Patients with chronic diseases often display poor health-related quality of life (HRQOL), perhaps due to frequent symptoms which can impact health perceptions and psychosocial functioning [1] Arthritis is a chronic disease that has been shown to impact physical, psychological, and social aspects of HRQOL [2‐4], anxiety and depression [5], worse pain and more fatigue and sleep problems [5], and poorer physical health [6]. While many arthritis studies focus on individual HRQOL domains, some research has shown that HRQOL domains are correlated. In the case of arthritis, anxiety and depression are associated with higher levels of pain intensity [7], as well as poorer physical function, vitality, and social function [8, 9]. Conversely, significant relationships have not been found between other HRQOL domains [8], pointing to the possibility that mixed profiles of HRQOL outcomes exist in arthritis, which has been examined in other chronic conditions [10, 11]. Understanding these mixed experiences may provide a more complete understanding of the differential effects of arthritis and how to provide the best, most well-rounded care. This is particularly important in the context of literature critiquing focus on domain-specific changes in HRQOL, which can hide heterogeneity across different HRQOL domains, potentially impacting treatment [12, 13].
The constellation of HRQOL outcomes could be influenced by multiple factors, including comorbid conditions [14, 15] and arthritis type; [16, 17] demographic variables like gender and age [18] or education; [19] and social support or isolation [20]. However, we currently only have evidence in arthritis-specific literature to suggest predictors of single HRQOL domain outcomes or HRQOL summary scores, which may obscure variability across domains. In addition, past research summarizes HRQOL impact in terms of overall means, which does not capture the heterogeneous HRQOL impacts on individuals with arthritis. As such, this paper aims to address gaps in the literature for individuals with arthritis by characterizing the heterogeneity in HRQOL outcomes through identification of HRQOL profiles in adults with arthritis and examination of biopsychosocial risk and protective factors that could be related to HRQOL outcomes.
Materials and methods
Participants and procedures
This cross-sectional observational study included data from participants in an online survey developed in a mixed methods study [21] and distributed by The Arthritis Foundation, a national non-profit organization in the US, between March 2019 and October 2022. Participants were recruited to complete the online survey through social media and email campaigns, using the support of a volunteer-driven community engagement network, and through a link on the Arthritis Foundation website. The non-incentivized survey was deployed using Qualtrics, an online survey platform. Participants were allowed to complete the survey multiple times, but this study only includes the most complete or (if equally complete) first instance of the survey from participants. The study was reviewed by the Institutional Review Boards of Advarra (Pro00032161) and Dartmouth College (#00031180) and deemed exempt from further review by both.
Participants in this study were English-speaking adults aged 18 years and older who had at least one form of self-reported arthritis, and who were able to provide consent. All included participants consented to participation.
Measures
The survey included the Patient-Reported Outcome Measurement Information System® (PROMIS®) measures that assessed different domains of HRQOL and emotional support, along with sociodemographic items and items about type and duration of arthritic conditions [21].
The PROMIS-29 Profile v2.1 is a 29-item measure that assesses seven HRQOL domains using four items for each domain, including physical function, pain interference, sleep disturbance, fatigue, anxiety, depression, and ability to participate in social roles. It also includes a single 0-to-10 item assessing pain intensity. The PROMIS-29 profile has strong evidence for its reliability and validity with individuals with arthritis [22]. It was used as the primary outcome. We also used the PROMIS Short Form v2.0 – Emotional Support 4a, which assesses “perceived feelings of being cared for or valued as a person” and “having confidant relationships.”
PROMIS items are answered using a five-point Likert-type scale. Each PROMIS domain is scored on a T-score metric, which is calibrated with the US general population mean of 50 and standard deviation of 10. The exception is Sleep Disturbance which is calibrated to a mixed sample of the general population and clinical samples. Higher PROMIS function T-scores reflect better functioning and higher symptom T-scores reflect worse symptom burden.
Participants self-reported their arthritis diagnosis with a check-all-that-apply question. Based on these responses, we created the following mutually exclusive groups of arthritis type to use in our analysis: osteoarthritis (“OA only”) alone (no IA or fibromyalgia); inflammatory arthritis (“IA only”) alone (no OA or fibromyalgia); fibromyalgia alone (“fibro;” no IA or OA); OA and IA (no fibro); OA and fibro; IA and fibro (no OA); OA, IA and fibro. Inflammatory arthritis (IA) included ankylosing spondylitis, gout, juvenile arthritis persistent to adulthood, psoriatic arthritis, and rheumatoid arthritis. Individuals in any group may have been diagnosed with other conditions not specified here.
Other data collected on the self-reported survey included physical comorbidities (which we converted to a count, see Appendix A), mental health comorbidities, age, gender (using a single-select item with the response options of “male,” “female,” and an option to describe their own gender), race and ethnicity, education level, work status, living situation, and major life events (like the death of a loved one) or lifestyle changes (change in diet or exercise routine). Participants also reported the year of their arthritis diagnosis which we used to calculate time since their arthritis diagnosis.
Analyses
We used latent profile analysis (LPA), a patient-centered modeling approach [23], to characterize the heterogeneity of HRQOL experiences among adults with arthritis [24‐26]. LPA groups individuals based on the similarity of their mean scores across the seven HRQOL domains of the PROMIS-29. We progressively increased the number of groups, starting with two profiles, and selected the ideal number of profiles based on both statistical model fit and clinical interpretability [27‐29]. Goodness-of-fit statistics included Akaike information criterion (AIC), Bayesian information criterion (BIC), adjusted BIC, Vuong-Lo-Mendell-Rubin likelihood ratio (VLMR-LR) test, Lo-Mendell-Rubin adjusted likelihood ratio (LMR-LR) test, and entropy. Lower AIC, BIC, and adjusted BIC values indicate better fit. Both the VLMR-LR and the LMR-LR tests compare one nested LPA model (e.g., three groups) to another LPA model with one additional group (e.g., four groups); with statistically significant improvement (p < .05), suggesting the model with more groups reflects better fit [30]. Entropy is a measure of uncertainty in the posterior classifications of the model with higher entropy values reflecting less uncertainty. Clinical interpretability was conducted among the research team (including JAS, a rheumatologist with more than 20 years of experience in providing care to patients with arthritis and other rheumatic diseases) based on a review of the profiles [31]. Participants were then assigned to an HRQOL profile based on highest posterior probability.
To identify potential risk and protective factors that might impact HRQOL profile membership, we then conducted a multinomial logistic regression model with HRQOL profile as the outcome to determine whether profile membership was accounted for by demographic and clinical biopsychosocial factors. Independent variables included self-reported diagnoses, demographic characteristics, and clinical factors, as well as the PROMIS Emotional Support T-score. To evaluate the overall classification accuracy of the regression model, we provided hit rate and hit rate by chance, which is the recommended effect size for determining whether the classification model performs better than chance [32]. We used M-Plus (V 8) to implement the LPA, and SAS (V 9.4) for the data summary and regression analysis.
Missing at random was assumed, and we used all available data to estimate the models using full information maximum likelihood.
Results
Table 1 provides demographic and clinical information from our 25,305 participants included in the study.
Table 1
Self-reported participant demographics
Mean (SD) | Range | ||
|---|---|---|---|
Age | 61.3 (14.0) | 18–106 | |
Time since diagnosis (estimated), in years | 17.0 (13.7) | 0–79 | |
PROMIS Emotional Support T-score | 48.8 (9.4) | 25.7–62.0 | |
Count of other physical health diagnoses | 1.4 (1.3) | 0–10 | |
Categories | Frequency | % | |
Gender | Male | 3490 | 13.8% |
Female | 21,677 | 85.7% | |
Other | 57 | 0.2% | |
No response/Prefer not to respond | 81 | 0.3% | |
Rurality | Urban | 21,087 | 83.3% |
Rural | 3149 | 12.4% | |
No response | 1069 | 4.2% | |
Education | Less than high school | 354 | 1.4% |
HS diploma/GED | 2881 | 11.4% | |
Some college (e.g., AA, technical degree) | 8394 | 33.2% | |
4-year college degree | 6671 | 26.4% | |
Graduate degree (e.g., Masters, Doctorate) | 5954 | 23.5% | |
No response | 1051 | 4.2% | |
Race and Ethnicity | White, non-Hispanic | 20,116 | 79.5% |
Black or African American, non-Hispanic | 1922 | 7.6% | |
Hispanic or Latino | 971 | 3.8% | |
Asian, non-Hispanic | 390 | 1.5% | |
American Indian or Alaska Native, non-Hispanic | 131 | 0.5% | |
Middle Eastern or No. African, non-Hispanic | 30 | 0.1% | |
Native Hawaiian or Pacific Islander, non-Hispanic | 27 | 0.1% | |
More than One Race or Ethnicity | 850 | 3.4% | |
No response / Prefer not to respond | 868 | 3.4% | |
Arthritis Type | OA only (no IA or fibro) | 8994 | 35.5% |
IA only (no OA or fibro) | 5911 | 23.4% | |
OA and IA (no fibro) | 4893 | 19.3% | |
Fibro (no IA or OA) | 254 | 1.0% | |
OA and fibro (no IA) | 1909 | 7.5% | |
IA and fibro (no OA) | 1076 | 4.3% | |
OA, IA, and fibro | 2268 | 9.0% | |
Work status | Working (full-time, part-time, full/part time student, homemaker) | 9561 | 37.8% |
Not working (laid off, on strike, unemployed) | 862 | 3.4% | |
Unable to work | 3312 | 13.1% | |
Retired | 10,284 | 40.6% | |
No response | 1286 | 5.1% | |
HRQOL profiles
Table 2 includes the model fit statistics for each model assessed. Model fit statistics slightly favored six profiles, though a five-profile solution was preferred due to clinical interpretability. Six profiles did not add substantial interpretability and included profiles that were not distinct; more specifically, it displayed two separate groups with PROMIS scores across domains falling within normal limits, while a five-profile solution included only one group falling within normal limits. Figure 1 provides a visual representation of the five-profile solution. Note that the y-axis for the symptom scores (Sleep Disturbance, Fatigue, Pain Interference, Anxiety, and Depression) have been inverted so that a higher position on the figure indicates better HRQOL across all PROMIS symptom and function domains.
Table 2
Latent profile analysis fit statistics by number of profiles in each tested model
Number of Profiles | AIC | BIC | Adj. BIC | VLMR LRT | LMR LRT | Entropy |
|---|---|---|---|---|---|---|
2 | 1196763.513 | 1196942.566 | 1196872.650 | 57612.16, p < .01 | 56910.514, p < .01 | 0.852 |
3 | 1175992.535 | 1176236.698 | 1176141.359 | 20786.978, p < .01 | 20533.818, p < .01 | 0.847 |
4 | 1166279.729 | 1166589.002 | 1166468.239 | 9728.806, p < .01 | 9610.321, p < .01 | 0.842 |
5 | 1161773.112 | 1162147.495 | 1162001.308 | 4522.617, p < .01 | 4467.537, p < .01 | 0.829 |
6 | 1157019.708 | 1157459.201 | 1157287.590 | 4769.404, p < .01 | 4711.319, p < .01 | 0.813 |
Fig. 1
Five profiles emerging from the latent profile analysis
×
![]()
Profile One (n = 2002, 8% of the sample; labelled “good HRQOL”), experienced better HRQOL outcomes across all domains. Overall, this group experienced functioning and symptoms within 1 standard deviation (SD, 10 points) above the US general population mean. Profile Two (n = 6504, 26% of the sample; “average HRQOL”), experienced average HRQOL scores comparable to the population mean, although physical function and pain interference were impacted and more than a 1/2 SD below the US general population mean. Profile Three (n = 2423, 10% of the sample; “poor HRQOL with non-impacted mental health”) experienced poor physical functioning and social functioning with high pain interference but did not appear to be impacted in the domains of depression or anxiety. Both Profile Four (n = 9395, 37%, “poor HRQOL”) and Profile Five (n = 4981, 20% of the sample, “very poor HRQOL”) experienced poor (about 1 SD below average population mean) and very poor (about 2 SDs below average population mean) HRQOL, respectively, across all domains. Profile Four was comparable to Profile 3, with the exception of the mental health domains.
Demographic and clinical factors associated with HRQOL profiles
The unadjusted frequency distributions of the independent variables across the five HRQOL profiles are provided in Table 3 with all variables statistically significantly associated (p < .01) except living situation (p = .34). The adjusted odds ratios (ORs) and 95% confidence intervals (CIs) from the multinomial logistic regression for all independent variables are provided in Table 4 using Profile One (“good HRQOL” profile) as the reference group. For continuous variables, the OR is interpreted as the change in log odds for every increase of 1 in the predictor. Risk factors associated with a higher likelihood of being in Profile Five (“very poor HRQOL”) compared with Profile One (“good HRQOL”) included:
Table 3
Frequency and percentage of categories of covariables by five profile groups
Category or Statistic | Overall (n = 25305) | Good HRQOL (n = 2002) | Average HRQOL (n = 6504) | Poor HRQOL (non-impacted MH, n = 2423) | Poor HRQOL (n = 9395) | Very poor HRQOL (n = 4981) | |
|---|---|---|---|---|---|---|---|
Emotional Support t-score | Mean | 48.8 | 54.9 | 51.4 | 51.9 | 46.9 | 44.9 |
SD | 9.4 | 8.3 | 8.8 | 8.9 | 8.5 | 9.5 | |
Time since diagnosis | Mean | 17.0 | 14.0 | 16.2 | 19.7 | 16.8 | 18.1 |
SD | 13.7 | 11.9 | 13.4 | 14.6 | 13.6 | 14.0 | |
Count of comorbidities | Mean | 1.4 | 0.7 | 1.0 | 1.7 | 1.4 | 2.0 |
SD | 1.3 | 0.9 | 1.1 | 1.4 | 1.3 | 1.5 | |
Recent surgery | No | 22,608 (93.9%) | 1832 (96.1%) | 5866 (95.2%) | 2165 (93.2%) | 8351 (93.6%) | 4394 (92.5%) |
Yes | 1458 (6.1%) | 74 (3.9%) | 298 (4.8%) | 158 (6.8%) | 571 (6.4%) | 357 (7.5%) | |
Arthritis type | OA only | 8994 (35.5%) | 969 (48.4%) | 2971 (45.7%) | 898 (37.1%) | 3080 (32.8%) | 1076 (21.6%) |
IA only | 5911 (23.4%) | 717 (35.8%) | 1748 (26.9%) | 387 (16.0%) | 2241 (23.9%) | 818 (16.4%) | |
OA + IA | 4893 (19.3%) | 269 (13.4%) | 1232 (18.9%) | 572 (23.6%) | 1868 (19.9%) | 952 (19.1%) | |
Fibro only | 254 (1.0%) | 9 (0.4%) | 41 (0.6%) | 16 (0.7%) | 103 (1.1%) | 85 (1.7%) | |
OA + fibro | 1909 (7.5%) | 17 (0.8%) | 228 (3.5%) | 200 (8.3%) | 794 (8.5%) | 670 (13.5%) | |
IA + fibro | 1076 (4.3%) | 9 (0.4%) | 114 (1.8%) | 82 (3.4%) | 465 (4.9%) | 406 (8.2%) | |
OA + IA + fibro | 2268 (9.0%) | 12 (0.6%) | 170 (2.6%) | 268 (11.1%) | 844 (9.0%) | 974 (19.6%) | |
Mental health comorbidities | No anxiety or depression | 13,987 (55.3%) | 1519 (87.3%) | 4588 (77.3%) | 1916 (83.9%) | 4321 (49.2%) | 1643 (34.8%) |
Anxiety, no depression | 2530 (9.9%) | 118 (6.8%) | 562 (9.5%) | 131 (5.7%) | 1218 (13.9%) | 501 (10.68%) | |
Depression, no anxiety | 2190 (8.7%) | 38 (2.2%) | 378 (6.4%) | 136 (6.0%) | 1078 (12.3%) | 560 (11.9%) | |
Both anxiety and depression | 4748 (18.8%) | 64 (3.7%) | 408 (6.9%) | 100 (4.4%) | 2157 (24.6%) | 2019 (42.7%) | |
Age | Mean | 61.3 | 61.0 | 62.6 | 65.5 | 60.4 | 59.6 |
SD | 14.0 | 15.0 | 14.1 | 11.6 | 14.4 | 12.8 | |
Gender | Male | 3490 (13.8%) | 454 (22.7%) | 1162 (17.9%) | 356 (14.7%) | 1120 (12.0%) | 398 (8.0%) |
Female | 21,677 (85.9%) | 1537 (77.0%) | 5311 (82.0%) | 2054 (85.1%) | 8219 (87.8%) | 4556 (91.7%) | |
Self-described | 57 (0.2%) | 5 (0.3%) | 6 (0.1%) | 5 (0.2%) | 25 (0.3%) | 16 (0.3%) | |
Race and ethnicity | American Indian/Alaska Native | 131 (0.5%) | 7 (0.3%) | 25 (0.4%) | 8 (0.3%) | 55 (0.6%) | 36 (0.7%) |
Asian | 390 (1.5%) | 90 (4.5%) | 126 (1.9%) | 26 (1.1%) | 116 (1.2%) | 32 (0.6%) | |
Black/African American | 1922 (7.6%) | 207 (10.3%) | 565 (8.7%) | 226 (9.3%) | 566 (6.0%) | 358 (7.2%) | |
Hispanic/Latino | 971 (3.8%) | 98 (4.9%) | 236 (3.6%) | 57 (2.4%) | 392 (4.2%) | 188 (3.8%) | |
Middle Eastern/No. African | 30 (0.1%) | 3 (0.1%) | 7 (0.1%) | 2 (0.1%) | 8 (0.1%) | 10 (0.2%) | |
Native Hawaiian/Pac. Islander | 27 (0.1%) | 1 (0.0%) | 10 (0.2%) | 2 (0.1%) | 9 (0.1%) | 5 (0.1%) | |
White | 20,116 (79.5%) | 1489 (74.4%) | 5127 (78.8%) | 1966 (81.1%) | 7571 (80.6%) | 3963 (79.6%) | |
More than one race | 850 (3.4%) | 45 (2.2%) | 167 (2.6%) | 63 (2.6%) | 344 (3.7%) | 231 (4.6%) | |
No response | 868 (3.4%) | 62 (3.1%) | 241 (3.7%) | 73 (3.0%) | 334 (3.6%) | 158 (3.2%) | |
Education | Less than high school | 354 (1.5%) | 36 (1.9%) | 66 (1.1%) | 49 (2.1%) | 104 (1.2%) | 99 (2.1%) |
High school/ GED | 2881 (11.9%) | 161 (8.3%) | 497 (8.0%) | 290 (12.4%) | 1132 (12.6%) | 801 (16.8%) | |
Some college | 8394 (34.6%) | 433 (22.4%) | 1653 (26.5%) | 825 (35.3%) | 3401 (37.8%) | 2082 (43.7%) | |
4-year college degree | 6671 (27.5%) | 615 (31.8%) | 2037 (32.7%) | 615 (26.3%) | 2369 (26.3%) | 1035 (21.7%) | |
Graduate degree | 5954 (24.5%) | 687 (35.6%) | 1975 (31.7%) | 559 (23.9%) | 1987 (22.1%) | 746 (15.7%) | |
Urban vs. rural | Urban | 21,087 (87.0%) | 1783 (92.3%) | 5615 (89.6%) | 2044 (88.1%) | 7711 (85.8%) | 3934 (83.1%) |
Rural | 3149 (13.0%) | 149 (7.7%) | 651 (10.4%) | 275 (11.9%) | 1276 (14.2%) | 798 (16.9%) | |
Work status | Working | 9561 (39.8%) | 1009 (52.9%) | 2862 (46.4%) | 792 (34.1%) | 3724 (41.8%) | 1174 (24.9%) |
Not working | 862 (3.6%) | 45 (2.4%) | 168 (2.7%) | 65 (2.8%) | 393 (4.4%) | 191 (4.0%) | |
Unable to work | 3312 (13.8%) | 15 (0.8%) | 118 (1.9%) | 272 (11.7%) | 1115 (12.5%) | 1792 (38.0%) | |
Retired | 10,284 (42.8%) | 839 (44.0%) | 3020 (49.0%) | 1193 (51.4%) | 3669 (41.2%) | 1563 (33.1%) | |
Life event: death of loved one | No | 23,544 (93.0%) | 1926 (96.2%) | 6146 (94.5%) | 2309 (95.3%) | 8695 (92.5%) | 4468 (89.7%) |
Yes | 1761 (7.0%) | 76 (3.8%) | 358 (5.5%) | 114 (4.7%) | 700 (7.5%) | 513 (10.3%) | |
Life change: started exercising | No | 22,062 (87.2%) | 1752 (87.5%) | 5547 (85.3%) | 2140 (88.3%) | 8041 (85.6%) | 4582 (92.0%) |
Yes | 3243 (12.8%) | 250 (12.5%) | 957 (14.7%) | 283 (11.7%) | 1354 (14.4%) | 399 (8.0%) | |
Life change: stopped exercising | No | 21,864 (86.4%) | 1883 (94.1%) | 5821 (89.5%) | 2126 (87.7%) | 7808 (83.1%) | 4226 (84.8%) |
Yes | 3441 (13.6%) | 119 (5.9%) | 683 (10.5%) | 297 (12.3%) | 1587 (16.9%) | 755 (15.2%) | |
Life change: changed diet | No | 20,728 (81.9%) | 1761 (88.0%) | 5455 (83.9%) | 2025 (83.6%) | 7501 (79.8%) | 3986 (80.0%) |
Yes | 4577 (18.1%) | 241 (12.0%) | 1049 (16.1%) | 398 (16.4%) | 1894 (20.2%) | 995 (20.0%) | |
Living situation | Live with 1 + people | 18,223 (76.0%) | 1476 (77.1%) | 4707 (76.2%) | 1722 (74.5%) | 6758 (76.1%) | 3560 (76.0%) |
Live alone | 5750 (24.0%) | 439 (22.9%) | 1472 (23.8%) | 589 (25.5%) | 2123 (23.9%) | 1127 (24.0%) |
Table 4
Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for independent variables from the multinomial logistic regression
Variable Domain | Variable (*reference category) | Adjusted odds ration (95% Wald CI) compared to Profile 1 | ||||
|---|---|---|---|---|---|---|
Profile 2 | Profile 3 | Profile 4 | Profile 5 | |||
Emotional support | Emotional support | 0.952 (0.945, 0.959) | 0.962 (0.954, 0.97) | 0.904 (0.897, 0.91) | 0.885 (0.878, 0.893) | |
Health status | Time since diagnosis | 1.008 (1.003, 1.013) | 1.015 (1.01, 1.021) | 1.010 (1.005, 1.015) | 1.011 (1.005, 1.016) | |
Count of physical health comorbidities | 1.325 (1.24, 1.417) | 1.898 (1.766, 2.039) | 1.582 (1.479, 1.692) | 1.913 (1.783, 2.054) | ||
Had recent surgery vs. did not have recent surgery* | 1.283 (0.945, 1.74) | 1.771 (1.266, 2.478) | 1.701 (1.252, 2.311) | 2.106 (1.518, 2.921) | ||
Arthritis type | Fibromyalgia vs. OA only* | 1.114 (0.501, 2.477) | 1.494 (0.599, 3.723) | 2.180 (1.002, 4.744) | 3.953 (1.758, 8.892) | |
IA and fibromyalgia vs. OA only* | 2.993 (1.435, 6.242) | 5.937 (2.781, 12.671) | 7.505 (3.642, 15.467) | 13.047 (6.261, 27.186) | ||
IA only vs. OA only* | 0.867 (0.75, 1.002) | 0.716 (0.592, 0.866) | 0.982 (0.843, 1.143) | 1.014 (0.843, 1.219) | ||
OA and IA vs. OA only* | 1.275 (1.073, 1.515) | 1.578 (1.295, 1.924) | 1.540 (1.291, 1.838) | 1.935 (1.586, 2.361) | ||
OA, IA and fibromyalgia vs. OA only* | 4.180 (2.033, 8.597) | 14.543 (7.069, 29.917) | 11.652 (5.714, 23.76) | 21.825 (10.640, 44.765) | ||
OA and fibromyalgia vs. OA only* | 2.917 (1.675, 5.081) | 6.887 (3.922, 12.094) | 6.210 (3.588, 10.748) | 10.131 (5.799, 17.699) | ||
Mental health comorbidities | Anxiety and depression vs. no anxiety or depression* | 1.792 (1.331, 2.413) | 0.736 (0.513, 1.055) | 6.933 (5.196, 9.25) | 11.97 (8.894, 16.109) | |
Depression only vs. no anxiety or depression* | 2.673 (1.868, 3.825) | 1.640 (1.103, 2.44) | 6.369 (4.467, 9.079) | 6.997 (4.838, 10.12) | ||
Anxiety only vs. no anxiety or depression* | 1.555 (1.238, 1.952) | 0.854 (0.643, 1.135) | 3.150 (2.512, 3.949) | 2.995 (2.331, 3.848) | ||
Demographics | Age | 0.997 (0.991, 1.003) | 1.007 (1.000, 1.015) | 0.988 (0.982, 0.994) | 0.982 (0.975, 0.989) | |
Male vs. Female* | 0.756 (0.654, 0.874) | 0.659 (0.549, 0.791) | 0.601 (0.515, 0.701) | 0.467 (0.385, 0.566) | ||
Race and ethnicity | American Indian or Alaska Native (non-Hispanic) vs. White (non-Hispanic)* | 1.586 (0.461, 5.457) | 1.386 (0.349, 5.499) | 1.486 (0.428, 5.162) | 1.355 (0.368, 4.987) | |
Asian (non-Hispanic) vs. White (non-Hispanic)* | 0.439 (0.321, 0.601) | 0.309 (0.186, 0.512) | 0.404 (0.285, 0.573) | 0.364 (0.216, 0.613) | ||
Black or African American (non-Hispanic) vs. White (non-Hispanic)* | 0.783 (0.642, 0.955) | 0.726 (0.57, 0.925) | 0.553 (0.448, 0.683) | 0.701 (0.551, 0.893) | ||
Hispanic or Latino vs. White (non-Hispanic)* | 0.850 (0.623, 1.16) | 0.696 (0.463, 1.045) | 0.891 (0.650, 1.222) | 0.911 (0.636, 1.306) | ||
Middle Eastern or North African (non-Hispanic) vs. White (non-Hispanic)* | 1.228 (0.245, 6.146) | 1.514 (0.196, 11.691) | 0.580 (0.096, 3.507) | 2.469 (0.405, 15.056) | ||
Native Hawaiian or Pacific Islander (non-Hispanic) vs. White (non-Hispanic)* | 3.368 (0.41, 27.678) | 0.884 (0.05, 15.57) | 1.779 (0.185, 17.118) | 1.193 (0.09, 15.748) | ||
More than one race and ethnicity vs. White (non-Hispanic)* | 1.017 (0.694, 1.491) | 0.972 (0.621, 1.52) | 1.109 (0.754, 1.631) | 1.272 (0.841, 1.925) | ||
Education | 4-year college degree vs. Graduate degree* | 1.150 (0.995, 1.329) | 1.248 (1.040, 1.498) | 1.254 (1.076, 1.461) | 1.242 (1.035, 1.49) | |
High school diploma vs. Graduate degree* | 1.182 (0.936, 1.493) | 2.316 (1.777, 3.019) | 2.507 (1.984, 3.168) | 3.541 (2.737, 4.58) | ||
Less than high school vs. Graduate degree* | 0.719 (0.436, 1.186) | 2.032 (1.178, 3.506) | 1.191 (0.719, 1.974) | 1.949 (1.115, 3.407) | ||
Some college vs. Graduate degree* | 1.293 (1.103, 1.516) | 1.985 (1.643, 2.398) | 2.189 (1.859, 2.577) | 2.616 (2.172, 3.15) | ||
Rural vs. Urban | Rural vs. Urban* | 1.206 (0.975, 1.492) | 1.309 (1.028, 1.668) | 1.437 (1.16, 1.779) | 1.592 (1.265, 2.003) | |
Work Status | Not working vs. Working* | 1.193 (0.814, 1.75) | 1.596 (1.018, 2.501) | 1.910 (1.309, 2.788) | 2.546 (1.692, 3.831) | |
Retired vs. Working* | 1.111 (0.951, 1.299) | 1.001 (0.829, 1.208) | 1.198 (1.018, 1.409) | 1.584 (1.312, 1.912) | ||
Unable to work vs. Working* | 2.524 (1.302, 4.894) | 13.060 (6.803, 25.07) | 12.206 (6.430, 23.17) | 41.838 (21.965, 79.69) | ||
Life events or changes | Recent death in the family vs. no recent death in the family* | 1.365 (1.027, 1.815) | 0.971 (0.690, 1.366) | 1.552 (1.164, 2.068) | 1.929 (1.422, 2.615) | |
Recently started exercising vs. did not recently start exercising* | 1.164 (0.976, 1.389) | 0.999 (0.806, 1.238) | 1.178 (0.982, 1.413) | 0.573 (0.461, 0.711) | ||
Recently stopped exercising vs. did not recently start exercising* | 1.788 (1.424, 2.245) | 2.01 (1.559, 2.593) | 2.98 (2.375, 3.741) | 2.446 (1.918, 3.12) | ||
Recently changed diet vs. did not recently change diet* | 1.463 (1.221, 1.752) | 1.761 (1.466, 2.114) | 1.564 (1.268, 1.928) | 1.742 (1.426, 2.128) | ||
Live alone vs. live with 1 or more people* | 0.810 (0.701, 0.936) | 0.874 (0.737, 1.036) | 0.753 (0.649, 0.873) | 0.719 (0.609, 0.849) | ||
-
Work status, specifically being unable to work (OR = 41.84, 95% CI [21.97–79.69]) or not working (OR = 2.55, 95% CI [1.69–3.83]) compared to working,
-
Arthritis type, specifically OA co-occurring with IA and fibromyalgia (OR = 21.83, 95% CI [10.64–44.77]), IA and fibromyalgia (OR = 13.05, 95% CI [6.26–27.19]), OA and fibromyalgia (OR = 10.13, 95% CI [5.80–17.70]), fibromyalgia (OR = 3.95, 95% CI [1.76–8.89]), or OA and IA (OR = 1.94, 95% CI [1.59 to 2.36]) compared to OA only,
-
Co-occurring anxiety and depression (OR = 11.97, 95% CI [8.89–16.11]), depression alone (OR = 7.00, 95% CI [4.84–10.12]), or anxiety alone (OR = 3.00, 95% CI [2.33–3.85]), compared to having neither,
-
Education, including having a high school diploma (OR = 3.54, 95% CI [2.74–4.58]), some college (OR = 2.62, 95% CI [2.17–3.15]), or less than high school (OR = 1.95, 95% CI [1.12–3.41]), compared to a graduate degree,
-
Having recently stopped exercising (OR = 2.45, 95% CI [1.92–3.12]) compared to not having stopped exercising,
-
Having had surgery recently (OR = 2.11, 95% CI [1.52–2.92]) compared to not having had surgery recently, and.
-
Having more physical health diagnoses (OR = 1.91, 95% CI [1.78–2.05], for an increase of 1 physical health diagnosis).
Protective factors associated with a lower likelihood of being in Profile Five (“very poor HRQOL”) compared to Profile One (“good HRQOL”) included identifying as male (OR = 0.47, 95% CI [0.39-0.57]) compared to identifying as female, identifying as Asian (OR = 0.36, 95% CI [0.22-0.61]) or Black/African American (OR = 0.70, 95% CI [0.55-0.89]) compared to identifying as white, having recently started exercising (OR = 0.57, 95% CI [0.46-0.71]) compared to not having recently started exercising, living alone (OR = 0.72, 95% CI [0.61-0.85]) compared to living with one or more other people, and having more emotional support (OR = 0.89, 95% CI [0.88-0.89], for a half SD increase in emotional support score).
While living alone was a protective factor against membership in Profile Five (“very poor HRQOL”) as compared to Profile One (“good HRQOL”), it was not a protective factor for membership in Profile Three (“poor HRQOL with non-impacted mental health”). Additionally, while recently having started exercising was a protective factor for membership in Profile Five, it was not a protective factor for membership in any other profile.
The classification hit rate was 48.27%, the hit rate by chance was 25.96%, indicating that the classification model performed better than chance. Huberty’s I index was 0.30, indicating a medium effect size [33].
Discussion
In this first study to examine constellations of HRQOL for adults with arthritis and/or fibromyalgia, five HRQOL profiles emerged: good HRQOL, average HRQOL, poor physical health with average mental health functioning, poor HRQOL, and very poor HRQOL. Similar to prior research, many participants with arthritis displayed HRQOL that is, in at least some domains, significantly poorer than the general population. This finding is particularly important given that we found this to be true across HRQOL domains, as opposed to single domains [2, 4, 6, 34, 35]. Specifically, three profile groups (67%) displayed mean scores that were at least 1 SD below the mean on all or most domains. In fact, regardless of profile, almost all adults with arthritis in this sample displayed poor physical functioning and higher pain than the general population. This emphasizes the almost universally negative impact of arthritis on these two HRQOL outcomes specifically, regardless of functioning and symptoms in other domains, a unique finding enabled by the analytic method we used.
While the impact of arthritis on HRQOL in these results is clear, it is notable that 10% of the sample displayed mental health outcomes similar to the general population despite poor physical HRQOL outcomes. This is unique to this LPA conducted with arthritis patients whereby a small but substantial group of adults possess some factors that protect against poorer mental health outcomes, even with impacted physical health. Previous studies using LPA with patients with other diseases have not identified a mixed profile with poor physical functioning and average mental health functioning [10, 11]. Examining if and how similar profiles exist in other chronic conditions and contributing factors may shed light on alternative strategies to enhance mental health outcomes in the face of poor physical HRQOL outcomes.
While some previous research has found no differences in HRQOL by arthritis type [4, 34], , our findings for membership in the very poor HRQOL group (Profile Five) are consistent with studies that show fibromyalgia and RA as significant contributors to poorer outcomes [16, 36]. We also found that mental health comorbidities were associated with a higher likelihood of membership in the very poor HRQOL group; specifically, depression with comorbid anxiety as compared to those without mental health comorbidities. Mental health factors commonly impact HRQOL, regardless of disease [10, 11]. Previous literature has shown that factors such as social and emotional support [37] can moderate the impacts of arthritis symptoms on quality of life. Additional research examining moderating or mediating factors that protect against poor mental health outcomes, even beyond the findings in this study, would be beneficial for identifying effective biopsychosocial treatments to protect against poor mental health outcomes.
Our findings regarding work status support recommendations put forth by organizations like the European League Against Rheumatism (EULAR), suggesting that “work participation may have beneficial effects on health outcomes of people with [rheumatic and musculoskeletal diseases, RMDs].” [38] Specifically, we found that working is associated with positive HRQOL outcomes, as evidenced by the much higher likelihood of being in the very poor HRQOL group for individuals who were unable to work or not working as compared to those who were currently working. Possible confounders in our results include that people who are unable to work may have worse disease activity, which could directly impact their HRQOL outcomes. However, since participants self-reported their inability to work on a single item, we do not have information as to why they were unable to work (e.g., due to physical limitations, mental health, etc.).
Some lifestyle factors, and factors that may be intervenable, were shown to be protective against the poorest outcomes, suggesting opportunities for intervention. For example, having recently started exercising was associated with a lower likelihood of membership in the very poor HRQOL group, consistent with numerous recommendations, particularly in OA [38]. Additionally, people who lived alone were less likely to be in the very poor HRQOL outcomes group (or in the average or poor HRQOL groups) than those who lived with one or more people. The direction of this finding suggests that this was more related to functioning than to receiving support from people with whom one lives. This finding does not necessarily mean that living alone promotes better functioning; it is more likely to suggest that participants who live alone tend to have higher functioning that allows them to live independently.
Having more emotional support was protective against membership in the very poor HRQOL group, the logical inverse of previous findings that social isolation can negatively impact HRQOL [20]. A scoping review of self-management support needs found that people with RA discuss needing emotional support from different people in their life, including other people with RA, peers, and even colleagues [39]. This review highlighted the importance of individualizing supports and allowing people with RA to choose the type of support they need when they need it [39]. Despite patient needs for individualized support with emotional or social support in their clinical care, some patients have reported receiving minimal education on self-management and nonpharmacological treatments [40]. Given that emotional support is protective against universally poor HRQOL outcomes, across pain, physical functioning, fatigue, and sleep, and these HRQOL domains are critically important to patients [41, 42], identifying emotional support status and needs early on could impact the success of treatments.
Importantly, our approach using profile analysis can inform the development of risk prediction models to identify subgroups of patients in a provider’s panel who may be at high risk for poor HRQOL. Specifically, programmed and automated data-mining or artificial intelligence (AI)-enabled models can scan a medical record to look for those demographic and clinical factors that have been shown, through this and related research, to be associated with membership in the poorer HRQOL groups and flag them for more priority clinical attention. Alternatively, if the HRQOL is assessed as part of the outpatient care, these scores can directly identify the subgroups of people with these profiles.
Each of the identified groups may have different needs for clinical care, including more psychological support for those with mental/emotional HRQOL deficits versus evidence-based pharmacological and non-pharmacological treatments to improve physical HRQOL in those with these deficits [43], and this framework can inform what care may be beneficial. Assessing HRQOL regularly in practice could help guide suggested treatments. For example, if a patient is trending towards poorer HRQOL, assessing emotional support and avenues for improving emotional support along with other interventions may be beneficial in preventing further decreases in HRQOL. Those with physical HRQOL deficits may benefit from walking programs, motivational interviewing to encourage physical activity, use of physical and occupational therapy to strengthen muscles and improve musculoskeletal health and balance, use of assistive devices (walking aids, prosthetics, wheelchair) to improve function and social participation.
While there are limitations associated with the use of self-report measures, we chose this method due to the benefit of being able to get a wider and larger sample through distribution methods that would not be possible using other types of data collection. These methods allowed us to collect responses from people with different forms of arthritis, a strength of the study. One notable limitation is in the arthritis type variable, which was not mutually exclusive (participants could select more than one diagnosis). We aimed to address this by developing clinically meaningful and mutually exclusive groupings, which were more reflective of how arthritic conditions are experienced by people in their everyday lives. We were not able to verify self-reported arthritis diagnosis, which is another limitation of the study. Additionally, due to our methods, our sample included participants more likely to respond to a survey, and thus may not be fully representative of the population.
Future directions for research could include further exploration of the profile group that displayed non-impacted mental health despite poor physical HRQOL, if and how this exists for other chronic conditions, and factors that mediate or moderate this relationship. Overall, our findings coalesce to provide evidence that understanding the constellation of HRQOL outcomes is important to consider in providing comprehensive biopsychosocial care for patients with arthritis.
Declarations
Financial interests
EMK, KLC, and KES report funding provided by contracts with the Arthritis Foundation to conduct related work. EMK received a travel bursary award to attend the 2023 EULAR Congress to present this work. KLM is Vice Chair of the Board for Health Research Alliance, which is an unpaid position. JAS reports consulting fees paid by AstraZeneca, Crealta/Horizon, Medisys, Fidia, PK Med, Two labs Inc., Adept Field Solutions, Clinical Care options, Clearview healthcare partners, Putnam associates, Focus forward, Navigant consulting, Spherix, MedIQ, Jupiter Life Science, UBM LLC, Trio Health, Medscape, WebMD, and Practice Point communications; and the National Institutes of Health and the American College of Rheumatology; participation on the speaker’s bureau of Simply Speaking for which he has received honoraria; membership on the steering committee of OMERACT, from which he received support to attend meetings every 2 years; participation on the FDA Arthritis Advisory Committee, with no financial support; participation as the Co-Chair of the Veterans Affairs Rheumatology Field Advisory Committee, from which he receives no financial support; participation as editor and Director of the UAB Cochrane Musculoskeletal Group Satellite Center on Network Meta-analysis, from which he receives no financial support; ownership of stock options in Atai life sciences, Kintara therapeutics, Intelligent Biosolutions, Acumen pharmaceutical, TPT Global Tech, Vaxart pharmaceuticals, Atyu biopharma, Adaptimmune Therapeutics, GeoVax Labs, Pieris Pharmaceuticals, Enzolytics Inc., Seres Therapeutics, Tonix Pharmaceuticals Holding Corp., and Charlotte’s Web Holdings, Inc.; and previous ownership of stock options in Amarin, Viking and Moderna pharmaceuticals.
Ethics approval
This works was reviewed by the Advarra Institutional Review Board and the Dartmouth College Committee for the Protection of Human Subjects. It was deemed exempt from IRB oversight by both committees.
Informed consent
Informed consent was obtained from all individual participants included in the study.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
