Abstract
Objective. To examine the association of income relative to the poverty threshold [poverty income ratio (PIR)] with self-reported physical functioning (PF) in a cohort of patients with systemic lupus erythematosus.
Methods. We used cross-sectional data on 744 participants from Georgians Organized Against Lupus (GOAL), and secondary analyses used data on 56 participants from a nested pilot study. Primary analyses used multivariable linear regression to estimate the association between PIR (categorized as < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00; lower PIR indicate higher poverty) and PF (scaled subscore from the Medical Outcomes Study Short Form–12 survey; range 0–100, higher scores indicate better functioning). Secondary analyses summarized complementary measures of PF as means or percentages by PIR (categorized as < 1.00, 1.00–1.99, and ≥ 2.00).
Results. The mean age of participants was 48.0 years; 6.7% were male; 80.9% were black; and 37.5%, 21.0%, 29.6%, and 12.0% had PIR of < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00, respectively. The overall mean PF score was 45.8 (36.2, 40.7, 55.5, and 61.2 for PIR of < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00). With adjustment, higher PIR remained associated with higher PF scores [2.00–3.99 vs 1.00–1.99: β = 10.9 (95% CI 3.3–18.6); ≥ 4.00 vs 1.00–1.99: β = 16.2 (95% CI 6.4–26.0)]. In secondary analyses, higher PIR was also associated with higher scores for objective physical performance.
Conclusion. Our results show that higher income relative to the poverty threshold is associated with better PF across multiple domains, warranting further research into multicomponent functional assessments to develop individual treatment plans and potentially improve socioeconomic disparities in outcomes.
Low socioeconomic status (SES) is an established risk factor for worse outcomes in patients with systemic lupus erythematosus (SLE), such as greater disease damage and worse depressive symptomatology1,2,3,4,5,6,7. Additionally, patients with SLE often experience work loss8,9, activity limitations10, and reduced health-related quality of life (HRQOL)11,12,13,14,15,16, and low SES has been associated with lower HRQOL among patients with SLE5,7,17. In particular, studies examining HRQOL and SES have shown that lower individual-level and neighborhood-level SES have been associated with poorer physical functioning (PF) among patients with SLE5,7,17. To our knowledge, no study has used the income-to-poverty ratio, commonly known as the poverty income ratio (PIR), with more than 2 categories. The PIR, the official poverty measure of the US Census18, not only reflects individual SES relative to the poverty threshold but also accounts for household size, resulting in an estimate that provides a more accurate picture of an individual’s poverty experience. For example, many patients with SLE may have no income because they are dependents, or they may have household members who act as full-time caretakers; whether these patients are “living in poverty” depends on the combined income of their entire household and number of household members depending on that income. Further, it may be important to use more than 2 PIR categories, because the experience of living just above the poverty threshold likely differs substantially from the experience of living far above the threshold.
Studies of HRQOL have targeted mostly white populations5,6,7, despite black individuals having a greater burden of disease (e.g., black women have > 3 times greater incidence of SLE than white women)19 and greater susceptibility for worse SLE-related outcomes1,3. We hypothesized that lower PIR would be associated with worse functioning among patients with SLE. Using cross-sectional data from the Georgians Organized Against Lupus (GOAL) cohort, an ongoing, population-based cohort predominantly comprising black participants, we examined the association between multiple categories of PIR and self-reported PF and whether the association differed by work status or race. In secondary analyses, using data from a nested ancillary pilot study, we also examined whether associations of PIR with functioning were consistent across a comprehensive set of measures related to PF, including objective measures of physical performance.
MATERIALS AND METHODS
Study populations and data sources
For primary analyses, we used data from the ongoing GOAL cohort study, a population-based sample of patients with SLE from metropolitan Atlanta, Georgia. Recruitment and data collection methods have been published20. Briefly, participants of GOAL were primarily recruited from the existing Georgia Lupus Registry, a population-based registry funded by the US Centers for Disease Control and Prevention, which estimated the incidence and prevalence of SLE in metropolitan Atlanta19. Patients not included in the registry but who were receiving SLE treatment at Emory University, Grady Memorial Hospital (a large safety-net hospital in Atlanta), or from community rheumatologists in metropolitan Atlanta at the time of recruitment were recruited to enrich the cohort. Additionally, recruitment emphasized incident patients (≤ 2 yrs since diagnosis) to minimize survival bias. All participants were recruited by mail, by telephone, or in person, with subsequent assessments performed annually since Wave 1 (baseline; September 2011–September 2012). A total of 850 participants who were aged ≥ 18 years at the time of enrollment with a documented diagnosis of SLE [≥ 4 revised American College of Rheumatology (ACR) criteria21 or 3 ACR criteria with a final diagnosis of SLE by a board-certified rheumatologist] were included in Wave 1. We used a cross-sectional design to describe the association of PIR with measures of PF, which were reported by questionnaire during a single wave of GOAL (Wave 5; June 2016–July 2017). There was a total of 814 adult participants in Wave 5 of GOAL. For primary analyses, participants were excluded if they were missing either the question comprising the PF summary score (n = 14), the PIR (n = 45), or any other covariates (n = 70), leaving 744 participants in the final models.
For secondary analyses, a cross-sectional design was used to examine the association of PIR with additional complementary measures of PF not identified in annual GOAL assessments (i.e., objective physical performance, reported activities of daily living, and history of falls), which were measured during study visits for a nested, GOAL-ancillary pilot study (October 2016–April 2017). Recruitment and data collection methods for the pilot have been described22. There were 60 participants in the pilot, and we excluded individuals missing information on PIR (n = 4) from analyses, yielding a sample of 56 participants.
The Emory University Institutional Review Board approved the main and ancillary study protocols (no. IRB00003656), and all participants provided informed consent.
Study variables
Self-reported PIR was estimated as the ratio of a household income, as reported by the participant, to the appropriate poverty threshold for household size23, as defined by the US Census Bureau. PIR was grouped into categories of < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00 for primary analyses. When examining the association of PIR with complementary measures of PF among the n = 56 included in these analyses, PIR was collapsed into categories of < 1.00 (household income below the poverty threshold), 1.00–1.99, and ≥ 2.00 (household income more than twice the poverty threshold) to maximize study power.
Details of PF
Self-reported PF, the primary outcome of interest, was ascertained from the self-administered Medical Outcomes Study Short Form-12 questionnaire (SF-12), a 12-item version of the SF-36 that is validated24 and recommended for use in SLE12. Scores for the PF subscale were calculated from responses to 2 items of the SF-12: “Does your health now limit you in moderate activities, such as moving a table, pushing a vacuum cleaner, bowling, or playing golf?” and “Does your health now limit you in climbing several flights of stairs?”, with possible responses for both items of “Yes, limited a lot,” “Yes, limited a little,” and “No, not limited at all.” The PF score was scaled 0–100, where higher scores represent better functioning25. In sensitivity analyses, PF was dichotomized as “Limited a lot” vs “Not limited a lot” for each question within the scaled PF subscore.
Complementary measures of PF in nested pilot
Physical performance was assessed using the Short Physical Performance Battery (SPPB)26. The SPPB assessed balance (ability to hold standing poses in different foot positions), gait speed (fastest of two 4-m walks at regular pace), and lower body strength (time taken to complete 5 chair stands without using arms), which were scored 0–4 (higher scores indicating better levels of physical performance). The physical performance score was the sum of these 3 individual scores (range 0–12)26.
Activities of daily living
Instrumental activities of daily living (IADL; e.g., food preparation and housework)27 and basic activities of daily living (BADL; e.g., bathing and dressing)28 were self-reported, yielding scores that were dichotomized as the ability to perform the activity independently or with minimal assistance versus the inability to perform the activity without assistance.
Falls
Participants were asked if they had fallen in the past year and how many falls they had had in the past year.
Other variables
All other variables were obtained with the Wave 5 GOAL questionnaires. SLE-related organ damage was assessed using the Self-Administered Brief Index of Lupus Damage (SA-BILD) score (range 0–30), where higher scores indicate greater levels of damage29,30. Depressive symptomatology was assessed with the 9–item Patient Health Questionnaire (PHQ–9; range 0–27), where higher scores indicate more severe depression symptomatology31. Current SLE activity was assessed using the Systemic Lupus Activity Questionnaire (SLAQ; range 0–44), with higher scores indicating greater SLE-related disease activity32. Participants self-reported their age at SLE onset, sex, race, ethnicity, years of education, work status, marital status, social support, and body mass index (BMI). Disease duration was calculated as the difference in age at survey and age at SLE onset.
Statistical analysis
Participant characteristics of GOAL were summarized overall and by PIR category using chi-square test, Fisher’s exact test, ANOVA, or nonparametric equality of medians tests, as appropriate. For the association between PIR and PF scores, slopes (β) and 95% CI were estimated with multivariable linear regression models. Adjustment was performed for age, race, sex, education, marital status, and disease duration, which were considered a priori confounders. Because SLE-related organ damage (SA-BILD), depression (PHQ-9), SLE-related disease activity (SLAQ), and BMI were considered potential mediators as well as confounders, separate adjustment for each of these factors was performed using the fully adjusted multivariable model. Interaction terms between PIR and race, current work status, and depression (PHQ-9 score ≥ 5 vs ≤ 4) were included to assess potential effect modification by these variables. To address the robustness of results to the scoring of PF, sensitivity analyses of the association between PIR and PF were performed using multivariable logistic regression models for each question of the PF subscore to estimate OR (95% CI). Logistic regression analyses used an identical modeling strategy for linear regression models, but did not address effect modification. Complementary measures of physical performance were summarized overall and by PIR category. Scores for physical performance and self-reported functioning were reported as means or percentages, as appropriate. Comparisons of scores across PIR categories were tested through Fisher’s exact test or nonparametric equality-of-means tests, as appropriate. All analyses were conducted using SAS version 9.4 (SAS), and the threshold for statistical significance was set at α = 0.05.
RESULTS
Characteristics of the SLE cohort
Overall, the mean age was 48.0 years, 6.7% were male, and 80.9% were black (Table 1). The prevalence of PIR < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00 among GOAL participants was 37.5%, 21.0%, 29.6%, and 12.0%, respectively. Participants with lower PIR were younger, had lower educational attainment, were less likely to be married, and were more likely to be black. The mean ages at onset of SLE for PIR of < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00 were 30.1, 34.3, 32.6, and 35.1 years, respectively. The mean years of disease duration at the time of the survey differed by PIR category, in that participants with a higher PIR were more likely to have longer disease duration. Both PHQ-9 and SLAQ scores differed significantly by PIR, in that participants with lower PIR were more likely to have higher depressive symptoms and disease activity scores. SA-BILD scores did not differ significantly by PIR category.
Characteristics of patients with systemic lupus erythematosus participating in the Georgians Organized Against Lupus cohort (June 2016–July 2017) overall and categorized by poverty income ratio (PIR).
Association of PIR with self-reported PF in GOAL
The overall mean scaled PF score for GOAL participants was 45.8; and PF scores by PIR category, < 1.00, 1.00–1.99, 2.00–3.99, and ≥ 4.00, were 36.2, 40.7, 55.5, and 61.2, respectively (Table 2). With adjustment for age, sex, and race, participants with a PIR < 1.00 had a PF score that was, on average, 7.0 points lower than participants with a PIR of 1.00–1.99. Participants with a PIR of 2.00–3.99 had a mean PF score that was 13.1 points higher and participants with a PIR ≥ 4.00 had a mean PF score 20.6 points higher than participants with a PIR of 1.00–1.99. Further adjustment for education, marital status, and disease duration did not substantially change these results. After multivariable adjustment and adjusting for SA-BILD, PHQ-9, and SLAQ scores individually, differences in PF scores by PIR were reduced; adjustment for SLAQ scores reduced differences in PF scores the most (Table 2). Interactions between PIR and work status, race, and depression were not statistically significant.
Association between poverty income ratio (PIR) and self-reported physical functioning (PF) among participants with systemic lupus erythematosus in the Georgians Organized Against Lupus cohort (analysis of complete data, n = 744).
Sensitivity analyses
Sensitivity analyses in which the 2 SF-12 questions of the scaled PF score were dichotomized (Table 3) revealed comparable results to primary analyses. In comparison to participants with a PIR of 1.00–1.99, participants with a PIR < 1.00 had 26% increased likelihood of reporting that their health limited moderate activities, while participants with a PIR of 2.00–3.99 had 51% reduced corresponding likelihood, and participants with a PIR ≥ 4.00 had 64% reduced corresponding likelihood. Adjustment for age, sex, race, education, marital status, and disease duration did not considerably change these results. Similarly, individual adjustment of SA-BILD, PHQ-9, and SLAQ scores with multivariable adjustment did not substantially change the association. Participants with a PIR < 1.00 were 30% more likely to report their health limiting their ability to climb several flights of stairs in comparison to those with a PIR of 1.00–1.99, whereas those with a PIR of 2.00–3.99 were 54% less likely and those with a PIR ≥ 4.00 were 61% less likely to report limited ability to climb stairs. Further multivariable adjustment did not substantially change the association of PIR with individuals’ health limiting their ability to climb several flights of stairs. Additional adjustment for SA-BILD, PHQ-9, SLAQ, and BMI scores separately gave similar estimates, but adjusting for SLAQ reduced differences in estimates closer to the null.
Association between poverty income ratio (PIR) and low vs high physical functioning (PF) among participants with systemic lupus erythematosus in the Georgians Organized Against Lupus cohort: sensitivity analysis (analysis of complete data, n = 744).
Complementary PF measures
The mean PF score for the nested pilot participants included in our study was 38.0, where participants with the highest PIR had the highest PF scores (Table 4). The overall mean balance score was 3.6, while the overall mean gait speed score was 3.4; however, neither balance nor gait speed scores differed statistically significantly by PIR category. The mean lower body strength scores for PIR of < 1.00, 1.00–1.99, and ≥ 2.00 were 1.6, 1.4, and 2.7, respectively. For PIR < 1.00, 1.00–1.99, and ≥ 2.00, the mean overall physical performance scores were 8.4, 8.2, and 10.2. Overall, 35.7% of participants reported difficulty with food preparation, 14.3% reported difficulty with housework, 41.1% reported difficulty with shopping, and 12.5% reported difficulty with transportation; yet the only IADL that differed statistically significantly by PIR was transportation, where 22.6% of participants with a PIR < 1.00 reported difficulty with transportation and 0.0% of participants with PIR of 1.00–1.99 and ≥ 2.00 reported difficulty with transportation. Overall, 19.6% of pilot participants included in our study reported difficulty with incontinence, which was the only BADL that differed statistically significantly by PIR: 25.8% of participants with a PIR < 1.00 reported difficulty with incontinence, 30.0% with a PIR of 1.00–1.99 reported difficulty with incontinence, and 0.0% with a PIR ≥ 2.00 reported difficulty with incontinence. The mean number of falls that participants reported in the year previous to the study was 2.1, and falls were less frequently reported among those with a PIR > 2.00 (26.7% vs 48.4% and 70.0% for PIR < 1.00 and 1.00–1.99, respectively; Table 4).
Physical performance and self-reported function overall and categorized by poverty income ratio (PIR) among participants with systemic lupus erythematosus in the Georgians Organized Against Lupus ancillary pilot study (October 2016–April 2017). Higher scores reflect better functioning for all scales.
DISCUSSION
In our study, self-reported PF scores were fairly low in a predominantly black cohort of individuals with SLE (the GOAL study), regardless of PIR category. The overall PF scores were below the mean of the healthy population in which the SF-12 was developed (50.0)33,34, similar to previous studies investigating predictors of PF that also show lower PF scores for individuals with SLE35. On average, participants with higher PIR had higher PF scores in this study. However, differences between PF scores by PIR category were greatest among participants with the highest income relative to poverty level, compared to those at or just above the poverty threshold; whereas those with income below the poverty level had scores similar to those with income at the poverty level.
Participants of the nested pilot were demographically and clinically similar to the overall cohort22, and while those in the nested pilot had lower PF scores than those in the overall cohort (38.0 vs 45.8), associations of complementary measures of PF with PIR in the pilot, on average, reflected associations similar to those observed in the overall GOAL cohort. Although PIR of 2.00–3.99 and ≥ 4.00 were collapsed into a single category for the nested pilot, the lowest and highest PIR categories of the pilot had PF scores similar to the lowest and highest PIR categories of GOAL (36.2 and 61.2 vs 32.3 and 60.0), indicating that levels of PF are not substantially different from the overall cohort from which participants were selected. Differences in physical performance scores, on average, were larger with higher PIR. Of the instrumental activities of daily living (IADL), a greater proportion of individuals with lower versus higher PIR reported difficulties with food preparation, housework, laundry, shopping, and transportation. Other IADL domains showed similar patterns, although they were not statistically significant. Statistically significant differences in the proportion of individuals reporting difficulties with BADL between PIR categories were observed only for incontinence, which was reported only among those at (30%) or below (26%) the poverty threshold.
Previous studies have shown that patients with SLE frequently have muscle weakness, high levels of fatigue, and low rates of physical activity36,37,38, resulting in reduced PF39,40,41, an important aspect of HRQOL. While perceived PF is important in addressing HRQOL, objective measures of PF have not been studied as thoroughly in SLE populations22. Additionally, studies examining PF in SLE often do not use multicomponent assessments of functioning, such as those applied in populations of older adults. Measures of PF in older adults, such as IADL and BADL, history of falls, gait speed, and chair stands42, are predictors of worse mortality and health outcomes26. We found that multiple components of functioning may be associated with socioeconomic status (SES). Further, while lower individual-level and neighborhood-level SES have been shown to be associated with poorer PF among patients with SLE5,7,17, to our knowledge, none have used multiple categories of the income-to-poverty ratio, commonly known as the PIR, which provides a more in-depth measurement of relative poverty.
Because our method of determining PF scores25 is not validated across studies of HRQOL, it is unknown whether the estimated differences reflect clinically important differences in PF. However, using the statistical definition of a minimally important difference in PF scores as half an SD of the PF scores43 from the overall GOAL cohort (= 18.0 points), we found that the range of mean unadjusted PF scores was 36.2–61.2, indicating a minimally important difference in PF scores across all PIR categories by this definition. However, pairwise differences in mean unadjusted PF scores between adjacent PIR categories were not meaningful.
In the ancillary pilot study, substantial levels of impairment in physical performance and self-reported functioning were found, irrespective of PIR category. For many domains, increasing PIR was associated with less impairment; however, we also found slightly greater impairment among participants with a PIR of 1.00–1.99 than among those with a PIR of < 1.00 for balance, lower body strength, and overall physical performance scores. Regardless of PIR, physical performance in this SLE cohort was comparable to, and sometimes lower than, that in the older adult population (age ≥ 70 years) in which the test was developed26. In a population-based sample of older adults born before 1947 in the United States, adults who had more sources of income had faster gait speed44, which corresponds to the similar association of increased gait speed scores with higher PIR found in our study.
Our study has limitations and strengths worth mentioning. First, the study was cross-sectional, which limits causal inference, and the lack of longterm followup data means that we do not know individual trajectories in PIR or PF over time. Exclusions due to missing data, especially with regard to PIR, may have led to selection bias. Because PF scores were determined using 2 questions from the SF-12 survey, the measure may not adequately represent PF, and misclassification may have occurred. Functioning may fluctuate over time with SLE activity, so a single measure of PF may not accurately portray participants’ functioning. As with all observational studies, it is possible that we have not accounted for unknown confounders, and thus have residual confounding. Because our cohort was predominantly black, the study was likely inadequately powered to examine effect modification by race. For the complementary outcomes measured only in our nested pilot study, the small sample size further decreased the power to examine factors that influence, confound, or modify functioning. Generalizability of the results beyond metropolitan Atlanta may be limited, because the cohort is a population-based sample reflecting the demographics of this specific area.
Despite these limitations, our study has several strengths, such as the large sample size of GOAL. A population-based sample of patients with SLE with adequate representation of black individuals yields an accurate portrayal of HRQOL in a diverse cohort. Sensitivity analyses showing that the association between PIR and PF remained after dichotomizing the outcome reduce concerns about whether the measurement of PF scores was too crude. Finally, the use of multidomain functional assessments is relatively novel in SLE populations, providing new insight that may allow for developing individual treatment plans and improving disparities in outcomes.
Lower income, relative to poverty thresholds and household size, may be associated with worse functioning across multiple domains in SLE. Given these results, future studies could include in-depth assessment of SES (taking relative poverty into account), multidomain functioning assessments in a larger cohort, and investigation of trajectories in both relative poverty and functioning. Further research into multicomponent functional assessments to develop individual treatment plans and potentially improve socioeconomic disparities in outcomes is warranted.
Acknowledgment
We thank the participants of GOAL. We also thank Benjamin Tift, Mechelle Lockhart, and Charmayne Dunlop-Thomas for invaluable assistance with recruitment and functioning assessments.
Footnotes
Supported in part by Public Health Service Grant UL1TR000454 from the Clinical and Translational Science Award Program, US National Institutes of Health, National Center for Advancing Translational Sciences. The GOAL cohort study is supported by the US Centers for Disease Control and Prevention Grant 1U01DP005119.
- Accepted for publication February 10, 2020.
REFERENCES
In this issue
