify patients at risk. EHRs are collections of highly inter-dependent records that include biological, anatomical, physiological, and behavioral observations. They comprise a patient’s clinical phenome, where each patient has thousands of date-stamped records distributed across many relational tables. Development of EHR computer-based phenotyping algorithms require time and medical insight from clinical experts, who most often can only review a small patient subset representative of the total EHR records, to identify phenotype features. In this research we evaluate whether relational machine learning (ML) using Inductive Logic Programming (ILP) can contribute to addressing these issues as a viable approach for EHR-based phenotyping. Methods Two relational learning ILP approaches and three well-known WEKA (Waikato Environment for Knowledge Analysis) implementations of non-relational approaches (PART, J48, and JRIP) were used to develop models for nine phenotypes. International Classification of Diseases, Ninth Revision (ICD-9) coded EHR data were used to select training cohorts for the development of each phenotypic model. Accuracy, precision, recall, F-Measure, and Area Under the Receiver Operating Characteristic (AUROC) curve statistics were measured for each phenotypic model based on independent manually verified test cohorts. A two-sided binomial distribution test (sign test) compared the five ML approaches across phenotypes for statistical significance. Results We developed an approach to automatically label training examples using ICD-9 diagnosis codes for the ML approaches being evaluated. Nine phenotypic models for each MLapproach were evaluated, resulting in better overall model performance in AUROC using ILP when compared to PART (p=0.039), J48 (p=0.003) and JRIP (p=0.003). Discussion ILP has the potential to improve phenotyping by independently delivering clinically expert interpretable rules for phenotype definitions, or intuitive phenotypes to assist experts. Conclusion Relational learning using ILP offers a viable approach to EHR-driven phenotyping.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Relational machine learning for electronic health record-driven phenotyping

Peissig

Costa

Caldwell

et al. 2014

Journal of Biomedical Informatics

Self Cite

View full text Add to dashboard Cite

show abstract

“…That is, we train them in a stage-wise manner at low and scalable costs following Friedman's Gradient Tree Boosting (GTB) (Friedman 2001). This boosting approach has been proven successful in a number of cases, see e.g., Ridgeway (2006), Kersting and Driessens (2008), Dietterich et al (2008), Elith et al (2008), Natarajan et al (2012Natarajan et al ( , 2014bNatarajan et al ( , 2013 and Weiss et al (2012), and since it estimates the parameters and the structure jointly, it is generally related to structure learning of graphical models, in particular to approaches that use the local neighborhood of each variable to construct the entire graph. For example, covariance selection is used in Gaussian models where edges are added to the graph until a stopping criterion is met.…”

Section: Related Work and A First Empirical Illustrationmentioning

confidence: 99%

Poisson Dependency Networks: Gradient Boosted Models for Multivariate Count Data

et al. 2015

Self Cite

View full text Add to dashboard Cite

Although count data are increasingly ubiquitous, surprisingly little work has employed probabilistic graphical models for modeling count data. Indeed the univariate case has been well studied, however, in many situations counts influence each other and should not be considered independently. Standard graphical models such as multinomial or Gaussian ones are also often ill-suited, too, since they disregard either the infinite range over the natural numbers or the potentially asymmetric shape of the distribution of count variables. Existing classes of Poisson graphical models can only model negative conditional dependencies or neglect the prediction of counts or do not scale well. To ease the modeling of multivariate count data, we therefore introduce a novel family of Poisson graphical models, called Poisson Dependency Networks (PDNs). A PDN consists of a set of local conditional Poisson distributions, each representing the probability of a single count variable given the others, that naturally facilitates a simple Gibbs sampling inference. In contrast to existing Poisson graphical models, PDNs are non-parametric and trained using functional gradient ascent, i.e., boosting. The particularly simple form of the Poisson distribution allows us to develop the first multiplicative boosting approach: starting from an initial constant value, alternatively a log-linear Poisson model, or a Poisson regression tree, a PDN is represented as products of regression models grown in a stage-wise optimization. We demonstrate on several real world Editors: João Gama, Indre Žliobaite, Alípio M. Jorge, and Concha Bielza. datasets that PDNs can model positive and negative dependencies and scale well while often outperforming state-of-the-art, in particular when using multiplicative updates.

show abstract

“…A number of studies have been since conducted purporting significant improvements over the Framingham Risk Score using different models or by collecting additional information [11]. In particular, the use of EHR data to predict heart attacks was previously addressed in [12]. However, in that work the temporal dependence of the outcome and its predictors was strictly logical and limited the success of their approach.…”

Section: Related Workmentioning

confidence: 99%

Forest-Based Point Process for Event Prediction from Electronic Health Records

Weiss

Page

2013

Advanced Information Systems Engineering

Self Cite

View full text Add to dashboard Cite

Abstract. Accurate prediction of future onset of disease from Electronic Health Records (EHRs) has important clinical and economic implications. In this domain the arrival of data comes at semi-irregular intervals and makes the prediction task challenging. We propose a method called multiplicative-forest point processes (MFPPs) that learns the rate of future events based on an event history. MFPPs join previous theory in multiplicative forest continuous-time Bayesian networks and piecewisecontinuous conditional intensity models. We analyze the advantages of using MFPPs over previous methods and show that on synthetic and real EHR forecasting of heart attacks, MFPPs outperform earlier methods and augment off-the-shelf machine learning algorithms.

show abstract

Machine Learning for Personalized Medicine: Predicting Primary Myocardial Infarction from Electronic Health Records

Cited by 70 publications

References 31 publications

Relational machine learning for electronic health record-driven phenotyping

Relational machine learning for electronic health record-driven phenotyping

Poisson Dependency Networks: Gradient Boosted Models for Multivariate Count Data

Forest-Based Point Process for Event Prediction from Electronic Health Records

Contact Info

Product

Resources

About