Health-related quality of life in adults reporting arthritis: analysis from the National Health Measurement Study

The six HRQOL indexes are described in Fryback et al. [23]. Briefly, the EQ-5D, which we computed using US weights [34], is a five-question index asking about mobility, self-care, restrictions of usual activities, pain/discomfort and anxiety/depression. The SF-6D [27] is constructed using 11 of 36 questions from the widely used SF-36v2™ [35]. The SF-6D uses questions concerning physical function, limitations in role functioning, social function, pain, mental health, and vitality. The HUI2 [25] and HUI3 [26] are related measures computed from the Health Utilities Index questionnaire. The HUI2 defines health status on six domains (sensation, mobility, emotion, cognition, self-care, and pain—we excluded an optional fertility dimension). HUI3 addresses vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain. The QWB-SA summarizes functioning and mobility, physical activity, and social activity, but also collects extensive information about 58 symptoms/problem complexes. The four domains into a single score QWB-SA [28]. Finally, the HALex [29] is an index-constructed post hoc to use data collected by the recurring NHIS. HALex combines level of physical activity restriction with self-rated health (excellent, very good, good, fair, poor) and derives single weighted score using preferences from the HUI.

Three indexes, EQ-5D, HUI2, and HUI3 allow scores less than 0.0 to reflect health states considered worse than dead by the general population samples from which those indexes’ weighting schemes were derived.

To allow analyses to reflect the target US population, observations in the NHMS data set are supplied with survey weights computed first as the inverse sampling probability for each participant based on the sampling scheme and then post-stratified to the US Census 2000 population by age, gender, and race [23]. Descriptive statistics (Table 1) and mean estimates of the six HRQOL indexes (Table 2, Fig. 1, “Appendix”) incorporated these weights using the “svyset” collection of commands available in STATA10.2 (Stata Corp., College Station, TX).

Unadjusted estimates of the six HRQOL indexes, stratified by age, gender, and arthritis category, were computed as the survey-weighted mean and standard error (SE) of the index for each age/gender/arthritis combination (Table 2). Age-groups were defined as 35–44, 45–54, 55–64, 65–74, and 75–89. Estimates stratified by age-group and arthritis category were then recomputed adjusting for demographic variables and comorbidities selected based on Table 1 (see Fig. 1, “Appendix”). Under this approach, the HRQOL index is modeled as a continuous variable using survey-weighted linear regression. Predictors in the model consist of dummy variables for age-group and arthritis category, the interaction of age-group and arthritis category, and the following: gender, race (White, African American, Other), level of education (less than high school, high school, some post-high school, 4-year college degree or higher), household income (<$20,000, $20,000–$34,999, $35,000–$74,999, >$75,000), body mass index (BMI) as a continuous variable, and the presence or absence of the following self-reported health care professional-diagnosed comorbidities: coronary heart disease, diabetes, stroke, COPD, and sleeping disorder. We did not adjust for depression as mental health is represented in each of 6 HRQOL indexes. The following covariates were centered to the US Census 2000 population proportions where possible and to survey means and proportions from NHMS when census data were not available (centering values provided in parentheses): male gender (0.474), BMI (27.6), history of coronary heart disease (0.092), history of stroke (0.039), diabetes (0.122), COPD (0.131), and sleeping disorder (0.080). For k-level categorical variables, k-1 indicator variables each centered to the population or survey proportion for the appropriate level were incorporated into the model as follows: race (African American: 0.105, other race: 0.086), education (high school: 0.286, some post-high school: 0.273, 4-year college degree or higher: 0.244), and income level ($20,000–$34,999: 0.183, $35,000–$74,999: 0.353, $75,000+: 0.249). Estimates of HRQOL mean and SE of the mean for each age-group/arthritis category were found by applying the appropriate post-estimation linear contrast involving the intercept term, dummy variables for age-group and arthritis category, and age-group-by-arthritis interaction terms.