In agreement with prior research [20
], results of this study show that arthritis has a significant impact on multiple dimensions of HRQOL. Specifically, older adults with OA and RA reported poorer general health, physical health, mental health, and sleep, as well as more activity limitation and pain, compared to those without arthritis. These results also confirm that arthritis differentially impacts specific aspects of HRQOL. Not surprisingly, pain showed the largest difference between subjects with arthritis and without arthritis. Although the sleep and mental health items showed statistically significant differences across groups, the magnitude of these differences was small (1–2 days) and of questionable clinical relevance. Mili et al. reported a similar difference in "not good" mental health days between subjects without arthritis and those reporting doctor-diagnosed arthritis in the BRFSS sample [24
]. Yet studies using other measures have shown that depression is a substantial problem among individuals with arthritis particularly RA [32
Subjects with OA and RA in this sample had considerably poorer HRQOL than individuals in the BRFSS sample who were ≥ 65 years of age and reported having arthritis [24
]. For example, 40% of the BRFSS sample reported fair or poor general health, compared with 55% of subjects with OA and 66% of subjects with RA in our sample. In the BRFSS group, the mean number of days of activity limitation was 3.5, compared with 10 days in our OA group and 13 days in our RA group. These differences may be partially due to the definitions of arthritis used in these two studies. In the BRFSS study, the arthritis sample was defined as those reporting either the presence of chronic joint symptoms or a doctor's diagnosis of arthritis. The present study used ICD-9 codes from medical records, so all patients were under physician care for OA. Since all participants in the BRFSS sample did not report having a doctor's diagnosis, this may have been a sample with a lower average level of OA severity than our sample. However, demographic and health characteristics of the samples may also be related to the observed differences. Even among those without arthritis, HRQOL was poorer in the current study compared to the BRFSS sample. Our sample consisted of older adults with fairly low income levels, and they may been in generally poorer health than those in the BRFSS. It is significant that even in our sample of older adults with relatively low income and multiple comorbid illnesses, a diagnosis of arthritis was associated with poorer HRQOL. This suggests an important independent impact of arthritis on HRQOL among older adults.
Also in agreement with prior studies [21
], this study revealed poorer HRQOL among individuals with RA compared to OA. Individuals with RA had poorer scores than those with OA on all HRQOL items, though this difference was not statistically significant for days of "not good" mental health. In general, RA is associated with greater disability than OA, though individuals with more severe OA and widespread pain may have comparable impairment [21
]. Although HRQOL may be lower for individuals with RA than OA, this study shows that OA also has a significant impact on HRQOL. Furthermore, the high prevalence of OA makes this a considerable public health concern.
Among subjects with OA and RA in this sample, we found that HRQOL differed according to demographic characteristics and comorbidity. These associations were largely in the expected direction and similar to other research [24
], with older age, non-white race, nursing home residence, lower income, and greater comorbidity being associated with poorer scores on at least one HRQOL item. In contrast to other research [24
], men in this sample reported a greater number of days of "not good" physical and mental health compared with women. The reason for this association is not clear, but men in this sample may have had greater arthritis severity or overall poorer health (which was not captured adequately by the Charlson Comorbidity Score) compared to women.
This study also confirmed the utility of the CDC's HRQOL items for mail distribution and among older adults. This scale differentiated between individuals with arthritis and without arthritis, and it also detected differences between those with OA and RA. Because the CDC HRQOL Modules are brief and easy to administer by telephone or mail, they are particularly useful for large-scale epidemiological studies.
There are several limitations so this study. Compared to the general U.S. elderly population, PACE enrollees are older and have a lower annual income. PACE enrollees also include a greater proportion of females, a smaller percentage of minorities, and a smaller proportion of married individuals. This sample was also restricted to subjects with complete Medicare data during the observation period. These individuals were also less likely to be married than subjects who were excluded because of incomplete Medicare data. These sample characteristics may limit generalizability, and additional studies are needed among different samples of older adults. Because of the very small numbers (<20) in some racial/ethnic groups, it was not possible to examine differences across specific groups. We used a broad race categorization (white vs. non-white) that may have masked differences in some specific racial/ethnic groups, and further study is needed on more racially diverse samples.
Another limitation concerns the specificity of ICD-9 codes for OA. Studies have reported the sensitivity of ICD-9 codes for OA to be between 0.32 and 1.0, and the positive predictive value between 0.76 and 0.90 [34
]. Sensitivity and positive predictive value are both greater when codes are used to identify OA in general (as was done in this study), rather than site-specific OA. Because of these known issues regarding ICD-9 code accuracy, there may have been some misclassification in this study. However, there are also inaccuracies associated with self-reported diagnosis of arthritis, and radiographic evidence is not feasible for samples of this size.
A limitation of all survey studies is the potential for non-response bias. The response rate for this survey was similar to the average response rate in other published studies involving mail surveys [36
]. Non-respondents were similar to respondents on a variety of demographic characteristics. Though this survey was brief (16 questions), older adults with poorer health or greater functional limitation may have been less likely to complete the survey. Therefore this sample may under-represent older adults with poorer health and possibly greater arthritis severity.
In summary, this study revealed a substantial impact of OA and RA on HRQOL, as measured by the CDC items. Both pharmacological and non-pharmacological treatments can yield improvements in arthritis symptoms and HRQOL [37
]. However, both medical treatment and non-pharmacological interventions are underutilized [38
], and some evidence suggests that medications are often prescribed at doses that may be too low to optimally affect symptoms [41
]. Efforts are needed to enhance access to medical care for OA, ensure optimal pharmacological management, and disseminate self-management interventions.