Prevalence, treatment patterns, and healthcare resource utilization in Medicare and commercially insured non-dialysis-dependent chronic kidney disease patients with and without anemia in the United States

Anemia is common in patients with non-dialysis-dependent (NDD) chronic kidney disease (CKD), and prevalence increases as CKD worsens [1]. The most current data on anemia prevalence in NDD-CKD patients are from the 2007–2010 National Health and Nutrition Examination Survey (NHANES) [2]. These data are representative of the United States population, but the sample size is relatively small and does not allow for subgroup analysis. Thus, more recent anemia prevalence information using a larger dataset is necessary to allow evaluation of trends by key factors such as age, sex, and race. Much attention has been focused on anemia treatment in patients with stage 5 CKD on dialysis. However, relatively little is known about anemia treatment patterns in NDD-CKD stage 3–5 patients. Thamer and colleagues retrospectively evaluated erythropoiesis-stimulating agent (ESA) therapy in commercially insured NDD-CKD patients before and after publication of TREAT (Trial to Reduce Cardiovascular Events with Aranesp Therapy) results in 2009 [3]. They showed that, post-TREAT, about 8% of CKD patients were prescribed ESAs by September 2011, with higher prescription rates in stage 4 than in stage 3 patients. They did not assess treatment patterns for iron therapy or red blood cell (RBC) transfusion use.

TREAT results showed that targeting anemia treatment to a hemoglobin level of 13 g/dL with darbepoetin vs. placebo (rescue darbepoetin if hemoglobin < 9 g/dL) in NDD-CKD patients with diabetes did not reduce risk of death or cardiovascular events, but was associated with increased stroke risk [4]. Subsequent to TREAT, and with evidence from other randomized trials evaluating ESA treatment in NDD-CKD patients [5, 6], the US Food and Drug Administration (FDA) revised ESA product labels in June 2011, adding new boxed warnings and recommending more conservative dosing [7]. In 2012, the Kidney Disease Improving Global Outcomes (KDIGO) guideline group released updated practice guidelines for anemia in CKD [8]. The FDA action and KDIGO guidelines were intended to influence anemia treatment among NDD-CKD patients.

Anemia has been identified as a risk factor for increased hospitalizations and/or hospital days in NDD-CKD patients [9, 10], but more research is needed regarding how anemia may affect other healthcare utilization, such as emergency department (ED) or clinic visits or specialty care.

Our study objectives were to: 1) update anemia prevalence estimates in Medicare-covered and commercially insured stage 3–5 NDD-CKD patients; 2) examine anemia prevalence in these patients by CKD stage, age, sex, and race; 3) evaluate treatment patterns in NDD-CKD patients in a contemporary period after TREAT and the enhanced FDA boxed warning, and around the time of the updated KDIGO anemia guidelines release; and 4) assess healthcare burden/utilization in NDD-CKD patients with and without anemia.

We identified 218,079 stage 3–5 NDD-CKD patients aged 66–85 years from the 20% Medicare sample and 56,188 patients aged 18–63 years from the commercially insured population during the study period (Additional file 2: Figure S1).

Our study greatly expands existing information on anemia prevalence in older Medicare-covered and younger commercially insured stage 3–5 NDD-CKD patients. Using two large datasets (Medicare and MarketScan), we showed that anemia prevalence was much higher (52%) in older than in younger (28%) patients. Anemia prevalence rose sharply with increasing CKD stage, from 22.4% in younger commercially insured stage 3 NDD-CKD patients to 53.9% in stage 5, and from 43.9% in older Medicare stage 3 patients to 72.8% in stage 5. Anemia prevalence was generally higher among women at each CKD stage. The most common anemia-related treatment across both data sets was RBC transfusion, followed by ESAs and IV iron. Only younger stage 5 NDD-CKD patients with anemia were more likely to be treated with ESAs than with RBC transfusion. After adjustment for patient case-mix, hospital admissions, nephrologist visits, and hematologist visits were significantly more frequent for anemic than for non-anemic patients.

Prior to our study, the only recent analysis of anemia prevalence in NDD-CKD patients was derived from NHANES 2007–2008 and 2009–2010 datasets [2]. Prevalence of CKD and anemia were defined by laboratory values (hemoglobin levels < 12 g/dL in adult women and < 13 g/dL in adult men), which is ideal; however, only 410 patients with CKD in the dataset could be evaluated, so subgroup analysis was not possible. This analysis showed that 15.4% of CKD patients (stages 1–5, including dialysis) had anemia. The authors found a weighted anemia prevalence of 17.4% (n = 231), 50.3% (n = 37), and 53.4% (n = 17) for CKD stage 3, 4, and 5 (likely including dialysis patients), respectively. In contrast, our evaluation included 109,251 older Medicare-covered and 15,716 younger commercially insured patients with CKD and anemia. We found CKD stage-specific anemia prevalence similar to NHANES findings in the younger (aged ≤63 years) commercially insured population, 22.4%, 41.3%, and 53.9% for CKD stages 3, 4, and 5 (not including dialysis patients), respectively, but higher anemia prevalence in the older Medicare-covered population, 46.2%, 65.8%, and 73.8% for CKD stages 3, 4, and 5 (not including dialysis patients), respectively. The age distribution was not included in the NHANES analysis, only the information that participants were adults aged older than18 years [2], so direct age-specific comparison to our data is not possible.

Anemia prevalence in stage 3–5 NDD-CKD patients increased monotonically as age increased, similar to the general population. Using NHANES data from 2003 to 2012, Le et al. showed that weighted anemia prevalence in the US increased by age from 3.8% for ages 15–29 years to 19.4% for ages 80–85 years [13]. Not surprisingly, and in contrast to Le’s findings, we found anemia prevalence several-fold higher in NDD-CKD patients than in the general population; prevalence was 25.9% and 54.7%, respectively, among patients aged 18–44 and 80–84 years. Variability of anemia prevalence by sex was much less (Medicare: 53.2%, women; 46.9%, men; commercially insured: 33.6%, women; 23.7% men) than in the general population, where women are more than twice as likely as men to have anemia [13]. We also found less racial variability among NDD-CKD patients with anemia (Medicare only: black 57.9%, white 48.9%, other 51.4%) than in the general population, where non-Hispanic blacks were 2.4–3.7 times more likely to have anemia than other race groups [13].

In our updated analyses (2011–2013), we showed that both younger and older anemic patients were most likely to be treated with RBC transfusion (22.2%, older; 11.7%, younger), followed by ESAs (12.7%, older; 10.8%, younger), and IV iron (6.7%, older; 9.4%, younger). Two earlier studies showed lower use of RBC transfusions than we found. Lawler et al. identified anemia treatment in a predominantly older male Veterans Administration CKD population from 2003 to 2005 following a first hemoglobin level under 11 g/dL. Compared with our data, ESA use was lower (about 7%), and RBC transfusion use was lower (16.5%) than in the older Medicare population we evaluated [14]. These data are not surprising, as many Veterans Administration patients had not been diagnosed with anemia despite lower hemoglobin levels. We looked for anemia treatment after a documented anemia diagnosis. Lawler also showed that iron (oral and IV) use was 30.9%, which is not comparable to our data as we could track only IV iron use. Fox et al. evaluated transfusion burden in 374 NDD-CKD adult patients with severe anemia (hemoglobin < 10 g/dL) and a mean age of 71 years in the Henry Ford Health System from 2004 to 2008 [15]. They found that RBC transfusions were administered in 20% of these patients during a mean follow-up period of 459 days. We found a relatively higher rate of transfusions (22% over 1 year) in the older Medicare population. Several ESA trials showing increased risk of death, cardiovascular events, and stroke in NDD-CKD patients treated to higher hemoglobin targets of 13–15 g/dL [4‐6], additional FDA boxed warnings for ESAs released in June 2011 [7], and KDIGO anemia guidelines released in 2012 [8] may have led to decreased ESA use in our study timeframe and affected anemia treatment selection.

Stauffer and Fan evaluated self-reported anemia treatment in NHANES 2007–2008 and 2009–2010 participants with CKD [2]. Of 410 CKD patients, 22.8% self-reported anemia treatment in the past 3 months (20.7%, 43.0%, and 22.8% for stage 3, 4, and 5, respectively), but specific treatments were not described. Winkelmayer et al. studied trends in anemia treatment, 1995–2010, among NDD-CKD patients aged 67 years or older 2 years before ESRD, using US Renal Data System (USRDS) data. They showed that ESA use peaked in 2007 at 40.8%, but declined to 35.0% in 2010, while IV iron and RBC transfusions were used in 12.3% and 40.3%, respectively, in 2010 [16]. Their data most closely compare to our data for stage 5 NDD-CKD patients; in a contemporary Medicare cohort (2011–2013), we showed lower rates with 21.8% receiving ESAs, 9.6% IV iron, and 31.2% RBC transfusions. Lower use of all treatments in 2012 may be attributable to less severe anemia in our NDD-CKD stage 5 patients than in Winkelmayer’s cohort, all of whom initiated dialysis. Lower ESA use by 2012 may also be due to additional FDA boxed warnings added to ESA package inserts in June 2011 [7].

Thamer et al. evaluated ESA use among stage 3 and 4 NDD-CKD patients before and after the TREAT study (2007–2011) using the MarketScan database [3]. In patients aged younger than 65 years, they showed that 7.6% and 22.9% with CKD stages 3 and 4, respectively, were treated with ESAs from September 2009 to October 2011; in a more contemporary similarly aged younger population, we show even lower ESA use (6.5% and 15.7%), suggesting that FDA action in June 2011 [7] may have further affected ESA use. We did not evaluate patients aged older than 65 years using MarketScan data because of potential missing data for patients with both commercial and Medicare insurances.

In our 2012 cohort of stage 3–5 NDD-CKD patients, RBC transfusions were used more than ESAs following anemia diagnosis; 22.2% of older Medicare and 11.7% of younger commercially insured patients had at least one RBC transfusion after anemia diagnosis (within means of 2.7 and 2.4 months, respectively). Only younger anemic patients were more likely to receive an ESA (19.9%) than an RBC transfusion (15.0%) after anemia diagnosis. RBC transfusions as a common treatment following anemia diagnosis in NDD-CKD patients is concerning. Leffell et al. merged serial HLA antibody data with USRDS data for 1324 patients on the kidney transplant waiting list, 2004–2010. Using a matched patient cohort and a cross-over cohort, they showed that 20% of patients who received leukodepleted RBC transfusions, compared with 3% who did not, displayed an antibody response (P = 0.001). The response was similar in the crossover cohort [17]. Notably, they showed increased panel-reactive antibody values in 26.3% of transfusion patients compared with 5.8% of non-transfusion patients. Yabu et al. showed similar results in a patient population with a smaller percentage of black patients [18]. The mean waiting time for a kidney transplant was 2 months longer, on average, for transfused than for non-transfused patients [19]. Thus, our finding that RBC transfusions are common in both older and younger stage 4 and 5 NDD-CKD patients has potential implications regarding antibody sensitization and increased waiting times if these patients meet requirements for future kidney transplant.

We found that that healthcare utilization was higher for stage 3–5 NDD-CKD patients with than without anemia. Other studies have identified anemia in NDD-CKD patients as a risk factor for increased hospitalizations and hospital days [9], although this relationship may be confounded by common underlying comorbid conditions such as CHF with fluid retention [20]. Garlo et al. found that stage 3–5 NDD-CKD patients (controlling for CHF) with hemoglobin levels 9 g/dL or lower compared with above 9 g/dL admitted to tertiary referral hospital had increased lengths of stay, but no difference in 30-day readmission rates [10]. We showed that after adjustment for multiple baseline comorbid conditions, including CHF, younger and older NDD-CKD patients with diagnosed anemia had significantly more hospitalizations, ED visits, and SNF admissions, along with increased hospital lengths of stay, than those without anemia. Outpatient visits, and specifically visits to nephrologists, hematologists, and primary care practitioners were also significantly higher in patients with diagnosed anemia. A small prospective observational study of nursing home residents with eGFR below 60 mL/min/1.73 m² observed that residents with anemia showed greater earlier decline in a mobility distance measure, but not in other short-term or longer-term performance or self-report measures, than those without anemia [21]. CKD-related anemia affects many body systems but in particular the cardiovascular system. Anemia contributes to increased cardiac output, development of left ventricular hypertrophy and chronic heart failure; thus, it is not unexpected that healthcare utilization is higher among CKD patients with than without anemia.

Interestingly, we showed that anemia in CKD was associated with slightly fewer cardiologist visits, which seems counterintuitive given higher rates of ASHD and CHF in CKD patients with anemia versus without. A recent analysis comparing incident hemodialysis patients pre- and post-changes in CMS reimbursement policy and FDA label action showed that the risks of major adverse cardiovascular events (MACE), death, hospitalization for heart failure, and venous thromboembolism were similar in the pre- and post-policy periods, with lower stroke and myocardial infarction risk despite lower ESA use and higher rates of RBC transfusions; black patients showed significant reductions in risks of MACE and death [22]. Whether the same effect will occur in NDD-CKD patients or patient subgroups, or whether treatment of NDD-CKD anemia to lower hemoglobin target levels with ESAs, or alternative therapies such as hypoxia-inducible factor (HIF) prolyl hydroxylase inhibitors, will reduce mortality, particular cardiovascular outcomes, or overall healthcare utilization is unclear.

Our study has several strengths. We used two large national datasets to evaluate anemia prevalence, treatment, and healthcare utilization in older Medicare-covered and younger commercially insured patients with stages 3–5 CKD. We conducted several subgroup analyses to evaluate anemia prevalence by CKD stage, age, and race, which were not possible using the NHANES dataset. We also considered more contemporary trends in anemia treatment and included ESAs, IV iron, and RBC transfusion use, as previous studies did not. We compared broad and specific key categories in inpatient and outpatient healthcare utilization for NDD-CKD patients with and without anemia, adjusted for multiple factors.

Our study also has several limitations. This is a retrospective analysis of anemia using claims-based data that are not collected for research purposes, so regression relationships should be considered associations and not causal. Although we used multiple covariates as adjusters in our models, residual confounding may be present. Hemoglobin values were not available for NDD-CKD patients in the Medicare dataset, and were available for less than 5% of MarketScan patients, so we could not confirm anemia diagnosis with information on anemia severity. We evaluated anemia prevalence across age categories using two different datasets (Medicare for older, MarketScan for younger patients); data from these databases are not directly comparable as demographic and comorbidity characteristics differ among the populations. We chose not to evaluate anemia prevalence and treatment in older MarketScan individuals because previous research showed that 100% of claims may not be identified for patients with both commercial and Medicare insurances. Since the anemia of CKD is often a diagnosis of exclusion and can be seen in combination with other anemia types, we are not certain that all of the anemia diagnoses we identified were CKD-related. However, we used a timeframe around the CKD index date to increase the probability that anemia was related to CKD. We did not assess oral iron treatment, as most oral iron therapy is available over the counter and use is not reliably identified in Medicare or commercial prescription claims databases. Oral iron therapy is likely to be the predominant iron therapy for NDD-CKD patients, compared with predominance of IV iron for hemodialysis patients. In addition to general lack of laboratory information, we did not have information on fluid status, which may confound the relationship between CKD anemia and outcomes.

Our study is the largest to date examining anemia prevalence in stage 3–5 NDD-CKD older Medicare-covered and younger commercially insured patients, allowing analysis by important subgroups such as age, race (Medicare only), and CKD-stage. In a 2012 patient cohort, both younger and older patients with stage 3–5 NDD-CKD and diagnosed anemia were more likely to be treated with an RBC transfusion than with ESAs or IV iron. This is concerning due to the potential for increased panel-reactive antibodies and increased waiting time for kidney transplant. Although we showed increased healthcare utilization for NDD-CKD patients with anemia versus those without, further research is needed to determine whether future therapies such as HIF prolyl hydroxylase inhibitors, anemia treatment to lower hemoglobin targets with ESAs and iron, or treatment of specific subgroups (e.g., CKD with heart failure) can improve health outcomes and reduce healthcare utilization.