Effects on Preferences of Violations of Procedural Invariance

Abstract

Background. In studies of health preferences, utilities for hypothetical health states cannot always be successfully measured. One marker for unsuccessful measurement is violation of "procedural invariance": when the ranking of two health states varies across assessment procedures. Using preference values based on unsuccessful mea surement may result in misinterpretation of patients' attitudes about health. Objective. The authors sought to determine whether people who violated procedural invariance had different preferences than people who satisfied it. Methods. They performed sec ondary analyses of three completed studies that used the same two assessment pro cedures, identifying participants who violated procedural invariance and comparing the mean standard gamble (SG) and visual analog scale (VAS) scores of violators and satisfiers. Participants. Experiment 1, 30 healthy volunteers and 30 patients with car diac arrhythmias; expenment 2, 139 patients with depressive illness; experiment 3, 98 family members of patients with schizophrenia. Results. Rates of violation of proce dural invariance ranged from 16% to 32%. Violation of procedural invariance was not associated with age, education level, race, or gender. Subjects with violations of pro cedural invariance had, in general, less ability to discriminate among states and less reliable VAS and SG measurements, and sometimes had different mean SG and VAS values. Conclusions. Violation of procedural invariance of preferences across scaling methods may be a signal for failure of the measurement process. Researchers should test for procedural invariance and consider reporting data separately for satisfiers and violators. Key words: utility; standard gamble; visual analog scale; cost-effectiveness analysis; decision analysis; computers; preferences. (Med Decis Making 1999;19: 473-481)