Age Specific Models to Capture the Change in Risk Factor Contribution by Age to Short Term Primary Ischemic Stroke Risk

Abstract

Age is one of the most important risk factors when it comes to stroke risk prediction. However, including age as a risk factor in a stroke prediction model can give rise to a number of difficulties. Age often dominates the risk score, and also not all risk factors contribute proportionally to stroke risk by age. In this study we investigate a number of common stroke risk factors, using Framingham heart study data from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center to determine if they appear to contribute proportionally by age to a stroke risk score. As we find evidence that there is some non-proportionality by age, we then create a set of logistic regression risk models that each predict the 5 year stroke risk for a different age group. The age group models are shown to be better calibrated when compared to a model for all ages that includes age as a risk factor. This suggests that to get better predictions for stroke risk it may be necessary to consider alternative methods for including age in stroke risk prediction models that account for the non-proportionality of the other risk factors as age changes.

Digital twins and hybrid modelling for simulation of physiological variables and stroke risk

“…The independent age-specific models, without the ensemble approach, were evaluated in (31). However, since we do a new ensemble-based combination of these age-specific models, it is necessary to look at the evaluation metrics of the ensemble model.…”

“…One of the new models (Figure 4) describes blood pressure as a function of age (Figure 5-7) and medications (Figure 7), and can both agree with estimation data (Figure 5,7), and correctly predict validation data (Figure 6). The second new model is a modification of the stroke risk model from (31), using a new ensemble approach (Figure 8). This ensemble approach was introduced to smooth out discontinuities, when moving from one age group model to another, and we show that the performance of the new model is as good as the previous model (31), both regarding discrimination (Table 1) and calibration (Table 2).…”

“…The second new model is a modification of the stroke risk model from (31), using a new ensemble approach (Figure 8). This ensemble approach was introduced to smooth out discontinuities, when moving from one age group model to another, and we show that the performance of the new model is as good as the previous model (31), both regarding discrimination (Table 1) and calibration (Table 2). We illustrate how the hybrid model can be used, by simulating and comparing different lifestyle scenarios: for example comparing the health trajectory for an individual who gains weight and develops diabetes (Figure 9, scenario 1) with that of an individual who loses weight and thereby improves their glucose meal response (Figure 9-10, scenario 2); other scenarios considered include an individual who gains weight but take blood pressure medication (Figure 10, scenario 3), an individual who smokes and develops atrial fibrillation (Figure 11, scenario 4), and an individual who loses weight and takes blood pressure medication (Figure 11, scenario 5).…”

“…The used machine learning risk model is based on a set of four independent logistic regression models that are designed to predict the 5-year risk of initial ischemic stroke for an individual, belonging to four different age groups, as already presented in (31). There is a logistic regression model for each of the following four age groups: under 50, 50-59, 60-69 and over 70.…”

“…Age-specific models have been shown to capture the non-proportionality of risk factors by age and are better calibrated to the younger and older populations compared to a model that is trained on all ages with age included as a risk factor. Thus, age-specific models are particularly suited to this hybrid modelling approach that aims to calculate the risk based on simulated scenarios across time, as they will identify and use the risk factors that are most important to the individual's stroke risk at a given time point (31).…”