An Information Visualization Approach to Classification and Assessment of Diabetes Risk in Primary Care

Chronic disease risk assessment is a common information processing task performed by primary care physicians with many at-risk patients. However, effectively integrating information about many risk factors across many patients is cognitively difficult. Methods for visualizing multidimensional data may augment clinical disease risk assessment by providing reduceddimensional displays which stratify patient data points according to risk level while providing additional insight into clinically important individual risk factor variables. This study combines medical evidence, dimensionality reduction techniques and information visualization to develop a new framework for visually classifying and interpreting patient data. This framework is then explored and analytically validated using a unique health information database from the American Diabetes Association that contains risk predictions made by the Archimedes model. Results show that the framework may generate models which visually classify a large patient population with accuracy comparable to common statistical methods. Further, the visualizations provide rich displays that give insight into (i) the relative importance of individual risk factors, (ii) confidence in individual patient risk predictions and (iii) overall distributions of risk in a population. The proposed approach may produce models that can be embedded in health information systems to provide interactive visual analysis tools that support physician decision making.