Predictors of Depression in a Multiethnic Cohort of Patients with Rheumatoid Arthritis

Abstract

Objective

Patients with rheumatoid arthritis (RA) who experience depression have worse health outcomes. This study identifies predictors of depression in an ethnically and racially diverse population of patients with RA.

Methods

Patients with RA in a prospective cohort at San Francisco General Hospital out-patient rheumatology clinic were included if they were ≥ 18 years of age, met the American College of Rheumatology classification criteria for RA, had a Health Assessment Questionnaire (HAQ) score collected, and had the RA-specific Disease Activity Scale 28 performed by a rheumatologist. The outcome variable was a depression score measured by the Patient Health Questionnaire-9 (PHQ-9), a self-report questionnaire validated to correlate with a diagnosis of major depression.

Results

Three hundred and forty-nine clinical visits for 172 patients were included in the analysis. Forty percent of patients scored ≥ 10 on the PHQ-9 during at least one clinic visit which corresponds to a symptom severity of at least moderate depression. The mean PHQ-9 score was 7, corresponding to symptom severity of mild depression. In the multivariate analysis, higher HAQ scores were associated with depression, and Asians have lower depression scores compared to Hispanic, white, and black subjects.

Conclusion

Identifying associated predictors of depression in a diverse population of patients with RA can help guide treatment which should include preventing disability and decreased function as well as targeting depressive symptoms more specifically in patients with RA.

Background

Of the 1.3 million adults in the United States (0.6%) with rheumatoid arthritis (RA)^{1, 2}, most will develop progressive functional limitation and physical disability³. Depression, a frequent disorder among patients with RA, is often unrecognized or under-treated^{4, 5}. The prevalence of major depressive disorder is 13–42% in patients with RA^6–13, at least double that in the general population.

In patients with RA, poor clinical characteristics and function are associated with subsequent depression. There are more RA related physician visits and hospitalizations in patients with depression as well as negative health and functional outcomes and increased health service utilization¹⁴. Patients with rheumatologic diseases and depression are less likely to be adherent with their medications¹⁵ and more likely to discontinue anti-TNF-α therapy when compared to non-depressed patients¹⁶, which may then have a negative effect on their health status. Furthermore, comorbid depression is an independent risk factor for death in patients with RA, with a hazard ratio of 2.2 (95% CI 1.2–3.9, p = 0.01)^{17, 18}.

Studies have shown that disability and decreased function as measured by the Health Assessment Questionnaire (HAQ) are associated with depression in Caucasian populations^19–22 but have not addressed potential predictors of depression in an ethnically diverse group of patients with RA. A recent study identified Hispanics with RA as more likely to have symptoms of depression than non-Hispanic subjects²³; however, important covariates such as disease activity and disease duration were not taken into account. Furthermore, patients of other race/ethnicity (African-Americans, Asians) were not included in this study for comparison. Since concomitant depressive symptoms are known to worsen health outcomes and confer independent risk for mortality in RA, it would be useful to discover associations between depression and rheumatoid arthritis in diverse populations.

Our objective is to identify demographic and disease-related predictors of depressive symptoms in a socioeconomically and ethnically distinct population of patients with RA at an urban, public hospital. The independent effects of race/ethnicity on depressive symptoms were of special interest in this multiethnic cohort of patients with RA.

Subjects and Methods

Study population

Study subjects are patients from the University of California, San Francisco (UCSF) RA Cohort at San Francisco General Hospital (SFGH) outpatient rheumatology clinic. SFGH is the major provider to the poor, minority, and uninsured population of the City and County of San Francisco. Patients were consecutively enrolled as they presented for their initial or return visits between 11/06–12/08, and the data were gathered at regular clinical visits. Patients were included in the analysis if they were ≥ 18 years of age, met the American College of Rheumatology (ACR) classification criteria for RA²⁴, had a depression score measured by the Patient Health Questionnaire-9 (PHQ-9), had the RA-specific Disease Activity Scale 28 (DAS 28) performed by a rheumatologist at the same clinical visit, and had a Health Assessment Questionnaire (HAQ) score collected within 6 months before or after the depression measure. The UCSF Committee on Human Research approved the study. Written informed consent was obtained and patients were interviewed in the language of their choice (English, Cantonese, Mandarin, or Spanish).

Assessment of clinical data

In addition to sociodemographic characteristics (age, gender, self-reported ethnicity) and disease duration, a wide range of clinical data was collected at enrollment and at regular intervals thereafter during routine clinical encounters. Disease activity was assessed with the DAS 28, calculated at each clinic visit. The DAS 28 includes scores of tender and swollen joint counts, patient global disease activity rating scale, and erythrocyte sedimentation rate (ESR)^{25, 26}. Functional limitation was evaluated with the 20-item HAQ score²⁷. The HAQ is a reliable and valid assessment of functional limitations in RA. HAQ scores collected within 6 months of the depression scores were included since HAQ scores are generally stable over a one-year period among patients with RA^28–30. The HAQ has been translated and validated in Mandarin/Cantonese³¹ and Spanish³². Treatment-related variables were prednisone use and other disease-modifying anti-rheumatic drugs (DMARDs) such as methotrexate, hydroxychloroquine, sulfasalazine, azathioprine, leflunomide, anti-TNFα medications, or other biologic medications.

Ascertainment of depression

The outcome variable was depressive symptoms measured by the PHQ-9. This self-report depression questionnaire yields a 27-point scale that correlates with a diagnosis of major depression^{33, 34}. The PHQ-9 is composed of ten items, corresponding to the nine diagnostic criteria of major depressive disorder, plus one item to assess functional impairment secondary to depression. A score of 5–9 is consistent with mild depressive symptoms, 10–14 moderate, 15–19 moderate-severe, and ≥ 20 is severe depressive symptoms. The PHQ-9 has established reliability and validity in Caucasian, African-American, Asian, and Hispanic populations^{35, 36}. To classify patients with at least moderate symptoms of depression, we used a clinically useful cut-point of 10, which distinguishes between no/mild and moderate/severe depressive symptoms for the univariate analysis. The term “depressed” will be used from this point forward to refer to participants meeting this threshold of symptom severity.

Data Collection

Study research staff who were not the patient’s physician administered standardized questionnaires (PHQ-9, HAQ) and interviews in English, Spanish, or Chinese. Research staff and treating physicians were blind to PHQ-9 scores and HAQ scores, as scores were tabulated independently

Statistical analysis

The analysis accommodated the repeated measures from an individual’s multiple clinical encounters and was conducted using STATA, version 10.0 (Statacorp, College Station, Texas). The unit of analysis is a subject’s clinical visit in which the patient had a PHQ-9 collected. A total of 46 visits (11%), involving 22 patients were excluded because of incomplete information. Most of these visits were dropped because a component of the DAS 28 was not collected and therefore could not be calculated.

Differences in baseline characteristics between depressed and non-depressed subjects were assessed using a two-sided t-test with equal variance or the Pearson’s chi-square test, as appropriate. For the multivariate analysis, we used generalized estimating equations (GEE) with robust standard errors to assess associations with depression scores while taking into account within-subject correlations between measurements over time and to account for the influence of potential confounding covariates. The PHQ-9 score was used as a continuous outcome for the multivariate analysis. Interpretation of the coefficient using repeated measures analysis with robust standard errors is the same as for linear regression³⁷. For the multivariate analysis, independent variables were determined a priori with a forward inclusion strategy. Variables were selected based on face validity and biologic plausibility to obtain a minimally confounded estimate of their effect and identify important independent predictors of depression in patients with RA. Given the modest-sized data set and the number of predictors, bootstrapping (re-sampling of subjects not observations) was employed to check the reliability of the standard errors and confidence intervals.

Results

Three hundred and forty-nine clinical visits for 172 patients were included in the analysis. The average number of outpatient rheumatology clinic visits per patient was 2 (range 1–5). In this study, 89% percent of included patients were female with a mean age of 52 years (Table 1). In regard to race/ethnicity, 54% of the patients were Latino/Hispanic, 34% were Asian/Pacific Islander, 7% were African-American, and 4% were Caucasian. Eighty-four percent of patients were rheumatoid factor (RF) positive and 81% were anti-cyclic citrullinated peptide (CCP) positive. Eighty-three percent of patients were on a DMARD, 24% were on anti-TNFα medications, and 62% of patients were taking prednisone with a mean dose of 4mg. Fifteen percent of patients were prescribed an anti-depressant medication. The mean HAQ score was 1.25, and the mean DAS 28 was 4.1 indicative of farily high levels of functional impairment and disease activity, respectively.

Table 1.

Baseline Characteristics of SFGH RA Cohort by Depression

Characteristic	All Subjects	Subjects with PHQ-9 < 10 at any visit	Subjects with PHQ-9 ≥10 at any visit	p-Value
	n = 172	n = 103	n = 69
Gender – n (%)				.68
Female	153 (89)	92 (89)	61 (88)
Male	19 (11)	11 (11)	8 (12)
Age – mean ± SD	52 ± 13	51	53	.26
Ethnicity – n (%)				.17
Latino/Hispanic	93 (54)	55 (53)	37 (54)
Black/African American	12 (7)	8 (8)	4 (6)
Asian/Pacific Islander	58 (34)	34 (33)	25 (36)
White/Caucasian	7 (4)	6 (6)	1 (1)
American Indian/Other	2 (1)	0 (0)	2 (3)
DAS 28*	4.1 ± 1.4	4.0	4.3	.04
Disease Duration (year)	7.5 ± 8.0	6.9	8.4	.08
HAQ¶ score	1.25 ± .81	1.1	1.4	< .0001
Antibody Positive
RF	144 (84)	87 (84)	58 (84)	.92
Anti-CCP	139 (81)	83 (81)	57 (83)	.66
DMARD† – no (%)	143 (83)	90 (87)	54 (78)	.03
Methotrexate – no (%)	104 (61)	66 (64)	39 (56)	.11
Prednisone – no (%)	107 (62)	65 (63)	41 (60)	.65
Prednisone Dose (mg)	4 ± 4.8	4.1	4.3	.76
Anti-TNF – no (%)	41 (24)	25 (24)	17 (25)	.84

^*Disease Activity Scale 28 (DAS 28) includes tender and swollen joint counts, patient global rating scale, and erythrocyte sedimentation rate

^¶Health Assessment Questionnaire

^†Includes any disease-modifying anti-rheumatic drug other than prednisone (i.e. methotrexate, plaquenil, sulfasalazine, azathioprine, leflunomide, anti-TNFα medications, and rituximab)

With regard to the prevalence of depressive symptoms in our cohort, 40% of patients scored ≥ 10 on the PHQ-9 during at least one clinic visit (n=138) which corresponds to a symptom severity of at least moderate depression. The mean PHQ-9 score was 7, corresponding to symptom severity of mild depression.

In univariate analyses of baseline data, there were no statistically significant differences between depressed and non-depressed patients with regards to age, ethnicity, disease duration, antibody status or prednisone treatment and dosage. However, depressed participants had higher HAQ and DAS 28 scores, and were less likely to be treated with a DMARD (Table 1).

After adjusting for age, gender, race/ethnicity, disease duration and activity, prednisone treatment, and RA disease-modifying medications in a GEE model (Table 2), higher HAQ scores were significantly associated with higher PHQ-9 scores. A clinically meaningful change in HAQ score is .24³⁸ so the GEE coefficient, which represents associations with a 1-point difference in HAQ, was divided by 4. With each one .24 unit increase in the HAQ score there was an increase in PHQ9 score of .98 (95% CI .73, 1.3). Asian ethnicity was associated with lower depression scores, specifically a decrease in PHQ9 score of 1.6 (95% CI −3.0, −.28), as compared to the reference group, Hispanic/Latinos (Table 2). The confidence intervals from this multivariate model remained similar when bootstrapping was employed. The effects of HAQ score and race/ethnicity on PHQ-9 scores did not differ when modeled for possible interactions.

Table 2.

Multivariate Model of Predictors of Depression

Characteristic	GEE Coefficient (95% CI)	P-value
Female Gender	−.41 (−2.6, 1.7)	.70
Age - years	−.31 (−.09, .02)	.27
Ethnicity
Latino/Hispanic	Reference
Black/African American	−.20 (−2.7, 2.4)	.88
Asian/Pacific Islander	−1.6 (−3.0, −.28)	.02
White/Caucasian	.37 (−2.1, 2.8)	.77
American Indian/Other	−1.8 (−6.1, 2.6)	.43
DAS 28*	.08 (−.31, .45)	.70
Disease Duration (year)	−.02 (−.10, .05)	.50
HAQscore¶	.98 (.73, 1.3)	< .0001
DMARD†	−.76 (−2.6, 1.1)	.43
Prednisone	−.42 (−1.7, .83)	.51

^*Disease Activity Scale 28 (DAS 28) includes tender and swollen joint counts, patient global rating scale, and erythrocyte sedimentation rate

^¶Health Assessment Questionnaire score

^†Includes any disease-modifying anti-rheumatic drug other than prednisone (i.e. methotrexate, plaquenil, sulfasalazine, azathioprine, leflunomide, anti-TNFα medications, and rituximab)

Discussion

This is the first study evaluating predictors of depressive symptoms in an underserved, ethnically diverse cohort of patients with rigorously defined rheumatoid arthritis that includes measures of functional disability and disease status as measured by the DAS 28. Several conclusions stem from our data.

First, the prevalence of clinically significant depressive symptoms in an urban, multiethnic sample of patients with RA is 40%. While within the range of previously reported results (13–42%), this cohort appears to be on the higher end of the spectrum compared to Caucasian populations with RA and depression.

Second, those with more disability are at greater risk for depression. These findings in a previously unstudied population corroborate with results in the literature. Given the high prevalence of depression in our subjects and the strong association with decreased function, this raises the question if depression is a product of increased disability as measured by the HAQ. A mean HAQ score of 1.25 is well above the previously reported mean HAQ of .71 in a population-based cohort of patients with RA²⁸. Why is there a higher mean HAQ score in our study sample? The most obvious answer is that health disparities in the RA cohort at SFGH associated with race/ethnicity and socioeconomic status lead to worse outcomes³⁹, in this case a higher mean HAQ score. Identifying pathways and mechanisms by which socioeconomic status and race/ethnicity affect HAQ scores would provide more options for intervention to eliminate disability and subsequent depression in patients with RA.

Third, it is surprising that disease activity as measured by swollen, tender joints, patient global disease activity scale, and the ESR is not associated with higher depression scores given previous research in this area^{40, 41}. In the multivariate analysis, the upper limit for the GEE coefficient of the DAS 28 is .45. The standard deviation for the DAS 28 in the cohort is 1.4 and a two-point change in the DAS 28 would be quite large clinically. Therefore, the maximum change in the depression score associated with a two-point increase in DAS 28 is only .9, which is quite a small clinical change. This suggests that it is not the acute pain or clinical manifestations of RA disease activity but more the long-term disability and joint damage associated with arthritis that leads to depression in patients with RA.

Lastly, Asians in the SFGH RA cohort had lower depression scores compared to Latino/Hispanic, White/Caucasian, and Black/African-American subjects. To our knowledge, this is the first study to find an association between depressive symptoms and Asian race/ethnicity in patients with RA. This finding is supported by previous literature other than that specific to RA which substantiates that there is a low level of reporting depressive symptoms in Asians⁴² owing to the stigma of mental illness, implications of “weakness of character,” and factors that may protect against depression such as stoicism and cultural support systems⁴³.

There are limitations to this study. Like other chronic conditions, assessing depressive symptoms in patients with RA can be difficult. It is well understood that somatic symptoms of depression (e.g., fatigue or decreased energy) overlap with symptoms of RA. Consequently, there is a risk that depression in RA may be overestimated⁴⁴. Conversely, there is also some evidence to suggest that using conventional assessments of depression does not overestimate depression in similar chronic conditions⁴⁵. In our approach to the assessment of depressive symptoms we chose the PHQ-9 as a measure for depression since it was developed for primary care, is in widespread use for a range of acute and chronic medical conditions, and was developed to closely parallel the diagnostic symptoms of Major Depressive Disorder^{33, 34, 46}.

Our study design was cross-sectional and therefore causality between greater HAQ scores and the associated, increased depressive symptoms cannot be established. Future studies should examine this relationship in a prospective manner. Also, as a cross-sectional study, there is a limit to the information that can be produced about the waxing and waning nature of depressive symptoms over time. The high prevalence of depressive symptoms in this cohort is either because they occur more frequently or because the depressive symptoms last longer.

Another potential limitation to this study is that ‘detection bias’ or ‘medical surveillance bias’ could influence the association between depressive symptoms and clinical predictors if depressed patients with RA were less likely to attend clinical appointments. To investigate this possibility, we compared the mean number of clinical visits of patients with and without depressive symptoms and found no significant difference.

Finally, the question of how applicable our results are to other RA patients must be asked. The high percentage of subjects with positive anti-CCP (81%), which portends a poor prognosis, is greater than reported in other clinical cohorts. Furthermore, our study population has relatively high disease activity with a mean DAS 28 of 4.1, and a mean HAQ score of 1.25. The fact that our study population has more erosive and disabling disease is likely a reflection of referral practices to a public, urban hospital clinic as well as our precise definition of RA for entry into the study. While our cohort was skewed toward more severe disease and included an ethnically diverse patient population, we believe that the heterogeneity of our cohort is a benefit and not a limitation.

Given that patients in this cohort are comprised of minorities who struggle with poverty and often have poor education, future studies should directly compare this population to those with higher socioeconomic status to determine associations between depression, disease activity, disability and health disparities in patients with RA. Since depression is known to confer risk of poor health outcomes, identifying associated predictors of depression in all patients with RA, not just Caucasians, is useful. The above findings can help guide treatment to include preventing disability and decreased function as well as targeting depressive symptoms more specifically in a vulnerable population of patients with rheumatoid arthritis.