Can cied-derived physical activity provide as a biomarker to predict all-cause-mortality? validation of previous predictive modelling approaches.

Physical activity (PA) is a well-established determinant of health outcomes yet monitoring deterioration of patients' health using cardiac implantable electronic device (CIED)-derived PA levels lacks external validation when predicting mortality. The primary aim was to investigate the relationship between CIED-derived PA and all-cause mortality. The secondary aim was to replicate and externally validate predictive models incorporating PA to evaluate their performance in predicting mortality. Methods: A retrospective longitudinal analysis was conducting using CIED-derived PA data in 513 cardiac patients to measure the prediction of all-cause-mortality. External validation was conducted by replicating eight previous predictive modelling approaches and measured discrimination and calibration scores across each model. PA was an independent predictor of all-cause mortality with low PA having higher risk of survival probability in univariate and multivariate analysis (p<0.001) when stratified by baseline PA levels. Predictive modelling approaches scored modest discrimination in univariate analysis (C-index = 0.65), good to excellent in multivariate cox regressions (C-index = 0.71 to 0.84) and perfect calibrations scores across all models (calibration slope = 1.00). Overall, baseline CIED-derived PA, stratified at baseline successfully predicts risk of mortality. Additionally, replication and external validation of predictive models incorporating PA demonstrated modest to excellent discrimination.

Author dc:creator

• Cowie, Greig

Advisor dc:contributor.advisor

• A. Hall and P. Swinton

• Physical activity (PA)
• Cardiac implantable electronic devices (CIEDs)
• All-cause mortality
• Predictive modelling
• External validation
• Cardiovascular disease

Language dc:language

en

Identifier

oai:rgu-repository.worktribe.com:3463773
https://doi.org/10.48526/rgu-wt-3463773

OAI identifier oai:identifier

oai:rgu-repository.worktribe.com:3463773