Logical inconsistency in comorbid health state preference assessment for prostate cancer

Abstract

6568 Background: With increasing need for efficient resource allocation, attention is turning to evidence-based decision making and cost-effectiveness analysis (CEA). Policy decisions depend on quality-adjusted outcomes from such analyses. Even small variations in the health state utilities used to quality adjust can have considerable effects, yet measuring utilities accurately is difficult. A particular challenge in the aging population with high cancer risk is maintaining logical consistency when evaluating comorbid utilities. We define this as ensuring that a single health state (SS) is rated no worse than a health state that adds another SS to that first one to make a comorbid health state (CS). Determining why or by whom this condition fails could improve elicitation practices and CEA for cancer and comorbid diseases. Methods: Men were surveyed at the time of prostate biopsy. SS and CS utility values were elicited via the time-trade off method. Demographic information, current health, and scores on anxiety subscales (HADS and MAX-PC) were collected also. The most prevalent health states associated with prostate cancer in the SEER database were chosen for assessment. CS were combinations of the SS “impotence” with each of incontinence, asymptomatic localized disease, and post-prostatectomy. Regression analysis showed the association of other survey elements to utility inconsistency. Mann-Whitney U-tests determined whether utility values differed by consistency. Results: 68% of respondents rated inconsistently at least once. Marriage and anxiety were correlated with rating inconsistently in univariate logistic analysis (p < 0.01, p < 0.05, respectively). Higher education was independently associated with more consistent answers (p < 0.05). Inconsistent raters had significantly lower utility values for all SS, but not for CS (p < 0.01). Conclusions: Over two thirds of men gave at least one logically inconsistent response across the three CS conditions. Being married, feeling anxious, and having less education were correlated with giving fewer consistent responses. Correcting these inconsistencies may change the results of quality-adjusted analyses. Strategies for eliciting more consistent responses are needed to accurately quality-adjust comorbid health states. No significant financial relationships to disclose.