Objectives: To estimate the prevalence of concurrent hypertensionand dyslipidemia among a general veteran population andseparately among patients with diabetes mellitus, and to comparethe prevalence of cardiovascular disease among groups with isolatedversus concurrent hypertension and dyslipidemia.
Study Design: Retrospective cohort study.
Patients and Methods: This study was conducted in 6 medicalcenters of the Department of Veterans Affairs and included371 221 patients seen for any reason from October 1, 1998, toSeptember 30, 2001. The proportion of patients with isolated orconcurrent hypertension and dyslipidemia was estimated based ondiagnostic, pharmacy, laboratory, and vital sign information, andthe age-adjusted proportions of individuals with cardiovasculardisease were compared between groups.
Results: We found that 57.8% of all patients had hypertensionor dyslipidemia; 30.7% had both. Sixteen percent of all patientshad diabetes mellitus, and 66.3% of these patients had concomitanthypertension and dyslipidemia. The prevalence of coronaryartery disease was often more than doubled among patients withconcomitant conditions compared with patients with either conditionalone. The prevalence of stroke and peripheral arterial diseasesimilarly increased among patients with both conditions. Theprevalence of these cardiovascular diseases was highest amongpatients with diabetes mellitus.
Conclusion: The prevalence of cardiovascular disease was highamong this population of older, predominately male US veterans.
(Am J Manag Care. 2004;10:926-932)
Cardiovascular disease (CVD) is estimated toaffect more than 64 million individuals in theUnited States. Furthermore, CVD is the leadingcause of death in the United States, resulting in approximately38.5% of all deaths, or about 931 000 deaths peryear, and contributing to more than $368 billion inannual economic costs.1 Abnormalities in plasmalipoprotein metabolism play a central role in the pathogenesisof atherosclerosis, and arterial hypertensionwith elevated systolic or diastolic blood pressure (BP) ispositively and independently associated with coronaryheart disease (CHD).2,3
Data from the Framingham Study demonstrated thathypertension tends to occur in association with otheratherogenic risk factors (eg, 78% of hypertensive menand 82% of hypertensive women had multiple cardiovascular[CV] risk factors).4 Patients with multiple CVrisk factors are at much greater risk for CVD eventsthan those with 1 risk factor. Indeed, the risk of CVDassociated with the presence of concomitant hypertensionand dyslipidemia is typically greater than the sumof the CVD risks for hypertension and dyslipidemiaalone.5 This has been recognized in recent treatmentguidelines that emphasize the need to quantify a person'soverall CVD risk.6,7
This study was undertaken to estimate the prevalenceof hypertension, dyslipidemia, or both conditionsamong a large region of the Department of VeteransAffairs (VA) healthcare system. We also examined theprevalence of these conditions among patients with andwithout diabetes mellitus. We hypothesized that theproportion of patients with CVD morbidities would bemuch greater in patients with concomitant hypertensionand dyslipidemia compared with patients with isolatedhypertension or dyslipidemia. The study was conductedto assist resource allocation and the provision of effectivedisease management programs in the VA.
A retrospective design was used to collect data longitudinallyduring 3 years. The prevalence of hypertension,dyslipidemia, and diabetes mellitus was estimatedby identifying the proportion of all individuals whohad these conditions at any time during the study.Therefore, this period prevalence estimate combinesprevalent and incident cases.
International Classification of Diseases, Ninth
Revision, Clinical Modification (ICD-9-CM)
Computerized data from 6 VA acute care medicalcenters located in a regional network (Mississippi,Louisiana, Arkansas, Oklahoma, and Texas) wereobtained from October 1, 1998, to September 30, 2001,corresponding to fiscal years 1999 to 2001. Three datasources were used: (1) diagnostic information in theform of codes; (2)prescription drug dispensing from pharmacy dispensingrecords; and (3) clinical factors from laboratory testresults and vital signs. All data were transferred over theVA intranet from the regional network data warehousecomputer system to the study team for analysis. Thestudy was authorized by the local institutional reviewboards and VA research review committees at each site.
All patients seen at these 6 hospitals and their affiliatedoutpatient clinics during the 3 years formed thedenominator for this study (N = 371 221). Patients wereidentified for inclusion in the study based on a combinationof the 3 sources of data. Hospital dischargeabstracts and outpatient clinic visit records weresearched to find diagnostic criteria. Pharmacy prescriptionrecords were searched for fill records from the drugclasses indicated for treatment of any of the 3 conditions(hypertension, dyslipidemia, and diabetes mellitus).Laboratory test values and vital signs weresearched for any low-density lipoprotein cholesterol(LDL-C), hemoglobin A1C (HbA1c), or BP values.Elevated BP was defined according to the Joint NationalCommittee on Prevention, Detection, Evaluation, andTreatment of High Blood Pressure8 (130/85 mm Hg or140/90 mm Hg depending on risk factors), and elevatedLDL-C was defined according to the NationalCholesterol Education Program (100 mg/dL, 130 mg/dL,or 160 mg/dL depending on risk factors). ElevatedHbA1c was defined as 6.5% or higher. The use of HbA1calone (> 6.2%) as a diagnostic criterion has been shownto improve the case identification of diabetes mellitusabove the rates achieved by American DiabetesAssociation criteria, and patients with this level havebeen shown to be at increased CV risk.9-13 This value isoutside the normal range of HbA1c among patients inthe VA (normal range at the time of this study, 4.6%-6.5%), is useful as a screening tool in large populationsto improve sensitivity among patients in whom the useof fasting glucose tests is impractical, and has been usedpreviously within the VA.14 Patients were then classifiedas having hypertension, dyslipidemia, or diabetes mellitusif they met specific criteria based on algorithmsusing all 3 sources of data.15
Patients were classified as having hypertension ifthey met any of the following 5 criteria: (1) at least 2outpatient diagnoses of hypertension, (2) at least 1 prescriptionof an antihypertensive drug plus at least 1outpatient diagnosis of hypertension, (3) at least 1 prescriptionof an antihypertensive drug plus at least 1 elevatedBP measurement, (4) at least 1 elevated BPmeasurement plus 1 outpatient diagnosis of hypertension,or (5) at least 2 elevated BP measurements.
Patients were classified as having dyslipidemia if theymet any of the following 3 criteria: (1) at least 2 outpatientdiagnoses of dyslipidemia, (2) at least 1 prescriptionof an antilipemic drug, or (3) at least 1 elevatedfasting LDL-C level.
Patients were classified as having diabetes mellitus ifthey met any of the following 4 criteria: (1) at least 2outpatient diagnoses of diabetes mellitus, (2) at least 1inpatient diagnosis of diabetes mellitus, or (3) at least 1prescription of an antidiabetic drug or monitoring supply,or (4) at least 1 elevated HbA1c level.
For example, the algorithm for diabetes mellitusidentified patients by the presence of inpatient or outpatientdiagnostic codes, by use of drugs totreat diabetes mellitus, or by ever reaching elevatedHbA1c values. Patients meeting any of the 3 criteriawere thus classified as having diabetes mellitus. Patientswith dyslipidemia were similarly identified based onoutpatient codes, lipid-lowering drugs, orever reaching elevated LDL-C levels. Patients withhypertension could be identified based on codes alone, but identification based on drugs or clinicalfactors required additional diagnostic information toimprove specificity. For example, the use of any CVantihypertensive drug alone was not sufficient andrequired the presence of a diagnostic code or elevatedBP reading, because antihypertensive drugs could alsobe prescribed for indications other than hypertension.Patients who met these criteria formed the numeratorsfor disease groups. Complete algorithms with definitionsof diagnostic, pharmacy, and clinical factors are availableon request from the first author.
Demographic data (age and sex) and information onCVD comorbidities were similarly obtained from theinpatient and outpatient records. Comorbidity informationincluded the following nonfatal CV events: priormyocardial infarction (MI), coronary artery disease(CAD), and other atherosclerosis, including peripheralarterial disease (PAD) and cerebrovascular disease (thecoding of comorbidities is available on request from thefirst author). We do not report data on race or ethnicity,because this information is recorded as missing orunknown in up to 40% of cases for outpatients.
The prevalence of isolated and concurrent hypertensionand dyslipidemia is reported. Demographic informationis presented for selected disease groups. Wepresent age-adjusted proportions of patients with CVcomorbidities among patients with hypertension alone,dyslipidemia alone, or both conditions for nondiabeticpatients and separately for patients with diabetes mellitus.Age adjustment was performed by stratifying thesamples into 4 age groups: younger than 45 years, 45 to64 years, 65 to 74 years, and 75 years or older. The differencein proportions of patients with CV comorbiditieswas compared between the 3 groups (hypertensionalone, dyslipidemia alone, or both conditions) and testedby the normal approximation to the binomial.
Overall Observed Prevalence
A total of 371 221 patients were seen in these VAmedical centers during the 3 years. About 90% of thesepatients were male, and their mean age was 57.7 years.A total of 214 497 patients (57.8%) met the criteria forhaving hypertension or dyslipidemia, and 113 803patients (30.7%) met criteria for having both conditions.Specifically, 52.1% of patients (n = 193 497) met the criteriafor hypertension, 36.3% of patients (n = 134 803)met the criteria for dyslipidemia, and 16.1% of patients(n = 59 900) met the criteria for diabetes mellitus.
Prevalence and Demographic InformationFor all subsequent analyses, we stratified the sampleinto 2 groups: nondiabetic patients and patients withdiabetes mellitus. Among nondiabetic patients, 21.7%had hypertension without dyslipidemia, 5.7% had dyslipidemiawithout hypertension, 23.8% had both conditions,and 48.8% had neither. The mean age was 60.2years for patients with hypertension only, 56.5 years forpatients with dyslipidemia only, and 62.6 years forpatients with concomitant hypertension and dyslipidemia(Table 1). Among patients with diabetes mellitus,rates of isolated conditions were similar to those of nondiabeticpatients, but the rate of concomitant hypertensionand dyslipidemia was 66.3% (Table 2), or morethan double the proportion in nondiabetic patients.
Risk for Cardiovascular Outcomes inDisease Groups
The proportions of patients with CV comorbiditiesgenerally increased with age within each CV risk factorgroup (Table 3). Proportions of all CV comorbiditieswere statistically significantly higher (<.05) inpatients with concomitant hypertension and dyslipidemiacompared with patients with isolated hypertensionor dyslipidemia, except for the following 3subgroups of patients with diabetes mellitus (thoseyounger than 45 years with PAD, those younger than 45years with cerebrovascular disease, and those 75 yearsor older with cerebrovascular disease). Among patientswith both hypertension and dyslipidemia, the proportionsof patients with MI were generally 2 to 3 times theprevalences among patients with isolated conditions.For example, in the group aged 45 to 64 years, 2.0% to4.7% of patients with hypertension or dyslipidemiaalone had a history of MI, rising to 9.3% to 12.0% amongpatients with both conditions.
The proportions of patients with each CVD variedbetween patients with hypertension, dyslipidemia, orboth conditions. In many cases, the proportions in eachdisease group were statistically significantly differentfrom those in the other groups. For example, amongnondiabetic patients, the prevalence of CAD was statisticallysignificantly lowest for all age categories amongthe hypertension-only group compared with the dyslipidemia-only group or patients with both conditions.Furthermore, the prevalence of CAD for all age categorieswas statistically significantly highest among theconcomitant hypertension and dyslipidemia group,compared with the dyslipidemia-only group. The prevalencesof PAD and cerebrovascular disease were lowestamong patients with dyslipidemia only (among patientswith and without diabetes mellitus) and were about 2 to3.5 times higher among patients with both conditionsversus those with hypertension or dyslipidemia alone.
Among patients with diabetes mellitus, proportionsmarked in Table 3 with a double dagger were statisticallysignificantly different (almost always higher) than thecorresponding proportions among nondiabetic patients.For example, compared with nondiabetic patients, foralmost all age groups, the prevalence of all CV comorbiditieswas much higher among patients with hypertension,implying that diabetes mellitus may confer anadditive risk to the risk from hypertension alone.However, among patients with dyslipidemia and diabetesmellitus, the rates were not statistically significantlydifferent (except for those younger than 45 yearsor 75 years or older with CAD), indicating that theremay be no additional risk conferred by diabetes mellitusor that dyslipidemia and diabetes mellitus are collinear.All rates were highest among patients with diabetes mellitusand concomitant hypertension and dyslipidemia.
This study found that nearly one third (30.7%) of allpatients had both hypertension and dyslipidemia. Asubstantial proportion (57.8%) of all patients seen atthese VA medical centers during 3 years had hypertensionor dyslipidemia. The overall prevalence of diabetesmellitus was 16.1%. Among patients with diabetes mellitus,almost two thirds (66.3%) of patients also hadhypertension and dyslipidemia, which is more thantwice the rate (23.8%) of concomitant hypertension anddyslipidemia observed in the nondiabetic population inthis study; 86.6% of patients with diabetes mellitus alsohad hypertension and 71.6% also had dyslipidemia.These proportions were considerably higher than therates of hypertension and dyslipidemia (45.5% and29.5%, respectively) in the nondiabetic population.Therefore, 92% of patients with diabetes mellitus alsohad hypertension or dyslipidemia or both.
The prevalence of age-adjusted CV comorbidities (MI,CAD, PAD, and cerebrovascular disease) increased dramaticallyamong patients with concomitant hypertensionor dyslipidemia compared with patients withisolated conditions. The prevalence of all CV comorbiditieswas highest among patients with all 3 conditions.
We sought to place our findings in the context ofother studies. Estimates from the National Health andNutrition Examination Survey III found that the prevalenceof hypertension was 32.8% and the proportion ofpatients with LDL-C above 130 mg/dL was 49% for menand 43% for women.1 We found a 52.1% prevalence ofhypertension and a 36.3% prevalence of dyslipidemia.The national estimate for the prevalence of diabetesmellitus was 7.3%, whereas we observed that 16.1% ofVA patients had diabetes mellitus. Our findings wereconsistent with the hypothesis that VA patients tend tohave a higher prevalence of chronic disease than non-VA populations and that VA patients bear a larger burdenof disease relative to a non-VA sample.
Our observations have several important implications.First, they illustrate the severe burden of chroniccomorbid conditions and associated CV risk among veterans.In a study by Wolff et al16 of Medicare beneficiaries,nearly two thirds of older persons were found tohave multiple chronic conditions, placing them atincreased risk for costly avoidable hospitalizations andpreventable complications. Nearly one third (30.7%) ofall patients in our study had both hypertension and dyslipidemia.Applying this estimate to the entire VA system,some 2 million veterans currently seeking carewould have both of these prevalent conditions. Thiswould impose an enormous burden on the VA systemoverall and on individual providers in particular, forexample, in providing complex medical management formultiple comorbid conditions during limited clinic visits.Patients with these 2 conditions were found to have3 to 4 times the prevalence of MI than patients witheither condition alone, and 2 to 3 times the prevalencesof CAD, PAD, and cerebrovascular disease. Interestingly,rates of MI and CAD were also higher among nondiabeticpatients with hypertension and dyslipidemiathan among patients with hypertension and diabetesmellitus, implying that dyslipidemia may confer agreater risk than diabetes mellitus, even though diabetesmellitus is a risk equivalent to CHD. Finally, thefinding that 92% of patients with diabetes mellitus alsohave hypertension or dyslipidemia or both is striking.Recently, Snow et al17 summarized clinical trial evidenceemphasizing the need for tight BP control inpatients with diabetes mellitus. A companion article byVijan and Hayward18 noted that more than 80% ofpatients with type 2 diabetes mellitus develop or die ofmacrovascular complications (CAD, cerebrovasculardisease, or PAD). The authors questioned whether treatmentefforts should focus on macrovascular control,rather than glucose control and microvascular complications.Clinicians need to be aware of the large proportionof patients who have both hypertension anddyslipidemia, as well as the large proportion of patientswith diabetes mellitus who also have hypertension ordyslipidemia or both.
All 3 of the conditions examined are chronic andrequire comprehensive community-based and healthcaresystem disease management strategies, includingmedication, lifestyle modifications, and patient self-management.Recognizing that the first step in resourceplanning and treatment requires a practical way to identifypatients and determine their comorbidities, we useda method based on existing computerized medicalrecord information. Although clinical examination andlaboratory testing are ideal for determining the presenceof disease, it is cost prohibitive and impractical toundertake in a large national sample. Therefore, the useof a combination of diagnostic, pharmacy, laboratory,and vital sign data sources is a reasonable strategy tomaximally identify the prevalence of disease among alarge cohort.
Recent studies19,20 have suggested that substantialreductions in the risk of CVD-related events can beachieved by targeting hypertension and dyslipidemia.For example, it has been calculated that 79% ofischemic heart disease events and 69% of strokes wouldbe prevented if LDL-C levels were reduced by 70 mg/dL(1.8 mmol/L) and diastolic BP by 11 mm Hg.19 Similarly,in patients with metabolic syndrome, control of LDL-C,high-density lipoprotein cholesterol, and BP to normallevels would result in preventing 51.3% of CHD eventsfor men without CHD or diabetes mellitus and 42.6% forwomen; control to optimal levels would prevent 80.5%and 82.1%, respectively, of these events. Data supportingthese calculations have been obtained from clinicaltrials, which have demonstrated that the intensivetreatment of modifiable CV risk factors can markedlyreduce the risk of CV events.21,22 In addition, a recentstudy by Khot et al,23 which showed that 80% to 90% ofpatients with CHD have conventional risk factors(smoking, hypertension, hyperlipidemia, and diabetesmellitus) in contrast to conventional thinking that morethan half of such patients lack them, provides strongrationale for focusing on these conditions. The aggressivetreatment of these common modifiable CV riskfactors, particularly in patients at high risk for CVevents such as those with concomitant hypertensionand dyslipidemia or diabetes mellitus, could preventsome of the increased risk of MI and stroke in patientswith multiple versus single CV risk factors observed inthis study.
A possible concern of our study is that we determineda period prevalence estimate, which may be lessdesirable than obtaining a simple point prevalence estimate,or distinguishing the prevalence from the incidencerate. For example, our prevalence may beoverestimated by inclusion of incident cases or may beconsidered unreliable if the population is unstable, thedisease prevalence varies, or both. Because these arechronic conditions, the prevalence rate is not subject tothe fluctuations that would occur in measuring an acutecondition over time. Similarly, the population is notfixed but includes immigration and emigration; hence,it is a stable but dynamic cohort. In such cases, andwhen the exact onset of disease is difficult to determineas is the case with these conditions, it is preferable toestimate a period prevalence.24
A limitation is that our estimates of prevalence maybe imprecise because of selection biases or operationalizationof the clinical definitions by computerized medicalrecords. For example, not all patients may beidentified, because the diagnosis, pharmacy, and laboratorydata capture only services provided by the VAsystem. The use of single measurements of LDL-C orHbA1c might overestimate the prevalence of dyslipidemiaand diabetes mellitus; this highlights the importanceof using supplementary information fromdiagnostic and pharmacy sources in addition to the clinicalfactors. Conversely, patients who had their LDL-Cor HbA1c measured outside the VA system could not beidentified based on their clinical factors alone. Patientsmay use non-VA pharmacies, resulting in underrepresentationin the pharmacy records. However, becauseveterans are often uninsured or underinsured25 and theVA pharmacy copayment was only $2 per prescriptionduring this study (increased to $7 per prescription inFebruary 2002), underidentification from pharmacyrecords is less likely. Nonetheless, the inclusion of diagnostic,laboratory, and vital signs information furthermitigates this concern. Therefore, the use of all 3sources of information helps to balance limitations ofrelying on a single source of information. Finally,although not separately validated in this study, the algorithmsto identify patients have been used in other managedcare systems based on computerized data15 and arenot very different from algorithms used successfully inthe VA. For example, the study by Miller et al26 found thata combination of prescription records for a diabetesmedication or codes for diabetes mellitus had93% sensitivity and 98% specificity against patient self-report.
Finally, the Joint National Committee onPrevention, Detection, Evaluation, and Treatment ofHigh Blood Pressure recently released the SeventhReport,7 which lowered the target BP to 130/80 mm Hg(from 130/85 mm Hg) in patients with diabetes mellitusor chronic kidney disease. We determined it wasappropriate to use the Sixth Report8 levels that werecurrent at the time of our study; however, our estimateof the prevalence of hypertension is therefore now aconservative one.
This study estimated the prevalence of 3 chronicconditions (hypertension, dyslipidemia, and diabetesmellitus) with significant CV risk and found them to becommon among patients in the VA. In fact, the VA populationbears a larger burden of CV conditions thannon-VA populations. Identification of the burden ofdisease is essential for clinicians and managers to properlyprovide healthcare and manage resources. Theprevalence of CVD increased dramatically amongpatients with more than 1 condition and appeared toincrease more than additively. Further research isneeded to quantitatively describe the increased risk ofpatients with multiple versus single CV conditions. Ourstudy used a practical method to identify patients usingexisting computerized sources of information. Such amethod also forms the basis for future work to describetreatment patterns and estimate attainment of treatmentgoals. In accord with recent therapeutic guidelines,our observations highlight the importance ofdiagnosing and treating all CV risk factors to reduce thedevelopment of CVD. The major policy implication ofthis research is that the accurate identification of thecomplete burden of disease, including the presence ofmultiple chronic conditions, is essential to providehealthcare systems with the necessary information forresource allocation and provision of comprehensivedisease management.
We acknowledge the support of the Veterans Integrated ServiceNetwork 16 data warehouse team, Susan Pendergrass, DrPH, and JackBates, MS; the individual principal investigators at each off-site VA medicalcenter, Marisue Cody, PhD, Philip Comp, MD, Ali Mansouri, MD, EliseGomez-Sanchez, MD, and Shirley Butts, PhD; Abeer Alsarraj, BS, for projectsupport; and Meei Ku-Goto, MS, and Raji Sundaravaradan, BS, forcomputer programming. We especially thank Dan Pettitt, MS, DVM, for hisearly contributions to the research design, and Jon Edwards, PhD, forhis editorial support.
From the Houston Center for Quality of Care and Utilization Studies, Michael E.DeBakey VA Medical Center, and Section of Health Services Research, Department ofMedicine, Baylor College of Medicine, Houston, Tex (MLJ, KP, RJB); Global OutcomesResearch, Pfizer Inc, New York, NY (DSB) at the time of this study; and Leonard DavisInstitute of Health Economics, University of Pennsylvania, Philadelphia (DSB).
This research was supported in part by the Health Services Research and DevelopmentService, Department of Veterans Affairs, and by a grant from Pfizer Inc.
Drs Johnson and Beyth (both career development awardees of the Health ServicesResearch and Development Service, Department of Veterans Affairs) and Dr Pietz wereemployed in the Medical and Research Care Lines, Michael E. DeBakey VA Medical Center,Houston, at the time of this work. The views expressed in this article are those of the authorsand do not necessarily represent the views of the Department of Veterans Affairs.
Address correspondence to: Michael L. Johnson, PhD, Michael E. DeBakey VAMedical Center (152), 2002 Holcombe Boulevard, Houston, TX 77030. E-mail: email@example.com.
Heart Disease and Stroke Statistics: 2004
1. American Heart Association. Dallas, Tex: American Heart Association; 2003.
The Framingham Study: An Epidemiologic Investigation of
Cardiovascular Disease, Section 26.
2. Shurtleff D. Washington, DC: US Government PrintingOffice; 1970.
3. SHEP Cooperative Research Group. Prevention of stroke by antihypertensivedrug treatment in older persons with isolates systolic hypertension: final resultsof the Systolic Hypertension in the Elderly Program (SHEP). 1991; 265:3255-3264.
J Hum Hypertens.
4. Kannel WB. Fifty years of Framingham Study contributions to understandinghypertension. 2000;14:83-90.
Curr Opin Nephrol Hypertens.
5. Borghi C. Interactions between hypercholesterolemia and hypertension: implicationsfor therapy. 2002;11:489-496.
6. Expert Panel on Detection, Evaluation, and Treatment of High BloodCholesterol in Adults. Executive Summary of the Third Report of the NationalCholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, andTreatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). 2001;285:2486-2497.
7. Chobanian AV, Bakris GL, Black HR, et al. The Seventh Report of the JointNational Committee on Prevention, Detection, Evaluation, and Treatment of HighBlood Pressure: the JNC 7 report. 2003;289:2560-2572.
Arch Intern Med.
8. Joint National Committee on Prevention, Detection, Evaluation, andTreatment of High Blood Pressure. The Sixth Report of the Joint NationalCommittee on Prevention, Detection, Evaluation, and Treatment of High BloodPressure. 1997;157:2413-2446.
9. Jesudason DR, Dunstan K, Leong D, Wittert GA. Macrovascular risk and diagnosticcriteria for type 2 diabetes: implications for the use of FPG and HbA1c forcost-effective screening. 2003;26:485-490.
10. Perry RC, Shankar RR, Fineberg N, McGill J, Baron AD. HbA1c measurementimproves the detection of type 2 diabetes in high-risk individuals with nondiagnosticlevels of fasting plasma glucose: the Early Diabetes Intervention Program(EDIP). 2001;24:465-471.
11. Anand SS, Razak F, Vuksan V, et al. Diagnostic strategies to detect glucoseintolerance in a multiethnic population. 2003;26:290-296.
12. European Diabetes Policy Group 1998. A desktop guide to type 1 (insulin-dependent)diabetes mellitus. 1999;16:253-266.
13. European Diabetes Policy Group 1999. A desktop guide to type 2 diabetesmellitus. 1999;16:716-730.
14. Edelman D, Olsen MK, Dudley TK, Harris AC, Oddone EZ. Quality of carefor patients diagnosed with diabetes at screening. 2003;26:367-371.
Am J Manag Care.
15. Selby JV, Peng T, Karter AJ, et al. High rates of co-occurrence of hypertension,elevated low-density lipoprotein cholesterol, and diabetes mellitus in a largemanaged care population. 2004;10(pt 2):163-170.
Arch Intern Med.
16. Wolff JL, Starfield B, Anderson G. Prevalence, expenditures, and complicationsof multiple chronic conditions in the elderly. 2002;162:2269-2276.
17. Snow V, Weiss KB, Mottur-Pilson C; Clinical Efficacy AssessmentSubcommittee of the American College of Physicians. The evidence base for tightblood pressure control in the management of type 2 diabetes mellitus. 2003;138:587-592.
18. Vijan S, Hayward RA. Treatment of hypertension in type 2 diabetes mellitus:blood pressure goals, choice of agents, and setting priorities in diabetes care. 2003;138:593-602.
19. Wald NJ, Law MR. A strategy to reduce cardiovascular disease by more than80%. 2003;326:1419-1423.
20. Wong ND, Pio JR, Franklin SS, et al. Preventing coronary events by optimalcontrol of blood pressure and lipids in patients with the metabolic syndrome. 2003;91:1421-1426.
N Engl J Med.
21. Gaede P, Vedel P, Larsen N, et al. Multifactorial intervention and cardiovasculardisease in patients with type 2 diabetes. 2003;348:383-393.
22. Sever PS, Dahlof B, Poulter NR, et al; ASCOT Investigators. Prevention ofcoronary and stroke events with atorvastatin in hypertensive patients who haveaverage or lower-than-average cholesterol concentrations, in the Anglo-Scandinavian Cardiac Outcomes Trial-Lipid Lowering Arm (ASCOT-LLA): a multicentrerandomised controlled trial. 2003;361:1149-1158.
23. Khot UN, Khot MB, Bajzer CT, et al. Prevalence of conventional risk factorsin patients with coronary heart disease. 2003;290:898-904.
Principles and Quantitative Methods.
24. Kleinbaum DG, Kupper LL, Morgenstern H. New York, NY: Van Nostrand Reinhold;1982.
Health Aff (Millwood).
25. Wilson NJ, Kizer KW. The VA health care system: an unrecognized nationalsafety net. 1997;16(4):200-204.
26. Miller DR, Safford MM, Pogach LM. Who has diabetes? Best estimates of diabetesprevalence in the Department of Veterans Affairs based on computerizedpatient data. 2004;27(suppl 2):B10-B21.