Computational signatures for post-cardiac arrest trajectory prediction: Importance of early physiological time series

Kim, Han B.; Nguyen, Hieu; Jin, Qingchu; Tamby, Sharmila; Romer, Tatiana Gelaf; Sung, Eric; Liu, Ran; Greenstein, Joseph L.; Suárez, José I.; Storm, Christian; Winslow, Raimond L.; Stevens, Robert

doi:10.1016/j.accpm.2021.101015

Cited by 10 publications

(8 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The DL-based survival method Dynamic-DeepHit and the data concatenation method had a slightly lower performance. This finding agrees with previous reports showing trend analysis from the time series is more important in discriminating cases from controls than just using the raw time series as inputs in predicting cardiac arrest [ 55 , 56 ]. However, these studies are classification problems, our study extends the application of time series massive feature extraction to survival analysis.…”

Section: Discussionsupporting

confidence: 92%

Multivariate longitudinal data for survival analysis of cardiovascular event prediction in young adults: insights from a comparative explainable study

Nguyen

Vasconcellos

Keck

et al. 2023

BMC Med Res Methodol

Self Cite

View full text Add to dashboard Cite

Background Multivariate longitudinal data are under-utilized for survival analysis compared to cross-sectional data (CS - data collected once across cohort). Particularly in cardiovascular risk prediction, despite available methods of longitudinal data analysis, the value of longitudinal information has not been established in terms of improved predictive accuracy and clinical applicability. Methods We investigated the value of longitudinal data over and above the use of cross-sectional data via 6 distinct modeling strategies from statistics, machine learning, and deep learning that incorporate repeated measures for survival analysis of the time-to-cardiovascular event in the Coronary Artery Risk Development in Young Adults (CARDIA) cohort. We then examined and compared the use of model-specific interpretability methods (Random Survival Forest Variable Importance) and model-agnostic methods (SHapley Additive exPlanation (SHAP) and Temporal Importance Model Explanation (TIME)) in cardiovascular risk prediction using the top-performing models. Results In a cohort of 3539 participants, longitudinal information from 35 variables that were repeatedly collected in 6 exam visits over 15 years improved subsequent long-term (17 years after) risk prediction by up to 8.3% in C-index compared to using baseline data (0.78 vs. 0.72), and up to approximately 4% compared to using the last observed CS data (0.75). Time-varying AUC was also higher in models using longitudinal data (0.86–0.87 at 5 years, 0.79–0.81 at 10 years) than using baseline or last observed CS data (0.80–0.86 at 5 years, 0.73–0.77 at 10 years). Comparative model interpretability analysis revealed the impact of longitudinal variables on model prediction on both the individual and global scales among different modeling strategies, as well as identifying the best time windows and best timing within that window for event prediction. The best strategy to incorporate longitudinal data for accuracy was time series massive feature extraction, and the easiest interpretable strategy was trajectory clustering. Conclusion Our analysis demonstrates the added value of longitudinal data in predictive accuracy and epidemiological utility in cardiovascular risk survival analysis in young adults via a unified, scalable framework that compares model performance and explainability. The framework can be extended to a larger number of variables and other longitudinal modeling methods. Trial registration ClinicalTrials.gov Identifier: NCT00005130, Registration Date: 26/05/2000.

show abstract

Section: Discussionsupporting

confidence: 92%

Multivariate longitudinal data for survival analysis of cardiovascular event prediction in young adults: insights from a comparative explainable study

Nguyen

Vasconcellos

Keck

et al. 2023

BMC Med Res Methodol

Self Cite

View full text Add to dashboard Cite

show abstract

“…The DL-based survival method Dynamic-DeepHit and the data concatenation method had a slightly lower performance. This nding agrees with previous reports showing trend analysis from the time series is more important in discriminating cases from controls than just using the raw time series as inputs in predicting cardiac arrest [55,56]. However, these studies are classi cation problems, our study extends the application of time series massive feature extraction to survival analysis.…”

Section: Added Predictive Value Of Longitudinal Datasupporting

confidence: 91%

High-Dimensional Multivariate Longitudinal Data for Survival Analysis of Cardiovascular Event Prediction in Young Adults: Insights from a Comparative Explainable Study

Nguyen

Vasconcellos

Keck

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Background: High-dimensional, longitudinal data are under-utilized for survival analysis compared to cross-sectional data (CS - data collected once across cohort). Particularly in cardiovascular risk prediction, despite available methods of longitudinal data analysis, the value of longitudinal information has not been established in terms of improved predictive accuracy and clinical applicability. Methods: We investigated the value of longitudinal data over and above the use of cross-sectional data via 6 distinct modeling strategies from statistics, machine learning, and deep learning that incorporate high-dimensional repeated measures for survival analysis of the time-to-cardiovascular event in the Coronary Artery Risk Development in Young Adults (CARDIA) cohort. We then examined and compared the use of model-specific interpretability methods (Random Survival Forest Variable Importance) and model-agnostic methods (SHapley Additive exPlanation (SHAP) and Temporal Importance Model Explanation (TIME)) in cardiovascular risk prediction using the top-performing models. Results: In a cohort of 3539 participants, longitudinal information from 35 variables that were repeatedly collected in 6 exam visits over 15 years improved subsequent long-term (17 years after) risk prediction by up to 8.3% in C-index compared to using baseline data (0.78 vs. 0.72), and up to approximately 4% compared to using the last observed CS data (0.75). Time-varying AUC was also higher in models using longitudinal data (0.86-0.87 at 5 years, 0.79-0.81 at 10 years) than using baseline or last observed CS data (0.80-0.86 at 5 years, 0.73-0.77 at 10 years). Comparative model interpretability analysis revealed the impact of longitudinal variables on model prediction on both the individual and global scales among different modeling strategies, as well as identifying the best time windows and best timing within that window for event prediction. The best strategy to incorporate longitudinal data for accuracy was time series massive feature extraction, and the easiest interpretable strategy was trajectory clustering. Conclusions: Our analysis demonstrates the added value of longitudinal data in predictive accuracy and epidemiological utility in cardiovascular risk survival analysis in young adults via a unified, scalable framework that compares model performance and explainability. The framework can be extended to a larger number of variables and other longitudinal modeling methods. Trial registration ClinicalTrials.gov Identifier: NCT00005130, Registration Date: 26/05/2000

show abstract

“…Recently, there has been growing interest in the possibility of leveraging machine learning (ML) to support classification and prediction tasks after cardiac arrest. We demonstrated that ML applied to physiological data recorded in the early phase after CA resuscitation could accurately and reliably predict postcardiac arrest outcomes [9]. Here we build on this prior work to predict outcomes associated with TTM.…”

Section: Introductionmentioning

confidence: 89%

Hyperacute Prediction of Targeted Temperature Management Outcome After Cardiac Arrest

Hsu

Kim

Gong

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Introduction:Targeted temperature management (TTM) has been associated with greater likelihood of neurological recovery among comatose survivors of cardiac arrest. However, the efficacy of TTM is not consistently observed, possibly due to heterogeneity of therapeutic response. The aim of this study is to determine if models leveraging multi-modal data available in the first 12 hours after ICU admission (hyperacute phase) can predict short-term outcome after TTM.Methods:Adult patients receiving TTM after cardiac arrest were selected from a multicenter ICU database. Predictive features were extracted from clinical, physiologic, and laboratory data available in the hyperacute phase. Primary endpoints were survival and favorable neurological outcome, determined as the ability to follow commands (motor Glasgow Coma Scale [mGCS] of 6) upon discharge. Three machine learning (ML) algorithms were trained: generalized linear models (GLM), random forest (RF), and gradient boosting (XG). Models with optimal features from forward selection were 10-fold cross-validated and resampled 10 times.Results:Data were available on 310 cardiac arrest patients who received TTM, of whom 183 survived and 123 had favorable neurological outcome. The GLM performed best, with an area under the receiver operating characteristic curve (AUROC) of 0.86 ± 0.04, sensitivity 0.75 ± 0.09, and specificity 0.77 ± 0.07 for the prediction of survival and an AUROC of 0.85 ± 0.03, sensitivity 0.71 ± 0.10, and specificity 0.80 ± 0.12 for the prediction of favorable neurological outcome. Features most predictive of both endpoints included lower serum chloride concentration, higher serum pH, and greater neutrophil counts.Conclusion:In patients receiving TTM after cardiac arrest, short-term outcomes can be accurately discriminated using ML applied to data routinely collected in the first 12 hours after ICU admission. With validation, hyperacute prediction could enable personalized approach to clinical decision-making in the post-cardiac arrest setting.

show abstract

Computational signatures for post-cardiac arrest trajectory prediction: Importance of early physiological time series

Cited by 10 publications

References 28 publications

Multivariate longitudinal data for survival analysis of cardiovascular event prediction in young adults: insights from a comparative explainable study

Multivariate longitudinal data for survival analysis of cardiovascular event prediction in young adults: insights from a comparative explainable study

High-Dimensional Multivariate Longitudinal Data for Survival Analysis of Cardiovascular Event Prediction in Young Adults: Insights from a Comparative Explainable Study

Hyperacute Prediction of Targeted Temperature Management Outcome After Cardiac Arrest

Contact Info

Product

Resources

About