Rationalizing back pain: the development of a classification system through cluster analysis.

Abstract

OBJECTIVE: The clinical variations in undifferentiated back pain pose problems for those attempting to develop strategies for care. The objective of this work was to test a methodology for the experimental generation of clinical subgroups of patients with such a complaint, so as to assist more structured study of its natural history and response to treatment. DESIGN: Cluster analysis of dichotomous symptomatic variables from computer-based case histories of three patient cohorts. SETTING: Chiropractic and Orthopedic outpatient clinics. PATIENTS: Three cohorts of new patients with back pain whose symptoms were recorded in a highly standardized way using an interactive computer interview system. CRITERIA ASSESSED: Twenty-four aggravating, relieving and cyclic features of the patients' back complaints assessed for degrees of association and formation of reproducible clusters. RESULTS: Two main patient categories were discerned: one with mechanical features and one that was cyclic. Most patients were assignable to a group. Groupings were largely consistent across all three cohorts and were not related to patient demographics. CONCLUSION: Reproducible and easily-recognized clinical subgroups of back pain patients are possible by cluster analysis using dichotomous case-history variables. More definitive categorization should be obtainable by refining the variable selection and repeating the analysis for additional patient cohorts. These subgroups have the potential to increase the relevance of natural history studies and clinical trials to the day-to-day management of the problem.