Development of an audiovisual taxonomy of pain behavior: reliability and discriminant validity

This article represents the final stages in the development of a pain behavior taxonomy that was empirically derived from patients', spouses', and health care providers' nominations of pain behaviors. Seventy-one chronic low back pain patients and 40 controls were videotaped while undergoing a structured sequence of movements involving sitting, standing, walking, bending, and exercising, and a brief standardized interview component. Two independent observers rated each subject's videotape employing a 16-category observational rating system that included both verbal and nonverbal pain behaviors. Seven of the pain behavior categories, including guarding, bracing, position shifts, partial movement, grimacing, limitation statements, and sounds, had acceptable reliabilities and frequencies and were therefore selected for inclusion in a discriminant analysis. The subject sample was randomly divided into two equal groups, and a stepwise discriminant function analysis on the first subsample indicated that partial movement, limitation statements, sounds, and position shifts accounted for 75% of the variance in group membership and correctly classified 94.4% of the patients and 95.2% of the controls. On cross-validation, the same four categories identified in the initial discriminant function correctly classified 88.9% of the subjects in the second subsample. The utility of this pain behavior observational recording system is discussed in reference to both patient assessment and treatment outcome research.