Systemic lupus erythematosus in three ethnic groups. XIV. Poverty, wealth, and their influence on disease activity

Authors


Abstract

Objective

To determine the impact of wealth on disease activity in the multiethnic (Hispanic, African American, and Caucasian) LUMINA (Lupus in Minorities, Nature versus nurture) cohort of patients with systemic lupus erythematosus (SLE) and disease duration ≤5 years at enrollment.

Methods

Variables (socioeconomic, demographic, clinical, immunologic, immunogenetic, behavioral, and psychological) were measured at enrollment and annually thereafter. Four questions from the Women's Health Initiative study were used to measure wealth. Disease activity was measured with the Systemic Lupus Activity Measure (SLAM). The relationship between the different variables and wealth was then examined. Next, the impact of wealth on disease activity was examined in regression models where the dependent variables were the SLAM score and SLAM global (physician). Variables previously found to impact disease activity plus the wealth questions were included in the models.

Results

Questions on income, assets, and debt were found to distinguish patients into groups, wealthier and less wealthy. Less wealthy patients tended to be younger, women, noncaucasian, less educated, unmarried, less likely to have health insurance, and more likely to live below the poverty line. They also tended to have more active disease, more abnormal illness-related behaviors, less social support, and lower levels of self reported mental functioning. None of the wealth questions was retained in the regression models, although other socioeconomic features (such as African American ethnicity, poverty, and younger age) did.

Conclusions

Wealth, per se, does not appear to have an additional predictive value, over and above traditional measures of socioeconomic status, in SLE disease activity.

INTRODUCTION

Systemic lupus erythematosus (SLE) has long been recognized as a disease with a less favorable outcome among noncaucasian populations. This is unrelated to whether patients are living in their native or an adopted country (1–13). This holds true for Asians, African Americans, and Hispanics in the US and Asians and African Caribbeans in Great Britain, but the precise explanation for these disparate outcomes remains elusive (1–13). A complex interaction between socioeconomic, environmental, biologic, and genetic factors is felt to account for these differences. To date, the precise influence of a group of factors or of an individual factor on disease activity, damage, and early death have not been demonstrated consistently.

We have previously reported factors predictive of disease activity early in the course of the disease among a cohort of patients with SLE (the LUMINA [Lupus in Minorities, Nature versus nurture] cohort). The 3 well-defined ethnicities studied were Hispanic, African American, and Caucasian (14). African American ethnicity, lack of private health insurance, abrupt disease onset, presence of anti-Ro antibodies, lack of HLA–DRB1*0301, higher level of helplessness, and abnormal illness-related behaviors were found to be predictors of disease activity (14). Disease activity early in the course of the disease and over time, the number of American College of Rheumatology (ACR) criteria at study entry, corticosteroid intake, and Hispanic ethnicity were found to be predictors of damage (15). In turn, damage, disease activity, and poverty were found to be predictors of early mortality (16). Social scientists have emphasized that income per household (used to determine if a family falls below the federally defined “poverty” line after adjusting income according to the number of people in the household) is not a precise indicator of socioeconomic status, because even within individuals falling below the poverty line, there are distinct levels of wealth or lack of wealth (17, 18). Given the importance of some socioeconomic indicators in predicting these 3 important outcomes in SLE, we set out to further explore this additional dimension of socioeconomic status: wealth and its possible impact on disease activity.

PATIENTS AND METHODS

Patients

Patients were from the LUMINA study, a multiethnic, longitudinal study of outcome. The constitution of this cohort, the variables obtained, and the frequency of study visits have been described previously (14–16, 19, 20). In short, patients with well-defined Hispanic, African American, or Caucasian ethnicity, SLE as per the ACR criteria (21, 22), disease duration ≤5 years, and from the catchment areas of the participating institutions (The University of Alabama at Birmingham, The University of Texas-Houston Health Science Center, and The University of Texas Medical Branch at Galveston) and their affiliated practices were eligible to participate in the study. As of December 31, 2001 nearly 400 patients have been recruited into LUMINA. For these analyses, 202 patients either presenting for a followup (n = 117) or an enrollment visit (n = 85), and in whom the wealth questionnaire (vide infra) was available, were included (this instrument was not part of the initial protocol; hence, not all LUMINA patients have completed it). These patients' socioeconomic, clinical, immunologic, immunogenetic, psychological, and behavioral features were comparable with those of the LUMINA patients not included in the analyses and to the overall LUMINA cohort (data not shown).

Variables

As previously reported, data from the following domains were obtained: socioeconomic–demographic, clinical, immunologic, immunogenetic, and psychologic and behavioral (14–16, 19, 20). The socioeconomic–demographic domain includes age, sex, ethnicity, marital status, education (number of years of formal education), occupation, housing (ownership and density), income, health insurance (type and degree of coverage), health-related behaviors (drinking, smoking, exercising), and perceived difficulty in accessing health care. Home density and income were used to define poverty according to the US federal government guidelines (from the Department of Commerce) (23).

Social support was measured with the Interpersonal Support Evaluation List (higher scores indicate better social support) (24). Learned helplessness (the perception that nothing the individual can do will change the disease) was measured with the Rheumatology Attitude Index (higher scores indicate higher levels of helplessness) (25). Abnormal illness-related behaviors were ascertained with the Illness Behavior Questionnaire (higher scores indicate more abnormal illness-related behaviors) (26). Self-reported physical and mental functioning were measured with the Short Form 36 (higher scores indicate better function) (27, 28).

Wealth was measured using the wealth questionnaire that consists of 4 questions taken from the Women's Health Initiative study (18, 29). Each of these questions explores a different dimension of wealth: 1) time able to sustain current standard of living in the event monthly income would cease, 2) food availability and preferences, 3) assets, and 4) accumulated debts (mortgage and car payments excluded; see Appendix A, [Copy available at the Arthritis Care & Research Web site at http://www.interscience.wiley.com/jpages/0004-3591:1/suppmat/index.html]). The 4 questions are not aggregated into a score but examined individually.

Disease activity was measured with the Systemic Lupus Activity Measure (SLAM), a validated instrument that identifies clinical manifestations and laboratory parameters attributable to SLE and present sometime during the preceding month (30). SLAM items obtained by interview, physical examination, and laboratory testing are graded on a scale of 0 (absent) to 3 (severe), as per prespecified definitions, to constitute the SLAM score. Patients and physicians also grade disease activity using a 10-cm visual analog scale (VAS) where 0 is no disease activity and 10 is the maximum disease activity possible. Patients complete their VAS at the time of the visit, whereas study physicians complete their VAS after the laboratory parameters are available.

Statistical analyses

First, the distribution of selected baseline (entry into the cohort) features were examined for those with less and more wealth as per arbitrary cutoffs of the wealth questions (see Appendix A [Copy available at the Arthritis Care & Research Web site at http://www.interscience.wiley.com/jpages/0004-3591:1/suppmat/index.html]). Next, multivariate analyses were performed. Independent variables for these analyses were from the socioeconomic–demographic, clinical, immunologic, immunogenetic, and psychological and behavioral domains that had previously been found to be predictive of disease activity early in the course of the disease. Items from the wealth questionnaire were also entered into the regression models. The dependent variables in these regressions were the SLAM score and the SLAM global (physician VAS), as obtained at the time the wealth questionnaire was applied. Because the SLAM includes fatigue as one of the items (which may or may not reflect disease activity), another regression was done excluding fatigue from the SLAM score. To eliminate the possible confounding effect of wealth and ethnicity, all regressions were also done with the wealth questions forced into these models.

RESULTS

The 202 LUMINA patients included in these analyses were predominantly women (90%) of middle age (40 ± 14 years). Twenty-three percent were Hispanic, 45% were African American, and 33% were Caucasian. For 85 patients, the wealth questionnaire was applied at the baseline visit and for the remaining 117, at a return visit. Tables 1 and 2 show the distribution of selected baseline socioeconomic–demographic, clinical, or psychological and behavioral features according to their wealth status as per questions 1, 3, and 4 of the wealth questionnaire. Question 2 was excluded because the large majority of our patients had no difficulty with either food availability or preferences. Overall, patients with less wealth tended to be noncaucasian, younger, women, less educated, less likely to be married, less likely to have health insurance, and more likely to fall below the poverty line as defined by the US federal government than the wealthier patients.

Table 1. Selected baseline sociodemographic features of LUMINA patients according to their level of wealth*
WealthAble to sustain standard of living without income for 2 monthsAccumulated assetsAccumulated debt
No n = 89Yes n = 113PLimited n = 92Significant n = 118PSignificant n = 116Negligible n = 92P
  • *

    Only P values ≤0.05 are noted. LUMINA = Lupus in Minorities, Nature versus nurture

Sex, female, %89.589.2 95.584.50.026293.886.7 
Ethnicity, %         
 Hispanic19.124.8 27.219.5 29.316.3 
 African American50.636.3 54.431.4<0.000142.243.5 
 Caucasian30.338.9 18.549.5 28.540.2 
Age, years, mean ± SD37.9 ± 10.740.9 ± 13.5 36.8 ± 12.041.9 ± 12.40.014539.1 ± 14.239.8 ± 9.9 
Education, years, mean ± SD12.4 ± 2.613.5 ± 3.40.00511.9 ± 3.013.9 ± 2.8<0.000112.7 ± 3.313.5 ± 2.70.0263
Marital status, married/living together, %49.461.1 43.565.30.006451.760.9 
Health insurance at present, %77.386.7 79.187.30.031186.281.30.0203
Poverty, below line, %37.917.90.000652.89.5<0.000135.720.00.0310
Table 2. Selected baseline clinical and behavioral features of LUMINA patients according to their wealth*
WealthAble to sustain standard of living without income for 2 monthsAccumulated assetsAccumulated debts
No n = 89Yes n = 113PLimited n = 92Significant n = 118PSignificant n = 116Negligible n = 92
  • *

    Data presented as means ± SD. Only P values ≤0.05 are noted. LUMINA = Lupus in Minorities, Nature versus nurture; ACR = American College of Rheumatology; SLAM = Systemic Lupus Activity Measure; SF-36 = Short Form 36; PCS = physical component summary; MCS = mental component summary.

  • SLAM score obtained at the time the wealth questionnaire was applied.

ACR criteria number5.6 ± 1.35.4 ± 1.2 5.6 ± 1.35.4 ± 1.2 5.5 ± 1.35.4 ± 1.2
SLAM score10.1 ± 5.18.3 ± 5.10.031010.2 ± 5.48.2 ± 4.60.01379.1 ± 4.88.7 ± 5.3
SLAM physician global2.6 ± 2.12.0 ± 1.80.00712.6 ± 2.32.1 ± 1.7 2.5 ± 2.12.2 ± 1.7
SLAM score9.0 ± 4.67.0 ± 4.70.00209.3 ± 6.46.6 ± 3.7<0.00017.8 ± 4.78.0 ± 4.2
SLAM physician global2.1 ± 1.81.7 ± 1.80.01882.5 ± 2.11.6 ± 1.60.00711.9 ± 1.82.1 ± 1.9
SF-36 PCS36.5 ± 12.036.7 ± 11.7 33.7 ± 10.336.4 ± 12.6 36.3 ± 11.634.8 ± 11.9
SF-36 MCS41.3 ± 10.548.5 ± 10.50.000441.4 ± 11.748.1 ± 10.40.000645.3 ± 12.346.8 ± 10.6
Helplessness40.5 ± 7.739.0 ± 7.3 40.7 ± 7.039.3 ± 7.1 39.5 ± 7.239.9 ± 7.2
Social support7.1 ± 1.98.3 ± 1.6<0.00017.1 ± 1.98.4 ± 1.6<0.00017.9 ± 1.87.8 ± 1.9
Abnormal illness-related behaviors19.3 ± 7.015.4 ± 7.30.001019.7 ± 6.814.9 ± 6.5<0.000116.8 ± 7.116.7 ± 6.9

The most consistent differences between the 2 groups, however, were observed with question 3 of the wealth questionnaire, which ascertains accumulated assets (Table 1). The differences were statistically significant for sex (95.5% women in the less wealthy versus 84.5% in the wealthier; P = 0.0262), ethnicity (27.2% Hispanic, 54.4% African American, and 18.5% white among the less wealthy as compared with 19.5% Hispanic, 31.4% African American, and 49.5% Caucasian among the wealthier; P < 0.0001), age in years (36.8 ± 12.0 for the less wealthy and 41.9 ± 12.4 for the wealthier; P = 0.0145), marital status (43.5% married or living together for the less wealthy and 65.3% for the wealthier; P = 0.0064), health insurance (79.1% having it for the less wealthy versus 87.3% for the wealthier; P = 0.0311), and poverty (52.8% below the poverty line for the less wealthy and 9.5% for the wealthier; P < 0.0001).

Similar trends were observed with regard to clinical, and behavioral and psychological variables where, by and large, less wealthy patients had lower levels of mental functioning, less social support, and more abnormal illness-related behaviors than their wealthier counterparts (Table 2). They also had more active disease both at baseline and at the time the wealth questionnaire was applied. These differences were significant for questions 3 (accumulated assets) and 1 (ability to sustain standard of living without income), but not for question 4 (accumulated debt). In contrast, the number of ACR criteria at baseline, levels of physical functioning, and helplessness were comparable for the 2 groups for all 4 questions (Table 2). When only the data for the 85 patients in whom the wealth questionnaire was applied at baseline were examined, the findings were very similar (data not shown).

Results of the multivariate analyses are shown in Table 3. Disease activity, as determined by the SLAM score, was associated with younger age, African American ethnicity, less formal education, number of ACR criteria at baseline, and more abnormal illness-related behaviors. These variables explain 25% of the variance in the model. Similar results were obtained when fatigue was excluded from the SLAM score and the regressions recalculated (data not shown). Neither poverty nor any of the 3 questions from the wealth questionnaire were retained in the model. When the SLAM global (physician VAS) was the dependent variable instead of the SLAM score, poverty and the interaction of poverty and African American ethnicity were associated with disease activity. These variables explain only 13% of the variance in the model. The results were comparable when the wealth questions were forced into the models (data not shown).

Table 3. Variables associated with disease activity in LUMINA patients*
VariablesFPModel's R2%
DependentIndependent
  • *

    Other variables included in the models were not statistically significant and are not listed (see text). LUMINA = Lupus in Minorities, Nature versus nurture; SLAM = Systemic Lupus Activity Measure.

SLAM scoreAge, younger6.10.014325
 Ethnicity, African American6.90.0013 
 Education, lower4.60.0325 
 Criteria number4.90.0274 
 Abnormal illness-related behaviors8.30.0043 
SLAM physician globalPoverty6.30.013013
 Ethnicity (African American) × Poverty3.10.0189 

DISCUSSION

To further elucidate the role of socioeconomic status in the course and outcome in SLE, we have included wealth in our analyses, in addition to such traditional socioeconomic indicators as ethnicity, income, education, marital status, health insurance, access to health care, and health behaviors. Even among the poor (as defined by the federal government), there are different levels of wealth (or lack of wealth) (17, 18). Therefore, the addition of this construct seemed to offer an advantage in sorting out the extent to which socioeconomic status, rather than biologic factors associated with ethnicity, may explain the less favorable course and outcome of SLE among noncaucasian populations.

Our analyses conducted in a multiethnic cohort of patients with SLE failed to show that any of the items from the wealth questionnaire had a direct impact on disease activity (as measured by the SLAM score and physician SLAM global); instead, African American ethnicity was consistently found to be associated with more disease activity, and this was the case even after the wealth questions were forced into the models. Although poverty failed to predict disease activity as measured by the SLAM score, other indicators of poor socioeconomic status, such as fewer years of formal education and more abnormal illness-related behaviors, were retained in the multivariate analyses. On the other hand, poverty and the interaction of African American ethnicity and poverty were important factors associated with disease activity as measured by the physician's global assessment. Given the fact that disease activity at disease onset and over time have been consistently found to be important predictors of the occurrence of damage (15), our data support the notion that patients of noncaucasian ethnicity with a lower socioeconomic status are at risk of experiencing more damage. Wealth, as measured in our patients, did not appear to offer an advantage over the traditional measures of socioeconomic status in terms of further determining the risk of poorer outcomes in SLE among disadvantaged ethnic minorities.

Some inherent limitations of the analyses presented should be noted. These analyses did not include all LUMINA patients and not all patients who answered the wealth questionnaire were included in the multivariate analyses because of missing data for some of the other variables in the models. Those patients not included in the analyses (n = 198) were, however, comparable in their main features with those included in the analyses (n = 202) and to the overall LUMINA cohort. Because the wealth variables were not even marginally significant in the multivariate analyses, we do not think the results would have been very different had all patients been included in the multivariate analyses. Because the number of patients in whom the wealth questionnaire was applied on the intake visit constituted less than half of the total, we could not perform the multivariate analyses on this group alone. Thus, we cannot rule out that wealth may influence disease activity if examined much earlier in the course of lupus. Likewise, we cannot rule out the possibility that wealth may influence disease activity at a later time because few patients have had a significant number of visits after the first wealth questionnaire was applied, and thus, such analysis was not carried out.

In summary, although data from our cohort support the relationship between low socioeconomic status and lupus outcomes (14–16), we have failed to show the wealth questionnaire as having additional predictive value over and above the traditional measures of socioeconomic status in disease activity in SLE. It is possible, however, that this questionnaire may provide additional insight into other SLE outcomes where poverty has been shown to be a stronger predictor (mortality, for example) of outcome (3, 4, 13, 16). Such studies are now under way.

Ancillary