Dataset Information

A machine learning approach to triaging patients with chronic obstructive pulmonary disease.

ABSTRACT: COPD patients are burdened with a daily risk of acute exacerbation and loss of control, which could be mitigated by effective, on-demand decision support tools. In this study, we present a machine learning-based strategy for early detection of exacerbations and subsequent triage. Our application uses physician opinion in a statistically and clinically comprehensive set of patient cases to train a supervised prediction algorithm. The accuracy of the model is assessed against a panel of physicians each triaging identical cases in a representative patient validation set. Our results show that algorithm accuracy and safety indicators surpass all individual pulmonologists in both identifying exacerbations and predicting the consensus triage in a 101 case validation set. The algorithm is also the top performer in sensitivity, specificity, and ppv when predicting a patient's need for emergency care.

SUBMITTER: Swaminathan S

PROVIDER: S-EPMC5699810 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A machine learning approach to triaging patients with chronic obstructive pulmonary disease.

Swaminathan Sumanth S Qirko Klajdi K Smith Ted T Corcoran Ethan E Wysham Nicholas G NG Bazaz Gaurav G Kappel George G Gerber Anthony N AN

PloS one 20171122 11

COPD patients are burdened with a daily risk of acute exacerbation and loss of control, which could be mitigated by effective, on-demand decision support tools. In this study, we present a machine learning-based strategy for early detection of exacerbations and subsequent triage. Our application uses physician opinion in a statistically and clinically comprehensive set of patient cases to train a supervised prediction algorithm. The accuracy of the model is assessed against a panel of physicians ...[more]

PMID: 29166411

Similar Datasets

Project description:BACKGROUND:Chronic obstructive pulmonary disease (COPD) is combination of progressive lung diseases. The diagnosis of COPD is generally based on the pulmonary function testing, however, difficulties underlie in prognosis of smokers or early stage of COPD patients due to the complexity and heterogeneity of the pathogenesis. Computational analyses of omics technologies are expected as one of the solutions to resolve such complexities. METHODS:We obtained transcriptomic data by in vitro testing with exposures of human bronchial epithelial cells to the inducers for early events of COPD to identify the potential descriptive marker genes. With the identified genes, the machine learning technique was employed with the publicly available transcriptome data obtained from the lung specimens of COPD and non-COPD patients to develop the model that can reflect the risk continuum across smoking and COPD. RESULTS:The expression levels of 15 genes were commonly altered among in vitro tissues exposed to known inducible factors for earlier events of COPD (exposure to cigarette smoke, DNA damage, oxidative stress, and inflammation), and 10 of these genes and their corresponding proteins have not previously reported as COPD biomarkers. Although these genes were able to predict each group with 65% accuracy, the accuracy with which they were able to discriminate COPD subjects from smokers was only 29%. Furthermore, logistic regression enabled the conversion of gene expression levels to a numerical index, which we named the "potential risk factor (PRF)" index. The highest significant index value was recorded in COPD subjects (0.56 at the median), followed by smokers (0.30) and non-smokers (0.02). In vitro tissues exposed to cigarette smoke displayed dose-dependent increases of PRF, suggesting its utility for prospective risk estimation of tobacco products. CONCLUSIONS:Our experimental-based transcriptomic analysis identified novel genes associated with COPD, and the 15 genes could distinguish smokers and COPD subjects from non-smokers via machine-learning classification with remarkable accuracy. We also suggested a PRF index that can quantitatively reflect the risk continuum across smoking and COPD pathogenesis, and we believe it will provide an improved understanding of smoking effects and new insights into COPD.

Project description:The study developed accurate explainable machine learning (ML) models for predicting first-time acute exacerbation of chronic obstructive pulmonary disease (COPD, AECOPD) at an individual level. We conducted a retrospective case-control study. A total of 606 patients with COPD were screened for eligibility using registry data from the COPD Pay-for-Performance Program (COPD P4P program) database at Changhua Christian Hospital between January 2017 and December 2019. Recursive feature elimination technology was used to select the optimal subset of features for predicting the occurrence of AECOPD. We developed four ML models to predict first-time AECOPD, and the highest-performing model was applied. Finally, an explainable approach based on ML and the SHapley Additive exPlanations (SHAP) and a local explanation method were used to evaluate the risk of AECOPD and to generate individual explanations of the model's decisions. The gradient boosting machine (GBM) and support vector machine (SVM) models exhibited superior discrimination ability (area under curve [AUC] = 0.833 [95% confidence interval (CI) 0.745-0.921] and AUC = 0.836 [95% CI 0.757-0.915], respectively). The decision curve analysis indicated that the GBM model exhibited a higher net benefit in distinguishing patients at high risk for AECOPD when the threshold probability was <0.55. The COPD Assessment Test (CAT) and the symptom of wheezing were the two most important features and exhibited the highest SHAP values, followed by monocyte count and white blood cell (WBC) count, coughing, red blood cell (RBC) count, breathing rate, oral long-acting bronchodilator use, chronic pulmonary disease (CPD), systolic blood pressure (SBP), and others. Higher CAT score; monocyte, WBC, and RBC counts; BMI; diastolic blood pressure (DBP); neutrophil-to-lymphocyte ratio; and eosinophil and lymphocyte counts were associated with AECOPD. The presence of symptoms (wheezing, dyspnea, coughing), chronic disease (CPD, congestive heart failure [CHF], sleep disorders, and pneumonia), and use of COPD medications (triple-therapy long-acting bronchodilators, short-acting bronchodilators, oral long-acting bronchodilators, and antibiotics) were also positively associated with AECOPD. A high breathing rate, heart rate, or systolic blood pressure and methylxanthine use were negatively correlated with AECOPD. The ML model was able to accurately assess the risk of AECOPD. The ML model combined with SHAP and the local explanation method were able to provide interpretable and visual explanations of individualized risk predictions, which may assist clinical physicians in understanding the effects of key features in the model and the model's decision-making process.

Project description:RationaleChronic obstructive pulmonary disease (COPD) is often unrecognized and untreated.ObjectivesTo develop a method for identifying undiagnosed COPD requiring treatment with currently available therapies (FEV1 <60% predicted and/or exacerbation risk).MethodsWe conducted a multisite, cross-sectional, case-control study in U.S. pulmonary and primary care clinics that recruited subjects from primary care settings. Cases were patients with COPD and at least one exacerbation in the past year or FEV1 less than 60% of predicted without exacerbation in the past year. Control subjects were persons with no COPD or with mild COPD (FEV1 ≥60% predicted, no exacerbation in the past year). In random forests analyses, we identified the smallest set of questions plus peak expiratory flow (PEF) with optimal sensitivity (SN) and specificity (SP).Measurements and main resultsPEF and spirometry were recorded in 186 cases and 160 control subjects. The mean (SD) age of the sample population was 62.7 (10.1) years; 55% were female; 86% were white; and 16% had never smoked. The mean FEV1 percent predicted for cases was 42.5% (14.2%); for control subjects, it was 82.5% (15.7%). A five-item questionnaire, CAPTURE (COPD Assessment in Primary Care to Identify Undiagnosed Respiratory Disease and Exacerbation Risk), was used to assess exposure, breathing problems, tiring easily, and acute respiratory illnesses. CAPTURE exhibited an SN of 95.7% and an SP of 44.4% for differentiating cases from all control subjects, and an SN of 95.7% and an SP of 67.8% for differentiating cases from no-COPD control subjects. The PEF (males, <350 L/min; females, <250 L/min) SN and SP were 88.0% and 77.5%, respectively, for differentiating cases from all control subjects, and they were 88.0% and 90.8%, respectively, for distinguishing cases from no-COPD control subjects. The CAPTURE plus PEF exhibited improved SN and SP for all cases versus all control subjects (89.7% and 78.1%, respectively) and for all cases versus no-COPD control subjects (89.7% and 93.1%, respectively).ConclusionsCAPTURE with PEF can identify patients with COPD who would benefit from currently available therapy and require further diagnostic evaluation. Clinical trial registered with clinicaltrials.gov (NCT01880177).

Dataset Information

A machine learning approach to triaging patients with chronic obstructive pulmonary disease.

Publications

A machine learning approach to triaging patients with chronic obstructive pulmonary disease.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets