Using a large nationally representative population-based survey, we show that adults with self-reported arthritis have lower HRQOL scores than adults who did not report arthritis after adjusting for age and sex and even after adjusting for several covariates. This was seen across the six HRQOL instruments. In addition, our study provides important national age-related reference averages for six HRQOL indexes for adults with self-reported arthritis that can be used in future decision and cost-effectiveness analyses.
Arthritis is very common in age groups 35–89 years and a major contributor to HRQOL in the US population. The prevalence of arthritis has been steadily increasing [1
]. We found a 31% prevalence of self-reported arthritis by the NHMS study. Previous surveys that differ in age-groups targeted and how arthritis was defined have shown similar prevalence. In the Medical Expenditures Panel Survey Household Component (MEPS-HC; computer-assisted personal interviewing), the prevalence of arthritis in 35–89 age-group was 30% in 2003 [37
] (Table ). In two other national representative surveys, the US Valuation of the EuroQol EQ-5D Health States Survey in 2002 (USVEQ; self-administered with interviewer present) [24
] and Joint Canada/United States Survey of Health from 2002 to 2003 (JCUSH; telephone survey) [38
], the prevalence of self-reported arthritis for ages 35–89 was consistently estimated at around 30% (Table ).
One of our interesting results is the declining HRQOL scores in the 35–64 age-group seen in both arthritis and non-arthritis groups and then a trend toward an improvement in their HRQOL in the 65–74-year age-group. This interesting finding was also reported for the whole cohort in the original publication from the NHMS [23
]. Fryback et al. suggest that this may relate to reporting bias of poor HRQOL in 55–64 age-group (baby boomers) or greater HRQOL in 65–74 age-group (recent retirees). Another potential reason may be selective non-participation in the survey by age and health around retirement age. Perhaps, more healthy people in the 55–64 age-group may be too busy, or perhaps, less healthy people in the 65–74 group may be too ill to participate.
Our data have important implications for public policy. First, our analysis provides age-related HRQOL scores for six commonly used measures that can be used in future decision and cost-effectiveness analyses. The US Public Health Service Panel on Cost-Effectiveness in Health and Medicine [39
] recommended using HRQOL scores based on preference weights derived from the general public, rather than from patients, for cost-effectiveness analyses. The indexes used here meet this requirement. We provide population-based estimates of HRQOL burden for each age-group-by-arthritis stratum. Second, previous studies have found that the minimally important difference in HRQOL preference-based scores—the smallest difference in scores that patients perceive as beneficial [40
]—is about 0.03, with a range from 0.01 to 0.10 [41
]. The differences in HRQOL scores between arthritis and non-arthritis groups in our study exceed the minimally important difference and are thus clinically meaningful. Third, as previously reported, each of the six HRQOL measures provides different scores associated with the impact of arthritis, thus choice of measure can affect estimates of quality-adjusted life years (QALYs) gained by an intervention and thus different incremental cost-effectiveness estimates. In an analysis of treatment of rheumatoid arthritis with combination of infliximab and methotrexate versus methotrexate alone, Marra et al. [44
] showed that the four HRQOL measures (EQ-5D, HUI2, HUI3, and SF-6D) provided different QALYs and therefore different incremental cost-effectiveness scores. In their analysis, HUI3 produced the largest incremental QALYs gain followed by EQ-5D, HUI2, and SF-6D. It may be necessary for analysts doing cost-effectiveness analyses to agree upon a common measure of HRQOL in order to standardize analytic results.
Our survey is not without limitations. NHMS was administered using a random digital dial telephone survey, which may have had somewhat higher response rates among people who are more educated and/or with higher household incomes compared with US Census figures for the year 2000. Thus, the NHMS may have reached somewhat more healthy persons [23
]. Although the question eliciting arthritis diagnosis has been widely used in similar surveys, cases were not verified by examination or medical records. People responding to telephone surveys tend to report slightly better health than those responding to self-administered, paper, and pencil questionnaires [45
]. Telephone-based health surveys were apparently little affected by cell phone usage in the timeframe of the NHMS [46
In conclusion, this study provides important national reference averages for six HRQOL indexes for adults with self-reported arthritis for future decision analysis and cost-effectiveness analysis.