A neural system dynamics modeling platform and its applications in randomized controlled trial data analysis

Hamid, Nadira; Sarkar, Joydeep; Redfors, Björn; Balani, Anisha; Ramaswamy, Rajagopalan; Ghosh, Abhijit; Alu, Maria; Crowley, Aaron; Zhang, Yiran; Leon, Martin B.; Stone, Gregg W.; Granada, Juan F.

doi:10.1016/j.imu.2021.100612

Cited by 1 publication

(1 citation statement)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Both models performed similarly on the hold-out dataset; we suspect in some instances, the dynamic model may outperform the static model if adequate follow-up data with a regular time interval was available. 27 Static models use data -such as medical history, index procedures, biomarkers and vitals at admission -from the index event and before as input to predict post-discharge MACE risk. such as real-time inference from dynamic models are largely similar to that of the static model.…”

Section: Use In Clinical Settingsmentioning

confidence: 99%

Major Adverse Cardiovascular Event Risk Prediction in Asian Patients After Myocardial Infarction: A Novel, Dynamic, Machine-learning Approach

Page

Tandi³

et al. 2023

J Asian Pac Soc Cardiol

Self Cite

View full text Add to dashboard Cite

Background: Cardiovascular disease remains the leading cause of death worldwide. Existing risk scores often focus on primary cardiovascular event prediction among mostly North American and European populations with predictors derived from baseline before and up to index events. Methods: We aim to build machine learning algorithms to model recurrent cardiovascular events among Asian post-MI patients via a retrospective observational cohort study. The study uses multicentre, real-world electronic health records from the Singapore Cardiovascular Longitudinal Outcomes Database (SingCLOUD), integrating nationwide clinical, administrative, laboratory, procedural and medication data from five public hospitals and 18 public outpatient clinics. Data from 4,575 patients admitted between 2011 and 2014 with MI as a primary diagnosis were included in the analysis. Their major adverse cardiovascular events (MACE) risk was modelled using a static approach with baseline and index event information and a dynamic approach which incorporates additional post-discharge data including medication adherence. Results: The static model achieved an area under the receiver operator curve (AUROC) of 0.77 (95% CI [0.72–0.82]) when predicting 1-year MACE risk and 0.80 (95% CI [0.76–0.85]) predicting 2-year MACE risk on a hold-out dataset (n=515); the dynamic model achieved a mean AUROC of 0.75 (95% CI [0.72–0.78]) in estimating 1-year MACE risk from 3, 6, 9 and 12 months post-index. Conclusion: We developed a focused risk model for an Asian population, incorporating post-MI time-variable information via the causal approach. The causal framework enables differentiation of immediate risk components from long-term disease progression, including disease management. The model outperforms existing benchmark risk scores and can be applied to simulate personalised disease trajectory and to inform treatment options.

show abstract

Section: Use In Clinical Settingsmentioning

confidence: 99%