Purpose: To evaluate the combined effect of marker states and a prompt on the reliability and validity of direct utility assessment.
Methods: In a structured interview, 141 prostate cancer (PC) patients provided rating scale (RS) and standard gamble (SG) utilities for their "own health" (RS-/SG-). Following self-completion of comparison instruments (two generic utility and one disease-specific psychometric), they answered a checklist of PC-related items (a prompt to consider self-health) and provided utilities for self-health and mild and severe PC marker states (RS+/SG+). The interview was repeated 5 weeks later, but without comparison instruments. Using Bayesian modeling, we computed and compared correlation coefficients to assess RS and SG test-retest reliability and validity and the effects of the prompt and marker states.
Results: RS and SG had acceptable test-retest reliability (intraclass correlation coefficients = 0.57-0.63). The prompt and marker states decreased the reliability of the RS by 0.01 (from 0.58 to 0.57) but increased the reliability of the SG by 0.05 (from 0.58 to 0.63). The probability that the reliability of the SG+ was greater than that of the SG- was very high (0.96). Correlations with comparison instruments were higher by 0.01-0.06 for RS+ vs RS-, and higher by 0.03-0.06 for SG+ vs SG-. The probabilities that the prompt and marker states improved validity ranged from 0.55 to 0.74 (RS), and from 0.61 to 0.70 (SG).
Conclusions: A self-health description prompt and marker states modestly improved the reliability and validity of direct utility elicitation.