Statistical evaluation of adding multiple risk factors improves Framingham stroke risk score

Area of Science:

• Cardiovascular epidemiology
• Biostatistics
• Public health

Background:

• The Framingham Stroke Risk Score (FSRS) is a widely used tool for assessing stroke risk.
• The original FSRS model omits several recognized stroke risk factors.
• This study addresses the need for a more comprehensive stroke risk prediction model.

Purpose of the Study:

• To develop and validate a novel stroke risk prediction model (NEW-STROKE).
• To integrate additional significant risk factors into stroke risk assessment.
• To compare the performance of the NEW-STROKE model against the FSRS using real-world data.

Main Methods:

• Utilized the Atherosclerosis Risk in Communities (ARIC) longitudinal dataset.
• Developed a new risk prediction model by synthesizing existing models with new risk factors.
• Evaluated model performance using discrimination (Modified C-statistics), calibration (Hosmer-Lemeshow Test), and reclassification (NRI).

Main Results:

• The NEW-STROKE model exhibited higher Modified C-statistics, indicating improved discrimination.
• The NEW-STROKE model showed better calibration with smaller Hosmer-Lemeshow chi-square values.
• Significantly positive classless and class Net Reclassification Improvement (NRI) values demonstrated superior reclassification ability for NEW-STROKE.

Conclusions:

• The NEW-STROKE model, incorporating seven additional risk factors, significantly outperforms the original FSRS.
• The inclusion of these seven factors markedly enhances stroke risk prediction accuracy.
• This improved model offers a more precise tool for identifying individuals at high risk of stroke.