Because Drs. Katz and Yelin are Editors of Arthritis Care & Research, review of this article was handled by the Editor of Arthritis & Rheumatism.
Association of socioeconomic and demographic factors with utilization of rheumatology subspecialty care in systemic lupus erythematosus
Version of Record online: 30 APR 2007
Copyright © 2007 by the American College of Rheumatology
Arthritis Care & Research
Volume 57, Issue 4, pages 593–600, 15 May 2007
How to Cite
Yazdany, J., Gillis, J. Z., Trupin, L., Katz, P., Panopalis, P., Criswell, L. A. and Yelin, E. (2007), Association of socioeconomic and demographic factors with utilization of rheumatology subspecialty care in systemic lupus erythematosus. Arthritis & Rheumatism, 57: 593–600. doi: 10.1002/art.22674
- Issue online: 30 APR 2007
- Version of Record online: 30 APR 2007
- Manuscript Accepted: 20 JUL 2006
- Manuscript Received: 3 APR 2006
- Rosalind Russell Medical Research Center for Arthritis
- US Public Health Service's National Center for Research Resources. Grant Number: 5-M01-RR-00079
- American College of Rheumatology/Rheumatology Education Foundation Physician Scientist Development award
- NIH. Grant Numbers: K24-AR-02175, R01-AR-44804
- Arthritis Foundation's State of California Lupus Fund
- Agency for Healthcare Research
- Quality/National Institute of Arthritis and Musculoskeletal and Skin Diseases. Grant Number: 1-R01-HS-013893
- Systemic lupus erythematosus;
- Health services;
- Office visits/utilization
To examine the role of sociodemographic factors (age, race/ethnicity, and sex) and socioeconomic factors (income and education) in the utilization of rheumatology subspecialty care in a large cohort of subjects with systemic lupus erythematosus (SLE).
Data were derived from a cohort of 982 English-speaking subjects with SLE. Between 2002 and 2004, trained survey workers administered a telephone survey to subjects eliciting information regarding demographics, SLE disease status, medications, health care utilization, health insurance, and socioeconomic status. We identified predictors of utilization of rheumatology subspecialty care, defined as at least 1 visit to a rheumatologist in the previous year. In addition, we examined factors associated with identifying any specialist as primarily responsible for SLE care.
Older age, lower income, Medicare insurance, male sex, and less severe disease were associated with lack of rheumatology care. However, race/ethnicity and educational attainment were not significantly related to seeing a rheumatologist. After multivariate adjustment, only older age, lower income, and male sex remained associated with absence of rheumatology visits. Those least likely to identify a specialist as primarily responsible for their SLE care included older subjects and those reporting lower incomes.
Although elderly subjects and those with lower incomes traditionally have access to health care through the Medicare and Medicaid programs, the presence of health insurance alone did not ensure equal utilization of care. This finding suggests that additional barriers to accessing rheumatology subspecialty care may exist in these patient populations.
Although survival for patients with systemic lupus erythematosus (SLE) has significantly improved over the last several decades, the relative burden of morbidity and mortality caused by the disease remains high. A report from the Centers for Disease Control and Prevention in 2002 indicated 22,861 deaths from SLE from 1979 to 1998, with many occurring in patients younger than 45 years (1). Previous literature suggests that racial and ethnic minorities (2–4) and individuals with low socioeconomic status (5, 6) may have increased disease severity and poorer health outcomes over time. The relative contribution of processes of care, such as utilization of subspecialty services, to these poorer health outcomes has not been adequately determined.
Research concerning other rheumatic diseases has found that specialty care is associated with improved patient outcomes (7–9), and that various demographic groups, including the elderly and racial/ethnic minorities, have decreased utilization of such care (10–12). Several studies have found that patients with rheumatoid arthritis (RA) treated by rheumatologists (versus nonrheumatologists) had slightly better functional status, number of painful joints, overall pain rating, and patient global assessment (7–9). A large historic cohort study of patients with RA found that for all processes of care examined, care that included relevant specialists was associated with significantly higher quality (13). Furthermore, care provided by specialists has not been associated with higher costs (14).
One study attempted to identify whether specific segments of the population are underserved with regard to SLE specialty care. Using Medicare data, the authors found that utilization of rheumatology subspecialty care among patients ages 65 years and older was limited in the 3 states examined (Colorado, Massachusetts, and Virginia) (15). Only 42% of elderly patients with SLE had a visit to a rheumatologist during the observed year. Visits to a rheumatologist were even less frequent for African American patients, particularly for women. These findings are of particular interest because a substantial body of literature has documented increased SLE disease burden among ethnic minorities (16). Minority populations experience a higher incidence of severe end-organ manifestations such as lupus nephritis (3,17), accrue damage more rapidly (18, 19), and may have higher mortality from SLE (6, 20).
Little is known about utilization of rheumatology subspecialty care in other demographic groups, such as younger patients with SLE, and certain racial and ethnic minorities, such as Hispanic/Latino and Asian patients. No studies of SLE have examined use of services by potentially important health system factors, such as type of medical insurance. Furthermore, prior studies have not adjusted for disease severity when reporting utilization. The current study therefore aimed to expand on previous work by examining a number of these important factors related to specialty care utilization in a large cohort of subjects with SLE.
SUBJECTS AND METHODS
Data were derived from the University of California, San Francisco (UCSF) Lupus Outcomes Study (LOS), a large observational cohort of 982 English-speaking subjects with SLE. Details regarding creation of the cohort are described elsewhere (21), but relevant aspects are summarized here. The study reenrolled subjects from a previous UCSF study of genetic factors in SLE. Subjects who had completed enrollment for the genetics study between 1997 and 2002 were asked to participate in the LOS, which involved an annual structured telephone interview and updates of participants' medical records. All subjects were confirmed to have SLE after chart review by either a rheumatologist or a registered nurse working under a rheumatologist's supervision. A total of 83% of subjects eligible for the LOS from the genetics study were successfully contacted, and 78% of those reached agreed to participate between 2002 and 2003. The 982 persons with SLE in the UCSF LOS were originally recruited from academic rheumatology offices (23%), community rheumatology offices (11%), and community-based sources such as SLE support groups, the Internet, and media advertisements (66%). The majority of the respondents reside in California (75%), but a total of 41 states are represented. The UCSF Committee on Human Research approved the study protocol.
Subjects completed a structured telephone interview after providing informed consent. Trained survey workers administered the hour-long interviews eliciting information about demographics, socioeconomic status, SLE disease status, medications, disability, general health and social functioning, health care utilization, and health insurance coverage.
The primary outcome in the study was having ≥1 visit to a rheumatologist within the past year. Predictors of interest included age, sex, disease status, race/ethnicity, income, education, type of health insurance, and subject recruitment source. Age was collected as a continuous variable and was later categorized into 5 groups to facilitate interpretation of age effects (<30 years, 31–40 years, 41–50 years, 51–64 years, and ≥65 years). Race/ethnicity was categorized as Hispanic/Latino, African American, Asian/Pacific Islander, white, or other. Disease was defined as severe if subjects required major immunosuppressive medications (mycophenolate mofetil, cyclophosphamide, methotrexate, intravenous steroids, chlorambucil, or azathioprine) or experienced major end-organ manifestations during the study period (kidney disease requiring biopsy, dialysis, or transplantation; bronchial or open-lung biopsy; hemoptysis; or venous thromboembolism). In additional analyses, we adjusted for SLE status using the patient's global assessment (a single question regarding presence and severity of lupus flare with flare categories including mild, moderate, or severe), a measure found to correlate with validated physician-assessed measures of disease activity (22, 23).
A more detailed SLE disease activity instrument was not available during the first interview year, but was subsequently introduced in the second interview year (Systemic Lupus Activity Questionnaire [SLAQ]) (22). The use of the first interview, which we thought would most accurately capture subject utilization patterns without contamination by the interview process itself, therefore precluded direct use of the SLAQ. However, we found that both disease status measures used in the study predicted higher SLAQ scores in the second interview year (for those reporting an SLE flare in year 1, SLAQ scores were 8.6 points higher [P < 0.0001]; for those with markers of severe disease, SLAQ scores were 2.3 points higher [P < 0.0001]). Furthermore, the patient global assessment in the first year was significantly associated with the patient global assessment and SLAQ score in the second year (r = 0.46 for patient global, P < 0.0001; r = 0.45 for SLAQ score, P < 0.0001). Therefore, disease status measures selected for the analysis were believed to represent reasonable proxies of SLE disease activity.
Questions regarding the type and source of insurance coverage were derived from the Medical Expenditures Panel Survey (24). For the analysis, insurance was categorized as employer based, Medicaid, Medicare, or other (e.g., privately purchased plans, Veterans Affairs). We excluded uninsured subjects because the small number of subjects in this category (n = 19) was inadequate to draw generalizable conclusions. Questions regarding family income and household size were modified from the National Health Interview Survey (25). Income categories included yearly household incomes <$40,000, $40,000–$60,000, $60,000–$80,000, $80,000–$100,000, and >$100,000. In additional analyses, we examined income as a dichotomous variable, defined as subjects reporting incomes <125% of the federal poverty level versus all others. Educational categories included subjects attaining up to a high school education, those with some college/vocational training, those with a bachelor's degree, and those with a postgraduate degree. Recruitment source was dichotomized to include subjects originally enrolled for the genetics study through rheumatology practices versus those enrolled through community-based efforts.
Univariate logistic regression was used to examine the relationship between visits to a rheumatologist within the past year and age, sex, race/ethnicity, income, education, type of health insurance, subject recruitment source, and disease severity. The primary analysis included 867 subjects (86% of the sample) with complete data on the examined predictors.
Multivariate logistic regression was used to examine the relationship between visits to a rheumatologist within the past year and the predictors of interest, including age, sex, race/ethnicity, income, education, subject recruitment source, type of health insurance, and disease severity. Additional models using poverty status, which considers income and household size, were analyzed to explore the income effect. Similarly, multivariate models using patient global assessment, rather than disease severity, were analyzed to explore the role of disease status. Based on theoretical considerations, we examined potential interactions between income and several variables, including age, sex, health insurance, and education.
Multivariate logistic regression was also used to examine predictors of identifying any specialist as primarily responsible for SLE care. Although most subjects reporting a specialist as primarily responsible for their SLE cited a rheumatologist or nephrologist, a small number also referenced dermatologists, cardiologists, hematologists/oncologists, endocrinologists, pulmonologists, or neurologists. Given that SLE may have manifestations limited to these areas, all of these specialists were included in the analysis.
Because of the large number of missing values (9.4%) for the income variable compared with other variables (0.2% for education; 2.5% for principal source of health insurance; and 0% for age, sex, race/ethnicity, or recruitment source), the potential for nonresponse bias for the income variable was considered. Compared with subjects reporting household income, those without income data were both more likely to have lower educational attainment and to have employer-based health insurance; however, both groups were similar with regard to other variables examined (age, sex, disease severity, and ethnicity). We performed a sensitivity analysis to address this concern using single variable imputation for income. A regression model with income as the specified outcome was used to predict missing values for that variable. Factors found by univariate analysis to be associated with income were included in this regression-based imputation, including age, ethnicity, employment status, education, and type of health insurance. The imputed income variable was then substituted for the original income variable in the logistic model. All statistical analyses were completed using STATA software, version 8.0 (StataCorp, College Station, TX).
The final study sample included 867 subjects with SLE (Table 1). Most were women (91.7%), and the mean ± SD age was 47 ± 13.2 years. Sixty-nine percent of subjects self-identified as white, 11% as Hispanic/Latino, 8% as African American, 10% as Asian/Pacific Islander, and 2% as another race/ethnicity. Forty-one percent of the sample reported medications or end-organ manifestations suggesting severe disease, and 49% reported experiencing a disease flare in the 3 months prior to the interview. Of those reporting a flare, 40% had a mild flare, 37% had a moderately severe flare, and 24% had a severe flare.
|Characteristic||Subjects (n = 867)||Subjects with ≥1 rheumatology visits|
|<30||87 (10.3)||73 (83.9)|
|31–40||161 (19.1)||139 (86.3)|
|41–50||255 (30.3)||204 (80.0)|
|51–64||257 (30.6)||183 (71.2)|
|≥65||81 (9.6)||53 (65.4)|
|Female||771 (91.7)||608 (78.9)|
|Male||70 (8.3)||44 (62.9)|
|White||584 (69.4)||443 (75.9)|
|Hispanic/Latino||91 (10.8)||71 (78.0)|
|African American||66 (7.9)||56 (84.9)|
|Asian/Pacific Islander||80 (9.5)||66 (82.5)|
|Other||20 (2.4)||16 (80.0)|
|Milder disease||495 (58.9)||369 (74.5)|
|Severe disease||346 (41.1)||283 (81.8)|
|Patient global assessment|
|No flare in past 3 months||409 (51.0)||297 (72.6)|
|Mild flare||157 (19.6)||130 (82.8)|
|Moderate flare||144 (18.0)||120 (83.3)|
|Severe flare||92 (11.5)||75 (81.5)|
|Employer based||519 (61.7)||413 (79.6)|
|Medicare||195 (23.2)||139 (71.3)|
|Medicaid||49 (5.8)||38 (77.6)|
|Other||78 (9.3)||62 (79.5)|
|Income per year|
|$0–$40,000||312 (37.1)||224 (71.8)|
|$40,000–$60,000||163 (19.4)||133 (81.6)|
|$60,000–$80,000||125 (14.9)||97 (77.6)|
|$80,000–$100,000||94 (11.2)||76 (80.9)|
|>$100,000||147 (17.5)||122 (83.0)|
|Below poverty†||99 (11.4)||69 (69.7)|
|Up to high school graduate||164 (19.5)||129 (78.6)|
|Some college/vocational||345 (41.0)||258 (74.8)|
|Bachelor's degree||213 (25.3)||166 (77.9)|
|Postgraduate||119 (14.2)||99 (83.2)|
|Rheumatology practice||271 (32.2)||212 (78.2)|
|Community‡||570 (67.8)||440 (77.2)|
A total of 22% of subjects reported no visits to a rheumatologist over the past year. However, a majority of the subjects who lacked rheumatologic care had seen a primary care provider (80%) or other physician such as a pulmonologist, nephrologist, dermatologist, or obstetrician/gynecologist (62%). Those who visited a rheumatologist had a mean ± SD of 3.7 ± 4.3 visits (median 3.0 visits), with a small number of subjects (n = 20) reporting >13 visits per year. Many subjects reported care by multiple providers: 78% had both a rheumatologist and a primary care provider, 16% had both a rheumatologist and a nephrologist, and 10% had both a rheumatologist and a pulmonologist.
The results of univariate analyses are displayed in Table 2. Older age was significantly associated with reporting no rheumatology visits in the last year, with subjects ≥65 years of age reporting the fewest visits. As anticipated, subjects with severe disease were more likely to have seen a rheumatologist in the previous year. Similarly, those reporting an SLE flare in the preceding 3 months on the patient global assessment were more likely to have seen a rheumatologist. Male subjects with SLE were less likely to visit a rheumatologist than female subjects, as were those with lower incomes. Medicare insurance was also associated with absence of rheumatology visits in the past year. Of note, Medicare insurance and older age were not perfectly correlated; only 26% of subjects reporting Medicare insurance were older than age 65, and 43% were actually younger than age 50. Medicare recipients under the age of 65 years qualify for this insurance as a result of entitlement to Social Security Disability Insurance. Education, subject recruitment source, and race/ethnicity were not significantly associated with reporting rheumatology visits in the univariate analysis.
|Characteristic||Odds of any rheumatology visits in the last year (n = 867)|
|Unadjusted OR (95% CI)||P||Adjusted OR (95% CI)||P|
|31–40||1.21 (0.59–2.51)||0.61||1.08 (0.51–2.28)||0.83|
|41–50||0.77 (0.40–1.47)||0.42||0.71 (0.37–1.39)||0.32|
|51–64||0.47 (0.25–0.89)||0.02||0.43 (0.22–0.83)||0.01|
|≥65||0.36 (0.17–0.76)||0.007||0.43 (0.19–0.97)||0.04|
|Female||2.20 (1.31–3.69)||0.003||2.61 (1.55–4.39)||< 0.0001|
|Hispanic/Latino||1.13 (0.66–1.92)||0.65||0.78 (0.44–1.38)||0.40|
|African American||1.78 (0.89–3.56)||0.11||1.23 (0.62–2.46)||0.56|
|Asian/Pacific Islander||1.50 (0.82–2.75)||0.19||1.09 (0.57–2.08)||0.79|
|Other||1.27 (0.42–3.87)||0.67||1.49 (0.73–3.01)||0.27|
|Severe disease||1.53 (1.09–2.15)||0.01||1.43 (1.00–2.03)||0.05|
|Medicaid||0.64 (0.44–1.79)||0.74||1.06 (0.65–1.74)||0.80|
|Medicare||0.89 (0.44–0.93)||0.019||1.27 (0.56–2.91)||0.56|
|Other||0.99 (0.55–1.79)||0.99||1.11 (0.59–2.06)||0.75|
|Income per year|
|$0–$40,000||0.57 (0.35–0.93)||0.03||0.52 (0.29–0.95)||0.03|
|$40,000–$60,000||0.93 (0.53–1.66)||0.81||0.92 (0.50–1.71)||0.80|
|$60,000–$80,000||0.72 (0.40–1.29)||0.27||0.70 (0.37–1.30)||0.26|
|$80,000–$100,000||0.80 (0.42–1.52)||0.49||0.74 (0.38–1.46)||0.39|
|Up to high school graduate||–||–||–||–|
|Some college/vocational||0.80 (0.52–1.26)||0.34||0.76 (0.48–1.20)||0.23|
|Bachelor's degree||0.96 (0.58–1.57)||0.87||0.76 (0.44–1.30)||0.32|
|Postgraduate||1.34 (0.73–2.47)||0.34||1.19 (0.62–2.28)||0.60|
|Community||0.94 (0.66–1.33)||0.74||1.08 (0.75–1.57)||0.67|
Multivariate logistic regression results are also displayed in Table 2. Older age, male sex, less severe disease, and low income were associated with a lower likelihood of reporting a rheumatology visit in the last year. The univariate analysis suggested that Medicare insurance was associated with decreased rheumatology visits, but this association did not hold after adjustment for other factors in the multivariate model. Education level and subject recruitment source did not play a large role in predicting rheumatology visits after adjustment for other factors.
The analysis was repeated adjusting for patient global assessment as a marker of SLE disease activity, and the results did not appreciably change (for ages 50–65: odds ratio [OR] 0.39, 95% confidence interval [95% CI] 0.20–0.75, P = 0.005; for ages ≥65: OR 0.40, 95% CI 0.18–0.91, P = 0.03; for female sex: OR 2.24, 95% CI 1.33–3.78, P = 0.002; for presence of disease flare in the preceding 3 months: OR 1.86, 95% CI 1.31–2.65, P = 0.001; for income <$40,000: OR 0.55, 95% CI 0.30–0.99, P = 0.05; data not shown). In addition, when poverty status was substituted for income, the results remained similar (for ages 50–65: OR 0.41, 95% CI 0.21–0.79, P = 0.008; for ages ≥65: OR 0.42, 95% CI 0.19–0.95, P = 0.04; for female sex: OR 2.36, 95% CI 1.38–4.02, P = 0.002; and for poverty status: OR 0.55, 95% CI 0.31–0.98, P = 0.04; data not shown). We did not find any significant interactions of income with potential effect modifiers, including age, sex, health insurance, or disease severity.
Because of a significant degree of nonresponse to the survey question about income, single imputation of income using a regression model was performed. A total of 92 income values were imputed, bringing the total number of subjects analyzed to 954. Although the magnitude of associations between certain predictor variables and the outcome differed slightly, substantive conclusions from the model remained the same. Older age, male sex, and lower income remained significantly associated with absence of rheumatology visits over the past year (for ages 50–65: OR 0.43, 95% CI 0.22–0.83, P = 0.01; for ages ≥65: OR 0.43, 95% CI 0.19–0.97, P = 0.04; for female sex: OR 2.61, 95% CI 1.55–4.39, P < 0.0001; for income <$40,000: OR 0.54, 95% CI 0.30–0.99, P = 0.05; data not shown).
Subjects in the LOS identified the physician primarily responsible for their SLE. A majority identified a rheumatologist (76.6%), but many reported other physicians, including nephrologists (4.2%), internists (11.3%), or general/family practitioners (6.0%). A small number of subjects reported another specialist as primarily responsible for their SLE; 2 reported a dermatologist, 3 reported a hematologist/oncologist, 2 reported a neurologist, and 1 reported each of the following: a pulmonologist, orthopedist, gastroenterologist, and endocrinologist.
Consistent with the results for rheumatology visits, univariate logistic regression results suggest that subjects who were older, reported lower incomes, or had Medicare insurance were less likely to identify a specialist (rheumatologist, nephrologist, or other specialist) as their primary SLE physician (Table 3). Subjects with lower educational attainment were also somewhat less likely to report a specialist as their primary SLE physician. African American and Asian subjects more often identified a specialist as primarily responsible for their SLE care. In the multivariate model, only older subjects, those with lower incomes, and those with milder disease were less likely to identify a specialist as primarily responsible for their SLE. Men were somewhat less likely to identify a specialist as primarily responsible for their SLE, but this did not reach statistical significance.
|Characteristic||Odds of identifying a subspecialist as primarily responsible for SLE care (n = 867)|
|Unadjusted OR (95% CI)||P||Adjusted OR (95% CI)†||P|
|31–40||0.74 (0.31–1.76)||0.49||0.68 (0.28–1.65)||0.39|
|41–50||0.57 (0.26–1.27)||0.17||0.59 (0.25–1.34)||0.21|
|51–64||0.29 (0.13–0.63)||0.002||0.32 (0.14–0.72)||0.006|
|≥65||0.20 (0.83–0.47)||< 0.0001||0.27 (0.11–0.70)||0.007|
|Female||1.44 (0.82–2.52)||0.20||1.57 (0.86–2.88)||0.14|
|Hispanic/Latino||0.71 (0.41–1.23)||0.23||0.64 (0.35–1.15)||0.14|
|African American||2.42 (1.02–5.74)||0.05||2.28 (0.93–5.61)||0.07|
|Asian/Pacific Islander||2.52 (1.13–5.63)||0.02||1.53 (0.66–3.53)||0.32|
|Other||1.25 (0.64–2.47)||0.51||1.26 (0.61–2.58)||0.53|
|Severe disease||2.09 (1.42–3.05)||< 0.0001||1.96 (1.31–2.94)||0.001|
|Medicaid||0.62 (0.31–1.22)||0.17||0.75 (0.33–1.70)||0.49|
|Medicare||0.57 (0.38–0.85)||0.006||1.06 (0.62–1.81)||0.82|
|Other||0.82 (0.44–1.53)||0.54||0.97 (0.49–1.91)||0.93|
|Income per year|
|$0–$40,000||0.40 (0.23–0.70)||0.001||0.47 (0.24–0.93)||0.03|
|$40,000–$60,000||0.81 (0.42–1.58)||0.54||0.91 (0.44–1.85)||0.79|
|$60,000–$80,000||0.69 (0.34–1.38)||0.30||0.78 (0.38–1.61)||0.50|
|$80,000–$100,000||0.59 (0.29–1.23)||0.16||0.60 (0.28–1.27)||0.18|
|Up to high school graduate||–||–||–||–|
|Some college/vocational||0.99 (0.63–1.56)||0.98||0.89 (0.55–1.43)||0.63|
|Bachelor's degree||1.62 (0.95–2.77)||0.08||1.18 (0.65–2.11)||0.59|
|Postgraduate||1.81 (0.95–3.44)||0.07||1.48 (0.73–2.98)||0.28|
|Community||0.79 (0.58–1.09)||0.15||0.71 (0.46–1.09)||0.12|
African American subjects were somewhat more likely to identify a specialist as primarily responsible for their SLE. Compatible with results from other studies, African American subjects in the cohort were more likely to report severe disease (OR 2.05, 95% CI 1.23–3.4, P = 0.006), have markers of kidney disease such as undergoing a renal biopsy (OR 2.99, 95% CI 1.25–7.17, P = 0.01), and to report seeing a nephrologist (OR 1.99, 95% CI 1.12–3.52, P = 0.02; data on relationship between race/ethnicity and disease manifestations not shown).
The UCSF LOS draws on a diverse population of subjects recruited largely from the community and includes detailed self-reported data. These attributes have allowed us to look at utilization of rheumatology physician services while controlling for important potential confounding factors not accounted for in the previous literature. We found that older subjects and those with lower incomes were less likely to visit a rheumatologist, even after controlling for sociodemographic variables, health insurance, and disease status. Although it is possible that elderly individuals or those with lower incomes either prefer not to see rheumatologists or tend to underreport utilization, it seems more plausible that these groups face barriers to accessing care. Such barriers may include lack of rheumatology physician services in geographic proximity to patient residences, unawareness of rheumatology subspecialty care, less frequent referrals to rheumatologists, or inadequate followup by rheumatologists once care is established.
In this study, the magnitude of the differences in utilization of rheumatology services remained sizeable after multivariable adjustment. Subjects who were older than 50 years or in the lowest income category were twice as likely to report no rheumatology visits in the past year. These findings are particularly striking given that participants in the LOS conceivably have greater access to care than the general population with SLE, and given that the sampling frame of our study would tend to underestimate any differences. A prior study using Medicare physician claims found significantly lower rates of utilization of rheumatology services among the elderly, and that African Americans with SLE had significantly fewer rheumatology visits compared with whites (15). However, because analyses in that study relied on administrative data, diagnoses of SLE were unconfirmed. Moreover, the authors could not adjust for disease activity or socioeconomic status. These factors may partly explain the larger magnitude of utilization differences found in this study compared with the LOS. Although we did not find a similar discrepancy in specialty care utilization by race/ethnicity, the limitations of sampling in the LOS, discussed below, cannot definitively rule out such a difference.
Subjects in the LOS reported differing principal sources of outpatient care, with approximately one-fourth identifying a provider other than a rheumatologist as primarily responsible for their SLE. More than 10% reported that an internist primarily cared for their disease, whereas others identified general or family practitioners, nephrologists, or other specialists. After adjustment for other factors, those least likely to identify a specialist as their primary SLE physician included older subjects and those with lower incomes. Men were also somewhat less likely to report a specialist as their primary SLE physician. Why might men be less likely to see a rheumatologist or to identify a specialist as primarily responsible for their SLE? The relatively low prevalence of SLE among men may explain the paucity of research in this area. Both physician referral patterns for men with SLE and male patients' perceptions of disease warrant further study in this predominantly female disease.
Although our analysis adjusted for disease status, this may serve as an imperfect surrogate for patient need for rheumatology subspecialty care. Do patients with quiescent SLE require monitoring by a rheumatologist? In a previous study examining the role of rheumatology care in patients with inactive SLE, Urowitz and colleagues found that, among patients with an SLE Disease Activity Index score of 0 over 2 consecutive visits, a substantial proportion (74%) required an intervention during a subsequent routine clinic visit (26). A majority of these interventions related to management of SLE itself, implying that ongoing monitoring of these patients may be required. The need for regular monitoring may also extend to older patients, a group that is often thought to have decreased SLE activity (27). Interestingly, many elderly subjects in our sample were classified as having severe disease largely based on therapy with major immunosuppressive medications such as azathioprine, mycophenolate mofetil, or methotrexate, all of which require regular monitoring.
In an observational study such as the LOS, some caution is required in interpreting the results. Nonrandom sampling of subjects may have led to selection bias. Although adjusting for the recruitment source of patients had no significant effect on the findings, it is likely that this cohort has increased access to care overall. In particular, members of racial/ethnic minority groups were more likely to be recruited from rheumatology practices in the years preceding eligibility for the LOS (OR 1.2, 95% CI 1.10–1.33), raising the possibility that our results may overestimate specialty utilization for these groups. Moreover, the proportion of African Americans in the LOS falls below the number expected based on the prevalence of disease in this population, suggesting that the distribution of race/ethnicity in this sample may not reflect the general population with SLE. In addition, although we attempted to adjust for disease severity using both the patient global assessment and self-reported disease manifestations and medications, these are likely imperfect surrogates for detailed clinical information, and may again lead to underestimates of utilization differences in minority participants who tend to have more severe disease.
Finally, we are unable to generalize our findings to certain population groups, such as non–English-speaking or uninsured patients. Members of minority groups who speak limited English, particularly Asian/Pacific Islanders and Hispanic/Latinos, may indeed have differential access to care that is not addressed in this study. Despite these limitations, our finding that English-speaking, insured racial/ethnic minorities have similar utilization patterns compared with whites seems consistent with recent literature; racial/ethnic disparities in health care utilization among insured populations often diminish after controlling for socioeconomic factors (28, 29).
In conclusion, elderly, lower-income, and male subjects with SLE had decreased utilization of rheumatology physician services in our cohort, even after adjusting for sociodemographic factors, health insurance, and disease severity. Although elderly patients and those with lower incomes traditionally have access to health care through the Medicare and Medicaid programs, the presence of health insurance alone did not ensure equal utilization of care. This finding suggests that additional barriers to accessing rheumatology subspecialty care may exist in these patient populations.
Dr. Yazdany had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Yazdany, Zell Gillis, Katz, Criswell, Yelin.
Acquisition of data. Trupin, Criswell.
Analysis and interpretation of data. Yazdany, Zell Gillis, Trupin, Katz, Panopalis, Criswell, Yelin.
Manuscript preparation. Yazdany, Zell Gillis, Katz, Panopalis, Criswell, Yelin.
Statistical analysis. Yazdany, Trupin.
- 1Centers for Disease Control and Prevention (CDC). Trends in deaths from systemic lupus erythematosus: United States, 1979–1998. MMWR Morb Mortal Wkly Rep 2002; 51: 371–4.
- 25National Center for Health Statistics. Survey description document. In: 2002 National Health Interview Survey. Hyattsville (MD): Centers for Disease Control and Prevention; 2003. URL: ftp://ftp.cdc.gov/pub/HealthStatistics/NCHS/DatasetDocumentation/NHIS/2002/srvydesc.pdf.