Abstract
Objective. To assess the effect of socioeconomic status (SES) on outcomes in patients with early inflammatory arthritis, using data from the Canadian Early Arthritis Cohort (CATCH) study.
Methods. In an incident cohort, 2023 patients were recruited, and allocated to low SES or high SES groups based on education and income. Outcomes at baseline and 12 months were analyzed in relation to SES including the 28-joint Disease Activity Score (DAS28), Simplified Disease Activity Index (SDAI), pain, patient’s global assessment scale (PtGA), the Health Assessment Questionnaire–Disability Index (HAQ-DI), and the SF12-v2 Health Survey, using the ANOVA, chi-squared test, and regression analyses.
Results. The CATCH population had 43% with high school education or less and 37% in the low-income group (< 50,000 Can$ per annum household income). The low-education group had higher DAS28 at baseline (p = 0.045), becoming nonsignificant at 12 months and lower physical component score on SF12-v2 at baseline (p = 0.022). Patients in the low-income group presented with higher HAQ-DI (p = 0.017), pain (p = 0.035), PtGA (p = 0.004), and SDAI (p = 0.022). Low-income versus high-income groups were associated with an OR above the median for HAQ-DI (1.20; 95% CI 1.00–1.45), PtGA (1.27; 95% CI 1.06–1.53), and SDAI (1.25; 95% CI 1.02–1.52) at baseline. The association with low income persisted at 12 months for HAQ-DI (OR 1.30; 95% CI 1.02–1.67), but not for other variables.
Conclusion. Low SES was initially associated with higher disease activity, pain, and PtGA, and poorer function. At 1 year, outcomes were similar to those with high SES, with the exception of HAQ-DI.
Lower socioeconomic status (SES) has been known to be associated with increased mortality, worse disease activity, poorer function, and more pain and disability in patients with rheumatoid arthritis (RA)1,2,3,4,5,6,7,8,9. SES has been defined in previous studies by neighborhood social deprivation and personal attributes such as education or occupation1,2,3,4,5,6,7,8,9. Whether at the community level or personal level, patients with low SES have been shown to have higher disease activity and lower quality of life at baseline1,2,3,4,5,6,7,8,9. Interestingly, several studies have shown that those with lower SES improve significantly more in outcome variables upon followup, although the differential between low and high SES remains2,4. There have been many hypotheses regarding the mechanism through which SES affects disease outcome. Low SES has been shown to be associated with less use of allied health resources and higher incidence of depression3. Learned helplessness is a concept that has been explored in a previous study, and found to be a potential mechanism by which SES associates with disease outcome8. The significance of SES on early RA disease outcome in the Canadian health system is still unclear because it has never been studied in a Canadian cohort. We evaluated the effect of SES on disease outcomes in patients with early RA or suspected RA defined as fixed joint symptom onset of less than 1 year.
MATERIALS AND METHODS
Patients
As of October 2012, there were 2023 patients enrolled in the Canadian Early Arthritis Cohort (CATCH) study. The CATCH study is a prospective cohort study of patients with early inflammatory arthritis (EIA) recruited at 17 sites across Canada since July 2007, with the objective of gathering longitudinal Canadian data to study treatment effectiveness. The inclusion criteria for CATCH have been published10: age > 16 years at time of enrollment, persistent synovitis for at least 6 weeks but less than 12 months, ≥ 2 swollen joints or 1 swollen metacarpophalangeal/proximal interphalangeal joint, and ≥ 1 of positive rheumatoid factor (RF), positive anticyclic citrullinated peptide (anti-CCP), morning stiffness > 45 min, response to nonsteroidal antiinflammatory drugs, or painful metatarsophalangeal squeeze test. Patients enrolled were evaluated at baseline and at followup visits every 3 months in the first year and every 6 months in the second year and annually thereafter, according to a standard protocol. Treatment options at any visit included monotherapy with methotrexate or another disease-modifying antirheumatic drug (DMARD), combination DMARD, biologics, and oral/intramuscular/intraarticular glucocorticoids. Treatment plans were left to the discretion of the treating physician; therefore, data pertaining to medications prescribed were not collected. Outcome measurements that were evaluated in the CATCH cohort at baseline, 3 months, 6 months, 9 months, and 12 months included the 28-joint disease activity score (DAS28), the pain visual analog scale (pain VAS), physician’s global assessment (PGA), patient’s global assessment (PtGA), Health Assessment Questionnaire–Disability Index (HAQ-DI)11, Simplified Disease Activity Index (SDAI), Clinical Disease Activity Index (CDAI), and the SF-12v2 Health Survey12.
Statistical analyses
SES was defined by either educational level or income. Educational level was categorized as low education (high school diploma and lower) and high education (postsecondary education and higher). Annual income, which was not specified on the questionnaires given to patients as household or individual income, was categorized as low income [≤ 50,000 (all dollars Canadian) below the median income] or high income (greater than $50,000). The median total family income for all family types in 2011 as provided by Statistics Canada is $50,610; therefore, the definition of low income in this study approximates below-median annual household income for an average Canadian family in 2011. The associations between SES and categorical variables such as ethnicity, employment status, education level, income level, smoking status, and marital status were studied using Pearson’s chi-squared analyses. Employment status was defined as employed (including full-time, part-time, or seasonal employment), retired, homemaker, student, disabled, on sick leave, on maternity leave, or unemployed. Smoking status was defined as current smoker, ex-smoker, or never smoker; a patient is considered a smoker in our analyses if he or she currently smokes. Marital status is defined as single, common law, or married.
Continuous variables were analyzed using 1-way ANOVA analyses. The associations between SES and outcomes were quantified using binary logistic and multiple linear regression models. The model analyzed OR for low SES and above versus below median disease activity and disability. The OR were adjusted for age, sex, symptom duration, meeting the American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) criteria for RA13, and smoking status, which were confounders known to influence disease activity in patients with RA. Available case analysis was performed instead of complete case analysis. Cases were not excluded because of some missing data; instead, all cases were analyzed if the variable of interest was present. Missing variables were excluded from analyses (i.e., a complete case analysis for the variables of interest was used). A subgroup analysis was performed for patients who satisfied the ACR 2010 criteria for RA13. Outcome and disease activity measures such as DAS28-erythrocyte sedimentation rate (ESR), HAQ-DI, pain, ESR, PtGA, SDAI, CDAI, and SF-12 were analyzed using ANOVA.
RESULTS
Cohort characteristics
A total of 2023 patients were enrolled in CATCH. Of those, 1991 patients self-reported an education level, and 1990 self-reported an income level. There were 357 without data at 12 months for various reasons (96 withdrew consent, 113 were lost to followup, 90 did not have RA, 15 had comorbidity, for 32 the physician had moved, 7 patients died, 4 had a language issue). The remainder of the patients (111) had not yet had the 12-month visit. Differences in withdrawal rates between income and education groups studied were not significant. The mean age of the cohort was 53; 73% were women, 82% were white, 71% were either married or had a common law partner, 54% were employed (includes full-time, part-time, and seasonal employment), 38% had less than $50,000 annual income, and 19% were current smokers. Mean symptom duration at presentation was 184 days, 78% met 2010 ACR/EULAR criteria for RA, 58% had positive RF, and 51% had positive anti-CCP. Cohort baseline characteristics are shown in Table 1.
SES as defined by education: comparison of baseline characteristics and outcomes between education cohorts
Forty-three percent of the cohort was allocated to the low-education group, and 55% was allocated to the high-education group. The low-education group was significantly older, had fewer women, had a higher proportion of whites, a lower employment rate, lower income, shorter symptom duration, and a higher proportion of smokers (Table 1). The 2 groups did not differ significantly in RF positivity, anti-CCP positivity, or the percentage meeting the ACR/EULAR criteria. The low-education cohort had significantly higher mean DAS28 and ESR at baseline compared to the high-education group, but not at 12 months. The physical component score (PCS) of SF-12 was worse in the low-education group. These were not statistically different between cohorts at baseline: HAQ, pain VAS, PtGA, SDAI, and CDAI. At 12 months, however, PtGA and CDAI had worse values in the low-education group.
The binary logistic regression analysis showed that low education was associated with above-median DAS28, but the OR was only significant at 6 months (OR 1.36, 95% CI 1.07–1.73; p = 0.01), and not at baseline or 12 months (Table 2).
SES as defined by income: comparison of baseline characteristics and outcomes between income cohorts
Thirty-seven percent of the cohort was allocated to the low-income group, and 61% to the high-income group. The low-income group had a significantly higher proportion of patients who were smokers, nonwhites, single, and unemployed; who met the ACR/EULAR criteria for RA; and who had less than post-high school education. The 2 groups were similar in age, sex, symptom duration, RF, and anti-CCP positivity (Table 1). With regards to disease outcomes, there was a significant difference observed in HAQ, pain VAS, PtGA, SDAI, and CDAI. The low-income group had a significantly higher HAQ at all timepoints (baseline, 6 mos, and 12 mos). Baseline pain, PtGA, CDAI, and SDAI were worse in the low income group (Table 1), and not statistically significant at 6 and 12 months.
The binary logistic regression analysis showed that low income was associated with above median HAQ at baseline (OR 1.20, 95% CI 1.00–1.45; p = 0.05) and at 12 months (OR 1.30, 95% CI 1.02–1.67; p = 0.04). In addition, low income was associated with above-median PtGA at baseline (OR 1.27, 95% CI 1.06–1.53; p = 0.01) and SDAI at baseline (OR 1.25, 95% CI 1.02–1.52; p = 0.03; Table 2).
Subgroup analysis of patients who satisfied the ACR 2010 criteria for RA at baseline
Characteristics and analyses of the subgroup of patients who satisfied the ACR 2010 criteria for RA are shown in Tables 3 and 4. Baseline DAS28-ESR was not significantly higher in patients with low education level (p = 0.09), whereas it was significantly higher (p = 0.045) in the original cohort. Baseline ESR was higher in patients with low education and was nearly statistically significant (p = 0.053). Baseline PCS of the SF-12 showed similar significant findings compared to the original cohort. HAQ-DI at baseline, 6 months, and 12 months was worse in patients with low income, which is similar to findings in the original cohort. Baseline pain and PtGA scores were also worse in patients with low income. However, baseline SDAI and CDAI, both of which were significantly worse in patients with low income in the overall cohort, were numerically but not significantly worse in this subgroup (p = 0.056 and p = 0.062, respectively).
Table 5A shows the characteristics of those who completed 1 year of followup and of the 17% who dropped out. One-quarter of those who dropped out did not have RA (and by the protocol, they were removed from the study). Thus, it was expected that fewer in the dropout group would meet RA criteria, including having RF and anti-CCP positivity. Table 5B shows reasons for dropping out. There were some differences between those who remained in the study versus those who did not, including income but not education.
DISCUSSION
Our study showed that patients with EIA, including those meeting ACR criteria for RA and low SES, presented with higher disease activity, more pain, worse function, and more disability from disease. Interestingly, most of these relationships became statistically nonsignificant over time enrolled in the study. HAQ is the only outcome that remained significantly worse in patients with low SES at 12 months of followup. In the subgroup of patients who satisfied the ACR 2010 criteria for RA, there were fewer associations that were significant between SES and disease variables. For instance, only HAQ-DI at baseline and 6 and 12 months was worse in the low SES stratum, whereas only baseline pain, PtGA, and ESR were also higher in patients with low SES and PCS.
Low SES has been shown in previous studies to be associated with worse disease outcomes in patients with known RA and early RA1,2,3,4,5,6,7,8,9. Our study is the first, to our knowledge, to study the effect of SES on early RA in the Canadian healthcare system. Few studies have investigated this in an incident cohort or followed patients to assess SES effect on disease outcome over time. The advantages of studying the effect of SES on the early RA population are that the SES of these patients likely has not yet been affected by disability from their disease, and response to treatment early in their disease can be assessed in the context of SES differences. However, it is important to note that not all these patients met the ACR/EULAR 2010 criteria for the diagnosis of RA; 78% of patients enrolled met these criteria. Binary regression models yielded similar results in the stratum that met RA criteria.
In the few studies that followed patients with early RA, SES was associated with higher disease activity at baseline, but the results of SES on disease outcome upon followup were conflicting2,5. Our study showed that the negative effect of low SES at baseline resolved by 1 year with the exception of HAQ. The access to healthcare and affordability of traditional DMARD in the Canadian healthcare system may also explain why our study saw reversal of the negative SES effect on disease outcomes. There may be patient factors such as ethnicity that affect the self-reporting of HAQ. For example, Hispanic patients with RA were found to score significantly higher in self-reported HAQ compared to whites and African Americans, despite similarity in joint scores, ESR, and PGA14. There were too few ethnic minorities to explore this in our dataset. Low literacy could also affect HAQ scores15. Although our study does not explore some of these possible factors, poor coping mechanism and depression have been explored in previous studies as reasons to explain poorer outcomes in patients with low SES7,8. It is also possible that functional requirements to return to work are different between the low-income and high-income populations. For example, there may be a higher prevalence of employment requiring manual labor within the low-income population. There may also be insurance differences in SES strata even in a publicly funded system, because medications are not universally covered. The public system may pay for certain medications only in the elderly or very poor populations, depending on each province. We did not study whether there were early medication differences because the DAS28 scores were not different between the groups, so it is unlikely that the lower SES group was undertreated. There could, however, be medication differences after the first year such as initiation of biologic medications (the rate in the first year of our cohort is very low but increases in the next couple of years).
Table 6 reviews the effects of SES on RA and early RA from the literature including worse effects on HAQ1,2,3,4,5,6,8. In our study, we used either education or income as 2 separate surrogates to define SES. The advantage of using education or income to define SES is that these factors are individualized, as opposed to a patient’s residence in a particular neighborhood. Many studies that investigated SES effects on RA and EIA outcomes defined SES in terms of SES scores assigned to each patient’s residence1,5,6,8. Social deprivation scores such as the Townsend Index of Disadvantage and Deprivation, the Carstairs Deprivation Score, and most recently, the Index of Multiple Deprivations take into account factors such as unemployment, not owning a car, not owning a home, household crowding, and social class to determine an overall deprivation score for each area across the United Kingdom. This makes the assumption that all persons residing within the same neighborhood are of the same SES. In our study, we had the advantage of access to patient self-reported income data, a statistic that carries significant weight in the determination of individual SES.
Ours is a study in early RA using 2 variables for SES in a large incident cohort, which helps us to understand the potential but weak effects of SES on disease.
Limitations of our study include nonstandardization of treatment, some missing data, variable length of followup, and using education and income as 2 separate surrogates to define SES instead of having a unified definition of SES. If we defined low SES as both lowest education and income strata, there were insufficient numbers to do analyses adjusting for confounders, because we had only about 200 patients who met those criteria. Also, low income cutoff was $22,000 to $25,000 annually, according to Statistics Canada, but if we used that cutoff, the numbers were too small. Using higher income biased our results toward a null effect, because the very lowest income group was not analyzed. The median income in Canada is $50,000 annually. The definition of poverty and low education may differ in various regions of Canada. Because this is an observational study of usual care, the treatment plan was left to the discretion of each treating rheumatologist and was not standardized. Also, coverage of medications is not universal across the nation and depends on different access in the provinces and various formularies. Available case analyses were performed, which explains the variations in sample size of different variables analyzed. Of the 2023 patients enrolled into this study at the time of analysis, there were 468 patients who had withdrawn or did not have 1 year of followup at the time of analysis. There were no between-group differences in withdrawal rates. We ensured that the variables studied had large sample population numbers to maintain power in our analyses, which made it impossible to combine education and income into 1 unified SES definition, because that would have meant losing power in our analyses. Also, the patient self-reported income did not differentiate between individual income and household income; therefore, this could have been biased by the lack of reporting spousal income. Finally, some differences in results such as the inconsistencies seen among low SES effects on CDAI, SDAI, and DAS28 can be due to the differences in components of composite scores that are used. For example, PGA and PtGA of disease activity are used in the CDAI and SDAI, whereas DAS28 uses the PtGA of health.
One American study showed that patients with RA and low income and other low SES factors were less likely to be prescribed DMARD16.
There are not many adverse effects of SES on disease activity and function by 1 year in this early RA cohort (early RA and EIA). Perhaps the patients included demonstrate a “best case” scenario, in which patients are literate and willing to enter our observational cohort. The 17 sites included rheumatologists with an interest in early RA and an interest in collaborating with other sites. Thus there is a very positive bias for good outcomes in our study. It is reassuring that SES is not a major determinant of outcomes in ERA over the critical first year of disease, but the most marginalized patients would likely not have timely access to a rheumatologist and these patients could not be identified by our cohort. This can affect the generalizability of the data where patients not seen in these centers (or at all) may not have optimal outcomes.
The results of our study emphasize the importance of identifying patients with low SES to enable identification of a population likely to present with higher disease activity. What could be interesting is determining whether future differences are present in this early inflammatory cohort over the next few years, because there may be differential access to biologic DMARD and because the rate of biologics increases in our cohort over years 2 and 3. This is perhaps where outcomes between SES groups may vary if access to expensive medications is delayed or reduced in the low SES population.
Our study is the first, to our knowledge, to report the association of SES as measured by education and income with disease outcome in a large prospective Canadian cohort of patients with early RA in a real-world setting. Low SES as defined by education was associated with significantly worse DAS28, ESR, and PCS of SF-12 in patients at baseline. The negative effects of SES are overcome with initiation of treatment, such that by 1 year the between-group differences in disease activity seen at baseline disappeared. Low SES as defined by income was associated with significantly worse HAQ, CDAI, SDAI, PtGA, and pain in patients at presentation. HAQ is persistently worse in low-income patients up to 1 year of followup. All other disease activity indices such as CDAI, SDAI, PtGA, and pain VAS are not significantly different between low-income and high-income groups by the 1-year followup visit. This study shows that SES worsens baseline disease activity in patients presenting with early RA. Although many of these differences resolved by 1 year, the persistence of worse HAQ in low SES patients was observed.
APPENDIX 1
CATCH investigators: Vandana Ahluwalia, Pooneh Akhavan, Murray Baron, William Bensen, Louis Bessette, Gilles Boire, Vivian P. Bykerk, Ines Colmegna, Boulos Haraoui, Carol A. Hitchon, Shahin Jamal, Edward C. Keystone, Alice Klinkhoff, Majed Kraishi, Maggie Larche, Chris Lyddell, Bindu Nair, Chris Penney, Janet E. Pope, Laurence Rubin, J. Carter Thorne, and Michel Zummer.
Footnotes
-
The CATCH study was designed and implemented by the investigators and financially supported initially by Amgen Canada Inc. and Pfizer Canada Inc. through an unrestricted research grant since inception of CATCH. As of 2011, further support was provided by Hoffmann-La Roche Ltd., United Chemicals of Belgium Canada Inc., Bristol-Myers Squibb Canada Co., AbbVie Laboratories Ltd., and Janssen Biotech Inc. (a wholly owned subsidiary of Johnson & Johnson Inc.).
- Accepted for publication September 17, 2014.