OBJECTIVES:   To assess construction of the primary HUI3 scoring system. METHODS:  Mean visual analogue scale (VAS) and standard gamble (SG) scores were collected in 2 face-to-face interview surveys: modeling survey (MS) for the Person-Mean utility scoring model to calculate community utilities for health states; direct survey (DS) for validation of the Person-Mean scoring model.  Survey results are assessed for completeness and consistency.  Completeness is evaluated by response rates and consistency by health state rankings. RESULTS: Completed interviews were obtained from 65% of contacted eligible subjects. Demographic distributions are similar to the underlying general population. There are 256 respondents in MS and 248 in DS.  MS and DS respondents use 71% of HUI3 attribute levels in describing their own health status.  Each of the 8 HUI3 attributes are reported by 35 or more respondents as being important in their preference measurements: pain (49%); vision (37%); cognition (34%); emotion (28%); ambulation (28%); hearing (17%); dexterity (17%); speech (7%). More than 70% of respondents focused on 2+ attributes. Other important preference measurement factors are self-care ability (89% of respondents), family life (76%), happiness of others (69%), ability to work current job (61%), leisure activities (42%).  83% of respondents report the interviewing did not change their opinions about the health states.  Consistency of health state rankings by mean VAS and SG scores between MS and DS was 100%: PH >MA>MB>MC>Dead>Pits. MS had a missing data rate of 0.29% (17/5920) for VAS and 0.10% (1/1024) for SG. CONCLUSIONS: The Person-Mean HUI3 utility function is founded on a survey that was well-constructed in terms of community and attribute representation, consideration of multiple attributes and day-to-day impacts, stable opinions, and consistency of health state rankings.  This evidence supports use of the primary HUI3 utility function for group-level analyses, such as allocation of societal resources.