Estimating Angina Prevalence in a Managed Care Population

October 1, 2004
Carol Zaher, MD, MBA, MPH

,
George A. Goldberg, MD

,
Pam Kadlubek, MPH

Supplements and Featured Publications, Chronic Angina: Clinical Management and Cost of Care, Volume 10, Issue 11 Suppl

Estimations of angina prevalence were calculatedusing managed care administrative data and applying3 angina-related definitions. The definitionscomprised angina pectoris diagnosis codes, diagnosisand procedure codes signifying the broader conditionof coronary artery disease (CAD), includingangina pectoris, and diagnosis codes for the symptomof chest pain. Prevalence rates were calculatedin 2000, 2001, and the combined period of 2000and 2001 for each definition based on the number ofmembers with at least 1 day of eligibility in eachperiod. Results were compared with published estimatesand projected to the US population.

The prevalence rates per 1000 people for anginapectoris in 2000, 2001, and 2000-2001 were 12.3,14.0, and 17.5, respectively. The prevalence rate ishigher in the combined 2-year period primarily becausethere is little duplication in patients with anginawho appear in both years, but there is significantoverlap in the overall (denominator) population eligiblein both years. For CAD the rates were 52.2,59.9, and 65.4, respectively, and for chest pain theywere 63.4, 75.8, and 93.4, respectively. Rates werehigher in men versus women and in each successiveage group. These gender and age results wereobserved in the projections to the US population. Bycomparison, the American Heart Association (AHA)estimates angina pectoris prevalence to be 35 per1000 in 2001. The lower managed care rate for anginapectoris may reflect differences in data capture(ie, self-reported data for AHA vs claims submittedfor reimbursement for managed care). AHA estimatesare higher for women versus men while themanaged care estimates show the opposite trend.

Prevalence of angina in the United States is substantial.With the aging of the US population, numbersof patients with angina presenting to thehealthcare system can be expected to increase, furtheradding to the cost burdens facing managed care.

(Am J Manag Care. 2004;10:S339-S346)

Coronary artery disease (CAD), whichincludes myocardial infarction, anginapectoris (chest pain), and atherosclerosisof the coronary arteries, isestimated to affect approximately 13.2 millionpeople in the United States. Angina, amanifestation of CAD, occurs in approximately6.8 million people.1 Healthcareresource utilization and costs in this patientpopulation are significant, with angina citedas the first-listed hospital discharge diagnosisin 82 000 hospitalizations and associatedwith 910 000 physician office visits in 1999.2An estimated 516 000 coronary arterybypass surgery procedures and 571 000 percutaneoustransluminal coronary angioplastieswere performed in 2001. The angioplastyfigures represent an increase of 266% from1987. When total angioplasties (with andwithout stent) for 2001 are combined, morethan 1 million procedures were performedin 2001. The mean charges for bypass are$60 853 and $28 558 for angioplasty.1 Withthe aging of the US population as a wholeand the epidemics of obesity and diabetesaffecting children and adults alike, the totalnumbers of patients with angina and the correspondingservice utilization will increase,further straining a healthcare systemalready struggling to manage costs.

Against this background, this article providesadditional insight by estimating anginaprevalence rates in a managed care populationthrough the use of administrativeclaims data, comparing these results withpublished national statistics, and projectingthe results to the US population. Administrativeclaims are an important resourcewithin managed care organizations. Thesedata represent the population of patientswho seek care from the system. This populationis typically a subset of the total populationof patients who actually have a diseaseor condition. Demographic trends can bequantified, providing information for managedcare to use in developing strategies tobetter target and manage patients.

Data Source. The study data source includeslongitudinal, member-linked data onmedical services provided through commercialhealth maintenance organizations, preferredprovider organizations, Medicare risk,and other indemnity products. Derived frommore than 25 different managed care organizationsthroughout the United States, thisprivate benefit plan information includes30 million members over time and approximately11 million members annually. Healthinsurance eligibility information is availablefor 100% of the membership. Data for theprivate benefit plans are complete and availablefor paid non-inpatient professional,non-inpatient facility, and inpatient facilityclaims. For the subset of the population thatalso has drug benefit coverage, non-inpatientpharmacy claims are available. Datafully comply with the Health InsurancePortability and Accountability Act of 1996(HIPAA) and meet the requirements for deidentificationof protected health informationas specified in Section 164.514 of theHIPAA privacy standard.

InternationalClassification of Diseases, NinthRevision, Clinical Modification (ICD-9-CM)

The database contains patient demographicand eligibility information, inpatientand non-inpatient diagnoses by codes, inpatient and non-inpatient procedureinformation, non-inpatient drugs dispensed,and dates of service for drug andmedical information.

ICD-9-CM

Claims Definition of Angina. In determiningprevalence of angina, choice of diagnosisand procedure codes will have animpact on the number of patients identifiedwith the disease. The particular disease thecode is intended to signify is usually obviousby the code's text description. In submittingclaims for reimbursement, however, it doesnot necessarily follow that a physician willuse these codes in all instances of a patientpresenting with the disease. For example,the codes most specific for anginapectoris are 413, 413.0, 413.1, and 413.9.However, a patient with angina also hasCAD, and, thus, a physician may select thisbroader code or may opt to code for only thesymptom (ie, chest pain). We estimatedangina prevalence using 3 related definitions(Table 1). Although it is likely that "chestpain" by itself will be too broad a term todefine angina, we included it as a referencepoint because angina prevalence is oftenestimated using self-reported data fromquestionnaires that query on chest pain.These definitions are not necessarily mutuallyexclusive (eg, angina pectoris codes areincluded in the CAD definition).

Identification of Patients. Applying the3 code definitions, we reviewed diagnosisand/or procedure claims in 3 separate timewindows: calendar year 2000, calendar year2001, and calendar years 2000-2001 combined.Thus, patients with claims only in2000 are represented in the counts for 2000and 2000-2001 combined. Similarly, patientswith claims only in 2001 are representedin the counts for 2001 and2000-2001 combined. Patients with claimsin both years are represented in each singleyear and also in 2000-2001 combined,where they are counted only once. Withineach time window, we identified patientswho had at least 1 claim containing any ofthe specified codes for a particular definition,were at least 35 years of age, and wereeligible for medical benefits for at least 1 dayduring the time window.

Managed Care Prevalence Estimatesfor Angina. Prevalence estimates per 1000members using each of the 3 disease definitionsare presented in Table 2. Prevalence iscalculated overall, by gender, and by agegroup. Regardless of definition, prevalenceestimates increased from 2000 to 2001.Increases might be the result of a number offactors. In addition to actual increases inclinical disease, greater patient awarenessand improved screening may impact prevalence.The higher numbers observed in thecombined 2000-2001 time window are likelybecause of the availability of longer timeframes during which a claim with the appropriatediagnosis may be submitted for reimbursement.

For the angina pectoris and CAD definitions,prevalence estimates were greater formen than women across all time windows.One explanation for this finding may be agenuinely higher prevalence rate amongmen, particularly in the younger age ranges.The incidence of CAD in women lags 10years behind that of men, attributable to thepositive effects of endogenous estrogen inpremenopausal women.1 Although it may betrue that angina is more prevalent amongmen than women, other explanations arepossible. In particular, physicians may havea tendency to more aggressively pursue certaindiagnoses in males, supporting the falsehistorical notion that heart disease is a"man's disease." This gender gap is particularly evident in the CAD definition, supportingthe hypothesis that screening for specificdiseases may be more likely to occur inmen. The gap is narrowed using the anginapectoris definition, for which patientsactively presenting with symptoms consistentwith angina are likely to be diagnosed,regardless of gender. The prevalence estimatesusing the chest pain definition didnot reflect the men:women pattern observedwith the other 2 definitions.Estimates were generally the same in bothgenders, with a slight female predominance.Given the nonspecificity of this definition,these results are difficult to interpret. Thediagnosis of chest pain can represent painor discomfort of cardiac or noncardiac origin.The lack of gender differentiation in thechest pain category may represent how menand women are evaluated equally at initialpresentation, but men are more likely toundergo more aggressive diagnostic evaluationfor an underlying cardiac etiology.However, the absence of a differencebetween men and women again suggeststhat a patient presenting with a specificsymptom is likely to have that symptomrecorded on the claim, regardless of gender.

Estimates for all 3 definitions showedincreased prevalence with each succeedingage group, consistent with the natural historyof the disease.

Table 3 presents results for the anginapectoris definition, broken out by age withingender. The trends did not change, withprevalence rates greater for men thanwomen for each age range and with increasedprevalence with increasing age.

Developing claims-based algorithms thatidentify patients with angina and balancesensitivity and specificity is challenging.Broader definitions based on CAD or chestpain will identify patients who may not actuallyhave angina. Using the angina pectorisdefinition provides the more specific definition,but will not identify all cases. Forexample, of the 107 933 patients with anginapectoris with drug coverage in the 2000through 2001 time window, 33 636 (31.2%)patients had a prescription for short-actingnitroglycerin (sublingual, translingual, orbuccal tablet or spray), the only medicaltreatment for an acute angina attack. If wereview the claims history of all patients inthis time window eligible for drug coverage,regardless of whether they were identifiedusing the angina pectoris definition, weidentify a total of 94 351 patients with ashort-acting nitroglycerin prescription.Thus, 60 715 patients with short-actingnitroglycerin use did not have a claim withan angina pectoris ICD-9-CM diagnosis codeand, thus, were not captured by this anginadefinition. Because many patients withchest pain of uncertain cause are given aprescription for short-acting nitroglycerin asa so-called "therapeutic trial," it is unrealisticto assume that all of these 60 715 doindeed have angina. Nevertheless, it is probablethat a portion of these patients haveangina. Some may have a past history ofangina or CAD but no symptoms in thestudy period and would therefore not be capturedas cases. Thus, the prevalence ratesemanating from the angina pectoris diagnosis-based definition are likely an underestimate.Similarly, the usage of short-actingnitroglycerin by patients with angina(31.2%) is likely an underestimate sincepatients may have had access to nitroglycerinobtained prior to the study period.

ICD-9-CM

Managed Care Prevalence Estimatesfor Specific Services and Drugs. Anginamanagement options include both revascularization procedures—coronary artery bypassgrafting and percutaneous transluminalcoronary angioplasty with or without stenting—and drug therapy with nitrates, beta-blockers,and/or calcium channel blockers.Although beta-blockers and calcium channelblockers can be used for indications otherthan angina, this is not generally the casewith nitrates. Thus we opted to measure therate of nitrate use, both long-acting andshort-acting, within the most specific definitionof angina, diagnosis code 413and all subcodes, to serve as an indicator ofmedical management of angina. Nitroglycerinpatches were included in the definitionof long-acting nitrates. We alsomeasured the occurrence of claims consistentwith revascularization procedures toserve as an indicator of interventional managementof angina.

Similar to the prevalence estimates forangina, comparison between genders inmost instances demonstrated higher ratesfor men across service and drug. Whereasthe long-acting nitrate rates were higher inwomen than in men (Table 4), this trendwas not evident after stratifying by age withingender, except at the greater age ranges(Table 5). The revascularization rates wereconsistently higher in men versus women,particularly in the younger age ranges.

The use of long-acting nitrates increasedwith each succeeding age group, regardlessof gender. Revascularization rates increaseduntil age 74 years in women and age 64 inmen, and then declined. These findings arenot unexpected because contraindicationsto surgical intervention are more likely toexist as the patient gets older.

Comparison With Published NationalStatistics. Table 6 presents prevalence estimatesfor angina from the American HeartAssociation (AHA),1 together with the managedcare estimates from this study. Theseestimates differ in 2 important respects.First, the managed care estimates for thetotal study population are somewhat lowerthan those reported by the AHA for the totalUS population. The AHA estimates arederived from the National Health andNutrition Examination Survey III (NHANESIII [1988-1994]), Centers for DiseaseControl and Prevention/National Center forHealth Statistics. A portion of the NHANESdata is self-reported, whereas the managedcare data are derived from claims submittedby providers and thus represent patientswho sought care. This may partially explainthe difference between the 2 estimates.Differences in the lower age boundary forangina between the 2 estimates may alsoimpact the results. The AHA estimates arebased on a lower age boundary of 20 years ofage while the managed care estimates have alower age boundary of 35 years of age.

Second, the AHA estimates show a higherprevalence in women than in men versusthe managed care data where prevalenceis higher in men than in women. As notedpreviously, one hypothesis is a physician'stendency to more aggressively screen andtreat men than women. Self-reported datawould not reflect this tendency, but mightinstead reflect cultural differences in thepropensity of men and women to reportsymptoms (eg, chest pain) or diseases (eg,angina or CAD).

Projections to the US Population. Themanaged care results were weighted toadjust for differences in characteristicsbetween the claims database population andthe US population. Specifically, a set ofweights based on age, gender, and regionwere constructed as the ratio of the 2000 USpopulation in each age/gender/region cell tothe number of members in the claims databasepopulation that also fell into the sameage/gender/region cells. These cell-specificweights were then applied to the claimsdatabase population and the rates wererecalculated.

The differences in prevalence estimatesbased on gender that were observed in themanaged care results were also observed inthe projected results. Thus, these differencescannot be explained by underlyingdifferences in gender distribution of theoverall managed care population versus theUS population. The same was true of thefindings relative to successive age groups(Table 7).

Discussion

The managed care data presented heresuggest that angina is a significant concernto managed care organizations in terms ofthe total numbers of patients who requiremanagement. A comparison with estimatesfrom the AHA indicates that the size of theangina population may be larger than thatidentified by managed care using conventionalpatient identification algorithms andthat not all patients who actually have anginaactively seek care.

The managed care prevalence estimatespresented in this study represent a high-levelexamination of CAD, angina, and chestpain based on the submission of a singleclaim with a diagnosis or procedure codeindicating one of the conditions. Furtherwork that examines the use of hierarchicaldefinitions that combine diagnosis, procedure,and drug markers of disease is needed.More complex algorithms might require thepresence of a hospitalization for angina, or2 or more claims with different dates ofservice each with a diagnosis code for angina,or 2 or more nitrate prescriptions submittedon different days. This greatercomplexity can provide an increased level ofcertainty that the claims algorithms areindeed identifying a patient with the soughtafter condition.

Although results were not adjusted forvariables other than gender and age, trendsrelating to gender differences in diagnosis andmanagement suggest the need for furtherinvestigation using a more rigorous approach.The higher medical and surgical treatmentrates in men versus women across almost allage groups suggest gender differences in patternsof care relating to the decision to treat.The differences are most striking in the revascularizationrates, further suggesting differencesin types of therapeutic decisions.

These findings are not unexpected,because gender differences in diagnosis,treatment, and outcomes of diseases havepreviously been cited in the literature.1,3-11Differences have been found in the referralof women versus men and blacks versuswhites for cardiac catheterization in chestpain, with women and blacks being less likelyto be referred.7 Lower rates of revascularizationin women have also been noted.3Gender differences in short-term managementversus long-term management ofmyocardial infarction have also been observed,with women less likely than men toreceive thrombolytic therapy within 60 minutesand aspirin within 24 hours of arrival atthe hospital and more likely to receiveangiotensin-converting enzyme inhibitors.8Gender differences have been found in thescreening and treatment of dyslipidemia10and also in noncardiovascular conditions.11Questions have been raised as to whethergender is an independent risk factor relativeto management and outcomes or whetherother variables may impact decisions regardingcare and resulting patient outcomes.9Further study is needed in these areas.

The use of administrative claims data hasinherent limitations. Physicians may notcode for the precise diagnosis or may notcode for a diagnosis at all. The number ofdiagnoses on claims is restricted. However,these data represent a ready source of informationthat are relatively inexpensive to useand that are viewed by managed care as representing"real-world" patterns and costs,and it is therefore important to analyze andunderstand these data.

By 2010, an estimated 40 million Americanswill be age 65 years and older.1 Giventhe implications to the healthcare system interms of angina prevalence and correspondinghealthcare costs, managed careorganizations will need to give careful considerationto strategies that can be implementedto care for these patients in acost-efficient manner that preserves qualityof care.

Acknowledgment

The authors gratefully acknowledge thecontributions of Ruby Vendiola in the preparationof this manuscript.

1. Heart Disease and Stroke Statistics–2004 Update.American Heart Association. Available at:http://www.americanheart.org/downloadable/heart/1072969766940HSStats2004update.pdf. Accessed December6, 2003.

2. Morbidity & Mortality: 2002 Chart Book on Cardiovascular,Lung, and Blood Diseases. National Institutesof Health, National Heart, Lung, and Blood Institute.Available at: http://www.nhlbi.nih.gov/resources/docs/02_chtbk.pdf. Accessed December 6, 2003.

Arch Intern Med.

3. Chang WC, Kaul P, Westerhout CM, et al. Impact ofsex on long-term mortality from acute myocardial infarctionvs unstable angina. 2003;163:2476-2484.

Womens Health Issues.

4. Grace SL, Abbey SE, Bisaillon S, Shnek ZM, Irvine J,Stewart DE. Presentation, delay, and contraindication tothrombolytic treatment in females and males withmyocardial infarction. 2003;13:214-221.

Prog Cardiovasc Dis.

5. Wenger NK. Coronary heart disease: the female heartis vulnerable. 2003;46:199-229.

Geriatrics.

6. Tecce MA, Dasgupta I, Doherty JU. Heart disease inolder women. Gender differences affect diagnosis andtreatment. 2003;58:33-39.

N Engl J Med.

7. Schulman KA, Berlin JA, Harless W, et al. The effectof race and sex on physicians' recommendations for cardiaccatheterization. 1999;340:618-626.

N Engl J Med.

8. Gan SC, Beaver SK, Houck PM, et al. Treatment ofacute myocardial infarction and 30-day mortality amongwomen and men. 2000;343:8-15.

Pharmacotherapy.

9. Bell DM, Nappi J. Myocardial infarction in women:a critical appraisal of gender differences in outcomes. 2000;20:1034-1044.

J GenIntern Med.

10. Kim C, Hofer TP, Kerr EA. Review of evidence andexplanations for suboptimal screening and treatment ofdyslipidemia in women. A conceptual model. 2003;18:854-863.

N Engl J Med.

11. Hawker GA, Wright JG, Coyte PC, et al. Differencesbetween men and women in the rate of use of hip andknee arthroplasty. 2000;342:1016-1022.