Using machine learning to develop smart reflex testing protocols.

OBJECTIVE: Reflex testing protocols allow clinical laboratories to perform second line diagnostic tests on existing specimens based on the results of initially ordered tests. Reflex testing can support optimal clinical laboratory test ordering and diagnosis. In current clinical practice, reflex testing typically relies on simple "if-then" rules; however, this limits the opportunities for reflex testing since most test ordering decisions involve more complexity than traditional rule-based approaches would allow. Here, using the analyte ferritin as an example, we propose an alternative machine learning-based approach to "smart" reflex testing. METHODS: Using deidentified patient data, we developed a machine learning model to predict whether a patient getting CBC testing will also have ferritin testing ordered. We evaluate applications of this model to reflex testing by assessing its performance in comparison to possible rule-based approaches. RESULTS: Our underlying machine learning models performed moderately well in predicting ferritin test ordering (AUC=0.731 in reference to actual ordering) and demonstrated promising potential to underlie key clinical applications. In contrast, none of the many traditionally framed, rule-based, hypothetical reflex protocols we evaluated offered sufficient agreement with actual ordering to be clinically feasible. Using chart review, we further demonstrated that the strategic deployment of our model could avoid important ferritin test ordering errors. CONCLUSIONS: Machine learning may provide a foundation for new types of reflex testing with enhanced benefits for clinical diagnosis.

Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study.

BACKGROUND: Large language models (LLMs) such as GPT-4 hold great promise as transformative tools in health care, ranging from automating administrative tasks to augmenting clinical decision making. However, these models also pose a danger of perpetuating biases and delivering incorrect medical diagnoses, which can have a direct, harmful impact on medical care. We aimed to assess whether GPT-4 encodes racial and gender biases that impact its use in health care. METHODS: Using the Azure OpenAI application interface, this model evaluation study tested whether GPT-4 encodes racial and gender biases and examined the impact of such biases on four potential applications of LLMs in the clinical domain-namely, medical education, diagnostic reasoning, clinical plan generation, and subjective patient assessment. We conducted experiments with prompts designed to resemble typical use of GPT-4 within clinical and medical education applications. We used clinical vignettes from NEJM Healer and from published research on implicit bias in health care. GPT-4 estimates of the demographic distribution of medical conditions were compared with true US prevalence estimates. Differential diagnosis and treatment planning were evaluated across demographic groups using standard statistical tests for significance between groups. FINDINGS: We found that GPT-4 did not appropriately model the demographic diversity of medical conditions, consistently producing clinical vignettes that stereotype demographic presentations. The differential diagnoses created by GPT-4 for standardised clinical vignettes were more likely to include diagnoses that stereotype certain races, ethnicities, and genders. Assessment and plans created by the model showed significant association between demographic attributes and recommendations for more expensive procedures as well as differences in patient perception. INTERPRETATION: Our findings highlight the urgent need for comprehensive and transparent bias assessments of LLM tools such as GPT-4 for intended use cases before they are integrated into clinical care. We discuss the potential sources of these biases and potential mitigation strategies before clinical implementation. FUNDING: Priscilla Chan and Mark Zuckerberg.

An open natural language processing (NLP) framework for EHR-based clinical research: a case demonstration using the National COVID Cohort Collaborative (N3C).

Despite recent methodology advancements in clinical natural language processing (NLP), the adoption of clinical NLP models within the translational research community remains hindered by process heterogeneity and human factor variations. Concurrently, these factors also dramatically increase the difficulty in developing NLP models in multi-site settings, which is necessary for algorithm robustness and generalizability. Here, we reported on our experience developing an NLP solution for Coronavirus Disease 2019 (COVID-19) signs and symptom extraction in an open NLP framework from a subset of sites participating in the National COVID Cohort (N3C). We then empirically highlight the benefits of multi-site data for both symbolic and statistical methods, as well as highlight the need for federated annotation and evaluation to resolve several pitfalls encountered in the course of these efforts.

Using Machine Learning to Develop Smart Reflex Testing Protocols.

Objective: Reflex testing protocols allow clinical laboratories to perform second line diagnostic tests on existing specimens based on the results of initially ordered tests. Reflex testing can support optimal clinical laboratory test ordering and diagnosis. In current clinical practice, reflex testing typically relies on simple "if-then" rules; however, this limits their scope since most test ordering decisions involve more complexity than a simple rule will allow. Here, using the analyte ferritin as an example, we propose an alternative machine learning-based approach to "smart" reflex testing with a wider scope and greater impact than traditional rule-based approaches. Methods: Using patient data, we developed a machine learning model to predict whether a patient getting CBC testing will also have ferritin testing ordered, consider applications of this model to "smart" reflex testing, and evaluate the model by comparing its performance to possible rule-based approaches. Results: Our underlying machine learning models performed moderately well in predicting ferritin test ordering and demonstrated greater suitability to reflex testing than rule-based approaches. Using chart review, we demonstrate that our model may improve ferritin test ordering. Finally, as a secondary goal, we demonstrate that ferritin test results are missing not at random (MNAR), a finding with implications for unbiased imputation of missing test results. Conclusions: Machine learning may provide a foundation for new types of reflex testing with enhanced benefits for clinical diagnosis and laboratory utilization management.

Clinical Artificial Intelligence: Design Principles and Fallacies.

Clinical artificial intelligence (AI)/machine learning (ML) is anticipated to offer new abilities in clinical decision support, diagnostic reasoning, precision medicine, clinical operational support, and clinical research, but careful concern is needed to ensure these technologies work effectively in the clinic. Here, we detail the clinical ML/AI design process, identifying several key questions and detailing several common forms of issues that arise with ML tools, as motivated by real-world examples, such that clinicians and researchers can better anticipate and correct for such issues in their own use of ML/AI techniques.

Who doesn't fit? A multi-institutional study using machine learning to uncover the limits of opioid prescribing guidelines.

BACKGROUND: Many U.S. institutions have adopted postsurgical opioid-prescribing guidelines to standardize prescribing practices, and yet there is inherent variability in patients' opioid consumption after surgery. The utility of these guidelines is limited by the fact that some patients' needs will inevitably exceed them, and yet there are no evidence-based tools to help providers identify these patients. In this study we aimed to maximize the value of these guidelines by training machine learning models to predict patients whose needs will be met by these smaller recommended prescriptions, and patients who may require an additional degree of personalization. The aim of the present study was to develop predictive models for determining whether a surgical patient's postdischarge opioid requirement will fall above or below common opioid prescribing guidelines. METHODS: We conducted a retrospective cohort study of surgical patients at one institution from 2017 to 2018. Patients were called after discharge to collect opioid consumption data. Machine learning models were used to identify outlier opioid consumers (ie, exceeding our institutional prescribing guidelines) using diagnosis codes, medical history, in-hospital opioid use, and perioperative factors as predictors. External validation was performed on opioid consumption data collected at a second institution from 2020 to 2021, and sensitivity analysis was performed using a third institution's prescribing guidelines. RESULTS: The development and external validation cohorts included 1,867 and 498 patients, respectively. Age, body mass index, tobacco use, preoperative opioid exposure, and in-hospital opioid consumption were the strongest predictors of postdischarge consumption. A lasso regression model exhibited an area under the receiver operating characteristic curve of 0.74 (95% confidence interval 0.67-0.81) in predicting postdischarge opioid consumption. External validation of a limited lasso model yielded an area under the receiver operating characteristic curve of 0.67 (0.60-0.74). Performance was preserved when evaluated on another institution's guidelines (area under the receiver operating characteristic curve 0.76 [0.72-0.80]). CONCLUSION: Patient characteristics reliably predict postdischarge opioid consumption in relation to prescribing guidelines for both opioid-naive and exposed populations. This model may be used to help providers confidently follow prescribing guidelines for patients with typical opioid responsiveness and correctly pursue more personalized prescribing for others.

Explainable deep learning in healthcare: A methodological survey from an attribution view.

The increasing availability of large collections of electronic health record (EHR) data and unprecedented technical advances in deep learning (DL) have sparked a surge of research interest in developing DL based clinical decision support systems for diagnosis, prognosis, and treatment. Despite the recognition of the value of deep learning in healthcare, impediments to further adoption in real healthcare settings remain due to the black-box nature of DL. Therefore, there is an emerging need for interpretable DL, which allows end users to evaluate the model decision making to know whether to accept or reject predictions and recommendations before an action is taken. In this review, we focus on the interpretability of the DL models in healthcare. We start by introducing the methods for interpretability in depth and comprehensively as a methodological reference for future researchers or clinical practitioners in this field. Besides the methods' details, we also include a discussion of advantages and disadvantages of these methods and which scenarios each of them is suitable for, so that interested readers can know how to compare and choose among them for use. Moreover, we discuss how these methods, originally developed for solving general-domain problems, have been adapted and applied to healthcare problems and how they can help physicians better understand these data-driven technologies. Overall, we hope this survey can help researchers and practitioners in both artificial intelligence and clinical fields understand what methods we have for enhancing the interpretability of their DL models and choose the optimal one accordingly. This article is categorized under: Cancer > Computational Models.

ATLAS: an automated association test using probabilistically linked health records with application to genetic studies.

OBJECTIVE: Large amounts of health data are becoming available for biomedical research. Synthesizing information across databases may capture more comprehensive pictures of patient health and enable novel research studies. When no gold standard mappings between patient records are available, researchers may probabilistically link records from separate databases and analyze the linked data. However, previous linked data inference methods are constrained to certain linkage settings and exhibit low power. Here, we present ATLAS, an automated, flexible, and robust association testing algorithm for probabilistically linked data. MATERIALS AND METHODS: Missing variables are imputed at various thresholds using a weighted average method that propagates uncertainty from probabilistic linkage. Next, estimated effect sizes are obtained using a generalized linear model. ATLAS then conducts the threshold combination test by optimally combining P values obtained from data imputed at varying thresholds using Fisher's method and perturbation resampling. RESULTS: In simulations, ATLAS controls for type I error and exhibits high power compared to previous methods. In a real-world genetic association study, meta-analysis of ATLAS-enabled analyses on a linked cohort with analyses using an existing cohort yielded additional significant associations between rheumatoid arthritis genetic risk score and laboratory biomarkers. DISCUSSION: Weighted average imputation weathers false matches and increases contribution of true matches to mitigate linkage error-induced bias. The threshold combination test avoids arbitrarily choosing a threshold to rule a match, thus automating linked data-enabled analyses and preserving power. CONCLUSION: ATLAS promises to enable novel and powerful research studies using linked data to capitalize on all available data sources.

Visceral Adiposity and Severe COVID-19 Disease: Application of an Artificial Intelligence Algorithm to Improve Clinical Risk Prediction.

BACKGROUND: Obesity has been linked to severe clinical outcomes among people who are hospitalized with coronavirus disease 2019 (COVID-19). We tested the hypothesis that visceral adipose tissue (VAT) is associated with severe outcomes in patients hospitalized with COVID-19, independent of body mass index (BMI). METHODS: We analyzed data from the Massachusetts General Hospital COVID-19 Data Registry, which included patients admitted with polymerase chain reaction-confirmed severe acute respiratory syndrome coronavirus 2 infection from March 11 to May 4, 2020. We used a validated, fully automated artificial intelligence (AI) algorithm to quantify VAT from computed tomography (CT) scans during or before the hospital admission. VAT quantification took an average of 2 ± 0.5 seconds per patient. We dichotomized VAT as high and low at a threshold of ≥100 cm2 and used Kaplan-Meier curves and Cox proportional hazards regression to assess the relationship between VAT and death or intubation over 28 days, adjusting for age, sex, race, BMI, and diabetes status. RESULTS: A total of 378 participants had CT imaging. Kaplan-Meier curves showed that participants with high VAT had a greater risk of the outcome compared with those with low VAT (P < .005), especially in those with BMI <30 kg/m2 (P < .005). In multivariable models, the adjusted hazard ratio (aHR) for high vs low VAT was unchanged (aHR, 1.97; 95% CI, 1.24-3.09), whereas BMI was no longer significant (aHR for obese vs normal BMI, 1.14; 95% CI, 0.71-1.82). CONCLUSIONS: High VAT is associated with a greater risk of severe disease or death in COVID-19 and can offer more precise information to risk-stratify individuals beyond BMI. AI offers a promising approach to routinely ascertain VAT and improve clinical risk prediction in COVID-19.

Artificial intelligence to assess body composition on routine abdominal CT scans and predict mortality in pancreatic cancer- A recipe for your local application.

BACKGROUND: Body composition is associated with mortality; however its routine assessment is too time-consuming. PURPOSE: To demonstrate the value of artificial intelligence (AI) to extract body composition measures from routine studies, we aimed to develop a fully automated AI approach to measure fat and muscles masses, to validate its clinical discriminatory value, and to provide the code, training data and workflow solutions to facilitate its integration into local practice. METHODS: We developed a neural network that quantified the tissue components at the L3 vertebral body level using data from the Liver Tumor Challenge (LiTS) and a pancreatic cancer cohort. We classified sarcopenia using accepted skeletal muscle index cut-offs and visceral fat based its median value. We used Kaplan Meier curves and Cox regression analysis to assess the association between these measures and mortality. RESULTS: Applying the algorithm trained on LiTS data to the local cohort yielded good agreement [>0.8 intraclass correlation (ICC)]; when trained on both datasets, it had excellent agreement (>0.9 ICC). The pancreatic cancer cohort had 136 patients (mean age: 67 ± 11 years; 54% women); 15% had sarcopenia; mean visceral fat was 142 cm2. Concurrent with prior research, we found a significant association between sarcopenia and mortality [mean survival of 15 ± 12 vs. 22 ± 12 (p < 0.05), adjusted HR of 1.58 (95% CI: 1.03-3.33)] but no association between visceral fat and mortality. The detector analysis took 1 ± 0.5 s. CONCLUSIONS: AI body composition analysis can provide meaningful imaging biomarkers from routine exams demonstrating AI's ability to further enhance the clinical value of radiology reports.

Hard for humans, hard for machines: predicting readmission after psychiatric hospitalization using narrative notes.

Machine learning has been suggested as a means of identifying individuals at greatest risk for hospital readmission, including psychiatric readmission. We sought to compare the performance of predictive models that use interpretable representations derived via topic modeling to the performance of human experts and nonexperts. We examined all 5076 admissions to a general psychiatry inpatient unit between 2009 and 2016 using electronic health records. We developed multiple models to predict 180-day readmission for these admissions based on features derived from narrative discharge summaries, augmented by baseline sociodemographic and clinical features. We developed models using a training set comprising 70% of the cohort and evaluated on the remaining 30%. Baseline models using demographic features for prediction achieved an area under the curve (AUC) of 0.675 [95% CI 0.674-0.676] on an independent testing set, while language-based models also incorporating bag-of-words features, discharge summaries topics identified by Latent Dirichlet allocation (LDA), and prior psychiatric admissions achieved AUC of 0.726 [95% CI 0.725-0.727]. To characterize the difficulty of the task, we also compared the performance of these classifiers to both expert and nonexpert human raters, with and without feedback, on a subset of 75 test cases. These models outperformed humans on average, including predictions by experienced psychiatrists. Typical note tokens or topics associated with readmission risk were related to pregnancy/postpartum state, family relationships, and psychosis.

A multidimensional precision medicine approach identifies an autism subtype characterized by dyslipidemia.

The promise of precision medicine lies in data diversity. More than the sheer size of biomedical data, it is the layering of multiple data modalities, offering complementary perspectives, that is thought to enable the identification of patient subgroups with shared pathophysiology. In the present study, we use autism to test this notion. By combining healthcare claims, electronic health records, familial whole-exome sequences and neurodevelopmental gene expression patterns, we identified a subgroup of patients with dyslipidemia-associated autism.

Three-Dimensional Neural Network to Automatically Assess Liver Tumor Burden Change on Consecutive Liver MRIs.

BACKGROUND: Tumor response to therapy is often assessed by measuring change in liver lesion size between consecutive MRIs. However, these evaluations are both tedious and time-consuming for clinical radiologists. PURPOSE: In this study, we sought to develop a convolutional neural network to detect liver metastases on MRI and applied this algorithm to assess change in tumor size on consecutive examinations. METHODS: We annotated a data set of 64 patients with neuroendocrine tumors who underwent at least two consecutive liver MRIs with gadoxetic acid. We then developed a 3-D neural network using a U-Net architecture with ResNet-18 building blocks that first detected the liver and then lesions within the liver. Liver lesion labels for each examination were then matched in 3-D space using an iterative closest point algorithm followed by Kuhn-Munkres algorithm. RESULTS: We developed a deep learning algorithm that detected liver metastases, co-registered the detected lesions, and then assessed the interval change in tumor burden between two multiparametric liver MRI examinations. Our deep learning algorithm was concordant in 91% with the radiologists' manual assessment about the interval change of disease burden. It had a sensitivity of 0.85 (95% confidence interval (95% CI): 0.77; 0.93) and specificity of 0.92 (95% CI: 0.87; 0.96) to classify liver segments as diseased or healthy. The mean DICE coefficient for individual lesions ranged between 0.73 and 0.81. CONCLUSIONS: Our algorithm displayed high agreement with human readers for detecting change in liver lesions on MRI, offering evidence that artificial intelligence-based detectors may perform these tasks as part of routine clinical care in the future.

Advancing PICO element detection in biomedical text via deep neural networks.

MOTIVATION: In evidence-based medicine, defining a clinical question in terms of the specific patient problem aids the physicians to efficiently identify appropriate resources and search for the best available evidence for medical treatment. In order to formulate a well-defined, focused clinical question, a framework called PICO is widely used, which identifies the sentences in a given medical text that belong to the four components typically reported in clinical trials: Participants/Problem (P), Intervention (I), Comparison (C) and Outcome (O). In this work, we propose a novel deep learning model for recognizing PICO elements in biomedical abstracts. Based on the previous state-of-the-art bidirectional long-short-term memory (bi-LSTM) plus conditional random field architecture, we add another layer of bi-LSTM upon the sentence representation vectors so that the contextual information from surrounding sentences can be gathered to help infer the interpretation of the current one. In addition, we propose two methods to further generalize and improve the model: adversarial training and unsupervised pre-training over large corpora. RESULTS: We tested our proposed approach over two benchmark datasets. One is the PubMed-PICO dataset, where our best results outperform the previous best by 5.5%, 7.9% and 5.8% for P, I and O elements in terms of F1 score, respectively. And for the other dataset named NICTA-PIBOSO, the improvements for P/I/O elements are 3.9%, 15.6% and 1.3% in F1 score, respectively. Overall, our proposed deep learning model can obtain unprecedented PICO element detection accuracy while avoiding the need for any manual feature selection. AVAILABILITY AND IMPLEMENTATION: Code is available at https://github.com/jind11/Deep-PICO-Detection.

Joint Modeling of Chest Radiographs and Radiology Reports for Pulmonary Edema Assessment.

We propose and demonstrate a novel machine learning algorithm that assesses pulmonary edema severity from chest radiographs. While large publicly available datasets of chest radiographs and free-text radiology reports exist, only limited numerical edema severity labels can be extracted from radiology reports. This is a significant challenge in learning such models for image classification. To take advantage of the rich information present in the radiology reports, we develop a neural network model that is trained on both images and free-text to assess pulmonary edema severity from chest radiographs at inference time. Our experimental results suggest that the joint image-text representation learning improves the performance of pulmonary edema assessment compared to a supervised model trained on images only. We also show the use of the text for explaining the image classification by the joint model. To the best of our knowledge, our approach is the first to leverage free-text radiology reports for improving the image model performance in this application. Our code is available at: https://github.com/RayRuizhiLiao/joint_chestxray.

Deep Learning Benchmarks on L1000 Gene Expression Data.

Gene expression data can offer deep, physiological insights beyond the static coding of the genome alone. We believe that realizing this potential requires specialized, high-capacity machine learning methods capable of using underlying biological structure, but the development of such models is hampered by the lack of published benchmark tasks and well characterized baselines. In this work, we establish such benchmarks and baselines by profiling many classifiers against biologically motivated tasks on two curated views of a large, public gene expression dataset (the LINCS corpus) and one privately produced dataset. We provide these two curated views of the public LINCS dataset and our benchmark tasks to enable direct comparisons to future methodological work and help spur deep learning method development on this modality. In addition to profiling a battery of traditional classifiers, including linear models, random forests, decision trees, K nearest neighbor (KNN) classifiers, and feed-forward artificial neural networks (FF-ANNs), we also test a method novel to this data modality: graph convolugtional neural networks (GCNNs), which allow us to incorporate prior biological domain knowledge. We find that GCNNs can be highly performant, with large datasets, whereas FF-ANNs consistently perform well. Non-neural classifiers are dominated by linear models and KNN classifiers.

High-throughput phenotyping with electronic medical record data using a common semi-supervised approach (PheCAP).

Phenotypes are the foundation for clinical and genetic studies of disease risk and outcomes. The growth of biobanks linked to electronic medical record (EMR) data has both facilitated and increased the demand for efficient, accurate, and robust approaches for phenotyping millions of patients. Challenges to phenotyping with EMR data include variation in the accuracy of codes, as well as the high level of manual input required to identify features for the algorithm and to obtain gold standard labels. To address these challenges, we developed PheCAP, a high-throughput semi-supervised phenotyping pipeline. PheCAP begins with data from the EMR, including structured data and information extracted from the narrative notes using natural language processing (NLP). The standardized steps integrate automated procedures, which reduce the level of manual input, and machine learning approaches for algorithm training. PheCAP itself can be executed in 1-2 d if all data are available; however, the timing is largely dependent on the chart review stage, which typically requires at least 2 weeks. The final products of PheCAP include a phenotype algorithm, the probability of the phenotype for all patients, and a phenotype classification (yes or no).

High-throughput multimodal automated phenotyping (MAP) with application to PheWAS.

OBJECTIVE: Electronic health records linked with biorepositories are a powerful platform for translational studies. A major bottleneck exists in the ability to phenotype patients accurately and efficiently. The objective of this study was to develop an automated high-throughput phenotyping method integrating International Classification of Diseases (ICD) codes and narrative data extracted using natural language processing (NLP). MATERIALS AND METHODS: We developed a mapping method for automatically identifying relevant ICD and NLP concepts for a specific phenotype leveraging the Unified Medical Language System. Along with health care utilization, aggregated ICD and NLP counts were jointly analyzed by fitting an ensemble of latent mixture models. The multimodal automated phenotyping (MAP) algorithm yields a predicted probability of phenotype for each patient and a threshold for classifying participants with phenotype yes/no. The algorithm was validated using labeled data for 16 phenotypes from a biorepository and further tested in an independent cohort phenome-wide association studies (PheWAS) for 2 single nucleotide polymorphisms with known associations. RESULTS: The MAP algorithm achieved higher or similar AUC and F-scores compared to the ICD code across all 16 phenotypes. The features assembled via the automated approach had comparable accuracy to those assembled via manual curation (AUCMAP 0.943, AUCmanual 0.941). The PheWAS results suggest that the MAP approach detected previously validated associations with higher power when compared to the standard PheWAS method based on ICD codes. CONCLUSION: The MAP approach increased the accuracy of phenotype definition while maintaining scalability, thereby facilitating use in studies requiring large-scale phenotyping, such as PheWAS.

Early Prediction of Acute Kidney Injury in Critical Care Setting Using Clinical Notes and Structured Multivariate Physiological Measurements.

The onset of acute kidney injury (AKI) during an intensive care unit (ICU) admission is associated with increased morbidity and mortality. Developing novel methods to identify early AKI onset is of critical importance in preventing or reducing AKI complications. We built and applied multiple machine learning models to integrate clinical notes and structured physiological measurements and estimate the risk of new AKI onset using the MIMIC-III database. From the clinical notes, we generated clinically meaningful word representations and embeddings. Four supervised learning classifiers and mixed-feature deep learning architecture were used to construct prediction models. The best configurations consistently utilized both structured and unstructured clinical features and yielded competitive AUCs above 0.83. Our work suggests that integrating structured and unstructured clinical features can be effectively applied to assist clinicians in identifying the risk of incident AKI onset in critically-ill patients upon admission to the ICU.

Use of machine-learning algorithms to determine features of systolic blood pressure variability that predict poor outcomes in hypertensive patients.

BACKGROUND: We re-analyzed data from the Systolic Blood Pressure Intervention Trial (SPRINT) trial to identify features of systolic blood pressure (SBP) variability that portend poor cardiovascular outcomes using a nonlinear machine-learning algorithm. METHODS: We included all patients who completed 1 year of the study without reaching any primary endpoint during the first year, specifically: myocardial infarction, other acute coronary syndromes, stroke, heart failure or death from a cardiovascular event (n = 8799; 94%). In addition to clinical variables, features representing longitudinal SBP trends and variability were determined and combined in a random forest algorithm, optimized using cross-validation, using 70% of patients in the training set. Area under the curve (AUC) was measured using a 30% testing set. Finally, feature importance was determined by minimizing node impurity averaging over all trees in the forest for a specific feature. RESULTS: A total of 365 patients (4.1%) reached the combined primary outcome over 37 months of follow-up. The random forest classifier had an AUC of 0.71 on the testing set. The 10 most significant features selected in order of importance by the automated algorithm included the urine albumin/creatinine (CR) ratio, estimated glomerular filtration rate, age, serum CR, history of subclinical cardiovascular disease (CVD), cholesterol, a variable representing SBP signals using wavelet transformation, high-density lipoprotein, the 90th percentile of SBP and triglyceride level. CONCLUSIONS: We successfully demonstrated use of random forest algorithm to define best prognostic longitudinal SBP representations. In addition to known risk factors for CVD, transformed variables for time series SBP measurements were found to be important in predicting poor cardiovascular outcomes and require further evaluation.