A Simple Method for Heuristic Modeling of Expert Knowledge in Chronic Disease: Identification of Prognostic Subgroups in Rheumatology

Identification of prognostic subgroups is of key clinical interest at the early stages of chronic disease. The aim of this study is to examine whether representation of physicians' expert knowledge in a simple heuristic model can improve data mining methods in prognostic assessments of patients with rheumatoid arthritis (RA). Five rheumatology consultants' experiences of clinical data patterns among RA patients, as distinguished from healthy reference populations, were formally represented in a simple heuristic model. The model was used in K-mean-clustering to determine prognostic subgroups. Cross-sectional validation using physician's global assessment scores indicated that the simple heuristic model performed better than crude data made in identification of prognostic subgroups of RA patients. A simple heuristic model of experts' knowledge was found useful for semi-automatic data mining in the chronic disease setting. Further studies using categorical baseline data and prospective outcome variables are warranted and will be examined in the Swedish TIRA-program.