Facility Variation in Utilization of Angiotensin-converting Enzyme Inhibitors and Angiotensin II Receptor Blockers in Patients With Diabetes Mellitus and Chronic Kidney Disease

, , , , ,
The American Journal of Managed Care, February 2007, Volume 13, Issue 2

Objective:To evaluate facility-level variation in prescription rates of angiotensin-converting enzyme inhibitor (ACEI) or angiotensin II receptor blocker (ARB) medications for patients with diabetes mellitus (DM) and chronic kidney disease (CKD).

Study Design: Retrospective database analysis from 143 Veterans Health Administration facilities.

Methods: Subjects with DM aged 18 to 75 years were identified as having stage 2-4 CKD using estimated glomerular filtration rate (eGFR) based on an index eGFR in 1999 and a subsequent eGFR 90-365 days later. Whether ACEI/ARB medications were prescribed within 1 year after the index eGFR was determined. Variation in facility-level rates was evaluated separately for subjects age <65 years and 65 to 75 years from facilities with more than 50 subjects per age group.


Results: A total of 103 853 subjects had stage 2 CKD; 51 728, stage 3; and 3233, stage 4. However, 25% of facilities had fewer than 50 patients age <65 years with either stage 3 or 4 CKD. The median (range) facility-level prescription rates of ACEI/ARB for stage 2 and combined stage 3-4 CKD were 58.5% (44.3%-71.2%) and 73.3% (51.7%-84.6%), respectively, for subjects age <65 years; and 56.5% (38.1%-71.4%) and 68.4% (51.6%-80.1%), respectively, for subjects aged 65 to 75 years. Spearman rank correlation between facility rankings by age group was 0.72 for stage 2 (139 facilities) and 0.49 for stage 3-4 (111 facilities) (< .001).

Conclusion: Although ascertainment of prescription rates of ACEI/ARB to CKD patients is feasible using electronic health records, small sample size at the healthcare-system level preclude their utility for public reporting.

(Am J Manag Care. 2007;13:73-79)

  • Guideline recommendations indicate that angiotensin-converting enzyme inhibitors (ACEIs) and/or angiotensin II receptor blockers (ARBs) should be used to delay progression of chronic kidney disease (CKD) to end-stage renal disease.
  • Of 143 Veterans Health Administration facilities, 25% had fewer than 50 patients with diabetes age <65 years with either stage 3 or stage 4 CKD.
  • Because of sample size considerations and possible selection biases in CKD identification, rates of ACEI/ARB administration in persons with diabetes mellitus and CKD should be used for internal quality improvement efforts by health plans, rather than for public reporting.

End-stage renal disease (ESRD) is a devastating condition with respect to both human and financial costs. In the year 2000 the point prevalence of ESRD in the United States was 1160 per 1 million persons at risk, with a cost of $17 billion; approximately 40% of these patients had diabetes.1 Between 1991 and 2001, Medicare spending for outpatient dialysis services increased about 10% per year, the fastest growing expenditure of the Medicare program. Moreover, a recent population-based study indicated that a far greater number of Americans, perhaps as many as 8 400 000 (5%), have pre-ESRD or chronic kidney disease (CKD), defined as an estimated glomerular filtration rate (eGFR) of less than 60 mL/min per 1.73 m2 for more than 3 months.2 Such individuals are at substantial risk for cardiovascular events and death before onset of dialysis3,4 because of a higher prevalence of cardiovascular risk factors.5,6 However, a substantial body of literature indicates that the progression of CKD to ESRD can be delayed through the use of angiotensin-converting enzyme inhibitors (ACEIs) and/or angiotensin II receptor blockers (ARBs).7-10

Although this evidence has been incorporated into guideline recommendations from the national Kidney Disease Outcomes Quality Initiative (K/DOQI),11 the absence of baseline data has hindered the development and evaluation of performance measures that reflect ACEI/ARB administration.12 Winkelmayer et al recently reported that about 50% of older (age >65 years) Medicare beneficiaries in Pennsylvania with diabetes and coexisting hypertension and/or proteinuria had an ACEI/ARB prescription in the first quarter of 2003.13 However, the feasibility of determining facility variation in ACEI/ARB prescription rates among healthcare organizations has not been reported previously.

The Veterans Health Administration (VHA) is an ideal healthcare delivery system in which to evaluate a performance measure for ACEI/ARBs in persons with diabetes mellitus (DM). The VHA is the nation's largest integrated healthcare system, with about 145 facilities, consisting of a lead secondary or tertiary medical center and associated community-based outpatient centers, administered by regional healthcare delivery systems referred to as Veterans Integrated Service Networks.14 The VHA also has a nationwide electronic medical record, permitting linkages of administrative, pharmacy, and laboratory data15 which facilitates the identification of individuals with DM16 and with CKD based on eGFR.1

The objective of this study was to evaluate VHA facility-level variation of ACEI/ARB prescription rates among veterans with DM and stage 2 or stage 3-4 CKD for both younger individuals (age <65 years) and seniors (aged 65-75 years). We hypothesized that facility-level adherence to guideline recommendations would vary both as a function of age and the stage of CKD.


Study Population

International Classification of Diseases, 9th Edition

We used an existing research registry, the VHA Diabetes Epidemiology Cohort, to identify subjects with DM who were alive as of September 30, 1999.16 Individuals were identified as having definite DM if they had 2 or more DM-specific codes (250, 357.2, 362.0, 366.41) from inpatient and outpatient visits to VHA physicians over the 24 months immediately preceding the study period, or if they had received an antiglycemic agent.15 Patient-level data elements in VHA records included demographics, visits and hospitalizations, providers, diagnosis and procedure codes, and values from serum creatinine tests. We also used Medicare inpatient and outpatient files (Part A) and physician care (Part B) that have the same data elements plus revenue center and beneficiary status codes, but without serum creatinine values. Medicare denominator files also included demographic data and reason for eligibility (including ESRD). Records were linked for individuals enrolled in both the VHA and Medicare systems.

We ascertained the completeness of data collection by examining facilities that had high incomplete transmittal of data for creatinine and antiglycemic medications, and eliminating them if their data completion rates were within the lower 5% of the normal distribution. Regarding the completeness of pharmacy data, we note that the percentage of subjects on antiglycemic medication in fiscal year 2000 was 81%,16 which compares with Centers for Disease Control and Prevention self-reported estimates of 85%.17

Predictor Variables

As previously described,18 we determined whether a subject with a baseline creatinine value in fiscal year 1999 had CKD based on an eGFR derived from the 4-variable (age, race, sex, and creatinine) Modification of Diet in Renal Disease study.19 The index eGFR was calculated for the first creatinine value in the data set, and the qualifying eGFR (if present) was calculated for the second creatinine value if it was obtained between 90 and 365 days after the index eGFR. Although we assigned individuals to stage 2 CKD if the eGFR was 60 to 89 mL/min per 1.73 m2, we recognize that such individuals cannot be distinguished from nonproteinuric individuals without CKD according to K/DOQI guidelines.11 Nonetheless, many of these subjects could represent individuals at high risk. Therefore, we reasoned that, even given the limitations of the electronic health record, information on ACEI/ARB utilization in this group could be of interest to healthcare plans.

Individuals were assigned to stage 3 or 4 CKD based on their index eGFR as long as their qualifying eGFR was less than 60 mL/min per 1.73 m2 using K/DOQI criteria: for stage 3 the eGFR was 30 to 59 mL/min per 1.73 m2, and for stage 4 the eGFR was 15 to 29 mL/min per 1.73 m2. Subjects with either stage 5 CKD (index eGFR <15 mL/min per 1.73 m2) or patients with any prior Medicare or VHA code for dialysis or transplantation were excluded, because of lack of clinical studies of efficacy in these populations. We also excluded individuals with only 1 available eGFR determination, because they could not be assigned to a definitive eGFR category.

Because the National Committee for Quality Assurance requires commercial managed care health plans to collect and report results separately by populations covered by different product lines, including Medicare, we categorized the population into 2 age groups, age <65 years and aged 65 to 75 years,20 to present our results in a format utilized for private sector performance measurement benchmarking.21

Outcome Variables

The outcome was at least 1 VHA pharmacy fill of either an ACEI or ARB agent within 365 days after the index eGFR. We determined the average percentage of subjects achieving the threshold measure within each age group. Individual patient data were then aggregated by facility. We determined the minimum, maximum, and median rates of administration for each age group.


Chi-square tests were used to compare difference in rates of ACEI/ARB prescription between stage 2 and combined stage 3-4 for all subjects stratified as <65 and 65 to 75 years of age.

League ranking tables are a commonly used format to stratify quality rankings, such as those for plans21 and states.22 Because use of the 90th percentile and 10th percentile of performance is an industry standard for identifying the best-performing and worst-performing plans,21 we ranked facilities into deciles by each age group for patients with stage 2 and either stage 3 or stage 4 CKD as our primary analyses. However, because the benefits of ACEI/ARBs are not as clearly demonstrated in stage 4 CKD, and the facility numbers were small, we evaluated stage 3 CKD as a sensitivity analysis. For comparison of facility rankings, we only included facilities with more than 50 patients in each age group; consequently, we did not present an analysis for stage 4 alone because almost all facilities had fewer than 50 patients.

Correlation between facility ranks for the 2 age groups was evaluated using the Spearman rank correlation coefficient. This is a nonparametric measure of correlation, and thus can measure correspondence between ordinal variables, such as rankings, and assess the significance of this correspondence.23 A correlation of 1.0 is perfect agreement; -1.0 is opposite agreement; and 0 is no correlation.

We also determined how decile rankings changed when facilities ranked in the best-performing and worst-performing deciles for the younger age group were ranked based on rates of ACEI/ARB prescription in the group aged 65 to 75 years. The VA New Jersey Health Care System Institutional Review Board approved the research.


We identified 412 929 persons with DM who had at least 1 serum creatinine value and whose race was reported. We excluded 143 678 individuals with a single serum creatinine value available, and 5421 individuals with ESRD in the 24 months immediately preceding the study period. There were, therefore, 263 830 veterans for whom eGFR could be determined during the study period. Our study population consisted of 158 814 individuals who had stage 2, 3, or 4 CKD and were 18 to 75 years of age. These individuals were largely male (98.1%), elderly (57.9% were older than age 65 years), and white (75.7%), with comparable distributions (within 5%) across the CKD stages (see Table 1).


Within each age stratum and stage of CKD, the facility-level maximum and minimum rates of ACEI/ARB prescription varied between 1.5-fold and nearly 2-fold, and subjects with stage 3 or 4 CKD had mean rates of ACEI/ARB prescription that were 10% to 15% higher than those for individuals with stage 2 CKD in each age group (<.001). The median (range) facility-level prescription rates of ACEI/ARB for stage 2 and combined stage 3-4 CKD were 58.5% (44.3%-71.2%) and 73.7% (51.7%-84.6%), respectively, for subjects age <65 years; and 56.5% (38.1%-71.4%) and 68.4% (51.6%-80.1%), respectively, for subjects aged 65 to 75 years (Table 2). We performed sensitivity analyses by removing subjects with stage 4 CKD from the combined stage 3 and 4 group and found that the results were unchanged (data not shown).

We determined Spearman rank correlations for all facilities with more than 50 subjects per age group for all facilities, and separately for the top and bottom deciles (Table 3). Of 143 facilities, 36 (25%) had fewer than 50 patients under age 65 years with only stage 3 CKD, and 32 had fewer than 50 individuals with combined stage 3 and 4 CKD. The overall correlations between age groups for all facilities were fair (0.49) to moderate (0.72) for combined stage 3-4 and stage 2, respectively. However, the correlations between age groups were more variable (0.15-0.65) and generally lower for facilities in the top and bottom deciles.

We further evaluated the changes in facility decile rankings by age group (Figure). Among facilities ranked in the top decile (higher rates) for stage 2 CKD for the group under age 65 years, 3 (out of 14) were ranked more than 2 deciles lower when compared against the group aged 65 to 75 years. Of facilities in the bottom decile (lower rates), 4 (out of 14) were ranked more than 2 deciles higher. Similarly, for combined stage 3-4 CKD, among the facilities ranked in the top and bottom deciles, 5 (out of 11) shifted more than 2 deciles lower and higher, respectively. Results were identical for stage 3 CKD.


Our study evaluated facility-level ACEI/ARB prescription rates, stratified by age (<65 and 65-75 years of age) for persons with DM and either stage 2 or combined stage 3-4 CKD. We observed a 30% difference in prescription rates at the facility level for each CKD stage, although the mean rates of ACEI/ARB prescription were clinically comparable (within 5%) for younger and older subjects for each CKD stage. Furthermore, facility-level rankings differed markedly when rates by age group were compared within a CKD stage, with only poor to fair correlation for facilities in the top and bottom deciles.

We can only speculate as to the reasons for the facility-level variation in a key measurement of evidence-based CKD quality of care. A prior study demonstrated that chronic comorbid conditions, including chronic lung disease, depression, dementia, and other mental illness, are associated with underutilization of ACEI/ARB in an elderly Medicare population with stage 2 CKD and proteinuria.13 This suggests the need for further evaluation of the impact of case mix adjustment on ACEI/ARB prescription rates in patients with stage 3 and stage 4 CKD, especially if public reporting or pay for performance were to be considered. Alternatively, adverse events, such as significant hyperkalemia, could impact patients with lower eGFRs in stage 3 and stage 4 CKD. We also speculate that organizational factors, such as nephrologist availability and primary care referral to nephrologists, also could impact adherence rates.

Independent of attribution as to the reasons underlying facility-level variation, our findings indicate the feasibility of using linked demographic, administrative, pharmacy, and laboratory data to monitor prescription rates of ACEI/ARBs for persons with DM and low eGFR, and indicate the importance of age-stratifying results. However, our study also indicates that there are unresolved feasibility and methodologic issues that would suggest that ACEI/ARB prescription rates should not be used as a public reporting measure at this time. First, in order to avoid the improper classification of patients into stages of CKD, 2 eGFRs between 90 and 365 days are needed. This could result in ascertainment biases if subsequent serum creatinine measurements are not done, or if they are obtained in another system and actual values are not available in the electronic medical record.24 Second, we could not fully operationalize the National Kidney Foundation criteria for stage 2 CKD because of the lack of valid urinary protein data. In addition, despite both a high burden of CKD among veterans with DM18 and a 20% prevalence of DM in the veteran population,16 there were fewer than 50 patients under age 65 years with stage 3 CKD at 36 of 143 VHA facilities, in contrast to only 4 facilities with fewer than 50 patients aged 65 to 75 years who had stage 3 CKD. Although we included facilities with at least 50 individuals in each age group in our analyses, prior work demonstrated that sample sizes larger than 100 may be more appropriate for performance measurement.25 Consequently, an ACEI/ARB prescription rate measure may not be appropriate for individual practices or small groups. Finally, further evaluation is necessary to determine whether risk adjustment for individuals with comorbid conditions is necessary.13 Our findings may have relevance for other researchers, in that they suggest the need for field testing in other systems of care.

On the other hand, our finding of marked variation among facilities, and the ability to identify "benchmark organizations," suggests that healthcare systems and larger physician practices with availability of electronic health records may find it feasible to use linked eGFR and ACEI/ARB prescription rates as useful internal quality improvement measures. One evolving area of inquiry suggests the importance of organizational factors. VHA medical centers with better performance had more frequent internal feedback to clinicians, identified frontline clinicians to lead (champions), and were more likely to accept the guidelines as applicable to their practice.26 Because analysis of administrative codes indicates that CKD is underdiagnosed on the basis of serum creatinine,18,27 presumably because physicians do not recognize that CKD can occur with normal or near-normal serum creatinine values, real-time availability of eGFR could improve ACEI/ARB administration rates through increased identification of patients with CKD at the time of the clinical encounter.28

Furthermore, studies have indicated that suboptimal ACEI/ARB prescription rates would be actionable by a healthcare system. Although our study did not evaluate whether individuals with stage 3 or stage 4 CKD had previously been placed on an ACEI or ARB but were unable to tolerate the medication, discontinuation of ACEIs or ARBs because of worsening creatinine levels or elevations in serum potassium can occur in at least 5% to 10% of individuals even in randomized clinical trials,8,10 and the discontinuation rate may be higher in practice.29 Because the use of bicarbonates and non-potassium-sparing diuretics may improve use of ACEI/ARB medications,30 nephrology referrals would be appropriate for, and could be monitored as a process measure, for patients with CKD not on ACEI/ARBs.

Our proposal has several strengths. We were able to not only accurately determine the presence of definite CKD in a defined population of persons with DM, but to evaluate variation in practice among a large number of healthcare administrative units, which is an essential component of performance measurement development.20

We also acknowledge several limitations. Our population was a largely male veteran population, which could limit the generalizability of the results. We were unable to ascertain proteinuria, because the reporting of this variable was not standardized. Consequently, individuals identified in this study as having stage 2 CKD may not have met evidence- based criteria for use of ACEIs or ARBs based on randomized clinical trials.7-10 On the other hand, we note that a recent publication by the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) group establishes that individuals with DM and eGFRs between 60 and 89 mL/min per 1.73 m2, without determination of proteinuria, have significantly higher progression to ESRD than individuals with higher eGFRs.31 For such individual patients, medical record review may be necessary. We were unable to ascertain serum potassium values as a possible contraindication to the use of ACEI/ARBs, nor could we determine whether ACEI/ARBs were administered before the index eGFR, which may result in unfair evaluation of clinicians.32 Future research, using linked longitudinal databases of patients with CKD, will be necessary to evaluate such factors.33


In conclusion, our study demonstrates the feasibility of using linked electronic medical records to assess healthcare plan performance in ACEI/ARB administration for persons with DM and CKD. However, we also identified sample size considerations and possible selection biases in CKD identification that would limit the usefulness of ACEI/ARB prescription rates for public reporting at this time. Until these issues and that of comorbid illness severity are resolved by further research and there is more widespread implementation of the electronic medical record, we recommend that ACEI/ARB prescription rates for persons with diabetes and CKD should be used for internal quality improvement efforts by plans, rather than for public reporting.34


We thank Christina Croft, MSW, for manuscript preparation.

Author Affiliations:

From the DVA-New Jersey Healthcare System, East Orange, NJ, and the University of Medicine and Dentistry, Newark, NJ (AT, CT, MM, LP); the Louis Stokes Cleveland Department of Veterans Affairs Medical Center and Case Western Reserve University Department of Medicine, Cleveland, Ohio (EFOK); and the DVA Bedford Center for Health Quality, Outcomes and Economic Research, Bedford, Mass, and Boston University School of Public Health, Boston, Mass (DRM).

Funding Sources:

This work was funded by VA Health Services Research grant IIR 04-205 to Dr Pogach and Dr Miller, and by a VA Medical Service Epidemiology Merit Review grant to Dr Pogach and Dr Miller.

The findings and opinions reported here are those of the authors and do not necessarily represent the views of any other individuals or organizations.

Correspondence Author:

Leonard Pogach, MD, MBA, DVA-New Jersey Healthcare System, Center for Healthcare Knowledge Management, 385 Tremont Ave, E Orange, NJ 07018. E-mail: leonard.pogach@va.gov.