SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Objective

Although studies have demonstrated suboptimal preventive care in RA patients, performance of primary lipid screening (i.e., testing before cardiovascular disease [CVD], CVD risk equivalents, or hyperlipidemia is evident) has not been systematically examined. The purpose of this study was to examine associations between primary lipid screening and visits to primary care providers (PCPs) and rheumatologists among a national sample of older RA patients.

Methods

This retrospective cohort study examined a 5% Medicare sample that included 3,298 RA patients without baseline CVD, diabetes mellitus, or hyperlipidemia, who were considered eligible for primary lipid screening during the years 2004–2006. The outcome was probability of lipid screening by the relative frequency of primary care and rheumatology visits, or seeing a PCP at least once each year.

Results

Primary lipid screening was performed in only 45% of RA patients. Overall, 65% of patients received both primary and rheumatology care, and 50% saw a rheumatologist as often as a PCP. Any primary care predicted more lipid screening than lone rheumatology care (26% [95% confidence interval (95% CI) 21–32]). As long as a PCP was involved, performance of lipid screening was similar regardless of the balance between primary and rheumatology visits (44–48% [95% CI 41–51]). Not seeing a PCP at least annually decreased screening by 22% (adjusted risk ratio 0.78 [95% CI 0.71–0.84]).

Conclusion

Primary lipid screening was performed in fewer than half of eligible RA patients, highlighting a key target for CVD risk reduction efforts. Annual visits to a PCP improved lipid screening, although performance remained poor (51%). Half of RA patients saw their rheumatologist as often or more often than they saw a PCP, illustrating the need to study optimal partnerships between PCPs and rheumatologists for screening patients for CVD risks.

Although patients with rheumatoid arthritis (RA) are most often cared for by both primary care providers (PCPs) and rheumatologists, preventive screening remains suboptimal (1, 2), and the mortality gap between RA patients and their peers has widened (3, 4). Cardiovascular disease (CVD) is the leading cause of death in patients with RA. These patients experience a 10-year risk of CVD events that is 50–60% higher than that in their age-matched peers (5, 6). The reductions in rates of death from cardiovascular causes seen in the general population in recent decades (7) have not been seen among patients with RA (3–5), and cardiovascular risk has not equilibrated even with aggressive RA treatments (8). Consequently, adequate screening for traditional CVD risk factors is strongly indicated for RA patients.

Primary preventive screening before the onset of CVD is key to identifying modifiable traditional CVD risk factors. To date, the performance of primary preventive lipid screening, meaning testing before the onset of CVD, CVD risk equivalents, or established hyperlipidemia, has not been systematically examined in a national RA sample. Compounding the challenge, no widely known RA-specific preventive guidelines for CVD exist despite increased CVD risk in RA. The European League Against Rheumatism (EULAR) has issued recommendations for an annual CVD risk review (9). For all adults, the National Cholesterol Education Program (NCEP) recommends lipid screening in those with CVD risk factors “more frequently than every 5 years” (10).

Prior reports suggested that RA patients frequently experience unidentified and uncontrolled traditional modifiable CVD risk factors, including hyperlipidemia and hypertension (11–13). Although it has not been studied specifically in RA, according to the recent Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER) study, elevations in the C-reactive protein level may also merit consideration of lipid-lowering therapy (14, 15). RA patients see multiple physicians annually, with rheumatology visits often outnumbering primary care encounters (16). The influence of multisource care and competing comorbidities raises questions of whether the process of care can be optimized to improve primary preventive screening for patients with RA and other chronic conditions (17). Older adults in particular are receiving aggressive treatment for RA (18) but are at greatest absolute risk of coronary events. They may also be most vulnerable to lapses in preventive care due to competing comorbidities and multisource care. As a result, older RA patients represent a key target population for CVD risk factor modification.

In the present study, we investigated the impact of rheumatology and primary care outpatient visit patterns upon primary preventive lipid screening among a group of older adults with RA. We specifically examined whether individual likelihood of lipid screening differed by types of providers seen each year as well as the relative proportions of visits to PCPs and rheumatologists. Reflecting the more conservative NCEP recommendation versus the EULAR recommendations, we examined lipid screening over a 3-year window.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Setting and participants.

In this retrospective cohort study, patients ages 65 and older who were continuously enrolled and alive from January 1, 2004 through December 31, 2006 were identified from a national 5% random sample of Medicare beneficiaries. Patients were determined to have RA if they had 2 or more International Classification of Diseases, Ninth Edition (ICD-9) codes for RA (714.0–714.33) on inpatient or outpatient claims at least 2 months apart during the 24-month period of January 1, 2004 through December 31, 2005, based upon a previously validated algorithm (1, 19). Exact correlation with the American College of Rheumatology (ACR) diagnostic criteria was not determined due to data limitations.

Enrollment and claims data for calendar years 2004–2006 were extracted for 29,425 patients meeting the RA definition. The Medicare denominator file was used to exclude beneficiaries without continuous Medicare Part A or B coverage, beneficiaries with supplemental health maintenance organization (HMO) or railroad benefits, or beneficiaries who died prior to December 31, 2006 (n = 3,970). The Institutional Review Board at the University of Wisconsin approved this study with a waiver of consent.

Given more strict standards for secondary lipid testing intervals, patients were excluded if they had preexisting CVD or diabetes mellitus, as indicated by a flag for those conditions dated before January 1, 2004 in the Chronic Condition Data Warehouse (CCW) Medicare data set (n = 17,707), or if they had lone hyperlipidemia at baseline (n = 3,675). The CCW database contains flags created using validated algorithms applied biannually since 1999 to define 21 chronic diseases (20–25). Exclusion flags from the CCW data set were preexisting CVD (myocardial infarction, stroke, heart failure, or ischemic heart disease) (21–24, 26) or diabetes mellitus, a CVD risk equivalent (20, 25). Baseline hyperlipidemia was identified and excluded based on more than one ICD-9 code (272.0–272.4) in the 24 months of years 2004–2005 (27). Given that the outcome of interest was outpatient lipid screening, we also excluded patients without any outpatient encounters during the years 2004–2006 (n = 775).

Ultimately, our remaining 3,298 patients who were eligible for primary prevention lipid screening from this 5% sample represented nearly 66,000 Medicare RA patients nationwide.

Variables.

All variables were obtained from the Medicare data. The main dependent variable was receiving screening for hyperlipidemia during the 3-year period of 2004–2006. This time period was selected based on existing recommendations, although no RA-specific cardiovascular screening guidelines exist. Lipid screening was identified by Current Procedural Terminology (CPT) codes indicating that the following tests had been performed: lipid panel (80061), low-density lipoprotein (LDL) cholesterol (83721), electrophoretic lipoprotein (83715), high-resolution lipoprotein (e.g., nuclear magnetic resonance analysis) (83716), electrophoretic or high-resolution lipoproteins (83700, 83701, or 83704), or calculated LDL components (82465, 83718, and 84478) (28). The patient was considered to have had a lipid screen if any of these CPT codes was present at least once over the 3 years.

The main explanatory variables were relative frequency of rheumatology and primary care visits or a dichotomous representation of seeing a PCP at least once each year. The first was a categorical representation of primary care and rheumatology visit patterns determined by examining the relative proportions of Medicare visit claims reflecting the type of practitioner. PCPs were defined as family medicine or internal medicine physicians, physician assistants, or nurse practitioners (29, 30). Combined care from a rheumatologist and a PCP was subdivided into “PCP ≥ Rheum” if annual average primary care visits equaled or outnumbered rheumatology visits and “Rheum > PCP” if rheumatology visits predominated. Patients who had primary care visits without any rheumatology visits over the 3-year period were considered “Lone PCP”; those with rheumatology visits and no primary care visits were considered “Lone Rheum.” A second analysis examined the likelihood of lipid testing in those who saw a PCP at least once during each of the 3 years from 2004 to 2006 as compared to those who did not.

Individual sociodemographic characteristics were included as other potential explanatory variables. These were age, sex, race, designation of ever receiving Medicaid, and residence grouped using US Department of Agriculture/US Census Bureau–based rural–urban commuting area codes (31). Claims for a gait assistance device or qualifying orthopedic surgery (32, 33) during the study period were used as RA disease surrogates to compare severity. Additionally, patient comorbidities were assessed using the Hierarchical Condition Categories (HCC) scale devised by the Centers for Medicare and Medicaid Services (34). This validated measure uses inpatient and ambulatory claims during the baseline year (in this case, 2004) to calculate predicted expenditures, reflecting comorbidities that increase health care utilization; a score of 1 represents the predicted cost for an average Medicare patient. Measures of utilization included average annual number of outpatient visits and total number of unique providers (primary care, rheumatology, and non-rheumatology specialists), as well as ever being hospitalized between 2004 and 2006.

Statistical analysis.

Logistic regression with robust estimates of the variance was used to analyze the relationship between explanatory variables and ever receiving lipid screening. Adjusted and unadjusted probabilities of screening were estimated by visit pattern. Age, sex, race, Medicaid status, prior hospitalization status, prior orthopedic surgery, prior gait-assistance device, HCC comorbidity, and average numbers of annual providers and annual visits were included within logistic models based upon theoretical importance. Given previous work demonstrating optimal cancer screening among RA patients with both rheumatology and primary care visits (1), the “PCP ≥ Rheum” category served as the referent group for the visit pattern variable.

Analyses were conducted using SAS version 9.1.3 (SAS Institute) and Stata version 10.0 (StataCorp) software. Results of logistic regression were reported as adjusted predicted probabilities, adjusted risk ratios (ARRs), and 95% confidence intervals (95% CIs) (35). Adjusted predicted probabilities were estimated based on the recycled predictions approach, using the Stata margins command. This approach predicts the outcome (lipid screening), assuming that everyone in the data set was treated as if they had a certain visit pattern. Confidence intervals were calculated using the delta method and allowed correlation among observations analogous to the robust option for estimating the logistic regression.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Descriptive characteristics.

As expected, our sample consisted of predominantly white women with RA who were generally healthy, although other patient characteristics varied significantly by visit pattern. Among the final sample of 3,298 RA patients, 58.4% were ages 65–74 years, 83.4% were female, and 90.0% were white (Table 1). During the 3-year period of January 1, 2004 through December 31, 2006, 39.0% were hospitalized at least once. The HCC scale comorbidity score at baseline was 0.81, suggesting that the RA patients in our primary CVD prevention study were predicted to have lower than average Medicare expenditures. Sixty-five percent of the patients saw both a PCP and a rheumatologist over the period of observation.

Table 1. Characteristics of Medicare beneficiaries with rheumatoid arthritis who were eligible for primary lipid screening, by visit pattern*
CharacteristicAll patients (n = 3,298)Visit pattern
Lone PCP (n = 886)PCP ≥ Rheum (n = 1,139)Rheum > PCP (n = 1,017)Lone Rheum (n = 256)
  • *

    See Patients and Methods for explanation of the visit patterns. PCP = primary care provider; Rheum = rheumatologist; HCC = Hierarchical Condition Categories.

Age, no. (%)     
 65–74 years1,927 (58.4)441 (49.8)639 (56.1)682 (67.1)165 (64.5)
 75–84 years1,163 (35.3)362 (40.9)424 (37.2)302 (29.7)75 (29.3)
 85+ years208 (6.3)83 (9.4)76 (6.7)33 (3.2)16 (6.3)
Female, no. (%)2,750 (83.4)728 (82.2)988 (86.7)837 (82.3)197 (77.0)
Race/ethnicity, no. (%)     
 White2,967 (90.0)760 (85.8)1,038 (91.1)938 (92.2)231 (90.2)
 Black212 (6.4)74 (8.4)70 (6.2)52 (5.1)16 (6.3)
 Other119 (3.6)52 (5.9)31 (2.7)27 (2.7)9 (3.5)
Medicaid, no. (%) ever341 (10.3)150 (16.9)95 (8.3)67 (6.6)29 (11.3)
Hospitalization, no. (%) ever1,285 (39.0)384 (43.3)516 (45.3)317 (31.2)68 (26.6)
Orthopedic surgery, no. (%) ever888 (26.9)202 (22.8)330 (29.0)306 (30.1)50 (19.5)
Gait device, no. % ever660 (20.0)206 (23.3)256 (22.5)161 (15.8)37 (14.5)
HCC scale comorbidity score, mean ± SD0.81 ± 0.450.80 ± 0.440.83 ± 0.500.80 ± 0.420.78 ± 0.37
Rural/urban commuting areas, no. (%)     
 Urban1,990 (60.5)484 (54.8)685 (60.3)677 (66.6)144 (56.5)
 Suburban327 (9.9)91 (10.3)117 (10.3)97 (9.6)22 (8.6)
 Large town452 (13.7)150 (17.0)168 (14.8)111 (10.9)23 (9.0)
 Small town522 (15.9)158 (17.9)167 (14.7)131 (12.9)66 (25.9)

As classified, the entire sample included 27% who received lone primary care (n = 886), 34% who received at least the same number of primary care visits as rheumatologist visits (n = 1,139), 31% who received fewer primary care visits than rheumatologist visits (n = 1,017), and 8% who received lone rheumatology care (n = 256). All patient characteristics varied by visit pattern. Patients without any primary care were younger, more rural, and had lower HCC risk adjustment scores than those with a PCP. They were less likely to have been hospitalized, to have undergone orthopedic surgery, or to have received gait devices.

Patients saw an average of 6 providers in an average of almost 9 annual visits (Table 2). The mean number of primary care visits exceeded the mean number of other non-rheumatology specialty visits (3.4 visits annually compared to 2.8). Both of these categories of visits outnumbered the average of 2.4 annual rheumatology visits among this RA group. Those without primary care saw fewer total providers (mean ± SD 3.3 ± 2.1 versus 6.2 ± 3.6) in fewer annual visits (mean ± SD 5.4 ± 4.7 versus 8.6 ± 5.5). The overall ratios of primary care visits to rheumatology visits varied widely, although approximately half of all patients saw their rheumatologist at least as often as they saw their PCP (Figure 1).

Table 2. Unique provider totals and annual average visits among rheumatoid arthritis patients eligible for primary lipid screening, by visit pattern*
CharacteristicAll patients (n = 3,298)Visit pattern
Lone PCP (n = 886)PCP ≥ Rheum (n = 1,139)Rheum > PCP (n = 1,017)Lone Rheum (n = 256)
  • *

    Unique provider totals reflect a 3-year total wherein physician assistant and nurse practitioner visits were attributed to primary care. Annual average visits indicates the annualized values determined by dividing the visit totals from 2004 to 2006 by 3 years. See Patients and Methods for explanation of the visit patterns. Values are the mean ± SD (range). PCP = primary care provider; Rheum = rheumatologist; NA = not available.

Total unique providers6.2 ± 3.6 (1–30)5.1 ± 3.0 (1–18)7.6 ± 3.8 (2–30)6.4 ± 3.4 (2–27)3.3 ± 2.1 (1–13)
Unique provider     
 PCPs2.1 ± 1.6 (0–17)2.3 ± 1.4 (1–16)2.8 ± 1.7 (1–17)1.8 ± 1.2 (1–10)NA
 Rheumatologists0.8 ± 0.7 (0–7)NA1.2 ± 0.5 (1–6)1.3 ± 0.6 (1–7)1.1 ± 0.3 (1–3)
 Other specialties3.2 ± 2.6 (0–21)2.8 ± 2.4 (0–13)3.7 ± 2.8 (0–21)3.3 ± 2.8 (0–12)2.2 ± 2.0 (0–12)
Annual average     
 Outpatient visits8.6 ± 5.5 (0.3–56)6.8 ± 4.6 (0.3–41)10.2 ± 5.8 (0.7–56)9.1 ± 5.3 (1.3–51)5.4 ± 4.7 (0.3–54)
 PCP visits3.4 ± 3.0 (0–35)4.4 ± 3.2 (0.3–28)4.9 ± 3.1 (0.3–35)1.8 ± 1.3 (0.3–8)NA
 Rheumatology visits2.4 ± 2.8 (0–51)NA2.2 ± 1.4 (0.3–10)4.4 ± 3.2 (0.7–46)3.3 ± 3.4 (0.3–52)
 Other specialty visits2.8 ± 3.1 (0–38)2.4 ± 2.7 (0–21)3.2 ± 3.3 (0–38)2.8 ± 3.1 (0–25)2.1 ± 2.5 (0–17)
thumbnail image

Figure 1. Distribution of the ratios of primary care visits to rheumatology visits among 3,298 patients with rheumatoid arthritis. The ratios were calculated over 3 years (from January 1, 2004 through December 31, 2006) and were rounded to the nearest integer. Ratio 1:0 (left) represents lone primary care provider (Lone PCP) visits, ratio 0:1 (right) represents lone rheumatology care (Lone Rheum) visits, and ratio 1:1 (center) represents individuals with equal proportions of visits. See Patients and Methods for further explanations.

Download figure to PowerPoint

Lipid screening.

The adjusted results showed that lipid screening occurred in 45.1% of eligible patients (Table 3). Adjusted predicted probabilities of lipid screening ranged from 26% (95% CI 21–32) of patients with lone rheumatology care to 44–48% (95% CI 41–51) of patients with at least some primary care visits. These percentages differed only slightly from the unadjusted results. When compared to the “PCP ≥ Rheum” referent group, patients with lone rheumatology care demonstrated significantly lower lipid screening (ARR 0.55 [95% CI 0.42–0.67]). There was no difference in lipid screening between the “Lone PCP,” “Rheum > PCP,” and the “PCP ≥ Rheum” groups by either statistical method.

Table 3. Analysis of the effect of visit pattern and seeing a PCP annually on multivariate adjusted predicted probability, adjusted risk ratio, and 95% CIs for logistic regression predicting lipid screening*
CharacteristicUnadjusted probability, %Adjusted predicted probability, % (adjusted 95% CI)Adjusted risk ratio (adjusted 95% CI)
  • *

    Final logistic models included either visit pattern or dichotomized annual primary care provider (PCP) visit status and all of the following: age category, sex, race, Medicaid status, Hierarchical Condition Category, residence, and annual total visit and provider quartiles. See Patients and Methods for explanation of the visit patterns. Rheum = rheumatologist.

  • Statistically significant within-group variation by 95% confidence interval (95% CI).

All patients (n = 3,298)45.145.1 (43.3–46.7) 
Screening by visit pattern   
 Lone PCP (n = 886)43.147.4 (44.0–50.8)0.99 (0.89–1.08)
 PCP ≥ Rheum (n = 1,139)50.748.1 (45.1–51.1)1.00 (reference)
 Rheum > PCP (n = 1,017)46.544.1 (41.0–47.1)0.92 (0.83–1.0)
 Lone Rheum (n = 256)21.526.3 (20.5–32.2)0.55 (0.42–0.67)
Screening by seeing PCP at least annually   
 <1 PCP visit/year (n = 1,671)39.439.4 (37.0–41.9)0.78 (0.71–0.84)
 ≥1 PCP visit/year (n = 1,627)50.950.8 (48.3–53.3)1.00 (reference)

In our second analysis, those who did not see a primary care provider at least annually were less likely to undergo lipid testing: 39.4% versus 50.9% (Table 3). ARR calculations predicted a 22% lower likelihood of lipid screening in RA patients who did not see a PCP at least once each year (ARR 0.78 [95% CI 0.71–0.84]).

Predictors of lipid screening.

Among our other explanatory variables, a higher provider quartile (reflecting more total unique providers) predicted a higher probability of lipid screening (Table 4). Older age, greater risk-adjustment scores (reflected by higher HCC quartile), large town residence, and lowest annual visit quartile (reflecting least number of outpatient visits) were associated with a lower likelihood of lipid screening (Tables 4 and 5).

Table 4. Multivariate adjusted predicted probability, adjusted risk ratio, and 95% CIs for logistic regression predicting lipid screening, by visit pattern, age group, sex, race/ethnicity, Medicaid status, and HCC comorbidity quartile (n = 3,298)*
CharacteristicUnadjusted probability, %Adjusted predicted probability, % (adjusted 95% CI)Adjusted risk ratio (adjusted 95% CI)
  • *

    Final logistic model included visit pattern, age category, sex, race, Medicaid status, Hierarchical Condition Category (HCC), residence, and annual total visit and provider quartiles. See Patients and Methods for explanation of the visit patterns. PCP = primary care provider; Rheum = rheumatologist.

  • Statistically significant within-group variation by 95% confidence interval (95% CI).

All patients45.145.1 (43.3–46.7) 
 Lone PCP43.147.4 (44.0–50.8)0.99 (0.89–1.08)
 PCP ≥ Rheum50.748.1 (45.1–51.1)1.00 (reference)
 Rheum > PCP46.544.1 (41.0–47.1)0.92 (0.83–1.0)
 Lone Rheum21.526.3 (20.5–32.2)0.55 (0.42–0.67)
Age group   
 65–74 years49.949.3 (47.1–51.6)1.00 (reference)
 75–84 years40.440.6 (37.9–43.5)0.82 (0.76–0.89)
 85+ years26.929.8 (23.3–36.2)0.60 (0.47–0.74)
Sex   
 Female44.944.6 (42.7–46.4)0.94 (84.9–1.03)
 Male46.047.6 (43.4–51.7)1.00 (reference)
Race/ethnicity   
 White45.144.8 (43.0–46.6)0.93 (0.75–1.11)
 Black44.846.5 (39.6–53.4)0.96 (0.74–1.20)
 Other46.248.3 (39.1–57.4)1.00 (reference)
Medicaid (ever)38.142.4 (36.8–47.9)0.93 (0.81–1.06)
HCC comorbidity quartile   
 Lowest quartile49.948.8 (45.2–52.4)1.00 (reference)
 Second quartile47.047.7 (44.4–51.0)0.98 (0.88–1.08)
 Third quartile45.144.8 (41.3–48.3)0.92 (0.82–1.02)
 Highest quartile39.339.8 (36.5–43.0)0.82 (0.72–0.91)
Table 5. Multivariate adjusted predicted probability, adjusted risk ratio, and 95% CIs for logistic regression predicting lipid screening, by residence category, annual total visit quartile, and total provider quartile (n = 3,298)*
CharacteristicUnadjusted probability, %Adjusted predicted probability, % (adjusted 95% CI)Adjusted risk ratio (adjusted 95% CI)
  • *

    Final logistic model included visit pattern, age category, sex, race, Medicaid status, Hierarchical Condition Category, residence, and annual total visit and provider quartiles.

  • Statistically significant within-group variation by 95% confidence interval (95% CI).

Residence category   
 Urban47.447.2 (44.9–49.3)1.00 (reference)
 Suburban43.142.3 (37.1–47.5)0.90 (0.78–1.02)
 Large town39.138.6 (34.2–42.9)0.82 (0.72–0.92)
 Small town42.344.7 (40.4–48.9)0.95 (0.85–1.05)
Annual total visit quartiles   
 Lowest quartile (1–5)32.640.7 (36.5–44.9)0.85 (0.75–0.95)
 Second quartile (>5–8)45.747.7 (44.5–51.0)1.00 (reference)
 Third quartile (>8–11)46.746.7 (43.0–50.4)0.98 (0.88–1.08)
 Highest quartile (≥11)52.944.8 (40.9–48.6)0.94 (0.83–1.05)
Total provider quartiles   
 Lowest quartile (0–3)37.235.1 (30.8–39.4)0.64 (0.53–0.74)
 Second quartile (>3–5)43.242.6 (39.2–45.9)0.77 (0.68–0.86)
 Third quartile (>5–8)50.146.8 (43.5–50.0)0.85 (0.76–0.93)
 Highest quartile (≥8)51.255.2 (50.9–59.5)1.00 (reference)

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Recognizing CVD as the leading cause of death in RA, we sought to examine primary preventive screening for hyperlipidemia as a modifiable cardiac risk factor among RA patients in relationship to primary care and rheumatology outpatient visits. Overall, lipid testing occurred in fewer than half of all those who were eligible over 3 years. Patients with lone rheumatologic care had substantially less lipid screening as compared to patients with any primary care: 22% for Lone Rheum versus 43–51% for other visit patterns that included at least some primary care. Those who saw a PCP at least once each year fared best, with 51% undergoing testing. This finding was consistent with a study done in 2000, in which lower routine cancer screening among RA patients without primary care contact was reported, and with a 2010 study examining multiple preventive services in RA (1, 26). However, in our study, even with primary care involvement, the observed rates of lipid screening remained poor regardless of visit proportions, suggesting a need to systematically improve preventive cardiovascular care for patients with RA.

We found that primary lipid screening for RA patients was significantly less frequent than the rates reported among average ambulatory Medicare beneficiaries, which has been estimated at 50–55% each year (36). Our low observed rate of screening performance also contrasts with the aforementioned 2010 RA study that reported 5-year lipid testing performance of 83.5% (26). However, that study did not separately examine primary versus secondary CVD risk screening populations or control for prevalent hyperlipidemia, CVD, or risk equivalents. Maintenance testing among secondary prevention populations, as compared with actual primary lipid screening, likely inflated the observed rates in that study, although longer observation may have also influenced the results. Performance in our primary prevention RA cohort was more analogous to screening in younger average-risk HMO populations, whose 3-year LDL testing frequency was ∼40% (37). The observed screening performance rates being below the rates in general HMO and Medicare populations suggest that RA patients were not receiving routine screening. This suggests possible impediments to routine care delivery, uncertainty regarding the complex relationship between lipids and CVD risk in RA (38, 39), or underrecognition of RA itself as a cardiovascular risk factor.

In our study, other predictors of poor rates of lipid screening demonstrate opportunities for improving care in RA patients who are older, sicker, and have the fewest outpatient visits. Patients with more visits and a higher number of unique providers were more likely to be screened, which is consistent with a study showing that relevant specialist involvement may improve screening in patients with complex conditions (40). Conversely, the finding of low rates of screening among patients from large towns may reflect lower provider density, or it may reflect smaller group practices that report lower quality than larger group practices (29).

Universally, addressing the elevated risk of CVD from inflammatory arthritis requires additional knowledge and vigilance to capture care delivery opportunities. In our sample, half of the RA patients saw their rheumatologist at least as often as they saw their PCP, and other studies have suggested that rheumatology encounters often outnumber PCP visits (16). Rheumatologists may feel that prevention is the role of primary care providers and may not want to interfere, even though they may be more familiar with CVD risk in relation to RA. PCPs may be stretched to invoke disease-specific prevention in limited encounters with patients who are also receiving specialty care. Coordinating the expertise of both the rheumatologist and the primary care provider may be useful in improving preventive cardiovascular care.

Collaboration with specialists has been shown to improve the quality of preventive care in patients with complex conditions (41, 42); this collaboration may mitigate the common finding that patients with competing comorbidities often receive less preventive care than healthier patients do (17, 40). One multispecialty health network with a well-integrated electronic health record reported superior lipid and osteoporosis screening among patients with RA as compared to the total network cohort, suggesting that optimal system support and multispecialty collaboration can enhance health care delivery to complex populations (43).

An optimal partnership between rheumatologists and primary care providers to address cardiovascular risk has not been defined. Rheumatologists are familiar with the disease-specific risks of RA and could play a more active role in this process. Rheumatologists could educate patients and primary care clinicians regarding increased CVD risk in RA or could actively order screening and/or co-manage modifiable risk factors. For instance, in contrast to the low frequency of lipid screening noted in our study, one academic rheumatology clinic that implemented routine screening practices reported 88% lipid testing at 5 years, highlighting the potential impact of specialty-driven protocols (44). A pivotal parallel example of shifting prevention roles is the move in recent years to include osteoporosis within the relevant scope of specialty practice. Studies have demonstrated that screening rates and treatment of routine and glucocorticoid-induced osteoporosis improve with rheumatologist collaboration (45, 46). Moreover, the 2010 study examining multiple evidence-based preventive services in RA showed that combined rheumatology and primary care predicted higher overall performance (26). Our finding of improved lipid testing among those who saw a PCP at least once each year may suggest a role for rheumatologists to advocate annual PCP visits for RA patients.

Formal specialist roles have also been examined amidst the expanding dialogue regarding the “patient-centered medical home” (47, 48). The American College of Physicians (ACP) identifies rheumatologists caring for RA patients as a possible specialty-based medical home if first-contact, whole-person, continuous, and integrated care is provided. However, the ACP Committee of Subspecialty Societies proposed an alternative specialist role as a “medical neighbor,” expanding the previous idea of a coordinated health care system as a “medical neighborhood” (49). As medical neighbors, specialists are not required to assume first-contact primary care responsibilities, but to promote co-management within the health system (47). As such, rheumatologists may advocate annual PCP visits or may co-manage cardiovascular prevention as good medical neighbors without assuming all primary care responsibilities. As research regarding the patient-centered medical home expands and health systems increasingly assume responsibility for promoting health among populations, the role of specialists as medical neighbors for cardiovascular preventive care should be explored further.

As with any scientific analysis, this study has some limitations. First, there is the potential for misclassification of RA and other diagnoses. To address this concern, previously validated algorithms were used (1, 19). Although the strictest validation study used only rheumatologist-reported RA coding (19), which demonstrated high correlation with audited ACR criteria, we adopted the convention of subsequent authors who used the criterion of more than 1 RA code in 24 months (1, 2, 18) to ensure that RA patients exclusively receiving primary care were included. Misclassification of osteoarthritis (OA) as RA may have occurred more frequently in the lone primary care group, but the low screening rates appear to be consistent with the rates among those receiving combined rheumatology and primary care, suggesting that if OA patients were included, it did not appear to have influenced the observed screening rates.

Second, there may be unmeasured differences between patients who see only a rheumatologist, such as patient preferences. We approached this concern by limiting our scope to primary prevention, stratifying and adjusting for a wide range of variables, including number of visits, unique providers, overall comorbidity, and RA severity surrogates, although data on RA disease activity measures and treatments were not available. We found that the lone rheumatology group was least likely to receive orthopedic surgery or gait devices, suggesting that they did not have historically worse RA to justify lapsed screening. However, in the absence of acute disease activity measures or medications, we cannot exclude rational delays in screening, given lipid fluctuations in patients with acute inflammation and steroid treatment (39).

We also acknowledge that ICD algorithms may underestimate hyperlipidemia if patients receive medications without the diagnosis being coded. Quality measures recommend annual lipid testing among such secondary prevention patients, even with statin treatment, so poor screening among potentially misclassified secondary risk patients receiving statins would reflect even more poorly.

Third, our study sample was limited to older adults with RA prior to 2006. It is unclear whether our results are generalizable to younger patients or to more recent years. However, it remains possible that with less comorbidity triggering health system contacts, younger RA patients may have even lower rates of lipid screening.

Finally, given that current RA-specific recommendations for CVD prevention are not explicit and that the exact role of lipids in CVD risk may be nonlinear, our choice of a 3-year versus a 5-year window for assessing lipid screening could be questioned (12, 14). It is unlikely, however, that the observed poor rates of screening would drastically improve by using a 5-year window. As a simple exercise, if we consider our screening rate over 3 years, the inclusion of 2 more years boosts screening rates to 71% at best, still leaving more than 1 in 4 RA patients unscreened. Future studies could examine a longer period and should include a comprehensive assessment of all traditional CVD risk factors, as well as actual CVD outcomes.

In this primary CVD prevention cohort of RA patients, lipid screening was frequently overlooked. When examining visit patterns, as long as a primary care provider was involved, no significant difference in screening probabilities emerged, regardless of the balance between primary care and rheumatology specialty care. A 22% improvement in testing among those seeing a PCP at least once each year suggests a role for advocating annual PCP visits for patients with RA, although performance improved only to 51% in the group seeing a PCP at least once a year. The remaining gap suggests that lapses in prevention may be one potential mechanism explaining why patients have not fully benefitted from declines in CVD seen in the general population despite aggressive treatment for the RA (3–5, 7). The observed gap in lipid screening highlights a key target for CVD risk reduction efforts. In addition, the finding that half of RA patients see their rheumatologist at least as often as they see their PCP suggests a need to study optimal partnerships between primary care providers and specialists for screening for CVD risk factors in high-risk populations within their medical homes and neighborhoods.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Bartels 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 conception and design.Bartels, Kind, Everett, Mell, Smith.

Acquisition of data.Bartels, Smith.

Analysis and interpretation of data.Bartels, Kind, McBride, Smith.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

The authors would like to thank Robert Purvis for meticulous data preparation, Lauren Fahey and Colleen Brown for manuscript preparation, and Dr. Alan Bridges for manuscript review.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES