Graphing comorbidity.

Abstract

INTRODUCTION: Coined in the1970s by physician, epidemiologist and applied statistician Prof. Alvan Feinstein MD (1925-2001), the term "comorbidity" may be defined as the presence of two or more disorders, such as alcohol use disorder and depression, or colorectal cancer and chronic obstructive pulmonary disease. Comorbidity or "multimorbidity" is generally associated with increased use of hospital services and poorer prognosis, due to additive (higher number of disorders generally indicates poorer health) and multiplicative (synergistic or interactive) relationships between the disorders. An example of such a synergistic relationship is patients with lower limb injuries and comorbid diabetes mellitus not being able to manage and control the latter, due to restricted movement. Instruments based upon weighted counts of chronic disorders such as the Charlson, and Elixhauser indices are predictive of mortality and other adverse outcomes but provide little information on specific synergistic co-relationships between comorbid disorders. Various regression methods, as well as decision trees (a favourite tool of Dr. Feinstein), can quantify such relationships and predict outcome, however there are very few methods of graphically representing co-morbidity. AIMS: To present a simple and easy to understand method of graphing comorbidity. METHADOLOGY: "Heat maps", a twentieth century development of a nineteenth century concept, can show in vivid colour or in monochrome the overall strength of relationships between two or more disorders in a sample of patients. Heat maps are less complex than network graphs, provide more information than pie charts and Venn diagrams and are readily implemented in R, SPSS, Stata and Microsoft Excel. Graphs such as heat maps allow effective communication with clinicians and may act as a conduit to further analysis, including factor analysis, latent class analysis, and self-organizing maps. RESULTS: Illustrative published medical data and custom charts and graphs will be presented. [See Poster].