Specialty Care: Heart Attack Survivors Treated by Cardiologists More
Likely to Take Recommended Drugs (Letter Report, 12/04/98,
GAO/HEHS-99-6).

Many studies comparing specialists and primary care physicians in
fee-for-service settings have found that specialists treat conditions
within their areas of expertise more intensively than primary care
doctors. Despite concerns that specialists may be more likely than
primary care physicians to provide costly and unnecessary care, studies
have shown that for some conditions, including heart attacks and severe
asthma, patients treated by specialists are more likely to receive
appropriate care and follow prescribed treatment regimens than patients
treated by other physicians. Less is known about differences in
treatment patterns between patients cared for by specialists and other
physicians in health maintenance organizations (HMO). Specialist care is
generally more expensive than care provided by other physicians, and
some studies have found that the proportion of appointments with
specialists is smaller for HMO patients than for those with private
fee-for-service insurance. This report examines potential differences in
treatment patterns for HMO patients treated by specialists and those
treated by generalist physicians. Specifically, GAO assesses follow-up
treatment for heart attack survivors enrolled in Medicare HMOs. GAO
chose this group because the differences in the quality of cardiac care
provided by cardiologists and generalists have been particularly
well-documented and the effectiveness of specific treatments for
coronary heart diseases has been clearly shown.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-99-6
     TITLE:  Specialty Care: Heart Attack Survivors Treated by 
             Cardiologists More Likely to Take Recommended Drugs
      DATE:  12/04/98
   SUBJECT:  Health maintenance organizations
             Health surveys
             Health care services
             Health insurance
             Elderly persons
             Physicians
             Cardiovascular diseases
             Drugs
             Comparative analysis
IDENTIFIER:  Medicare Health Maintenance Organizations Program
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Report to Congressional Requesters

December 1998

SPECIALTY CARE - HEART ATTACK
SURVIVORS TREATED BY CARDIOLOGISTS
MORE LIKELY TO TAKE RECOMMENDED
DRUGS

GAO/HEHS-99-6

Heart Attack Survivors in Medicare HMOs

(973800)


Abbreviations
=============================================================== ABBREV

  ACC - American College of Cardiology
  AHA - American Heart Association
  CCP - Cooperative Cardiovascular Project
  HCFA - Health Care Financing Administration
  HDL - high-density lipoprotein
  HMO - health maintenance organization
  LDL - low-density lipoprotein
  NHANES - National Health and Nutrition Examination Survey
  NHLBI - National Heart, Lung, and Blood Institute

Letter
=============================================================== LETTER


B-276534

December 4, 1998

The Honorable Ben Nighthorse Campbell
The Honorable John H.  Chafee
The Honorable Bob Graham
The Honorable James M.  Jeffords
The Honorable Joseph Lieberman
United States Senate

The Honorable Ken Bentsen
The Honorable Jim Greenwood
The Honorable James H.  Maloney
The Honorable Earl Pomeroy
The Honorable Pete Stark
House of Representatives

Many studies comparing specialists and primary care physicians in
fee-for-service settings have found that specialists treat conditions
within their areas of expertise more intensively than primary care
doctors.  Concerns have been raised that specialists may be more
likely than primary care physicians to provide costly and unnecessary
care.  In addition, some believe that, in certain situations,
patients benefit more from the integrated care that primary care
physicians are trained to provide.  Yet for a number of specific
conditions, including heart attacks and severe asthma, studies have
shown that patients treated by specialists are more likely to receive
appropriate care and follow prescribed treatment regimens than
patients treated by other physicians. 

Less is known about differences in treatment patterns between
patients cared for by specialists and other physicians in health
maintenance organizations (HMO).  Specialist care is generally more
expensive than care by other physicians, and HMOs often use primary
care physicians as gatekeepers to refer patients to specialty
services.  According to the National Ambulatory Medical Care Survey
and other sources, the proportion of physician appointments with
specialists is smaller for HMO enrollees than for those with private
fee-for-service insurance.\1

To learn more about the effect of physician speciality on the care
provided in HMOs, you asked us to examine potential differences in
treatment patterns for HMO patients treated by specialists and those
treated by generalist physicians.  To explore these differences, we
assessed follow-up treatment for heart attack survivors enrolled in
Medicare HMOs.  We chose to focus our examination on this group
because the differences in the quality of cardiac care provided by
cardiologists and generalists have been particularly
well-documented\2 and the effectiveness of specific treatments for
coronary heart disease has been clearly demonstrated.  Coronary heart
disease is the leading cause of death for the Medicare population,
and nearly 6 million Medicare beneficiaries--15 percent of the
eligible population--are currently enrolled in HMOs.\3

To conduct our analysis, we applied three standards of care--each
centered on a drug therapy--whose efficacy in reducing subsequent
morbidity and death for heart attack survivors has been well
established:  long-term use of cholesterol-lowering medications,
beta-blockers, and aspirin.  Specifically, we determined (1) the
proportion of Medicare heart attack survivors enrolled in HMOs who
take cholesterol-lowering drugs, beta-blockers, and aspirin and (2)
whether Medicare heart attack survivors in HMOs regularly treated by
a cardiologist are more likely to take cholesterol-lowering drugs,
beta-blockers, and aspirin than those who do not have regular
cardiology appointments.  We also examined the influence of
background patient characteristics and other medical conditions on
the use of these drugs. 

We surveyed Medicare HMO beneficiaries who were already a part of a
larger study on heart attack treatment--the Cooperative
Cardiovascular Project (CCP), conducted by HCFA.  CCP provided
detailed clinical data for each of our respondents from their initial
heart attack hospitalization in 1995.  The survey was conducted in
1997, about 2 years after the reported heart attack, and the sample
was restricted to individuals aged 65 to 84 when the heart attack
occurred.  CCP data allowed us to identify respondents with possible
contraindications for beta-blockers or aspirin, but it did not
include measurements of blood cholesterol levels. 

Because physician specialty data were not available for heart attack
survivors under fee-for-service Medicare, this report does not
compare the experiences of HMO patients and those in the traditional
Medicare program.  In addition, our report does not consider other
aspects of care provided to heart attack survivors, the
cost-effectiveness of care, or the procedures by which HMOs provide
specialist care to enrollees.  (For more details on our scope and
methodology, see app.  I; for a description of our statistical
analyses, see app.  II.) We conducted our work from January 1997 to
November 1998 in accordance with generally accepted government
auditing standards. 


--------------------
\1 C.  M.  Clancy and P.  Franks, "Utilization of Specialty and
Primary Care:  The Impact of HMO Insurance and Patient-Related
Factors," The Journal of Family Practice, Vol.  45, No.  6 (1997),
pp.  500-508; A.  B.  Flood and others, "How Do HMOs Achieve Savings? 
The Effectiveness of One Organization's Strategies," HSR:  Health
Services Research, Vol.  33, No.  1 (1998), pp.  79-99. 

\2 M.  T.  Donohoe, "Comparing Generalist and Specialty Care: 
Discrepancies, Deficiencies, and Excesses," Archives of Internal
Medicine, Vol.  158 (1998), pp.  1596-1608. 

\3 Throughout this report, the term "Medicare HMOs" refers solely to
so-called "risk plans"--which, in return for a specified monthly
capitated fee, assume full responsibility for the costs of patient
care--and not to other types of Medicare managed care plans in which
the Health Care Financing Administration (HCFA) reimburses for a
portion of the costs incurred by their enrollees. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

The ongoing use of cholesterol-lowering drugs and beta-blockers
reported by Medicare heart attack survivors enrolled in HMOs
generally parallels the patterns for heart attack survivors in the
U.S.  health care system overall.  As others have found for the
general patient population, we found a much smaller proportion of our
respondents reported taking cholesterol-
lowering drugs (36 percent) or beta-blockers (40 percent) than would
be expected if everyone who would benefit from using these drugs were
taking them. 

Medicare HMO heart attack survivors with regular cardiology care--40
percent of our survey respondents--were more likely to take the
recommended drugs than those without regular appointments with a
cardiologist.  Enrollees who saw cardiologists regularly for their
cardiac care were approximately 50-percent more likely to take
cholesterol-
lowering drugs and beta-blockers--a finding consistent with other
comparisons of care provided by cardiologists and generalists. 
Although factors such as age, education, self-reported health status,
and the presence of other illnesses also influenced who took
cholesterol-lowering drugs and beta-blockers, they did not account
for the higher use levels observed among patients who had routine
cardiology appointments.  Still, even patients of cardiologists often
did not take one or both of these drugs.  By contrast, the overall
use of aspirin was much higher--71 percent--and while regular
patients of cardiologists were still more likely to take aspirin, the
difference between them and other patients was smaller and not
statistically significant (75 versus 68 percent). 

On the whole, our results for heart attack survivors treated by
cardiologists and generalist physicians in Medicare HMOs are
consistent with those of other studies of physician specialty
differences in the United States.  Our finding that patients under
the regular care of cardiologists are more likely to take recommended
medications reinforces the findings of the small number of other
studies of physician specialty differences that are specifically
concerned with HMO members and extends those findings to an older
population and to a different medical condition. 


   BACKGROUND
------------------------------------------------------------ Letter :2


      STUDIES COMPARING CARE
      PROVIDED BY CARDIOLOGISTS
      AND PRIMARY CARE PROVIDERS
---------------------------------------------------------- Letter :2.1

Specialist physicians, by virtue of their narrower focus, can more
readily keep up with changes in clinical knowledge as they occur. 
This appears to be especially true for cardiac care, where changes in
treatment paradigms occur frequently.  Cardiologists also have the
advantage of seeing a larger number of patients with heart
conditions, so they have more experience with the range of variation
in presenting symptoms and responses to therapy. 

Numerous studies comparing the performance of cardiologists and
primary care physicians, or generalists, in providing patient care
tend to support the view that cardiologists provide a higher level of
cardiac care.  For example, researchers have found that cardiologists
demonstrate a better understanding of the appropriate use and
relative efficacy of alternative treatments for heart attacks and
congestive heart failure than generalists.\4 Moreover, cardiologists
are generally quicker to put successful innovations into practice and
to discontinue using therapies shown to be less effective.  This has
been found in the treatment of unstable angina as well as heart
attacks.\5 Studies have also demonstrated that cardiologists are more
likely to follow well-established treatment guidelines than
generalists.\6 Several studies report that cardiologists are more
likely to prescribe cholesterol-lowering drugs to patients with
elevated cholesterol levels and beta-blockers to heart attack
survivors.\7 A smaller group of studies has found that cardiologists
achieve better outcomes--including for inpatient care for heart
attacks.\8

Similar differences in practice patterns between specialists and
generalists have been found in the treatment of noncardiac conditions
as well, such as ulcers and strokes.\9

The findings of these studies do not mean that cardiologists always
provide superior care.  First, each study reports an overall
tendency, with considerable variation in performance among both
cardiologists and noncardiologists.  Moreover, some noncardiologists
do better than others.  For example, in several studies, the
performance of internists comes closer to that of cardiologists
(cardiology is actually a subspecialty within internal medicine) than
family practitioners.\10 Nonetheless, within cardiac care, studies
reveal a fairly consistent pattern--as physician specialization
increases, so does the overall level of adherence to established
standards of care. 

These studies, however, generally do not address the extent to which
HMOs affect the pattern of care provided by cardiologists compared
with that provided by noncardiologists.  The handful of studies
looking at physician specialty differences within an HMO setting have
focused on other medical conditions.  Specifically, we found two
recent studies by researchers employed by HMOs that compared the
treatment of asthma sufferers cared for by primary care physicians
and allergy and asthma specialists.\11 Statistically adjusting for
disease severity and patient characteristics, both studies found that
patients of specialists received more thorough and appropriate care. 
Specialists' patients more often reported taking medications
recommended by national treatment guidelines, had improved day-to-day
functioning, and had fewer asthma exacerbations requiring emergency
room treatment.  These findings suggest that treatment differences
across specialties can persist within an HMO structure.  However, in
cardiac care, comparable differences in care provided by primary care
providers and cardiologists might not be found if, for example, HMOs
placed a higher priority on standardizing care for cardiac patients. 


--------------------
\4 See J.  Z.  Ayanian and others, "Knowledge and Practices of
Generalist and Specialist Physicians Regarding Drug Therapy for Acute
Myocardial Infarction," The New England Journal of Medicine, Vol. 
331, No.  17 (1994), pp.  1136-42, and M.  H.  Chin and others,
"Differences in Generalist and Specialist Physicians' Knowledge and
Use of Angiotensin-Converting Enzyme Inhibitors for Congestive Heart
Failure," Journal of General Internal Medicine, Vol.  12, No.  9
(1997), pp.  523-30. 

\5 See T.  L.  Schreiber and others, "Cardiologist Versus Internist
Management of Patients With Unstable Angina:  Treatment Patterns and
Outcomes," Journal of the American College of Cardiology, Vol.  26,
No.  3 (1995), pp.  577-82; M.  A.  Hlatky and others, "Adoption of
Thrombolytic Therapy in the Management of Acute Myocardial
Infarction," The American Journal of Cardiology, Vol.  61 (1988), pp. 
510-14; and J.  Z.  Ayanian, "Knowledge and Practices of Generalist
and Specialist Physicians Regarding Drug Therapy for Acute Myocardial
Infarction."

\6 See M.  E.  Edep and others, "Differences Between Primary Care
Physicians and Cardiologists in Management of Congestive Heart
Failure:  Relation to Practice Guidelines," Journal of the American
College of Cardiology, Vol.  30, No.  2 (1997), pp.  518-26, and S. 
E.  Reis and others, "Unstable Angina:  Specialty-Related Disparities
in Implementation of Practice Guidelines," Clinical Cardiology, Vol. 
21 (1998), pp.  207-10. 

\7 See J.  J.  Whyte and others, "Treatment of Hyperlipidemia by
Specialists Versus Generalists as Secondary Prevention of Coronary
Artery Disease," The American Journal of Cardiology, Vol.  80 (1997),
pp.  1345-47; R.  S.  Stafford and others, "Variations in Cholesterol
Management Practices of U.S.  Physicians," Journal of the American
College of Cardiology, Vol.  29, No.  1 (1997), pp.  139-46; H.  M. 
Krumholz and others, "National Use and Effectiveness of
-Blockers for the Treatment of Elderly Patients After Acute
Myocardial Infarction," The Journal of the American Medical
Association, Vol.  280, No.  7 (1998), pp.  623-29. 

\8 See J.  G.  Jollis and others, "Outcome of Acute Myocardial
Infarction According to the Specialty of the Admitting Physician,"
The New England Journal of Medicine, Vol.  335, No.  25 (1996), pp. 
1880-87, and
I.  S.  Nash and others, "Do Cardiologists Do It Better?" Journal of
the American College of Cardiology, Vol.  29, No.  3 (1997), pp. 
475-78. 

\9 See A.  M.  Fendrick and others, "Differences Between Generalist
and Specialist Physicians Regarding Helicobacter Pylori and Peptic
Ulcer Disease," The American Journal of Gastroenterology, Vol.  91,
No.  8 (1996), pp.  1544-48; R.  A.  Hirth and others, "Specialist
and Generalist Physicians' Adoption of Antibiotic Therapy to
Eradicate Helicobacter Pylori Infection," Medical Care, Vol.  34, No. 
12 (1996), pp.  1199-1204; and L.  B.  Goldstein and others, "U.S. 
National Survey of Physician Practices for the Secondary and Tertiary
Prevention of Ischemic Stroke:  Carotid Endarterectomy," Stroke, Vol. 
27, No.  5 (1996), pp.  801-6. 

\10 See M.  A.  Hlatky, "Adoption of Thrombolitic Therapy in the
Management of Acute Myocardial Infarction"; R.  S.  Stafford,
"Variations in Cholesterol Management Practices of U.S.  Physicians";
and
H.  M.  Krumholz, "National Use and Effectiveness of
-Blockers for the Treatment of Elderly Patients After Acute
Myocardial Infarction."

\11 See W.  M.  Vollmer and others, "Specialty Differences in the
Management of Asthma:  A Cross-Sectional Assessment of Allergists'
Patients and Generalists' Patients in a Large HMO," Archives of
Internal Medicine, Vol.  157, No.  11 (June 9, 1997), pp.  1201-8,
and A.  P.  Legoretta and others, "Compliance With National Asthma
Management Guidelines and Specialty Care," Archives of Internal
Medicine, Vol.  158 (1998), pp.  457-64. 


      ASSESSING THE
      APPROPRIATENESS OF CARE FOR
      HEART ATTACK SURVIVORS
---------------------------------------------------------- Letter :2.2

Our study compares the use of three specific pharmacological
treatments among Medicare heart attack survivors who saw
cardiologists regularly and those who did not.  Although use of these
drugs represents only a portion of the post-heart-attack care
available, we chose to focus our analysis on this subset of
treatments because (1) there is strong scientific evidence that these
treatments are beneficial for a large proportion of heart attack
survivors and (2) other data indicate that many patients who would
benefit from these drugs are not using them. 

These two conditions do not apply to nearly the same extent to other
aspects of care provided to heart attack survivors.  For example,
while there is considerable variation in the extent to which invasive
procedures--such as cardiac catheterizations, angioplasty, and
coronary artery bypass graft surgery--are performed on heart attack
survivors, the evidence for these procedures is not as definitive as
the evidence supporting the use of cholesterol-lowering drugs,
beta-blockers, and aspirin.  As a result, existing clinical
guidelines for their use rest primarily on expert judgment.  For many
cases, that judgment is either equivocal or divided.\12 Thus, it is
more difficult to determine whether any given group of patients is
getting either too many or too few of these procedures. 

The value of cholesterol-lowering drugs, beta-blockers, and aspirin
for heart attack survivors has been widely publicized through
practice guidelines as well as numerous articles in prominent medical
journals.  It is therefore reasonable to expect physicians to know
about these therapies and to provide them to most of their patients,
while recognizing that the general benefits of these drugs may not
apply to certain individual patients. 

Since we limited the scope of our study to these drugs, we cannot
assume that our findings are indicative of relative performance in
other aspects of care, such as the appropriate use of invasive
procedures.  However, restricting the scope of this study to a set of
well-defined and well-supported therapies means that we can identify
with greater certainty a substantial number of patients who stood to
benefit from the treatments in question. 


--------------------
\12 J.  P.  Kahan and others, "Variations by Specialty in Physician
Ratings of the Appropriateness and Necessity of Indications for
Procedures," Medical Care, Vol.  34, No.  6 (1996), pp.  512-23. 


         PHARMACOLOGICAL
         TREATMENTS KNOWN TO BE
         BENEFICIAL
-------------------------------------------------------- Letter :2.2.1

Multiple, large-scale randomized clinical trials support the
widespread use of three pharmacological treatments in caring for
heart attack survivors. 

  -- Cholesterol-lowering medications:  A series of large-scale
     clinical trials have demonstrated the substantial therapeutic
     benefit of using "statin" drugs (HMG CoA reductase inhibitors)
     and other medications (in addition to proper diet and exercise)
     to lower the cholesterol level of people with coronary heart
     disease\13 --including those who have had a heart attack.  These
     studies show a reduction in subsequent coronary-related deaths
     for heart attack survivors ranging from 20 percent (for those
     with normal cholesterol levels) to 42 percent (for those with
     high cholesterol).  These studies have also shown a reduction in
     strokes of about 30 percent for both normal- and
     high-cholesterol patients.  These trials have been published in
     prominent journals and extensively described in the national
     media.  In 1993, the National Heart, Lung, and Blood Institute
     (NHLBI) issued practice guidelines that spelled out the
     implications of these trials for follow-up care of heart attack
     survivors.\14

Whether a patient should get such therapy depends on his or her
baseline level of low-density lipoprotein (LDL) cholesterol.  The
guidelines set an LDL goal for coronary heart disease patients of 100
mg/dL, well below the average level for the population as a whole. 
Those with baseline LDL levels of 130 mg/dL and above are definite
candidates for cholesterol-lowering medications, although specific
factors in individual cases can provide countervailing reasons not to
initiate drug therapy.  For those with baseline readings between 129
and 101, the guidelines recommend that physicians carefully weigh the
expected benefits and risks of cholesterol-lowering therapy for each
patient. 

  -- Beta-Blockers:  A second drug therapy whose benefits for heart
     attack survivors are well established in the clinical literature
     involves long-term use of beta-blockers.  This class of drugs
     inhibits stimulation of the heart and reduces the force of heart
     muscle contractions, thereby decreasing both the workload placed
     on the heart and arrhythmias that can lead to sudden death. 
     Beginning in the early 1980s, a series of large-scale clinical
     trials demonstrated that beta-blockers reduced overall mortality
     among heart attack survivors by about 25 percent.\15 Subsequent
     studies provided additional confirmation of these effects.\16
     Another study found that among the elderly patients surveyed,
     those receiving beta-blockers were 43-percent less likely than
     nonrecipients to die in the 2 years following their heart
     attacks.\17

In August 1990, the American College of Cardiology (ACC) and the
American Heart Association (AHA) jointly issued guidelines on the
management of heart attacks that cited these studies in support of a
general recommendation to treat heart attack survivors with
beta-blockers for at least 2 years, with the exception of patients
who had specific contraindications.  Six years later, ACC and AHA
issued revised guidelines that repeated this recommendation, while
reducing somewhat the scope of the stipulated contraindications.\18

In the years since the first beta-blocker trials were published, the
proportion of heart attack patients considered eligible to use them
has expanded.  In particular, the therapeutic value of beta-blockers
for many patients with moderately severe heart failure has become
more evident over time.\19 Thus, current ACC and AHA practice
guidelines list only relative contraindications, meaning that in each
case, the specific risks posed by beta-blockers for these patients
should be weighed against the general benefits. 

  -- Aspirin:  The 1990 practice guidelines for treating heart
     attacks issued by ACC recommended long-term aspirin therapy for
     all post-heart-attack patients "who could tolerate it." In its
     1996 revised guidelines, ACC specified that daily aspirin
     therapy should be continued indefinitely, with substitution of
     other antiplatelet agents only in the case of a "true aspirin
     allergy."

As with cholesterol-lowering medications and beta-blockers, multiple
randomized clinical trials provided the basis for these
recommendations.  A pooled analysis of these trials indicated that
long-term aspirin therapy led to a 13-percent reduction in vascular
mortality, a 31-percent reduction in recurrent nonfatal heart
attacks, and a 42-percent reduction in nonfatal strokes.\20


--------------------
\13 See J.  E.  Rossouw and others, "The Value of Lowering
Cholesterol After Myocardial Infarction," The New England Journal of
Medicine, Vol.  323, No.  16 (1990), pp.  1112-19; Scandinavian
Simvastatin Survival Study Group, "Randomised Trial of Cholesterol
Lowering in 4,444 Patients With Coronary Heart Disease:  The
Scandinavian Simvastatin Survival Study (4S)," Lancet, Vol.  344
(1994), pp.  1383-89; R.  P.  Byington and others, "Reduction in
Cardiovascular Events During Pravastatin Therapy:  Pooled Analysis of
Clinical Events of the Pravastatin Atherosclerosis Intervention
Program," Circulation, Vol.  92, No.  9 (1995), pp.  2419-25; F.  M. 
Sacks and others, "The Effect of Pravastatin on Coronary Events After
Myocardial Infarction in Patients With Average Cholesterol Levels,"
The New England Journal of Medicine, Vol.  335, No.  14 (1996), pp. 
1001-9; and P.  R.  Hebert and others, "Cholesterol Lowering With
Statin Drugs, Risk of Stroke, and Total Mortality," The Journal of
the American Medical Association, Vol.  278, No.  4 (1997), pp. 
313-21. 

\14 See National Cholesterol Education Program, "Second Report of the
Expert Panel on Detection, Evaluation, and Treatment of High Blood
Cholesterol in Adults (Adult Treatment Panel II)," Publ.  No. 
93-3095 (Bethesda, Md.:  National Institutes of Health; National
Heart, Lung, and Blood Institute; Sept.  1993). 

\15 See the Norwegian Multicenter Study Group, "Timolol-Induced
Reduction in Mortality and Reinfarction in Patients Surviving Acute
Myocardial Infarction," The New England Journal of Medicine, Vol. 
304 (1981), pp.  801-7, and Beta-Blocker Heart Attack Trial Research
Group, "A Randomized Trial of Propranolol in Patients With Acute
Myocardial Infarction (I):  Mortality Results," The Journal of the
American Medical Association, Vol.  247 (1982), pp.  1707-14. 

\16 See M.  M.  Bassan and others, "Improved Prognosis During
Long-Term Treatment With Beta-Blockers After Myocardial Infarction: 
Analysis of Randomized Trials and Pooling of Results," Heart and
Lung, Vol.  13, No.  2 (1984), pp.  164-68, and J.  Lau and others,
"Cumulative Meta-Analysis of Therapeutic Trials for Myocardial
Infarction," The New England Journal of Medicine, Vol.  327, No.  4
(1992), pp.  248-54. 

\17 S.  B.  Soumerai and others, "Adverse Outcomes of Underuse of
Beta-Blockers in Elderly Survivors of Acute Myocardial Infarction,"
The Journal of the American Medical Association, Vol.  277, No.  2
(1997), pp.  115-21. 

\18 See ACC/AHA Task Force, "Guidelines for the Early Management of
Patients With Acute Myocardial Infarction," Journal of the American
College of Cardiology, Vol.  16, No.  2 (1990), pp.  262-63, and T. 
J.  Ryan and others, "ACC/AHA Guidelines for the Management of
Patients With Acute Myocardial Infarction:  A Report of the American
College of Cardiology/American Heart Association Task Force on
Practice Guidelines (Committee on Management of Acute Myocardial
Infarction)," Journal of the American College of Cardiology, Vol. 
28, No.  5 (1996), pp.  1381-82, 1397-98. 

\19 See M.  J.  Kendall and others, "-Blockers and Sudden
Cardiac Death," Annals of Internal Medicine, Vol.  123, No.  5
(1995), pp.  361-64, and M.  Packer and others, "The Effect of
Carvediolol on Morbidity and Mortality in Patients With Chronic Heart
Failure," The New England Journal of Medicine, Vol.  334, No.  21
(1996), pp.  1349-55. 

\20 See ACC/AHA Task Force, "Guidelines for the Early Management of
Patients With Acute Myocardial Infarction," pp.  272-73; T.  J. 
Ryan, "ACC/AHA Guidelines for the Management of Patients With Acute
Myocardial Infarction," p.  1344; and R.  C.  Becker, "Antiplatelet
Therapy in Coronary Heart Disease:  Emerging Strategies for the
Treatment and Prevention of Acute Myocardial Infarction," Archives of
Pathology and Laboratory Medicine, Vol.  117 (1993), p.  93. 


         DRUG USAGE AS AN
         INDICATOR OF APPROPRIATE
         CARE
-------------------------------------------------------- Letter :2.2.2

The measure of appropriate care used in this study is whether
patients reported that they were actually taking cholesterol-lowering
drugs, beta-blockers, and aspirin about 2 years after their heart
attack occurred--not if these drugs were prescribed.  While tallying
prescriptions would be a more direct measure of one aspect of
physician behavior, none of the potential benefits of the drugs are
realized unless the patient is actually taking them.  Moreover,
research has demonstrated that self-reported drug use is strongly
related to more proximate measures of medication compliance, such as
pharmacy records of prescriptions filled and counts of pills
taken.\21

While it is ultimately the patient who decides how faithfully to
adhere to a treatment regimen, research has shown that physicians
strongly influence patient behavior by, among other actions,
prescribing certain medications, closely monitoring patient
compliance, and by simplifying and adjusting regimens to encourage
compliance.\22


--------------------
\21 See R.  E.  Grymonpre and others, "Pill Count, Self-Report, and
Pharmacy Claims Data to Measure Medication Adherence in the Elderly,"
The Annals of Pharmacotherapy, Vol.  32 (1998), pp.  749-54. 

\22 See W.  Insull, "The Problem of Compliance to Cholesterol
Altering Therapy," Journal of Internal Medicine, Vol.  241 (1997),
pp.  317-25. 


   PROPORTION OF MEDICARE HMO
   ENROLLEES WHO TAKE DRUG
   THERAPIES IS SIMILAR TO GENERAL
   POPULATION BUT BELOW DESIRED
   LEVELS
------------------------------------------------------------ Letter :3

Despite the strength of the clinical evidence, many patients who
would benefit from drug therapies to treat coronary heart disease do
not take the drugs.  While we were unable to test directly for
differences between the Medicare HMO enrollees in our sample and the
general population of Medicare fee-for-service heart patients, the
drug usage rates reported by our sample are both broadly comparable
to those found in studies by others of the fee-for-service population
and below the rates suggested by clinical guidelines. 

Just 36 percent of our sample reported taking any of the statin drugs
or another type of cholesterol-lowering drug.  NHLBI estimates that
only about one-third of patients in the general population with
coronary heart disease are receiving medications to lower their
cholesterol.\23 Further, based on cholesterol levels in the general
population of elderly Americans, we estimate that 57 percent of our
sample has LDL cholesterol levels of 130 or higher, and are therefore
clear candidates for cholesterol-lowering drugs given current
treatment guidelines for patients with established coronary heart
disease.\24 Our respondents' 36-percent usage rate falls considerably
short of that standard. 

Similarly, only 40 percent of our sample reported taking
beta-blockers.  As one comparison, 32 percent of Medicare
fee-for-service heart attack survivors in the CCP study received
prescriptions for beta blockers when they were discharged from the
hospital.\25 For the subset of our respondents identified in the CCP
study as ideal candidates for beta-blockers, the usage rate was
somewhat higher at 49 percent.\26 The finding that only one-half of
the ideal candidates took beta-blockers shows that these drugs are
underused as well. 

Usage rates for aspirin were much higher but still below recommended
levels.  At the time of our survey, 71 percent of our respondents
reported that they regularly took aspirin.  By comparison, CCP found
that 66 percent of Medicare fee-for-service heart attack survivors
were instructed to take aspirin when discharged from the hospital.\27

Similarly, 78 percent of our respondents identified as ideal
candidates for aspirin therapy in the CCP study took aspirin. 


--------------------
\23 Cited in S.  M.  Grundy and others, "When to Start
Cholesterol-Lowering Therapy in Patients With Coronary Heart Disease: 
A Statement for Healthcare Professionals From the American Heart
Association Task Force on Risk Reduction," Circulation, Vol.  95, No. 
6 (1997), p.  1683. 

\24 The procedures and data sources we used to derive this estimate
are described in appendix II. 

\25 See T.  A.  Marciniak and others, "Improving the Quality of Care
for Medicare Patients With Acute Myocardial Infarction," The Journal
of the American Medical Association, Vol.  279, No.  17 (1998), pp. 
1351-57. 

\26 Ideal candidates are those whose clinical records provided a
definitive indication that they did not have any of a number of
conditions that may be contraindications for a particular therapy. 
These criteria were established by a panel of medical experts
convened by HCFA and the American Medical Association as the CCP
study got under way.  (See T.  A.  Marciniak, "Improving the Quality
of Care for Medicare Patients With Acute Myocardial Infarction," p. 
1353.) The clinical information used to identify ideal candidates was
collected during the initial acute myocardial infarction
hospitalization--not at the time of our survey, approximately 2 years
later.  While this means that the identification of ideal candidates
is not strictly accurate for our sample, we believe that it is
unlikely that any measurement discrepancies would be large enough to
invalidate the finding that a substantial proportion of ideal
candidates did not take an appropriate drug.  (See app.  I for more
information about ideal candidate variables.)

\27 See T.  A.  Marciniak, "Improving the Quality of Care for
Medicare Patients With Acute Myocardial Infarction," p.  1355. 


   MEDICARE HMO ENROLLEES WHO SEE
   A CARDIOLOGIST REGULARLY ARE
   MORE LIKELY TO REPORT TAKING
   CHOLESTEROL-LOWERING DRUGS AND
   BETA-BLOCKERS
------------------------------------------------------------ Letter :4

Approximately 2 years after their heart attack, 41 percent of our
sample reported that they saw a cardiologist regularly.  For the
remainder, 19 percent reported that they visited a cardiologist only
occasionally--when they felt ill or when they were referred by their
primary care physician--and 40 percent told our interviewers that
they did not see a cardiologist about their heart (37 percent saw
only a primary care physician, and 3 percent saw a specialist
physician other than a cardiologist).  We compared the drug usage of
the 41 percent under the regular care of a cardiologist with that of
the 59 percent who saw a cardiologist occasionally or not at all. 

We found clear differences in the use of cholesterol-lowering drugs
and beta-blockers--and a smaller difference in aspirin usage--between
patients under the regular care of a cardiologist and all others.  As
table 1 shows, both cholesterol-lowering drugs and beta-blockers were
taken 50-percent more often by respondents who routinely saw a
cardiologist compared to those without regular cardiology
appointments.  In both cases, this is a statistically significant
difference.\28 For aspirin, we found that the tendency for patients
with regular cardiology appointments to have higher usage rates was
not statistically significant.\29



                          Table 1
          
           Drug Usage Rates for Patients With and
          Without Regular Cardiology Appointments

                                                  Patients
                                    Patients       without
                                with regular       regular
                                  cardiology    cardiology
                                appointments  appointments
Drug category          Overall         (41%)         (59%)
----------------  ------------  ------------  ------------
Cholesterol-               36%           45%         30%\a
 lowering
Beta-blocker                40            50          34\a
Aspirin                     71            75            68
----------------------------------------------------------
\a The difference between those with and without regular cardiology
appointments is statistically significant. 


--------------------
\28 Among patients without regular cardiology appointments who
occasionally saw a cardiologist, 29 percent took cholesterol-lowering
medications and 29 percent took beta-blockers.  Among patients
without regular cardiology appointments who received heart care from
a physician other than a cardiologist, 31 percent took
cholesterol-lowering medications and 36 percent took beta-blockers. 

\29 Among patients without regular cardiology appointments, 69
percent who occasionally saw a cardiologist and 68 percent who
received heart care from a physician other than a cardiologist took
aspirin. 


   PATIENT CHARACTERISTICS DO NOT
   EXPLAIN WHY PATIENTS WITH
   REGULAR CARDIOLOGY CARE TAKE
   APPROPRIATE DRUGS MORE OFTEN
------------------------------------------------------------ Letter :5

Our analysis shows that Medicare HMO heart attack survivors are more
likely to take appropriate heart-related medications if they have
regular follow-up appointments with a cardiologist.  The most direct
explanation for this finding is that cardiologists treat heart attack
survivors differently than physicians who are not heart specialists. 
However, taking medications is an outcome that involves patient as
well as physician behaviors, and differences in patient use of drug
therapies could be due more to differences in patient characteristics
than to differences in the treatment patterns of physicians.  For
example, patients who are most steadfast in their pharmaceutical
regimens may also be the most likely to seek specialty care. 

We tested this alternative explanation by conducting multivariate
statistical analyses to identify the variables associated with taking
each type of drug and with having regular cardiology appointments. 
These analyses included variables known from the work of other
researchers to influence the use of physician services or medication
compliance, including self-reported current health status; background
variables (such as education, current income, age, and race); and
clinical variables measured at the time of hospitalization (such as
heart attack severity and major comorbidities).\30 Because these
analyses found that the variables associated with having regular
cardiology appointments and with taking heart drugs are different, it
is unlikely that our finding--that patients with regular cardiology
appointments take these drugs more often--is due to systematic
differences between the patients who see cardiologists regularly and
those who do not.  However, as with any analysis of this type, it is
possible that patient attributes that are statistically unrelated to
any of the factors we examined could affect the relationship between
regular cardiology care and recommended drug therapy. 


--------------------
\30 See, for example, C.  M.  Clancy, "Utilization of Specialty and
Primary Care:  The Impact of HMO Insurance and Patient-Related
Factors," and J.  C.  Y.  Sung and others, "Factors Affecting Patient
Compliance with Anti-Hyperlipidemic Medications in a HMO Population,"
presented at the 144th Annual Meeting of the American Pharmaceutical
Association, 1997.  These analyses are described more fully in
appendix II. 


      HEALTHIER PATIENTS ARE MORE
      LIKELY TO REPORT TAKING
      CHOLESTEROL-
      LOWERING DRUGS,
      BETA-BLOCKERS, AND ASPIRIN
---------------------------------------------------------- Letter :5.1

In general, we found that healthier patients were more likely to take
all three types of drugs, although the specific predictive factors
varied among the drug categories.  For example, we found that
cholesterol-lowering drugs were taken more often by those who told
our interviewers that their current health was very good or excellent
(52 percent, compared to 31 percent of those in poor, fair, or good
health) and by those without other major illnesses at the time of the
heart attack (43 percent, compared to 29 percent of those with at
least one comorbidity).\31 Similarly, both beta-blockers and aspirin
were taken more often by those with fair to good heart function
measurements, compared to those with poor measurements.\32


--------------------
\31 We coded the presence of a comorbidity for heart attack survivors
with at least one of the following conditions at the time of their
hospitalization:  congestive heart failure, chronic obstructive
pulmonary disease, a previous stroke, dementia, or any form of
diabetes. 

\32 The category of those with fair to good heart function
measurements includes individuals with a left ventricular ejection
fraction of 35 or greater.  Forty-four percent of them reported that
they took beta-blockers and 79 percent aspirin, compared to 21
percent and 49 percent, respectively, of those with ejection
fractions less than 35.  Although originally viewed as a
contraindication for beta-blockers, the congestive heart failure
often associated with an ejection fraction less than 35 is
increasingly viewed as a condition that can be treated with
beta-blockers. 


      SOCIOECONOMIC VARIABLES
      AFFECT THE USE OF
      BETA-BLOCKERS AND ASPIRIN
---------------------------------------------------------- Letter :5.2

The use of beta-blockers and aspirin, but not of cholesterol-lowering
drugs, was also associated with variables reflecting socioeconomic
status.  Respondents with some postsecondary education, compared to
those whose education did not extend beyond high school, reported
greater use of beta-blockers (50 percent, compared to 34 percent) and
greater use of aspirin (76 percent, compared to 67 percent). 
Patients with incomes greater than the median for our sample also
used beta-blockers more often (48 percent, compared to 33 percent);
income did not affect aspirin use. 

We also found that cholesterol-lowering drugs were taken more often
by younger respondents (48 percent of those in the younger half of
our sample, aged 67 to 73 when they were interviewed, compared to 25
percent of those aged 74 to 86).  Respondent age, however, did not
affect the use of beta-blockers or aspirin.  In addition, gender and
race had no effect on usage rates for any of the three categories of
drugs. 


      PATIENTS WITH REGULAR
      CARDIOLOGY CARE ARE WHITE,
      YOUNG, AND SUFFERED MORE
      SEVERE HEART ATTACKS
---------------------------------------------------------- Letter :5.3

We conducted a separate analysis to identify patient-related
variables associated with having regular cardiology appointments.  We
found that those with regular cardiology appointments were more
likely to be white (43 percent had regular appointments, compared to
22 percent of nonwhites); relatively young (47 percent of those aged
73 or younger had regular appointments, compared to 34 percent of
those aged 74 to 86); and to have had relatively severe heart
attacks.\33 Regular cardiology care was not associated with gender,
educational attainment, current income, the presence of
comorbidities, or self-reported health status. 


--------------------
\33 We measured heart attack severity with an interval variable that
counted the presence of three indicators:  a transmural myocardial
infarction, a previous myocardial infarction, and angina more than 24
hours after arrival at the hospital.  Fifty percent of patients with
the most severe heart attacks by this measure (that is, with all
three severity indicators) had regular cardiology appointments, as
did 47 percent of those with two severity indicators, 41 percent of
those with one severity indicator, and 31 percent of those with no
severity indicators. 


      PATIENT CHARACTERISTICS
      ASSOCIATED WITH DRUG USE AND
      REGULAR CARDIOLOGY CARE ARE
      DIFFERENT
---------------------------------------------------------- Letter :5.4

We reexamined our analysis of factors associated with patients taking
cholesterol-lowering medications, beta-blockers, and aspirin, making
sure to include those variables that predicted regular care by a
cardiologist (race, age, and heart attack severity).  If the
relationship of regular care by a cardiologist to appropriate drug
therapy actually reflected differences in these patient
characteristics, then the inclusion of these factors in the analysis
would diminish greatly the statistical association of specialty care
with those treatments.  This did not occur.  Even with these factors
included in the analysis, the effect of regular visits with a
cardiologist did not change.  Neither race nor heart attack severity
was associated with taking any of the three types of drugs, and
patient age was associated only with taking cholesterol-lowering
medication.  Further, among the younger patients--those more likely
to have regular cardiology appointments--the usage rate of
cholesterol-lowering drugs was much higher among those with regular
cardiology appointments--60 percent, compared to 38 percent for those
without a regular cardiologist.\34


--------------------
\34 This difference is statistically significant.  Patients in the
older half of our sample, aged 74 and older at the time of the
interview, had regular cardiology appointments less frequently and
took cholesterol-lowering drugs less often; this lower usage rate was
not higher for those with regular cardiology appointments. 


   OBSERVATIONS
------------------------------------------------------------ Letter :6

On the whole, our conclusion that patients under the regular care of
a cardiologist are more likely to take recommended medications
parallels the findings of other studies of physician specialty
differences in the United States.  Our results also reinforce the
findings of the small number of other studies specifically concerned
with HMO members.  The pattern we found for older heart attack
patients in Medicare HMOs is the same as that reported by other
researchers for younger HMO members with asthma. 

One characteristic of medical care in the United States is that the
patients of specialist and generalist physicians sometimes receive
different treatments for the same medical condition.  Studies have
documented this phenomenon in both fee-for-service and HMO settings. 
However, it is both a special problem and a unique opportunity for
HMOs and their members.  It is a special problem because HMOs can
restrict access to specialists, perhaps leading some enrollees to
feel that they have been denied necessary care.  It is a unique
opportunity because these differences are not immutable and because
HMOs, unlike fee-for-service insurers, can actively manage care. 
Thus, HMOs can educate the physicians they employ about treatment
guidelines, review clinical records to ensure that patients are
taking appropriate medications, or take other organizational actions
to improve the quality of care provided by all types of physicians
that are not possible in fee-for-service settings. 


   AGENCY AND OTHER COMMENTS
------------------------------------------------------------ Letter :7

We provided a draft of this report to HCFA and a panel of experts for
their review.  Based on their comments, we expanded the number of
drugs we examined and explicitly addressed the possible confounding
effects of patient characteristics.  We also incorporated technical
changes where appropriate.  Several other issues that the reviewers
raised are addressed here. 

First, some reviewers were concerned that our survey sample had the
potential to introduce selection biases.  In general, enrollees in
Medicare HMOs who develop chronic conditions are more likely to
revert to standard fee-for-service Medicare.\35 Our sample, however,
was limited to heart attack survivors enrolled in Medicare HMOs who
remained enrolled for the roughly 2-year period from their heart
attack until we interviewed them.  If many patients were excluded
from our sample because they had left HMOs between their heart attack
and our survey, then our respondents could represent HMO enrollees
who were disproportionately healthy and satisfied with medical care
provided by HMOs.  However, we found that the potential effect of any
such selection bias was minimal because few patients in our initial
sample--less than 4 percent--were dropped from the study because they
had returned to fee-for-service Medicare between their heart attack
hospitalizations and the survey period.  Thus, because so few members
of our sample left HMOs, we believe that it accurately reflects the
population of Medicare patients who survived heart attacks that
occurred while they were enrolled in HMOs. 

Second, some reviewers pointed out that our finding that heart attack
survivors with regular cardiology appointments have more appropriate
drug treatment may be the result of having regular physician
appointments, not that the appointments are with a cardiologist. 
This explanation hypothesizes that the respondents in our comparison
group have fewer physician contacts overall.  Because we were
interested specifically in heart-related medical care, our survey
questions did not attempt to measure the overall level of physician
contacts.  Consequently, we are unable to rule out this explanation
with direct evidence.  However, two other aspects of our work--a
separate sensitivity analysis and the multivariate analyses--provide
indirect evidence that this alternative explanation is unlikely. 

We conducted a sensitivity analysis to judge the plausibility of this
alternative explanation.  For this analysis, we estimated how much
lower the rate of regular physician contacts would have to be among
those who did not see a cardiologist at all in order to explain their
lower use of cholesterol-lowering drugs and beta-blockers.  We found
that a lower rate of regular physician visits could explain the lower
use of cholesterol-lowering drugs among patients who had not seen a
cardiologist only if no more than 1 in 10 of them had regularly seen
their primary care doctor or another noncardiologist physician for
the treatment of any medical condition.  Similarly, to explain their
lower use of beta-blockers, the proportion seeing a noncardiologist
regularly would have to be no more than one-third.  By contrast,
among those who saw a cardiologist, two-thirds reported having
regular appointments.  Since these groups did not differ in
self-reported health status and incidence of major comorbidities, we
believe that it is implausible that such a high proportion of heart
attack survivors who did not see a cardiologist would also lack
regular contact with even their primary care provider. 

Our multivariate analyses included variables other than health that
are known to be associated with the use of physician services,
especially education, income, age, and gender.  If frequency of
physician contacts explained our findings, then including these
variables in the multivariate analyses should have greatly diminished
the statistical association between regular specialty care and drug
usage.  This did not occur.  (See app.  II for a description of our
sensitivity and multivariate analyses). 

In addition, some reviewers noted that more care is not always better
care.  That is, while our results are consistent with the finding
from the research literature that specialists provide more intensive
care than generalists, there is the possibility that specialists may
provide heart-related medications to patients whom the drug will not
help more often than generalists, which would account for at least
part of this difference.\36 We agree that it is likely that some
individual patients in our survey were not helped by these
medications; however, we do not believe that our results can be
attributed to a systematic tendency for patients with regular
cardiology care to take these drugs inappropriately.  The drugs we
selected as indicators of appropriate care have been demonstrated to
have great clinical benefits and few absolute contraindications. 
Moreover, for beta-blockers and aspirin, our statistical analyses
documenting the importance of regular cardiology care controlled for
the degree to which patients were ideal candidates for the therapy. 
Further, our results show that even patients under the regular care
of cardiologists took these drugs at rates below the recommended
guidelines--a finding that is more consistent with the position that
cardiologists provide too little appropriate care than it is with the
view that they provide too many inappropriate treatments. 

Finally, some reviewers also suggested that our results may be due to
differences in the out-of-pocket expenditures for these drugs between
respondents with regular cardiology care and those without regular
cardiology appointments.  If patients with regular cardiology care
systematically paid less for these drugs for any reason, their
increased usage rates may be due to lower costs instead of to the
care provided by cardiologists.  While we do not know how much these
drugs would have cost each respondent, we were able to identify heart
attack survivors who belonged to HMO plans with pharmacy benefits
and, thus, who presumably have lower drug costs.  We found that the
presence of a pharmacy benefit was not related to the self-reported
use of any of these three drugs or to having regular cardiology care. 
Moreover, in a comparable study of heart attack survivors treated by
the Department of Veterans Affairs--where none of the patients had to
pay more than minimal amounts for their drugs--patients under regular
cardiology care received cholesterol-lowering drugs much more often
than those cared for by primary care physicians.\37


--------------------
\35 See Medicare:  Fewer and Lower Cost Beneficiaries With Chronic
Conditions Enroll in HMOs (GAO/HEHS-97-160, Aug.  18, 1997). 

\36 See, for example, S.  Greenfield and others, "Outcomes of
Patients With Hypertension and Non-Insulin-Dependent Diabetes
Mellitus Treated by Different Systems and Specialties:  Results From
the Medical Outcomes Study," The Journal of the American Medical
Association, Vol.  274, No.  18 (1995), pp.  1436-44. 

\37 See J.  J.  Whyte, "Treatment of Hyperlipidemia by Specialists
Versus Generalists as Secondary Prevention of Coronary Artery
Disease."


---------------------------------------------------------- Letter :7.1

As we arranged with your staff, unless you publicly announce the
report's contents earlier, we plan no further distribution until 30
days after it is issued.  We will then send copies to the Secretary
of the Department of Health and Human Services and other interested
parties.  We will also make copies of this report available to others
upon request.  Please call me or Marsha Lillie-Blanton, Associate
Director, at (202) 512-7119 if you have any questions about this
report.  Martin T.  Gahart and Eric A.  Peterson are the major
contributors to this report. 

Bernice Steinhardt
Director, Health Services Quality
 and Public Health Issues


METHODOLOGY
=========================================================== Appendix I


   SURVEY PROCEDURES
--------------------------------------------------------- Appendix I:1


      THE SURVEY SAMPLE
------------------------------------------------------- Appendix I:1.1

The heart attack survivors sampled for this survey were all enrolled
in Medicare HMOs at the time they were hospitalized for an acute
myocardial infarction (between May and July 1995) and at the time the
survey was conducted (between April and July 1997).  They were
identified as part of a larger study, the Cooperative Cardiovascular
Project (CCP), conducted by HCFA.  For this study, HCFA abstracted
clinical data from hospital records for approximately 224,000
Medicare heart attack survivors.  CCP sampled acute myocardial
infarction admissions that occurred between February 1994 and July
1995.  Each hospital was sampled for only a subset of the months
during that period, and patients were included in the CCP data set
only if they were hospitalized during the time their hospital was
sampled. 

HMO patients are underrepresented in HCFA's claims data, from which
the CCP sampling frame was constructed.  In return for the fixed, per
month amount that HMOs receive for each Medicare enrollee, they
assume full responsibility for patient hospital bills.  Hospitals are
still supposed to submit "no pay" bills to HCFA for Medicare HMO
patients, but this requirement is frequently not followed.  As a
result, there is often no record in HCFA's claims files for
hospitalizations of HMO enrollees.  To compensate for this deficiency
in the original CCP sample, we contacted all Medicare HMOs with 1,000
or more enrollees as of August 1995 and asked the HMOs to send us
information on any enrollee who had been hospitalized with an acute
myocardial infarction during the CCP study period.  We passed this
information on to HCFA; HCFA then determined if the patients reported
by the HMOs belonged in CCP based on the sampling time frame for the
hospital where the patient was treated.  As a result, the CCP data
file now includes about 13,000 HMO patients. 

We then limited the sample to residents of seven states that together
totaled 72 percent of the Medicare HMO population in 1995: 
California, Florida, Massachusetts, New York, Ohio, Pennsylvania, and
Texas.  We limited our sample to these states to allow us to compare
our survey data to survey data that researchers at Harvard Medical
School collected on a subset of CCP patients treated under
fee-for-service Medicare in those states.  We also restricted our
sample to those aged 65 to 84 years at the time of their heart attack
to match the Harvard survey's selection criteria. 

We excluded Medicare beneficiaries known to have died by February
1997.  We also excluded individuals who were no longer in an HMO at
the time of the survey, even though they had been HMO members when
they suffered the heart attack.  Finally, we included in our sample
all the remaining patients who had been hospitalized at the end of
the CCP time period--May through July 1995--to make the interval
between heart attack and interview as close as possible to that of
the patients in the Harvard survey.  The final sample size was 578. 


      ADMINISTERING THE SURVEY
------------------------------------------------------- Appendix I:1.2

HCFA provided us with the mailing address of each member of our
sample; we then used publicly available directories to locate the
phone numbers of as many individuals as possible.  Next, we sent to
all selected beneficiaries letters that explained the study, asked
for their participation, and provided a list of heart-related drugs
for the interview.  The letters advised those for whom we found phone
numbers that an interviewer would be calling and asked those without
phone numbers to call a toll-free telephone number to participate in
the survey.  A second round of mailings was sent to nonrespondents
midway through the study period.  In the end, we were unable to
locate 112 individuals. 


      FINAL DISPOSITION
------------------------------------------------------- Appendix I:1.3

Of the 578 individuals in our sample, 19 died between February 1997
and the end of the survey period.  The survey was completed by 362
respondents--65 percent of the remaining 559.  We were unable to
contact, or could not make satisfactory arrangements to complete the
interview with, 118 individuals (21 percent).  Only 14 percent of the
sample (79 individuals) refused to participate.  Seventy-seven
percent of the completed interviews were with respondents reached
directly by phone by our interviewers, while 23 percent were with
respondents who contacted us through the toll-free telephone number. 
Eighty-eight percent of our respondents were interviewed within 2
years of their acute myocardial infarction hospitalization, and all
of the interviews were completed within 26 months of the
hospitalization. 


   ANALYSIS OF RESPONDENT
   CHARACTERISTICS
--------------------------------------------------------- Appendix I:2

To see how our respondents compared to the sample as a whole, we
analyzed demographic information from HCFA's administrative data
bases.  The two groups had similar distributions for gender, age, and
state of residence.  However, relative to their proportions in the
sample, whites completed the interview slightly more often
(accounting for 76 percent of the sample but 79 percent of completed
interviews) and Hispanics somewhat less often (12 percent of the
sample but only 9 percent of the completed interviews).  We do not
believe that these small differences affect the validity of our
findings, although they mean that we cannot generalize our findings
to Hispanic Medicare beneficiaries.\38


--------------------
\38 The race distributions here are different from those we report in
table I.1.  This is because the percentage of respondents coded as
Hispanics in HCFA's administrative data base (9 percent) is lower
than the percentage of our respondents who identified themselves as
Hispanic during the interview (15 percent).  One-third of our
Hispanic respondents were categorized as white by HCFA. 


   VARIABLE DESCRIPTIONS
--------------------------------------------------------- Appendix I:3


      CHOLESTEROL-LOWERING DRUGS
------------------------------------------------------- Appendix I:3.1

Several different categories of drugs can be used to lower
cholesterol levels.  The statins (HMG CoA reductase inhibitors) are
effective and have few short-term side effects, but they are
relatively expensive and lack a long-term track record.  Bile acid
resins are inexpensive and have a long safety record but are more
complicated to take and can produce unpleasant gastrointestinal
symptoms.  Nicotinic acid is also inexpensive.  However, it can be
fairly toxic when taken in higher doses.  Fibric acids are especially
potent in lowering triglycerides but have more limited effect on both
low- and high-density lipoprotein (LDL and HDL) cholesterol levels. 
To boost the cholesterol-lowering effect, drugs from several of these
categories can be combined. 

Prior to contacting respondents by telephone, we mailed each a
comprehensive list of drugs prescribed to heart attack survivors. 
During the interview, respondents were asked to tell the interviewer
the code number next to each drug that they were currently taking. 
For respondents who did not have the coded list--because they had not
received it or had misplaced, lost, or otherwise did not have the
list--were asked to tell the interviewers the names of the heart
drugs they took.  In addition, all respondents were asked if they
were taking any heart drugs not on the list. 

Respondents were coded as taking a cholesterol-lowering drug if they
said that they took any one of the 24 drugs on the list.  (The list
of 24 drug names actually measured only 11 distinct pharmaceuticals,
as each of 11 drugs was listed with both a generic name and at least
one trade name.) The list included 5 statins with both generic and
trade names (totaling 10 drugs):  atorvastatin (Lipitor), fluvastatin
(Lescol), lovastatin (Mevacor), pravastatin (Pravachol), and
simvastatin (Zocor).  The list also included 14 other
cholesterol-lowering drugs (6 distinct drugs with both generic and
trade names):  cholestyramine (Questran); clofibrate (Atromid-S);
colestipol (Colestid); gemfibrozil (Lopid); niacin (Niacor, Nicobid,
and Nicolor); and probucol (Lorelco). 

For respondents reporting that they took an anticholesterol drug, 82
percent reported taking only statin drugs, 13 percent only nonstatin
drugs, and 5 percent both statin and nonstatin drugs. 


      BETA-BLOCKERS
------------------------------------------------------- Appendix I:3.2

Beta-adrenergic blocking agents, or beta-blockers, inhibit
stimulation of the heart and reduce the force of heart muscle
contractions.  As a result, they reduce the patient's heart rate and
blood pressure, which in turn lowers the heart's workload and
consequent need for blood and oxygen.  These conditions increase the
likelihood that sufficient blood will flow through the coronary
arteries to prevent a new heart attack.  In addition, beta-blockers
reduce the incidence of arrhythmia, which can lead to sudden cardiac
death. 

Respondents were coded as taking a beta-blocker if they said that
they took any one of the 38 such drugs listed or if they volunteered
the name of a beta-blocker when asked about their heart drugs.  The
38 drug names referred to 13 distinct pharmaceuticals, with both
generic and one or more trade names listed.  We also included
formulations that combined several of these beta-blockers with
diuretics.  The list included acebutolol (Sectral), atenolol
(Tenormin), betaxolol (Kerlone), bisoprolol (Zebeta), carteolol
(Cartrol), labetalol (Normodyne and Trandate), metoprolol (Lopressor
and Toprol XL), nadolol (Corgard), penbutolol (Levatol), pindolol
(Visken), propranolol (Inderal), sotalol (Betapace), timolol
(Blocadren). 


      ASPIRIN
------------------------------------------------------- Appendix I:3.3

A separate survey question asked respondents if they took aspirin
every day or every other day.  We coded respondents as taking aspirin
if they answered "yes" to this question. 


      REGULAR APPOINTMENTS WITH A
      CARDIOLOGIST
------------------------------------------------------- Appendix I:3.4

We asked respondents the name and office location (city or town) both
of the physician they saw for general health care and of the doctor
mainly responsible for treating their heart condition.  For the
physician mentioned as primarily responsible for heart treatment, we
asked if they had regular appointments or only saw the doctor when
they were ill or when referred by a primary care physician.\39 For
these questions, 59 percent of the respondents provided the names of
two physicians, and 41 percent the name of one doctor. 

We then used physician directories from the American Medical
Association to identify the practice specialty of the physician named
as treating the respondent's heart condition.  We coded as
cardiologists any physician who listed cardiology as his or her
primary practice specialty, who listed cardiology as a secondary
practice specialty, or who had completed a residency in cardiology. 
Nearly 90 percent of the physicians we coded as cardiologists listed
cardiology as their primary practice specialty.  Some respondents
identified a cardiologist by name and office location but then
volunteered that they had not seen that physician for some time. 
Those respondents were coded as not having a cardiologist. 

Our criteria for identifying cardiologists were permissive.  That is,
if the physician and office location noted by the respondent could
plausibly identify a cardiologist, we coded that physician as a
cardiologist.  In practice, this meant that (1) physicians with
common names were counted as cardiologists if any one doctor with
that name was a cardiologist (for example, if 1 of the 10 Dr.  Smiths
in a city was a cardiologist, any Dr.  Smith there was coded as a
cardiologist) and (2) physicians in nearby towns were included (for
example, if Dr.  Jones the cardiologist was not found in the city
given by the respondent but practiced in an adjacent suburb, Dr. 
Jones was coded as a cardiologist).  Any bias that may have been
introduced by this practice worked against our major finding; the
most likely error in this method involves coding a noncardiologist as
a cardiologist, and to the extent that cardiologists prescribe
cholesterol-lowering drugs more often than noncardiologists, this
error would reduce the difference between the specialties that we
have reported. 


--------------------
\39 More specifically, we asked, "Do you have regularly scheduled
visits with Dr.  (name of heart doctor) (for example, every 3, 6, or
12 months), or do you only see this doctor when you are not feeling
well?" (Only when referred by a primary care doctor should be coded
as not feeling well.)


      BACKGROUND VARIABLES
------------------------------------------------------- Appendix I:3.5

Our analysis included a number of other variables, including the
following demographic and health-related variables. 

  -- Gender:  Gender was coded from a question on the survey. 

  -- Race:  Based on responses to the survey, we categorized each
     respondent as Hispanic, non-Hispanic white, or other. 

  -- Age:  Age in years at the time of the acute myocardial
     infarction was obtained from HCFA's administrative records.  We
     grouped the respondents into two age categories, each with about
     one-half of the total:  67 to 73 years at the interview date (65
     to 71 at the time of the heart attack) and 74 to 86 years (72 to
     84 at the time of the heart attack).  Individuals aged 85 and
     older at the time of the heart attack were excluded from the
     sampling frame. 

  -- Some College Education:  From a survey question, we measured
     education attainment by assigning a positive value to this
     variable for all respondents who said that they had completed at
     least 1 year of college, were college graduates, or who had some
     post-graduate education. 

  -- High Current Income:  Based on responses to a question on the
     survey, we coded individuals reporting a total yearly family
     income of $20,000 or more (not quite one-half of the
     respondents) as having a high current income.  The comparison
     group includes individuals with less income and those with
     missing values on this question. 

  -- Residency:  State of residence at the time of the interview was
     ascertained from a survey question.  We divided this group into
     three categories:  California residents (44 percent of
     respondents), Florida residents (32 percent), and residents of
     the five other states eligible for our sample (Massachusetts,
     New York, Ohio, Pennsylvania, and Texas). 

  -- Spanish-Language Interview:  Interview language was coded by the
     interviewers at the completion of the interview.  Thirty-three,
     or 9 percent, of the respondents completed the interview in
     Spanish. 

  -- Called in for the Interview:  In our contact letters, we asked
     beneficiaries for whom we could not find telephone numbers to
     call our interviewers on a toll-free telephone number.  About
     one-quarter of the completed interviews came from individuals
     who called in.  Compared to the sample as a whole, those who
     called in were disproportionately female and California
     residents.  We included this variable in our multivariate
     analysis to take account of these differences between those who
     were called and those who called in. 

  -- Very Good Current Health:  The survey included a self-reported
     health status measure.  Individuals reporting that their health
     was very good or excellent received a "1" on this variable;
     respondents reporting good, fair, or poor health were coded "0."

  -- Confirmed Acute Myocardial Infarction:  This variable was
     obtained from HCFA.  Based on information abstracted from each
     patient's clinical records as part of the CCP, HCFA determined
     if a heart attack could be confirmed.  Lack of confirmation may
     mean either that a heart attack did not occur or that
     information about relevant clinical measurements was missing
     from a patient's file. 

  -- Any Major Comorbidities:  From the abstracted clinical records
     provided by HCFA, we coded individuals as having a major
     comorbidity if they had any one of these conditions at the time
     of their heart attack hospitalization:  congestive heart
     failure, chronic obstructive pulmonary disease, dementia, any
     form of diabetes, or a previous stroke. 

  -- Heart Function:  The abstracted clinical records included
     measures of the left ventricular ejection fraction taken during
     the heart attack hospitalization for two-thirds of our
     respondents.  For our multivariate statistical analyses, we
     grouped this interval variable into three categories:  below 35,
     35 to 50, and above 50.  In the text and in some appendix
     tables, we categorized respondents with ejection fractions of
     less than 35 as having poor heart function, with the comparison
     group comprised of individuals with a fraction of 35 or greater. 

  -- Ideal Candidate for Beta-Blockers and Aspirin:  CCP data on our
     survey respondents allowed us to identify whether or not
     respondents were likely candidates for beta-blocker or aspirin
     therapy.  As part of CCP, HCFA determined which patients would
     be eligible for these therapies when they were discharged from
     the hospital and which among those were "ideal" candidates. 
     Since this status depended in large part on the presence or
     absence of chronic diseases--such as heart failure, diabetes,
     and chronic obstructive pulmonary disease--it would likely
     remain unchanged 2 years later for most (though probably not
     all) of our respondents.  Patients who are not ideal candidates
     may have evidence of one of the potential contraindications or
     have missing data for one of the contraindications.\40

  -- Heart Attack Severity:  We measured heart attack severity with
     an interval variable derived from the abstracted clinical
     records that counted the presence of three indicators:  a
     previous myocardial infarction, a transmural myocardial
     infarction, and angina more than 24 hours after arrival at the
     hospital.  Four percent of our sample had all three of these
     indicators, 25 percent had two indicators, 45 percent had one,
     and 26 percent had none. 


--------------------
\40 See T.  A.  Marciniak, "Improving the Quality of Care for
Medicare Patients With Acute Myocardial Infarction," p.  1353. 


   ANALYSIS OF EXCLUDED CASES
--------------------------------------------------------- Appendix I:4

The findings described in this report are based on our analysis of
data from a subset of the completed interviews.  We excluded cases
with missing data on the main explanatory variable (whether or not
the patient had regular appointments with a cardiologist) and
respondents who completed the interview in Spanish.  Twenty-two
percent of the respondents (or 78 individuals) were dropped for these
reasons.  The purpose of this section is to describe why and how we
made these exclusions, describe the differences between those kept in
the analysis and the excluded cases, and discuss the implications for
our conclusions. 

Fifty-one cases (14 percent of the entire sample) were excluded
because we could not determine if they had regular visits with a
cardiologist or not.  These individuals either did not answer the
physician contact questions on the survey or listed doctors we could
not find in the physician directories.  We excluded these cases
because they did not provide information that would help us answer
our research questions. 

Of those with complete physician data, an additional 27 cases (or 8
percent of the entire sample) with Spanish-language interviews were
excluded because their results were implausibly different from those
of the rest of the sample; we believe that these differences, at
least in part, may have been caused by our survey procedures.\41 For
example, only 6 percent of the Spanish-language interviews reported
taking cholesterol-lowering drugs, compared to 33 percent for the
sample as a whole and to 32 percent for the 22 Hispanic respondents
who completed the interview in English.  The Spanish-language
interviews also reported lower usage rates for beta-blockers and
aspirin than the other Hispanic respondents.  We believe that our
failure to provide a drug list in Spanish may have contributed to
this low level of self-reported drug use.  We also found that while
70 percent of those with Spanish-language interviews reported having
regular cardiology appointments, only 43 percent of the sample as a
whole and 19 percent of Hispanics who completed the interview in
English reported having such appointments.  We suspect that our
physician coding scheme led us to substantially overestimate the
proportion of these respondents with regular cardiology care.  Almost
all of the Spanish-language cases reside in southern Florida, an area
with many physicians with similar last names practicing in close
proximity.  In such circumstances, our physician specialty coding
rules were likely to have coded many generalist physicians as
cardiologists. 

As table I.1 shows, our decision to exclude some cases from the
analysis slightly increased our estimates of the proportion of
respondents taking cholesterol-lowering drugs and beta-blockers and
slightly decreased the percentage of respondents with regular
appointments with a cardiologist (from 43 percent for all respondents
to 40 percent for the analysis subset).  Both of these differences
result from excluding the low drug use but high cardiology
appointment set of respondents who completed the interview in
Spanish.  These decisions somewhat limit the generalizability of our
results.  In particular, we are unable to reach any conclusions about
Spanish-speaking Medicare HMO enrollees. 



                         Table I.1
          
          Percent of All Respondents, Respondents
             Included in the Analysis File, and
           Respondents Excluded from the Analysis
             File, by Variable Characteristics

                                                  Excluded
                           All   Included in          from
                   respondents      analysis      analysis
Characteristic         (N=362)       (N=284)        (N=78)
----------------  ------------  ------------  ------------
Cholesterol-                33            36            19
 lowering drugs
Beta-blockers               39            40            32
Aspirin                     71            71            72
Regular                     43            40            70
 cardiology
 appointments
Male                        64            62            71
White                       77            87            38
Hispanic                    15             6            49
Other race                   8             7            13
Aged 67 to 73               48            49            45
California                  41            45            24
 resident
Florida resident            37            32            58
Other state                 22            23            18
 resident
Confirmed acute             73            76            62
 myocardial
 infarction
Called in for               23            26            14
 interview
Some college                35            40            17
High current                44            49            23
 income
Very good                   24            24            26
 current health
Any major                   46            48            38
 comorbidities
Poor heart                  22            21            26
 function
Ideal candidate             12            12            13
 for beta-
 blockers
Ideal candidate             48            46            51
 for aspirin
Heart attack              1.06          1.10           .92
 severity
----------------------------------------------------------
Appendix II

--------------------
\41 An additional six respondents who completed the interview in
Spanish did not provide usable physician information and were
excluded by that criterion. 


STATISTICAL ANALYSES
=========================================================== Appendix I


   ESTIMATE OF SAMPLE CHOLESTEROL
   LEVELS
--------------------------------------------------------- Appendix I:5

Ideally, we would have taken into account each patient's baseline LDL
cholesterol level in determining the clinical appropriateness of
cholesterol-lowering medications for that patient.  Unfortunately,
these data were not part of the CCP data set.  However, recent data
on the distribution of LDL levels in the national population are
available from the Third National Health and Nutrition Examination
Survey (NHANES III).\42 Our analysis of data from this survey
indicates that 53 percent of men and 64 percent of women over age 65
have baseline LDL levels of 130 mg/dL or above.  These figures are
comparable for those that either have or have not had a heart attack. 

We used figures from NHANES III to estimate the proportion of our
survey respondents who were likely to benefit from
cholesterol-lowering drugs, based on the estimated incidence of
threshold levels of LDL cholesterol specified in NHLBI guidelines and
the proportion of men and women in our sample.  We estimate that
approximately 57 percent of our sample had LDL levels of 130 mg/dL or
greater.  This figure provides an estimate of the proportion of heart
attack survivors who should receive cholesterol-lowering drugs,
assuming that some patients with somewhat lower baseline LDL levels
would benefit from this therapy, while others with high LDL levels
would not, due to extreme frailty or terminal illness, for example. 


--------------------
\42 Department of Health and Human Services, National Center for
Health Statistics, "Third National Health and Nutrition Examination
Survey, 1988-1994 (NHANES III)," Laboratory Data File and Household
Adult Data File (CD-ROM), Public Use Data File Doc.  No.  76200
(Hyattsville, MD:  Centers for Disease Control and Prevention, 1996). 


   SENSITIVITY ANALYSIS FOR
   REGULAR PHYSICIAN APPOINTMENTS
--------------------------------------------------------- Appendix I:6

Some reviewers of a draft of this report explained the greater drug
usage rates among respondents with regular cardiology appointments as
possibly the result of those patients having regular appointments
with any physician, not necessarily to any aspect of care provided
specifically by cardiologists.  Although we are unable to directly
test this alternative explanation because we did not ask our
respondents about the regularity of their contacts with physicians
other than cardiologists, we addressed this concern by conducting a
rough sensitivity analysis of the effects of having regular physician
appointments on the use of cholesterol-lowering drugs and
beta-blockers. 

The sensitivity analysis starts with the assumption that the use of
cholesterol-lowering drugs and beta-blockers is equally appropriate
for each of our three patient groups:  those who saw cardiologists
regularly, those who saw cardiologists occasionally, and those who
saw only noncardiologist physicians.  While there are specific
reasons why a relatively small proportion of our respondents might
not benefit from one or the other therapy (for example, an unusually
low baseline LDL cholesterol level without drugs, or a specific
clinical contraindication for beta-blockers, such as asthma), we do
not expect these characteristics would affect the regularity of
physician contacts for these patients.  For instance, we know that
the respondents seeing cardiologists regularly did not differ from
other respondents in self-reported health status or incidence of
comorbidities.  Further, while those seeing cardiologists regularly
did tend to have more severe heart attacks, lower heart attack
severity does not make beta-blockers and cholesterol-lowering drugs
any less beneficial for heart attack survivors.  A heart attack of
any severity puts a patient in the high-risk group for future heart
attacks, according to NHLBI guidelines. 

Because of the structure of our survey, we know whether respondents
who saw a cardiologist had regular or occasional appointments, but we
do not have this information for respondents who saw only
noncardiologists.  That is why we cannot directly assess the effect
of regular visits compared to that of physician specialty with
respect to taking cholesterol-lowering medications and beta-blockers. 
However, by regrouping data from the main analysis to consider just
those patients who saw a cardiologist at least occasionally
(two-thirds regularly and one-third only occasionally), we can derive
an estimate of the magnitude of the effect of having regular
physician appointments for that subset of our respondents.  Thus, we
observed that 45 percent of those with regular appointments with
cardiologists used cholesterol-lowering drugs, compared to 29 percent
of those who saw cardiologists only occasionally.  For beta-blockers,
the comparable usage figures are 50 percent and 29 percent.  (See
table II.1.)



                         Table II.1
          
                Proportion of Patients Using
            Cholesterol-Lowering Medications and
                       Beta-Blockers

                                  Proportion
                                       using
                                cholesterol-    Proportion
Grou                                lowering   using beta-
p                            N         drugs      blockers
----  --------------  --------  ------------  ------------
A     Patients with        115           45%           50%
       regular
       cardiology
       appointments
B     Patients with         55            29            29
       occasional
       cardiology
       appointments
C     Patients with        114            31            36
       no cardiology
       appointments
D     Patients who         169            30            34
       did not see a
       cardiologist
       or saw one
       only
       occasionally
       (B+C)
----------------------------------------------------------
Our main analysis compared group A with group D (see table 1); this
analysis compares group A with group B to make inferences about group
C.  If, as suggested by the alternative explanation, the principal
determinant of drug use is the regularity of physician appointments
regardless of the physician's specialization, then one would expect
the same proportion of patients who did not see a cardiologist to
receive these drugs depending on whether they saw any other physician
regularly or not.  Thus, hypothetically, 45 percent of those
respondents who saw their primary care doctor or other physician
regularly should be taking cholesterol-lowering drugs and 50 percent
of them should be taking beta-blockers.  Similarly, among those with
only occasional appointments with any physician, 29 percent should be
taking cholesterol-lowering medications and (coincidentally) 29
percent of them should be taking beta-blockers. 

At the same time, we know from the survey responses what proportion
of the group not seeing cardiologists actually used these drugs
overall:  31 percent for cholesterol-lowering medications and 36
percent for beta-blockers.  Working from these figures, we can derive
what proportion of the group would have had to have seen any
noncardiologist physician on a regular basis in order for these two
assumptions to hold.  If that estimated proportion is implausibly
low, it would make it unlikely that the observed differences in drug
use we found reflect simply the effect of regular visits and not
physician specialty. 

Thus, for cholesterol-lowering drugs, respondents who did not see a
cardiologist had a usage rate of 31 percent.  Given the presumed
usage rates--29 percent for respondents with occasional visits and 45
percent for those with regular physician appointments--one can reach
the observed aggregate level for respondents not seeing cardiologists
only if the large majority--90 percent--of this group saw physicians
only occasionally:  (29 percent x .90) + (45 percent x .10) = 31
percent overall.  To the extent that more than 10 percent of this
group saw their primary care physician regularly (and therefore had a
45-percent usage rate for these drugs), the overall rate of use would
have to rise above the 31-percent level that we observed. 

The result of this calculation for beta-blockers is similar, though
less dramatic.  Thus, if respondents with regular noncardiology
appointments are presumed to use beta-blockers at a rate of 50
percent, and those with occasional physician visits at a rate of 29
percent, then to reach the observed overall rate of 36 percent, 32
percent of this group would have to have regular physician visits and
68 percent occasional appointments:  (29 percent x .68) + (50 percent
x .32) = 36 percent overall.  This would mean that two out of three
of these respondents--none of whom were seeing a cardiologist even
occasionally and all of whom had been hospitalized for a heart attack
within the last 2 years--were not seeing even a primary care
physician on a regular basis. 

For both types of drugs, the estimated rates of regular physician
appointments from our sensitivity analysis (one-tenth and one-third,
respectively) are considerably below the actual regular visit rate
for patients who saw a cardiologist (two-thirds).  Since the overall
health of our respondents with regular cardiology care does not
differ from that of the other members of our sample, we do not
believe that differences of this magnitude are plausible.  For that
reason, it seems quite unlikely that our findings about the influence
of regular cardiology care on the use of cholesterol-lowering drugs
and beta-blockers can be explained by differences in regular
physician contacts among the heart attack survivors in our sample. 

As a further check on the robustness of these conclusions, we tested
the potential impact of sampling error in our relatively small
sample.  All of the figures we used in the above calculations reflect
the responses provided by the particular sample HCFA drew from the
population of Medicare heart attack survivors in HMOs.  The extent to
which any other comparable sample might provide different results is
captured by the standard error for the rates of drug use for each of
the three respondent subgroups (those with regular cardiologist
visits, occasional cardiologist visits, and no cardiologist visits). 
Testing for the effect of changes in each of these parameters, we
found that variation in the rate of drug use by the group that had no
contact with cardiologists had the largest impact on the derived
estimate of regular physician visits for that group.  If the use of
cholesterol-lowering drugs was actually one standard error higher for
the group that had not seen a cardiologist (that is, 35 percent
instead of 31 percent), then this would imply that 37 percent--not 10
percent--of these patients had regular contact with a physician of
some sort.  Similarly, the estimated rate of regular visits increased
from 32 percent to 54 percent if overall use of beta-blockers by this
group was raised by one standard error.  There is one chance in six
that the "true" mean is greater than the sum of the observed sample
mean and the standard error.  In other words, even with sampling
error, there is a five in six chance that the estimated rate of
regular physician visits for patients who did not see a cardiologist
would be, at most, 37 percent in the analysis of cholesterol-lowering
drugs and 54 percent for beta-blockers.  Thus, the rate of inferred
regular visits for patients who did not see a cardiologist is still
clearly lower than that observed in our sample for patients who did
see one at least occasionally (67 percent). 


   MULTIVARIATE ANALYSES FOR
   CHOLESTEROL-
   LOWERING DRUGS, BETA-BLOCKERS,
   ASPIRIN, AND REGULAR CARDIOLOGY
   APPOINTMENTS
--------------------------------------------------------- Appendix I:7

For our major analyses, we compared the usage rates of
cholesterol-lowering drugs, beta-blockers, and aspirin for
respondents who had regularly scheduled cardiology visits with the
rates for those who do not see a cardiologist regularly.  As a
necessary step in this analysis, we also examined the overall rates
of taking these heart drugs and of receiving regular care from a
cardiologist.  In addition, we conducted multivariate statistical
analyses to ensure that any differences we found did not change when
we took into account the effects of other background and
health-related factors influencing the use of cholesterol-lowering
drugs, beta-blockers, and aspirin.  Finally, we conducted a
multivariate statistical analysis to identify variables associated
with having regular cardiology appointments.  All of our analyses
excluded respondents with missing physician information or who
completed the interview in Spanish. 

Table II.2 presents the results of a logistic regression analysis
predicting use of cholesterol-lowering drugs.  The outcome variable
is dichotomous:  "1" indicates that the respondent takes
cholesterol-lowering drugs; "0" indicates that he or she does not. 
The regression uncovered four statistically significant
factors--cholesterol-lowering drugs were taken more often by
respondents with regular cardiology appointments, by respondents aged
67 to 73 (or 65 to 71 at the time of the heart attack), by
respondents claiming that their health was very good or excellent,
and by respondents without major comorbidities at the time of the
heart attack. 



                                        Table II.2
                         
                             Logistic Regression Analysis for
                                Cholesterol-Lowering Drugs

                                                 Odds ratio\b
                                                         (95%
                                   Coefficient\    confidence          Chi-   Probability
Variable                                      a     interval)      square\c       level\d
---------------------------------  ------------  ------------  ------------  ------------
Regular cardiologist (versus not)           .68          1.97          6.00           .01
                                                  (1.15-3.40)
Confirmed heart attack (versus              .26          1.30           .66           .42
 not)                                              (.69-2.43)
Called in for interview (versus             .35          1.42          1.27           .26
 reached by phone)                                 (.77-2.59)
Male (versus female)                       -.02           .98           .01           .94
                                                   (.56-1.71)
White (versus nonwhite)                     .23          1.26           .28           .60
                                                   (.53-3.00)
Aged 67 to 73 years (versus aged           1.08          2.95         15.10          <.01
 74 to 86)                                        (1.71-5.09)
California resident (versus other           .30          1.35          1.19           .28
 six states)                                       (.79-2.31)
Very good current health (versus            .80          2.22          6.32           .01
 good, fair, or poor)                             (1.19-4.13)
Major comorbidity (versus none)            -.74           .48          6.81          <.01
                                                    (.28-.83)
Constant\e                                -1.91
-----------------------------------------------------------------------------------------
Note:  N=284. 

\a Coefficients are from a logistic regression analysis with the
SAS-PC software package. 

\b The odds ratio is the exponentiated coefficient (e\coefficient ). 
The odds ratio indicates the change in the odds of taking
cholesterol-lowering drugs relative to that of the group left out. 

\c Chi-square values test the statistical significance of the
coefficients. 

\d Probability level refers to the chances that the coefficient
equals zero in the population.  By convention, coefficients with a
probability level less than or equal to 5 percent (.05) are regarded
as statistically significant. 

\e To control for background factors, the first seven variables were
kept in the equation regardless of their statistical significance. 
The original regression equation also included other variables that
were dropped from this final analysis because none were statistically
significant.  The variables that were dropped, along with their
coefficients and probability levels in the original equation, are as
follows:  high current income (.20, p=.52), some college education
(.56, p=.16), heart attack severity (.03, p=.89), and heart function
(-.13, p=.61). 

Table II.3 shows the results of a logistic regression analysis
predicting use of beta-blockers.  The outcome variable is
dichotomous:  "1" indicates that the respondent takes beta-blockers;
"0" indicates that he or she does not.  The regression uncovered four
statistically significant factors--beta-blockers were taken more
often by respondents with regular cardiology appointments, by
respondents with current income above the median for our sample, by
respondents who had attended college, and by respondents with
relatively good heart function measurements.  In addition, the
control variable indicating a valid heart function measurement was
also statistically significant.  The variable identifying ideal
candidates for beta-blockers did not influence the actual use of
beta-blockers. 



                                        Table II.3
                         
                          Logistic Regression Analysis for Beta-
                                         Blockers

                                                 Odds ratio\b
                                                         (95%
                                   Coefficient\    confidence          Chi-   Probability
Variable                                      a     interval)      square\c       level\d
---------------------------------  ------------  ------------  ------------  ------------
Regular cardiologist (versus not)           .86          2.37         10.31          <.01
                                                  (1.40-4.02)
Confirmed heart attack (versus             -.30           .74           .96           .33
 not)                                              (.41-1.35)
Called in for interview (versus             .34          1.40          1.23           .27
 reached by phone)                                 (.77-2.54)
Male (versus female)                       -.19           .83           .43           .51
                                                   (.48-1.45)
White (versus nonwhite)                    -.31           .74           .57           .45
                                                   (.33-1.62)
Aged 67 to 73 years (versus aged            .03          1.03           .01           .91
 74 to 86)                                         (.61-1.73)
California resident (versus other          -.46           .63          2.56           .11
 six states)                                       (.36-1.11)
High current income (versus not)            .67          1.96          5.66           .02
                                                  (1.13-3.42)
Some college education (versus              .76          2.14          7.27          <.01
 not)                                             (1.23-3.71)
Heart function\e                            .53          1.69          5.05           .02
                                                  (1.07-2.68)
Ideal candidate for beta-                   .36          1.44           .75           .39
 blockers                                          (.63-3.25)
Valid heart function measure              -1.46           .23          5.98           .01
 (versus missing data)\f                            (.07-.75)
Constant\g                                 -.55
-----------------------------------------------------------------------------------------
Note:  N=284. 

\a Coefficients are from a logistic regression analysis with the
SAS-PC software package. 

\b The odds ratio is the exponentiated coefficient (e\coefficient ). 
The odds ratio indicates the change in the odds of taking
beta-blockers relative to that of the group left out. 

\c Chi-square values test the statistical significance of the
coefficients. 

\d Probability level refers to the chances that the coefficient
equals zero in the population.  By convention, coefficients with a
probability level less than or equal to 5 percent (.05) are regarded
as statistically significant. 

\e Heart function has three values, with the levels indicating left
ventricular ejection fractions below 35, 35 to 50, and above 50. 

\f Sixty-five percent of the cases have valid measures of the left
ventricular ejection fraction, the measure of heart function used
here.  Individuals without a valid ejection fraction were coded "0"
on the heart function variable. 

\g To control for background factors, the first seven variables were
kept in the equation regardless of their statistical significance. 
The original regression equation also included other variables that
were dropped from this final analysis because none were statistically
significant.  The variables that were dropped, along with their
coefficients and probability levels in the original equation, are as
follows:  very good current health (-.19, p=.81), heart attack
severity (.21, p=.22), and major comorbidity (.07, p=.80). 

Table II.4 presents our logistic regression analysis for aspirin. 
The outcome variable is dichotomous:  "1" indicates that the
respondent took aspirin; "0" indicates that he or she does not.  The
regression uncovered four statistically significant factors--aspirin
was taken more often by respondents who had attended college, by
respondents with relatively good heart function measurements, and by
respondents identified as ideal candidates for aspirin therapy.  The
control variable indicating a valid heart function measurement was
also statistically significant.  The variable for regular cardiology
appointments approached statistical significance (probability level =
.10) but did not reach the required threshold. 



                                        Table II.4
                         
                         Logistic Regression Analysis for Aspirin

                                                 Odds ratio\b
                                                         (95%
                                   Coefficient\    confidence          Chi-   Probability
Variable                                      a     interval)      square\c       level\d
---------------------------------  ------------  ------------  ------------  ------------
Regular cardiologist (versus not)           .49          1.63          2.78           .10
                                                   (.92-2.91)
Confirmed heart attack (versus              .16          1.17           .22           .64
 not)                                              (.61-2.23)
Called in for interview (versus            -.02           .98           .00           .96
 reached by phone)                                 (.52-1.86)
Male (versus female)                       -.39           .68          1.60           .21
                                                   (.38-1.24)
White (versus nonwhite)                     .46          1.59          1.39           .24
                                                   (.74-3.42)
Aged 67 to 73 years (versus aged           -.06           .95           .04           .85
 74 to 86)                                         (.54-1.67)
California resident (versus other          -.21           .81           .50           .48
 six states)                                       (.45-1.45)
Some college education (versus              .63          1.89          4.35           .04
 not)                                             (1.04-3.42)
Heart function\e                            .72          2.05          9.36          <.01
                                                  (1.29-3.25)
Ideal candidate for aspirin                 .80          2.22          7.29           .01
                                                  (1.24-3.97)
Valid heart function measure              -1.30           .27          5.30           .02
 (versus missing data)\f                            (.09-.82)
Constant\g                                 -.16
-----------------------------------------------------------------------------------------
Note:  N=284. 

\a Coefficients are from a logistic regression analysis with the
SAS-PC software package. 

\b The odds ratio is the exponentiated coefficient (e\coefficient ). 
The odds ratio indicates the change in the odds of taking aspirin
relative to that of the group left out. 

\c Chi-square values test the statistical significance of the
coefficients. 

\d Probability level refers to the chances that the coefficient
equals zero in the population.  By convention, coefficients with a
probability level less than or equal to 5 percent (.05) are regarded
as statistically significant. 

\e Heart function has three values, with the levels indicating left
ventricular ejection fractions below 35, 35 to 50, and above 50. 

\f Sixty-five percent of the cases have valid measures of the left
ventricular ejection fraction, the measure of heart function used
here.  Individuals without a valid ejection fraction were coded "0"
on the heart function variable. 

\g To control for background factors, the first seven variables were
kept in the equation regardless of their statistical significance. 
The original regression equation also included other variables that
were dropped from this final analysis because none were statistically
significant.  The variables that were dropped, along with their
coefficients and probability levels in the original equation, are as
follows:  high current income (.16, p=.61), very good current health
(.60, p=.15), heart attack severity (.03, p=.86), and major
comorbidity (-.50, p=.09). 

Table II.5 presents our logistic regression analysis for regular
cardiology appointments.  The outcome variable is dichotomous:  "1"
indicates that the respondent had regular appointments with a
cardiologist; "0" indicates that he or she did not.  The regression
uncovered three statistically significant factors--respondents who
were white, younger, or who had suffered relatively severe heart
attacks had regular appointments with a cardiologist more often than
other respondents. 



                                        Table II.5
                         
                         Logistic Regression Analysis for Regular
                                 Cardiology Appointments

                                                 Odds ratio\b
                                                         (95%
                                   Coefficient\    confidence          Chi-   Probability
Variable                                      a     interval)      square\c       level\d
---------------------------------  ------------  ------------  ------------  ------------
Confirmed heart attack (versus             -.30           .74          1.05           .31
 not)                                              (.41-1.31)
Called in for interview (versus            -.03           .97           .01           .91
 reached by phone)                                 (.55-1.71)
Male (versus female)                       -.13           .88           .23           .63
                                                   (.53-1.47)
White (versus nonwhite)                     .90          2.45          4.37           .04
                                                  (1.06-5.69)
Aged 67 to 73 years (versus aged            .56          1.74          4.84           .03
 74 to 86)                                        (1.06-2.86)
California resident (versus other          -.02           .98           .01           .94
 six states)                                       (.59-1.63)
Heart attack severity\                      .33          1.39          4.49           .03
                                                  (1.03-1.89)
Constant\e                                -1.51
-----------------------------------------------------------------------------------------
Note:  N=284. 

\a Coefficients are from a logistic regression analysis with the
SAS-PC software package. 

\b The odds ratio is the exponentiated coefficient (e\coefficient ). 
The odds ratio indicates the change in the odds of having regular
cardiology appointments relative to that of the group left out. 

\c Chi-square values test the statistical significance of the
coefficients. 

\d Probability level refers to the chances that the coefficient
equals zero in the population.  By convention, coefficients with a
probability level less than or equal to 5 percent (.05) are regarded
as statistically significant. 

\e To control for background factors, the first seven variables were
kept in the equation regardless of their statistical significance. 
The original regression equation also included other variables that
were dropped from this final analysis because none were statistically
significant.  The variables that were dropped, along with their
coefficients and probability levels in the original equation, are as
follows:  high current income (-.08, p=.78), very good current health
(-.29, p=.42), some college education (-.01, p=.98), major
comorbidity (.13, p=.61), and heart function (-.22, p=.31). 


EXTERNAL REVIEWERS
========================================================= Appendix III

In addition to obtaining official agency comments from HCFA, we asked
the following individuals to review an early draft of this report. 
Their comments prompted us to expand the scope of our analyses and to
consider more fully several alternative explanations for our
findings.  We gratefully acknowledge their assistance. 

  -- John Ayanian, M.D., M.P.P., Assistant Professor, Division of
     General Medicine, Brigham and Women's Hospital, and Department
     of Health Care Policy, Harvard Medical School

  -- Carolyn Clancy, M.D., Director, Center for Outcomes and
     Effectiveness Research, and Acting Director, Center for Primary
     Care Research, Agency for Health Care Policy and Research

  -- James Cleeman, M.D., Coordinator, National Cholesterol Education
     Program; National Heart, Lung, and Blood Institute; National
     Institutes of Health

  -- Robert Hurley, Ph.D., Associate Professor, Department of Health
     Administration, Medical College of Virginia

  -- Charles Alan Lyles, Sc.D., Assistant Professor, Department of
     Health Policy Management, School of Hygiene and Public Health,
     Johns Hopkins University

  -- Barbara Starfield, M.D., Professor, Department of Health Policy
     and Management, School of Hygiene and Public Health, Johns
     Hopkins University


*** End of document. ***