A review of risk concepts and models for predicting the risk of primary stroke

Abstract

Predicting an individual's risk of primary stroke is an important tool that can help to lower the burden of stroke for both the individual and society. There are a number of risk models and risk scores in existence but no review or classification designed to help the reader better understand how models differ and the reasoning behind these differences. In this paper we review the existing literature on primary stroke risk prediction models. From our literature review we identify key similarities and differences in the existing models. We find that models can differ in a number of ways, including the event type, the type of analysis, the model type and the time horizon. Based on these similarities and differences we have created a set of questions and a system to help answer those questions that modelers and readers alike can use to help classify and better understand the existing models as well as help to make necessary decisions when creating a new model.

Keywords: epidemiology; machine learning; predictive modeling; risk; stroke.

Figure 1

Representation of absolute risk, relative risk, and odds ratios. These are three of the common risk or association types presented in modeling studies and are often used interchangeably or misinterpreted when there are distinct differences between them. Absolute risk is often presented as a percentage and is the number of individuals with an event, number of strokes, over the total number of individuals in a group, number of strokes and non-strokes. Relative risk is a ratio of absolute risks, often treatment over control. Odds ratios are a measure of association rather than a risk. Odds are the number of times an event occurred in a group, number of strokes, over the number of times the event did not occur, number of non-strokes. Odds ratios are a ratio of two odds, often the ratio of the odds in a treatment group over the ratio of odds in a control group.

Figure 2

A visualization of the interconnections between analysis focus, event type, time horizon and model type that are often found in the literature. Each of the four characteristics is split into two main decisions to be made when creating a model (e.g., long- or short-time horizon). A decision in each category will often help to decide on one of the other categories (e.g., deciding to model stroke in event type will likely lead to the use of a short time horizon). However, in some cases, the decisions do not restrict any other categories (e.g., deciding to look at risk factors in the analysis does not restrict the model type).