Ensemble Age Models for Short Term Ischemic Stroke Risk Prediction
## Description
This is a modification of the project described below for a lab in Systems biology, digital twins, and AI.
Models created for the Precise4Q project to help in prevention of initial ischemic stroke. The ensemble age models are created from a set of logistic regression models that are designed to predict indivdiual 5-year ischemic stroke risk. The age models are designed to capture the non-proportionallity in risk factor contribution to an individual's stroke risk by age. For more information on the development of the age models see:
Hunter, E., & Kelleher, J. D. (2022). Age specific models to capture the change in risk factor contribution by age to short term primary ischemic stroke risk. Frontiers in Neurology, 13. Retrieved from https://www.frontiersin.org/article/10.3389/fneur.2022.803749 doi: 10.3389/fneur.2022.803749
...
...
@@ -12,13 +13,10 @@ Although the independent age models are shown to be better calibrated than a sin
## Usage
The ensemble_age_model.R script can be run in any R environment. The script will load in the four independent age models and weight the risk scores for an indiviudal by the distance from the individual's age to the distance of maxium and minimum ages used in model training and testing. The weighted risk scroes will be combined and a single average risk will be produced. The script produces odds ratios, relative risk and absolute risk.
The ensemble_age_model.R script can be run in any R environment. The script will load in the four independent age models and weight the risk scores for an indiviudal by the distance from the individual's age to the distance of maximum and minimum ages used in model training and testing. The weighted risk scroes will be combined and a single average risk will be produced. The script produces odds ratios, relative risk and absolute risk.
The models use the following risk factors: sex (coded as 1 = male and 2 = female), age, BMI, average cigarettes per day, diastolic blood pressure, systolic blood pressure, diabetes (0 = no diabetes, 1 = diabetes), and Atrial fibrillation diagnosed before stroke (0 = no atrial fibrillation, 1 = atrial fibrillation). To predict one or more individual's risk scores from the ensemble age model, a csv file with column headings SEX, AGE, BMI, CPD, DBP, SBP,DMRX, AF_beforestroke is needed. Each row of the file should correspond to the risk factors of different individual or the same individual's risk factors at a different ages.
## Support
For support contact elizabeth.hunter@tudublin.ie
## Authors and acknowledgment
This work was dones as part of the Precise4Q project that recieved funding from the EU's Horizon 2020 research and innovation programme under grant agreement No. 777107.
