Improved Care and Outcomes Among Elderly Medicare Managed Care Beneficiaries With Diabetes

, ,
The American Journal of Managed Care, April 2005, Volume 11, Issue 4

Objective: To examine individual- and plan-level factors relatedto improved diabetes care and outcomes between 1999 and2001 among elderly Medicare managed care beneficiaries withdiabetes.

Study Design: Retrospective analysis of Medicare and individual-level Health Plan and Employer Data Information Set data.

Methods: We merged Medicare demographic information withperson-level data regarding 6 comprehensive diabetes care measuresprovided by the National Committee on Quality Assurance forMedicare managed care enrollees for 1999. Then we identifiedthose beneficiaries for whom comprehensive diabetes care informationwas also reported in 2001, and determined the improvementfor each measure. Data on persons not meeting thecomprehensive diabetes care criteria in 1999 were analyzed todetermine the factors associated with appropriate care and outcomesin 2001.

Results: Of the 174 combinations of individual- or plan-levelfactors and comprehensive diabetes care measures, 167 showedsignificant improvements. Nonetheless, for most of the 6 comprehensivediabetes care measures, poor care occurred more frequentlyfor black patients than for white patients, among individuals infor-profit versus not for profit plans, and among individuals in independentpractice association plans compared with group and staffmodel plans. Among the beneficiaries whose healthcare failed tomeet the comprehensive diabetes care guidelines in 1999, by 2001,care guidelines were met in approximately three fourths for hemoglobinA1c and low-density lipoprotein cholesterol testing, but inonly one half for eye examinations, low-density lipoprotein cholesterolcontrol, and nephropathy monitoring.

Conclusions: Between 1999 and 2001, care of elderly Medicarepatients with diabetes improved, including among individuals whopreviously had not received appropriate care. However, moreimprovement is needed to achieve equality among members of allrace groups and plan types.

(Am J Manag Care. 2005;11:213-222)

The results of the Diabetes Control andComplication Trial (DCCT),1 the United KingdomProspective Diabetes Study (UKPDS),2 other studies,and consensus meetings have led to the developmentof guidelines for the treatment of diabetes.3-5 Anexample is the 6 comprehensive diabetes care measuresof the Health Plan and Employer Data and InformationSet (HEDIS) measures used by the National Committeeon Quality Assurance (NCQA) for evaluating the performanceof managed care plans.

From 1998 to 2001, the Centers for Medicare andMedicaid Services (CMS) has received from NCQA individual-level information for one (1998) to all (1999-2001) of the HEDIS comprehensive diabetes caremeasures from almost all managed care plans providingservices to Medicare beneficiaries. We and others6-9have performed cross-sectional analyses that showedlower rates of HEDIS measures among black elderlybeneficiaries compared with white individuals. For elderlydiabetic patients, we reported significantly higherrates of poor hemoglobin A1c (HbA1c) control amongblack and Hispanic persons compared with white individuals,as well as among those on Medicaid or havingthe lowest income. Plan-level information associatedwith poor control included being in a profit managedcare plan, an independent practice association (IPA)model plan, or a larger plan, and having a low percentageof minority group members in a plan.8

Using 1996 HEDIS data, Himmelstein et al showedthat investor-owned and IPA plans consistently performedless well on 14 HEDIS indicators than nonprofitplans and that either group or staff model plans generallyperformed better than IPA model plans.10 YetSchneider et al, using 1997 HEDIS data, failed to finddifferences in the rates of high-cost operative proceduresbetween elderly Medicare beneficiaries in forprofitand nonprofit plans.11

Summary national information available on theNCQA Web site indicates that between 1999 and 2001,the comprehensive diabetes care measures improvedbetween 2.5 and 12.6 percentage points among Medicarebeneficiaries with diabetes enrolled in managedcare.12 However, it is not known if these improvementswere the same for all racial or other groups of beneficiariesor members of all types of plans.

Combining individual-and plan-level informationfrom CMS and NCQA, we report a more in-depth analysisof the changes in the 6 comprehensive diabetes caremeasures among Medicare elderly beneficiaries whowere enrolled in managed care in 1999 and 2001 and onwhom information was provided by a managed careplan in both years. In particular, we investigated if differencesby race persisted and if the tax status and planmodel type influenced the quality of diabetes carereceived in this cohort. In addition, we report on thoseindividual-and plan-level factors that were associatedwith improved care for individuals who were not receivingappropriate care in 1999. This study was approvedby the Institutional Review Board: Human SubjectsCommittee of the University of Minnesota (study number9608S11640).


Data and Data Sources

We received the HEDIS information for 1999 and 2001from CMS. This information included both plan-levelsummaries and individual-level data. Individual-leveldata included each person's Health Insurance Claimnumber (HIC), which we used to link the person to theMedicare 1999 Denominator file. The 1999 Denominatorfile was the source of the age, sex, race, and monthly stateMedicaid "buy-in" status. The latter indicates whether abeneficiary was enrolled in a state Medicaid-administeredprogram for each month of the year.

Information regarding each managed care plan wastaken from Medicare's monthly Medicare Managed CareHealth Plan reports for December 1999 and 2001. Thesereports contained information on tax status (profit ornonprofit), type of managed care model (staff, group,IPA, or other), geographic location, and number ofMedicare managed care enrollees, as well as the planidentification number, which allowed us to determinewho switched plans between 1999 and 2001. Finally,the 1990 US Census STF 3b file was used to obtain themedian household income for heads of household 65years of age or older by ZIP code.

Case Identification

The criteria used by the managed care plans to identifyenrollees with diabetes, as well as the criteria usedto identify individuals who received each comprehensivediabetes care measure are described in the HEDIS2000 Technical Specifications and on the NCQA Website.12-14 These criteria are similar to those used by othersto identify persons with diabetes in either medicalrecords or administrative data.15-17 In addition, eachperson had to be continuously enrolled in 1999 or 2001;continuous enrollment was defined as no gap in coverageof more than 45 days, and enrolled as of December31, 1999 or 2001.

Outcome Measures

The 6 comprehensive diabetes care measures were(1) an HbA1c test, (2) poor control of HbA1c (a level>9.5%), (3) an eye examination, (4) a test for serum lowdensitylipoprotein cholesterol (LDL-C), (5) LDL-C controlledto less than 130 mg/dL, and (6) a screening forkidney disease (microalbuminuria test) or documentationof neuropathy or its treatment in the medicalrecord, also called nephrology monitoring or nephrologyscreening.14 For the HbA1c and LDL-C control, if novalue was recorded, then the HbA1c was counted as poorcontrol or the LDL-C was considered to equal or exceed130 mg/dL. Whereas each measure was evaluated for themeasurement year (1999 or 2001), it was only for theHbA1c test and poor HbA1c control that just 1 year wasexamined. The LDL-C test may have been done, andLDL-C control demonstrated, during the measurementyear or the year prior. The eye examination and microalbuminuriascreening may also have been conducted duringeither the measurement year or the prior year ifcertain criteria (generally, milder diabetes) were met.

Study Cohort

The methods used to create the study cohort weresimilar to those reported previously.7-9 In 1999, 293plans submitted individual-level data for 1 or more ofthe comprehensive diabetes care measures, andbetween 269 and 287 submitted plan-level summarydata, depending on the measure. For 2001, the numbersdecreased to 161 and 159 to 161, respectively. Thesedecreases reflected withdrawal of plans from theMedicare+Choice (M+C) program during that time period.We merged the 157 630 HICs for the beneficiarieswith diabetes from 1999 with the M+C plan enrollees in2001. This merger resulted in 95 515 people on whominformation was submitted by M+C plans for any HEDISmeasure, including 22 637 persons for whom informationon diabetes measures was submitted. The decreasein the number of persons in the 1999 cohort for whomdata were also reported in 2001 was due to the almostexclusive use of the "hybrid method" to report the comprehensivediabetes care measures.14 The hybridmethod involves manual searching of the clinical orother records by the plans of a sample of the populationthat has diabetes, generally 411, in order to find theresult for the comprehensive diabetes care measure.Because most plans had more than this number of persons with diabetes in each year, different persons withdiabetes were selected each year. Thus, the mean numberwith information reported for both years was 34 to59, depending on the measure. Because the upper agelimit for the comprehensive diabetes care measure is 75years, beneficiaries aged 74 and 75 years in 1999 eitherhad no information provided for 2001 or had to beexcluded. Persons who switched plans were maintainedin the cohort. Their individual-level information wasbased on the 1999 information, and their plan-levelinformation was that of the plan in 2001.

To create a study cohort for whom we thought thedata would be most accurate, we applied the same dataquality checks that we have described previously.7-9 Ifmore than 5% of the HICs submitted by a plan were inthe wrong format, we eliminated all members (4 planswith 156 members). Five-hundred seventy-six personsfrom 8 plans were deleted because control informationfor HbA1c or LDL-C was provided for 10% or more of themembers, but the plan did not indicate that the personhad had the test. Additional plans were omitted if thesummary value submitted by the plan for the comprehensivediabetes care measure was 10 or more absolutepercentage points different form the value we calculatedusing the individual data submitted by the plan.Three (nephropathy monitoring) to 15 plans (poorHbA1c control) with 213 to 2379 members were deleted.(For all but poor HbA1c control, the range was 3-5 plansand 213-335 persons.) At the individual level, dependingon the measure, 142 to 145 persons were eliminatedbecause they were older than 75 years; 441 to 1404persons did not have a value for a measure in 2001; and1455 to 1684 had race indicated as other, unknown, orNorth American Native. As a result of these steps, datafor between 16 654 (poor HBA1c control) and 19 422(HbA1c testing) persons were available for analysis.

Finally, we noted that, depending on the measure,between 10 843 and 11 832 of the above beneficiarieswere enrolled in 1 plan in 2001. This large plan used theadministrative method to report. Thus, it reported on allbeneficiaries with diabetes. Almost all other plans hadused the hybrid method in at least 1 year or were muchsmaller in size. To prevent this 1 plan from overwhelmingthe results, we randomly selected a number of beneficiarieswith diabetes from this plan equal to thenumber contributed by the plan reporting on the nextlargest number of beneficiaries for that measure, andincluded them in the cohort (134-306 depending on themeasure). This tactic resulted in the following populationsfor analysis: HbA1c testing = 6742; poor HbA1c control= 3946; eye examination = 6653; LDL-C testing =7316; LDL-C control = 4959, and nephropathy monitoring= 6116, representing a total of 8875 beneficiaries.

Data Analysis


We tabulated the number and percent of beneficiariesfor each individual-and plan-level variable that metthe criterion for each of the comprehensive diabetescare measures in 1999 and 2001, and calculated thepercentage point difference between the 2 years.Changes in the percentage point difference between1999 and 2001 for each subgroup within an individualor plan-level variable for each measure were testedusing the test. We tested the differences in the percentagepoint increase between 1999 and 2001 betweensubgroups of each independent variable. Because wehad repeat measures on the same individuals in our"1999-2001" cohort, a generalized estimating equationwas used. For the variables with more than 2 categories,multiple comparisons were made against a referencegroup.


For persons whose care did not meet the comprehensivediabetes care standard in 1999, we measuredthe improvement in care in 2001 by individual-andplan-level characteristic, and compared differencesbetween the subgroups of each using the test, also.

Multivariate logistic modeling was also done on bothcohorts using the individual-level covariates of age, sex,Medicaid status, income, and whether the personswitched plans, plus the plan-level covariates of plan taxstatus, model of managed care (group, staff, or IPA), USCensus region location of the health plan, and quartileof total plan size based on Medicare enrollment. In addition,model-derived adjusted odds ratios and 95% confidenceintervals (CI) were calculated for the populationthat did not meet the comprehensive diabetes caremeasure criteria in 1999.


Between 1999 and 2001, statistically significantimprovements of between 6.0 and 15.7 percentagepoints (HbA1c testing and poor HbA1c control) were seenfor the comprehensive diabetes care measures (Table 1).Of the 174 assessments (29 subgroups of the 10 personandplan-level variables × 6 comprehensive diabetescare measures), statistically significant improvementwas noted in 167. For HbA1c testing the only subgroupthat showed no improvement was Hispanic patients.Otherwise, HbA1c testing improvements ranged from 3.1to 9.5 percentage points. Only 10.2% of the 1999-2001cohort had poor control of their diabetes, down from25.9% in 1999 (data not shown). Statistically significantdecreases in poor HbA1c control were seen among allsubgroups except those in a Medicaid-administered program.An 8.7 percentage point increase in eye examinations was noted in the total cohort, and statistically significantincreases in all subgroups. Significant increasesin LDL-C testing and control were seen among all subgroupsexcept Hispanic persons. The increases for thesemeasures tended to cluster around the value for theentire cohort, 13.7 percentage points for LDL-C testingand 15.0 percentage points for LDL-C control.Nephropathy monitoring improved by 12.6 percentagepoints for the entire cohort; however, no improvementwas seen for individuals in Medicaid-administered programsand those in plans of the second smallest size.

In general, the generalized estimating equationanalysis comparing the increases from 1999 and 2001between subgroups of each characteristic indicated similarincreases between the subgroups. For the characteristicsof particular interest—race, plan tax status,and plan model type—however, greater increases werenoted in LDL-C control and nephropathy monitoringamong black individuals (20.4 and 17.5 percentagepoints, respectively) than among white individuals(14.1 and 11.9 percentage points, respectively), and eyeexamination rates increased more among Asian patientsthan among white patients (27.2 vs 8.7 percentagepoints). However, the increase in the eye examinationrate among black persons (4.3 percentage points) wasless than that among white individuals, and testing forLDL-C increased less for Hispanic persons than forwhite individuals (5.5 vs 13.4 percentage points).Differences in percentage point changes between for-profitand nonprofit plans were seen only in nephropathymonitoring, which improved by 10.9 percentagepoints in the for-profit plans and 15.2 percentage pointsin the nonprofit plans. Group model plan enrollees hada bigger percentage change in nephropathy monitoring(14.5) compared with IPA plan enrollees (11.0). Agreater increase was noted among plan switchers inHbA1c control and nephropathy testing (23.2 and 23.0percentage points, respectively) than among nonswitchers(13.8 and 10.5 percentage points, respectively.

Comparisons of the unadjusted rates of the comprehensivediabetes care measures in 2001 in the 1999-2001 cohort between subgroups indicated severalstatistically significant differences between race subgroups(Table 1). Black persons had significantly lowervalues (3.5 and 7.9 percentage points) than white individualsfor 5 of the 6 measures, and Asian individualshad statistically significantly better values than whiteindividuals for 5 of the 6 measures (range 6.6 to 19.9percentage points). Patients with diabetes in for-profitplans had consistently lower rates of care than those innonprofit plans, and patients with diabetes in IPA modelplans had consistently lower rates of care than those ineither group-or staff-model plans. The ranges in thepercentage point difference were 2.5 to 13.9 for profitversus nonprofit plans, 2.9 to 11.3 for group versus IPAmodel, and 4.9 to 18.3 for staff versus IPA model,respectively. Patients with diabetes who did not switchplans had higher levels of eye examinations, LDL-C control,and nephropathy monitoring: 4.7, 9.6, and 5.4 percentagepoints, respectively. The characteristics of age,sex, plan size, and geographic location did not have frequentor consistent patterns of differences among thesubgroups.

The multivariate analysis confirmed that the greatmajority of the differences in the unadjusted rateswere between subgroups of the covariates (Table 2).After adjusting for individual-and plan-level covariates,significantly poorer outcomes were seen for black personscompared with white individuals for the same 5comprehensive diabetes care measures noted in theunadjusted rates; additionally, Asian individualsreceived better treatment than white persons for 3 ofthe 5 measures. Members of for-profit plans did less wellon 4 of the measures. Members of IPAs had poorer carefor 5 of the 6 measures and the difference in the sixth,poor HbA1c control, almost achieved statistical significancein the multivariate model. A trend was noted forpeople who switched plans to be less likely to receiveappropriate care, although the difference between themand individuals who did not switch was statistically significantonly for eye examination and nephropathymonitoring.

Among those persons who had not met the comprehensivediabetes care criteria in 1999, in 2001 74.8%had their HbA1c tested, 16.2% had poor control of theirHbA1c, 55.1% had eye examinations, 75.6% had theirLDL-C tested, 51.8% had their LDL-C under control,and 44.8% were monitored for nephropathy (Table 3).Bivariate analyses of the different subgroups showedpatterns similar to those seen for the whole study populationin Table 1, although because of the smallernumber of beneficiaries, the differences in the unadjustedrates were not as frequently statistically significant.Black patients who did not meet acomprehensive diabetes care criterion in 1999 had significantlylower rates of HbA1c testing and eye examination(69.5% and 45.5%, respectively) than whitepatients (75.8% and 56.3%, respectively). Hispanic personswere more likely to have poor HbA1c control(24%) than white individuals (15.4%), and Asian individualshad the highest rates of eye examination(73.5%) and LDL-C control (71.2%). As in Table 1,those in Medicaid-administered programs, for-profitplans and IPA model plans did less well. Individualswho switched plans tended to do better than thosewho did not, although among the unadjusted ratesonly the differences for eye examination were statisticallysignificant (60.0% vs 53.7% for nonswitchers).

The multivariate analysis did not confirm most ofthe differences by race group or tax status seen in thecomparison of the unadjusted rates. Among the comparisonswith white individuals, the only odds ratio thatwas significantly different from 1 was for the comparisonbetween black persons and white individuals foreye examinations (odds ratio = 0.75, 95% CI = 0.58-0.97). Also, for only 1 comprehensive diabetes caremeasure, the rate of poor HbA1c control, was the relativeodds worse in profit compared with nonprofit plans(odds ratio = 0.49, 95% CI = 0.32-0.75). However, theodds ratios for members of staff or group model planswere significantly more than 1 compared with IPAmembers for HbA1c (odds ratios = 2.54, 95% CI = 1.12-5.78; and odds ratio = 1.75, 95% CI = 1.27-2.42, respectively)and LDL-C testing (odds ratio = 1.56, 95% CI =1.22-1.98, for group plans), eye examination (oddsratio = 1.63; 95% CI = 1.11-2.41, for staff model plans),and nephropathy monitoring (odds ratio = 2.03, 95% CI= 1.50-2.75; and odds ratio = 1.40, 95% CI = 1.19-1.64,respectively).


Our findings of substantial improvement between1999 and 2001 in the 6 comprehensive diabetes caremeasures for almost all individual-and plan-level characteristicsevaluated is consistent with the informationpublished by NCQA for the entire Medicare population.12 Despite these improvements, some importantdifferences were still noted between subgroups of thepopulation in 2001 across several of the measures.Among the 4 race groups, black persons showed thegreatest improvement in 5 of the 6 measures, but theunadjusted and multivariate analysis showed that in2001 they only achieved the performance levels seenin white individuals for 1 measure. The slightlygreater, but not significantly different, increases in 5 ofthe 6 comprehensive diabetes care measures seenamong the for-profit plan members compared with thenonprofit plan members were not sufficient to closethe gaps in diabetes care between these types of plansfor 4 of the measures. After adjusting for all the individual-andplan-level covariates, the relative odds for-profitplans were less than 1 for rates of eyeexamination, LDL-C testing, nephropathy monitoring,and poor HbA1c control. Similarly, despite improvementsin all 6 measures between 1999 and 2001, theresults of the multivariate analysis showed that compliancewas lower for 5 of the 6 comprehensive diabetescare measures among patients with diabetes inIPA model plans compared with individuals in eitherstaff or group model plans. These findings of differencesby ownership and plan model are consistentwith those reported by Himmelstein et al,10 and raiseconcerns that differences they initially pointed outusing data from 1996 were continuing, at least in thearea of diabetes care. Persons who switched plans hadthe same or greater (poor HbA1c control and nephropathymonitoring) percentage point improvements in all6 measures than did those who did not switch plans.However, in 2001, the unadjusted rates were lower inthe switchers. This trend was seen in the multivariateanalysis, also, which indicated that the 2001 valueswere significantly lower for eye examination andnephropathy monitoring among the switchers. Theinference is that these members had poorer quality ofcare before they switched, which was confirmed byexamining the measure values for 1999 (data notshown). Further, in 2000 and 2001, 41 and 65 plansstopped offering services to Medicare beneficiaries,respectively.18,19 Thus, because switching was notalways voluntary, plans that withdrew may haveoffered poorer quality diabetes care. Interestingly,among persons who received poor care in 1999, thetendency was for those persons who switched plans tohave had better care and test results in 2001, althoughonly 1 measure, eye examination, reached statisticalsignificance (Table 3 and odds ratio = 1.31, 95% CI =1.07-1.61). This finding may be explained by thenewly adopted plan providers carrying out initialworkups and screenings in new plan members.

Whether beneficiaries switched plans or not, thosewho had poor quality care in 1999 did not achieve thesame care levels in 2001 as those receiving adequatecare in 1999 (data not shown). However, approximately50% had eye examinations and had their LDL-Cunder control or had been screened for nephropathy,75% had been tested for HbA1c and LDL-C levels, andonly 16% had poor HbA1c control (Table 3).

The improvement we report is likely the result ofseveral factors, beginning with the reports of the DCCTtrial1 and the UKPDS2 that provided strong evidence tosupport the rigorous treatment of diabetes. Subsequently,NCQA and CMS have actively supportedimproving diabetes care and monitoring that care inboth the managed care and fee-for-service setting.5,20-22In particular, NCQA has partnered with the AmericanDiabetes Association in promoting the DiabetesPhysician Recognition Program, which assesses physicianperformance in the care of patients with diabetes.23 The hope is that all of these activities will exerta continuing influence on improving diabetes care.

One concern about our results could be that the beneficiarieswho were reported on in 1999 and 2001 mightrepresent a biased sample of the managed care populationregarding the care received. We compared the distributionof the individual-and plan-levelcharacteristics for each of the comprehensive diabetescare measures among those who were included in theanalysis and those who were excluded (data not shown,but available from the corresponding author). In ouropinion, the differences, although statistically significantdue to the large sample size, were not sufficient toharm the generalizability of the results to all Medicaremanaged care beneficiaries, with 1 exception, poorHbA1c control. As described in the Methods section, 15plans with 2379 members were excluded because thesummary information on HbA1c control submitted bythe plan was more than 10 percentage points differentfrom the value we computed from the individual-leveldata they submitted. We believe that those plans withthe higher quality data may provide either higher qualitydiabetes care or have mechanisms in place to assistin the monitoring of diabetes care. While this differencelikely affected the generalizability of the summaryvalue for HbA1c control, it would not necessarilyinvalidate the subgroup comparisons, particularly afterthe multivariate analysis.

We have previously commented on the weaknessescaused by the limitations of data collected according toHEDIS Technical Specifications.8 Briefly summarized,the HEDIS definition of poor HbA1c and LDL-C controlare liberal. The American Diabetes Association recommendsa HbA1c care goal level of less than 7%, and of lessthan 100 mg/dL for LDL-C control.24,25 Also, personswho die in the measurement year are excluded. Finally,the HEDIS measures apply only to diabetic individualsyounger than 76 years, and approximately 44% of theelderly with diabetes are 75 years or older.26

Despite these shortcomings, the information we havepresented confirms that the improvement in diabetescare reported by NCQA for the Medicare managed careplan members in general is also found among almost allof the population subgroups we analyzed. However,despite these improvements, for most of the comprehensivediabetes care measures, poor care continues tooccur more frequently among black patients thanamong white patients, among those in for-profit versusnonprofit plans, and among those in IPA plans comparedwith group and staff model plans. Finally, andsimilar to the NCQA national data, in 2001, only 75% ofMedicare managed care beneficiaries had eye examinationsthat met HEDIS criteria, fewer than two thirds hadtheir LDL-C under control, and only about one half hadbeen monitored for diabetic nephropathy.


We thank Dorothea Musgrave, MPH, formerly of CMS for assistanceobtaining the data, and Zhen Huang, MS, for early analysis of the 1999 data.

From the Division of Health Services Research and Policy, University of MinnesotaSchool of Public Health, Minneapolis, Minn.

This work was supported in part by HCFA contract HCFA 500-96-0023 and CMS contractCMS 500-01-0043.

Address correspondence to: A. Marshall McBean, MD, MSc, Division of HealthServices Research and Policy, University of Minnesota School of Public Health, MMC 97,Mayo Memorial Building, 420 Delaware Street, SE, Minneapolis, MN 55455.

N Engl J Med.

1. The Diabetes Control and Complications Trial Research Group. The effect ofintensive treatment of diabetes on the development and progression of long-termcomplications in insulin-dependent diabetes mellitus. 1993;329:977-986.



2. UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose controlwith sulphonylureas or insulin compared with conventional treatment and riskof complications in patients with type 2 diabetes (UKPDS 33) [published correctionappears in 1999;354:602]. 1998;352:837-853.

Diabetes Care.

Diabetes Care.

3. American Diabetes Association. Standards of medical care for patients with diabetesmellitus [published correction appears in 2003;26:972].2003;26(suppl 1):S33-S50.

4. Diabetes Quality Improvement Project Initial Measurement Set (Final Version).Available at: Accessed February10, 2005.

5. National Committee for Quality Assurance. HEDIS 1999: volume 2, TechnicalSpecifications. Washington, DC: National Committee for Quality Assurance; 1998.


6. Schneider EC, Zaslavsky AM, Epstein AM. Racial disparities in the quality ofcare for enrollees in Medicare managed care. 2002;287:1288-1294.

Health Aff (Millwood).

7. Virnig BA, Lurie N, Huang Z, Musgrave D, McBean AM, Dowd B. Racial variationin quality of care among Medicare+Choice enrollees. 2002;21(6):224-230.



8. McBean AM, Huang Z, Virnig BA, Lurie N, Musgrave D. Racial variation in thecontrol of diabetes among elderly Medicare managed care beneficiaries. 2003;26:3250-3256.

Arch Gen Psychiatry.

9. Virnig B, Huang Z, Lurie N, Musgrave D, McBean AM, Dowd B. DoesMedicare managed care provide equal treatment for mental illness across races?2004;61:201-205.


10. Himmelstein DU, Woolhandler S, Hellander I, Wolfe SM. Quality of care ininvestor-owned vs not-for-profit HMOs. 1999;282:159-163.

N Engl J Med.

11. Schneider EC, Zaslavsky AM, Epstein AM. Use of high-cost operative proceduresby Medicare beneficiaries enrolled in for-profit and not-for-profit healthplans. 2004;350:143-150.

The State of Health Care Quality: 2003.

12. National Commission for Quality Assurance (NCQA). Comprehensive diabetescare. In: Washington, DC: NCQA;2003:34-35. Available from Accessed November 7, 2003.

Health Plan Employer Data Information Set (HEDIS) 3.0.

13. NCQA. Washington,DC: National Committee for Quality Assurance; 1997.

HEDIS 2000 Technical Specifications.

14. NCQA. Washington, DC: NationalCommittee for Quality Assurance; 1999.

Am J Med Qual.

15. Hebert PL, Geiss LS, Tierney EF, Engelgau MM, Yawn BP, McBean AM.Identifying persons with diabetes using Medicare claims data. 1999;14:270-277.



16. Hux JE, Ivis F, Flintoft V, Bica A. Diabetes in Ontario: determination of prevalenceand incidence using a validated administrative data algorithm. 2002;25:512-516.

Diabetes Care.

17. Blanchard JF, Ludwig S, Wajda A, et al. Incidence and prevalence of diabetesin Manitoba, 1986-1991. 1996;19:807-811.

Medicare Managed Care Plans: Many Factors

Contribute to Recent Withdrawals; Plan Interest Continues.

18. General Accounting Office. Washington, DC:General Accounting Office; 1999.

Report to the Congress: Medicare

Payment Policy.

19. Medicare Payment Advisory Commission. Washington, DC: Medicare Payment Advisory Commission;March 2002. Available at: Accessed September 22, 2004.


20. Jencks SF, Cuerdon T, Burwen DR, et al. Quality of medical care delivered toMedicare beneficiaries: a profile at state and national levels. 2000;284:1670-1676.



21. Jencks SF, Huff ED, Cuerdon T. Changes in the quality of care delivered toMedicare beneficiaries, 1998-1999 to 2000-2001 [published correction appears in2002;289:2649]. 2003;289:305-312.

Quality Improvement

Organizations' 7th Scope of Work, Section C - Statement of Work.

22. Center for Medicare and Medicaid Services. Available from Accessed March 7, 2003.

Diabetes Physician Recognition Program.

23. NCQA. Available at: Accessed January 24, 2004.

Diabetes Care.

24. American Diabetes Association. Standards of medical care in diabetes.2004;27(suppl 1):S15-S35.

Diabetes Care.

25. Haffner SM, American Diabetes Association. Dyslipidemia management inadults with diabetes. 2004;27(suppl 1):S68-S71.

26. Center for Disease Control Diabetes Surveillance System. Data and trends.Available at: Accessed January 3,2004.