Marciniak TA, Ellerbeck EF, Radford MJ, Kresowik TF, Gold JA, Krumholz HM, Kiefe CI, Allman RM, Vogel RA, Jencks SF. Improving the Quality of Care for Medicare Patients With Acute Myocardial InfarctionResults From the Cooperative Cardiovascular Project. JAMA. 1998;279(17):1351-1357. doi:10.1001/jama.279.17.1351
From the Health Care Financing Administration, Baltimore, Md (Drs Marciniak and Jencks); the Iowa Foundation for Medical Care, Des Moines (Drs Ellerbeck and Kresowik); the University of Kansas Medical Center, Kansas City (Dr Ellerbeck); the Connecticut Peer Review Organization, Middletown (Drs Radford and Krumholz); the Division of Cardiology, University of Connecticut, Farmington (Dr Radford); the Department of Surgery, University of Iowa, Iowa City (Dr Kresowik); MetaStar Inc, Madison, Wis, and the Department of Preventive Medicine and the Health Policy Institute, Medical College of Wisconsin, Milwaukee (Dr Gold); the Cardiovascular Section, Yale University School of Medicine, New Haven, Conn (Dr Krumholz); the Alabama Quality Assurance Foundation and the Department of Veterans Affairs, Birmingham (Drs Kiefe and Allman); the Division of Preventive Medicine (Dr Kiefe) and the Center for Aging (Dr Allman), University of Alabama at Birmingham; and the Department of Cardiology, University of Maryland, Baltimore (Dr Vogel).
Context.— Medicare has a legislative mandate for quality assurance, but the effectiveness
of its population-based quality improvement programs has been difficult to
Objective.— To improve the quality of care for Medicare patients with acute myocardial
Design.— Quality improvement project with baseline measurement, feedback, remeasurement,
and comparison samples.
Setting.— All acute care hospitals in the United States.
Patients.— Preintervention and postintervention samples included all Medicare patients
in Alabama, Connecticut, Iowa, and Wisconsin discharged with principal diagnoses
of acute myocardial infarctions during 2 periods, June 1992 through December
1992 and August 1995 through November 1995. Indicator comparisons were made
with a random sample of Medicare patients in the rest of the nation discharged
with acute myocardial infarctions from August 1995 through November 1995.
Mortality comparisons involved all Medicare patients nationwide with inpatient
claims for acute myocardial infarctions during 2 periods, June 1992 through
May 1993 and August 1995 through July 1996.
Intervention.— Data feedback by peer review organizations.
Main Outcome Measures.— Quality indicators derived from clinical practice guidelines, length
of stay, and mortality.
Results.— Performance on all quality indicators improved significantly in the
4 pilot states. Administration of aspirin during hospitalization in patients
without contraindications improved from 84% to 90% (P<.001),
and prescription of β-blockers at discharge improved from 47% to 68%
(P<.001). Mortality at 30 days decreased from
18.9% to 17.1% (P=.005) and at 1 year from 32.3%
to 29.6% (P<.001). These improvements in quality
occurred during a period when median length of stay decreased from 8 days
to 6 days. Performance on all quality indicators except reperfusion was better
in the pilot states than in the rest of the nation in 1995, and the differences
were statistically significant for aspirin use at discharge (P<.001), β-blocker use (P<.001),
and smoking cessation counseling (P=.02). Postinfarction
mortality was not significantly different between the pilot states and the
rest of the nation during the baseline period, although it was slightly but
significantly better in the pilot states during the follow-up period (absolute
mortality difference at 1 year, 0.9%; P=.004).
Conclusions.— The quality of care for Medicare patients with acute myocardial infarction
has improved in the Cooperative Cardiovascular Project pilot states. Performance
on the defined quality indicators appeared to be better in the pilot states
than in the rest of the nation in 1995 and was associated with reduced mortality.
HEALTH CARE quality is a topic of current concern. Quality has been
the focus of recent articles in the medical literature1,2
and is being addressed by a presidential commission. These current concerns
about quality are partially motivated by speculations about the effect of
managed health care on quality and access, and may be related to a realization
that quality cannot be assumed but rather must be understood and engineered.
Since 1992 the Health Care Financing Administration (HCFA) has been
implementing a continuous quality improvement approach to ensuring the quality
of care for its Medicare beneficiaries. HCFA's Health Care Quality Improvement
Initiative is implemented by its contractors, the peer review organizations
(PROs).3 The first project of this program
is the Cooperative Cardiovascular Project (CCP), which focuses on treatment
of patients with acute myocardial infarction (AMI).
The CCP began with the development of quality indicators for the treatment
of AMI. A steering committee convened by HCFA and the American Medical Association
(AMA) drafted quality indicators heavily based on clinical practice guidelines
developed by the American College of Cardiology (ACC) and the American Heart
Association (AHA).4 The PROs in 4 states, Alabama,
Connecticut, Iowa, and Wisconsin, refined the quality indicators and developed
data collection instruments and computer algorithms for them. These PROs abstracted
data from medical records of Medicare patients in their states who were discharged
with a principal diagnosis of AMI from June 1992 through February 1993 and
evaluated the rates for each quality indicator. The results of this baseline
data collection have been reported previously.5
The 4 PROs provided the results to the practitioners in their states
during 1994 and encouraged the initiation of quality improvement activities
for the treatment of AMI. To evaluate the effects of these activities, we
collected a follow-up sample of AMI cases in the pilot states and a comparison
sample of AMI cases from the rest of the nation in 1995.
For all samples we identified cases using hospital bills (UB-92 claims
form data) in the Medicare National Claims History File. The National Claims
History File includes all claims submitted for Medicare patients treated under
fee-for-service arrangements but does not include UB-92 submissions for all
patients treated under Medicare managed care risk contracts. We sampled only
claims with an ICD-9-CM (International
Classification of Diseases, Ninth Revision, Clinical Modification)
principal diagnosis code of 410 (AMI), excluding codes with a fifth digit
of 2, which designates a subsequent episode of care. For the baseline sample
we identified all AMI claims submitted by hospitals in the 4 pilot states
that had discharge dates between June 1, 1992, and December 31, 1992. We used
cases from January 1993 and February 1993 for the quality improvement feedback,
but we excluded these months from our analyses because the sampling for them
was incomplete because of timing. We used cases with discharge dates between
August 1, 1995, and November 30, 1995, for the follow-up sample. We selected
this latter sampling time frame to allow 6 months of elapsed time from the
completion of the feedback sessions and to accumulate a sample size sufficient
for a second round of feedback sessions.
For the nonpilot indicator comparison sample we randomly selected cases
from all states other than the pilot states from the follow-up period of August
1, 1995, through November 30, 1995. We randomly sampled 2400 cases to provide
90% power for detecting a 5% absolute difference in a typical quality indicator.
This sample size provides low power (40%) for detecting even substantial (10%)
improvement in mortality.
For mortality comparisons we sampled all AMI claims from June 1, 1992,
through May 31, 1993, and from August 1, 1995, through July 31, 1996. We selected
only the first AMI episode for patients with more than 1 AMI during these
periods. We matched the patients to the Medicare Enrollment Database to determine
Copies of medical records corresponding to discharge dates were requested
from hospitals. For the baseline sample the copies were sent to and abstracted
at the 4 state PROs. For the follow-up and nonpilot samples the record copies
were abstracted at 2 clinical data abstraction centers. The 2 abstraction
centers were established by HCFA through contracts with private organizations
experienced in medical abstractions to maintain abstraction consistency and
efficiency for its national quality improvement projects, including CCP.
Trained abstracters extracted predefined variables from the medical
record copies. Data were entered directly into computers using interactive
software with online data definitions, help, and edits. The clinical data
abstracted included all variables needed to calculate the quality indicator
rates as well as measures of risk, comorbidity, and complications. The CCP
abstraction software, which includes the operational definitions of all variables,
is available through the CCP Internet home page at
Data quality was monitored and maintained by random reabstractions,
calculation of reliability statistics such as interrater agreement rates and κ
values,6 determinations of reasons for discrepancies,
and improvement actions such as focused abstracter training. For the baseline
sample, 912 cases were randomly selected and reabstracted by a second abstracter.
For the CCP pilot follow-up sample and for other CCP samples abstracted at
the clinical data abstraction centers, cases were randomly reabstracted on
a continuous basis to generate reasonable confidence intervals (CIs) for reliability
statistics on a quarterly basis (eg, 95% CI, 85-94 for an observed agreement
rate of 90%). For all CCP samples, more than 3000 cases were reabstracted
for quality control purposes at the clinical data abstraction centers.
Interrater reliability of the abstractions was good. The reliability
of the baseline data has been reported previously,5
with agreement rates exceeding 94% and κ values ranging from 0.72 to
0.88 for determining treatments. For the clinical data abstraction center
abstractions, overall variable agreement averaged about 95% for the time periods
during which the follow-up and nonpilot samples were abstracted. Individual
variable abstraction reliability varied depending on the nature of the variable.
Simple laboratory or drug variables that were well documented in medical records
were highly reliable, whereas softer clinical history variables that had inconsistent
documentation patterns in medical records were less reliable (eg, agreement
rates of about 80% for the time of symptom onset prior to hospital arrival).
We supplemented the clinical abstractions with other data, such as diagnoses
and procedures, extracted from Medicare billing records. We also extracted
dates of death from the Medicare Enrollment Database. Dates of death in the
Medicare Enrollment Database are derived from both the discharge dates of
billing records indicating a discharge disposition of death and from the Master
Beneficiary Record obtained from the Social Security Administration. The Medicare
Enrollment Database has accurate records of the vital status of Medicare beneficiaries,7 but entries from Social Security records include unverified
dates of death recorded as the last day of the month when the exact date is
unavailable from a death certificate. We eliminated cases with unverified
dates of death from the mortality analyses if the case could not be classified
with certainty at the evaluation time. For instance, for 30-day mortality
statistics, we eliminated cases with unverified dates of death if survival
was between 30 and 59 days after admission as described in an earlier report.8 We found unverified dates of death for 168 patients
with confirmed AMI (ie, about 1% of such patients or 2.2% of deaths).
The CCP focuses primarily on quality indicators, measurable aspects
of care that are presumed on the basis of evidence or consensus to be related
to the quality of the care delivered and to favorable outcomes. The CCP indicators
were developed by a panel of experts convened by HCFA and the AMA, as reported
previously.5 Validation studies of the indicators
also were reported previously.9 The definitions
of the quality indicators are listed in Table 1.
Most quality indicators are defined as rates, eg, the ratio of the number
of eligible patients who received a treatment to the total number of patients
eligible for that treatment. We calculated indicator rates for 2 different
levels of eligibility. For an eligible candidate
indicator, we included all patients who met minimum eligibility requirements
(eg, for aspirin use at discharge, we included all patients with confirmed
AMI who were discharged alive). For an ideal candidate
indicator, we considered the same set of eligible candidates but we excluded
all patients with possible contraindications (eg, for aspirin use at discharge,
we excluded patients with contraindications such as recent bleeding or warfarin
use). Minimum eligibility requirements are shown in Table 1 under the heading Eligible and
contraindications are listed under the heading Exclusions. Ideal indicators are frequently used in PRO feedback sessions because
clinicians understand them to represent cases that should have been treated.
Because the ideal indicators exclude many cases and may be more susceptible
to measurement error and because case mix adjustment is less critical for
population estimates, we analyzed both levels of indicator eligibility and
compared the results for consistency. The only exception is for the smoking
cessation counseling indicator. We used 1 major eligibility condition (current
smoker status) and did not analyze this indicator by different levels of eligibility.
The definitions in Table 1
were converted to computer algorithms programmed in the Stata programming
language.10 The indicator eligibility requirements
involve evaluating many different variables. Some of these variables, such
as the occurrence of significant bleeding, are soft clinical variables subject
to variability in interpretation. Because the abstractions were done by 2
different sets of organizations at 2 different times, we evaluated the frequency
distributions of the exclusion variables and avoided using those that appeared
to have different interpretations between the baseline and follow-up abstractions
or among the 4 pilot state baseline samples. We analyzed all data sets summarized
in this article with the same Stata programs. Because the indicator algorithms
are complex, we have made them available through the CCP Internet home page.
The PROs in the 4 pilot states provided the results from the baseline
sample to practitioners in their states from 1994 through January 1995. The
PROs used consistent definitions of the indicators and similar approaches
to data presentation based on continuous quality improvement theory. Results
were presented in a positive manner emphasizing approaches for improvement.
The PROs collaborated with appropriate professional organizations such as
local chapters of the ACC, and requested quality improvement plans from hospitals
if the indicator results suggested opportunities for improvement. The details
of the feedback materials and sessions were left to the discretion of the
individual PROs. Samples of speaker notes for typical hospital presentations
are contained in the CCP data analysis package available through the CCP Internet
The major setting for feedback sessions was the acute care hospital
treating AMI patients. Of 390 hospitals represented in the baseline sample,
379 received CCP feedback. Fifty-four percent of the hospitals received feedback
by on-site presentations by the PROs' physicians, 18% received feedback at
regional seminars, 13% received feedback from telephone conferences, and 16%
received feedback from report mailings. Seventy-three percent of the hospitals
submitted formal responses to the PROs. Responses included creating or revising
standing orders and critical pathways, additional educational efforts and
data dissemination, and follow-up data monitoring. Aspirin use, smoking cessation
counseling, and thrombolytic administration were the indicators most frequently
Because our sampling used billing record diagnoses of AMI that may not
be accurate, we restricted most analyses to cases with confirmation of AMI
based on clinical data. We considered cases to be clinically confirmed if
the creatine kinase–MB fraction was greater than 0.05, if the lactate
dehydrogenase (LDH) level exceeded 1.5 times the upper limit of normal with
LDH1 greater than LDH2, or if 2 of the following 3 conditions
were documented: chest pain, a 2-fold elevation of creatine kinase level,
or a report of a new AMI on an electrocardiogram.
Different units of analysis were used for different end points. For
indicators we used hospitalizations, for mortality we used patients, and for
invasive procedures we used episodes. For episodes we followed the ICD-9-CM convention and defined an episode as all admissions occurring
within 8 weeks of the initial admission. One patient, if multiple hospitalizations
or episodes were sampled, could contribute more than 1 value to an indicator
or invasive procedure rate calculation. Each patient was counted only once
for mortality calculations, with survival calculated from the admission date
of the first hospitalization in a sample period.
We designed our sampling and data collection approach for quality improvement
purposes, so our samples include cases that should be excluded from analyses
for other purposes. For indicator rates we excluded cases based on transfer
status at either admission or discharge if the reliability of data collection
could be adversely affected. For example, for aspirin use at discharge, we
excluded transfer patients because hospital records frequently do not include
information about planned postdischarge medication use for patients transferred
to another facility. All such exclusions are noted in the indicator definitions
in Table 1. For mortality statistics
we excluded cases transferred from another hospital because we did not have
an estimate of the date of occurrence for the AMI, and we also excluded cases
with unverified dates of death.
Analyses of the abstracted data were performed using the Stata statistical
package,10 and analyses of claims data for
mortality comparisons were performed using SAS.11
For comparisons among all 3 samples, we used the χ2 test for
significance of changes in categorical variables such as indicator results,
and the Kruskal-Wallis test for significance of differences in the distributions
of continuous variables. For 2 sample comparisons we used the χ2 test for categorical variables and the Wilcoxon rank sum test for continuous
variables. Confidence intervals were calculated for the rate differences based
on normal approximations. We also plotted Kaplan-Meier survival curves and
calculated log-rank and Wilcoxon tests of significance of differences in the
survival curves. All CIs are 95% and all P values
are 2-tailed and unadjusted for multiple comparisons.
Overall, 23535 (96%) of 24509 records for patient discharges sampled
were obtained and abstracted successfully. The abstraction completion rate
was slightly higher for the follow-up period. Table 2 shows summary statistics for the samples, the sizes of the
analytic subsets (units of analysis) we used, and the sizes of some intermediate
subsets used to derive them.
Patient characteristics are shown in Table 3. Demographics of the patients in the 2 pilot samples are
similar, whereas the nonpilot sample has a slightly higher proportion of blacks.
For Medicare AMI discharges, women are represented almost as frequently as
men. The percentage of women is slightly higher if only those aged 65 and
older are included (eg, 49.8% female in the follow-up sample). The small increase
in the female representation from baseline to follow-up shown in Table 3 is not statistically significant.
However, we intentionally restricted data in Table 3 to patients who were not transferred from another hospital
so these data report characteristics for the same subset of patients for whom
we tabulated mortality. For all patients with confirmed AMI, the increase
in the percentage of women in the follow-up sample is nominally statistically
We sampled all Medicare AMI discharges, including those occurring in
patients younger than 65 years who were eligible for Medicare because of disability
or end-stage renal disease. The percentage of Medicare AMI patients younger
than 65 years is approximately 6% to 7% in these samples. We included these
cases in most analyses, but excluding them does not significantly change any
of the differences noted herein.
Table 4 shows comparisons
of the quality indicator rates in the 3 samples. The results for the pilot
baseline to follow-up comparison are consistent. All indicators show significant
improvements from baseline to follow-up in both the eligible and the ideal
candidate versions, except for reperfusion in ideal candidates. All improvements
other than reperfusion in ideal candidates are statistically significant and
magnitudes of improvements in the 2 versions of the same indicator are similar.
The reperfusion in ideal candidates indicator may be susceptible to
variations in abstraction, because its denominator incorporates a number of
soft variables such as the time from symptom onset. In addition to improvement
in the reperfusion indicator in eligible candidates, the speed with which
thrombolysis was accomplished improved significantly from baseline to follow-up.
The median time from arrival to administration of thrombolytics decreased
from 56 minutes to 41 minutes in ideal candidates (P<.001).
The percentage of ideal candidates who received thrombolytics and who were
administered the drugs within 1 hour after arrival improved from 57.1% to
70.8% (P<.001), and the percentage who received
thrombolytics within 30 minutes (the National Heart Attack Alert Program goal)
increased from 17.6% to 30.1% (P<.001).
Performance on the quality indicators was better in the pilot states
in 1995 than in the rest of the nation, with all but 2 of the 13 indicator
comparisons suggesting higher rates in the pilot states. The differences for
2 of the indicators, aspirin use at discharge and β-blocker use at discharge,
are statistically significant (P<.001), and the
difference for a third indicator, smoking cessation counseling, is also statistically
significant (P=.02). Time to thrombolysis did not
differ significantly between the 2 samples, eg, median time to thrombolysis
in ideal candidates was 41 minutes for the pilot states and 43.5 minutes for
the nonpilot states (P=.99).
Table 5 compares mortality
in the 3 abstraction samples. Both short-term (30-day) and longer-term (1-year)
mortality improved significantly between the baseline and follow-up periods,
with about a 10% relative reduction in mortality at both evaluation times.
Survival improvement did not appear to diminish with increasing length of
follow-up, as shown by the Kaplan-Meier survival curves in Figure 1. Differences in survival for all ages and for those 65
years and older are statistically significant (P<.001)
by both the log-rank and Wilcoxon tests.
Table 6 compares mortality
from the claims data. There were no significant differences in either 30-day
or 1-year mortality between the pilot state patients and those in the rest
of the nation during the baseline period. Mortality was significantly better
in the pilot states during the follow-up period, and the difference in 1-year
mortality between the pilot states and the rest of the nation was statistically
significant (P=.004). The differences in survival
during the follow-up period were also statistically significant by both the
log-rank test (P=.003) and the Wilcoxon test (P=.004).
Rates for invasive cardiology procedures by episode increased significantly
from baseline to follow-up. Angioplasty rates increased from 15.1% to 21.9%
(P<.001) and catheterization rates increased slightly
from 44.5% to 47.3% (P=.001). Coronary artery bypass
surgery rates remained stable from 11.2% to 12% (P=.14).
These rates do not include procedures performed during nonsampled hospitalization
or on an outpatient basis. Invasive procedure rates by episode cannot be estimated
for the nonpilot sample because multiple discharges per episode were infrequently
sampled. Procedure rates for the sampled nonpilot discharges were not significantly
different from the pilot follow-up rates.
Length of stay decreased from the baseline period to the follow-up period.
For patients with confirmed MI who were not transferred in or out and who
did not die in the hospital, the mean length of stay decreased from 9.8 days
to 7.5 days and the median decreased from 8 days to 6 days (P<.001). Length of stay for the nonpilot cases did not differ significantly
from the pilot follow-up cases.
The quality of care for Medicare patients with AMI has improved in the
CCP pilot states between the baseline and follow-up periods. Improvements
noted in the CCP quality indicators (process measures) are consistent and
appear to be associated with an improvement in at least 1 important outcome
measure, mortality. The magnitudes of the changes are sufficiently reassuring
that meaningful improvements were accomplished, rather than small changes
made significant by large sample sizes. Although a recent article has cautioned
against attributing improvements in quality to large-scale programs like CCP,12 the differences between the pilot and nonpilot indicator
and mortality rates at follow-up, combined with no differences in mortality
rates at baseline and severity indexes at follow-up, suggest that CCP contributed
to the improvement.
The improvement in 1 indicator is particularly striking. The use of β-blockers
at discharge in minimally eligible patients increased from 32% to 50%, an
18% absolute increase or more than a 50% relative increase. In contrast, while
the speed with which reperfusion was delivered improved dramatically, only
modest improvements were noted in reperfusion rates. Only about 20% of patients
65 years and older with AMI received early reperfusion therapy.
The data show that Medicare patients with AMI have significant comorbidity,
eg, diabetes in about 30% and a history of hypertension in about 60%. High
comorbidity in patients 65 years and older may be one factor contributing
to the low reperfusion rates. The data in Table 3 indicate slightly more comorbidity in the follow-up sample
cases than in the baseline. Although some of the differences in comorbid factors
are highly statistically significant, the magnitudes of the differences are
usually small and could represent variations in abstracter interpretation
rather than actual differences in patient characteristics. The closeness of
the composite risk indexes (Medicare Mortality Predictor System [MMPS] and
Acute Physiology and Chronic Health Evaluation [APACHE] III) for the pilot
follow-up and nonpilot samples suggests that these samples are similar with
regard to severity of illness. The data in Table 3 also suggest that the follow-up sample does not represent
a healthier population and that comorbidity does not explain the lower mortality
for the follow-up patients.
The sex distribution for elderly Medicare AMI patients does not show
the male predominance typical of clinical trials. The distribution between
men and women is close to equal for confirmed AMI in patients 65 years and
older and, as the population ages, the sex distribution will shift to a female
predominance in the next decade. Ischemic heart disease is very much a women's
health issue for the Medicare population.
The results of the eligible and ideal indicators are consistent. Although
we do not provide any evidence in this study regarding the superiority of
either indicator for quality improvement feedback or for hospital rankings,
our results suggest that the simpler eligible indicators are adequate for
evaluating changes in large populations. The eligible indicators have an advantage
of requiring much less data collection. For example, for aspirin use during
hospitalization, we evaluated 22 more variables (details regarding possible
contraindications) for the ideal indicator than for the eligible indicator.
Eligible indicators may be more efficient for follow-up studies in populations
over the course of several years or for other situations in which case mix
While we believe that these data suggest that CCP has contributed to
a measurable improvement in the quality of AMI care in the pilot states, we
acknowledge that our study has some limitations. The 4 pilot states are not
a random subsample of the nation. The data from the follow-up period suggest
that the patient characteristics are not significantly different in the pilot
states than in the rest of the nation, but we lack a comparison group to confirm
that the processes of care were similar in the pilot states and the rest of
the nation at baseline. The similarity in mortality between the pilot states
and the rest of the nation at baseline suggests that differences in care processes
at baseline were minimal or noneffective.
Nonpilot follow-up results are better than pilot baseline results, suggesting
that there have been improvements in care in the rest of the nation. Many
factors, such as dissemination of clinical trial results, professional society
educational activities, commercially sponsored programs such as the National
Registry of Myocardial Infarction,13 other
national initiatives such as the National Heart Attack Alert Program,14 and state projects such as the Connecticut Medicare
Hospital Information Project15 probably have
contributed to the changes in both pilot and nonpilot states.
We have not yet analyzed any effects on costs. While we do not have
formal cost impact studies, we are reassured by the observations that the
2 treatments with the widest applicability that improved significantly, administration
of aspirin and administration of β-blockers, are inexpensive therapies,
and that the process and mortality improvements were accompanied by a decreased
average length of stay. We suspect that the decreased length of stay is predominantly
caused by non-CCP forces, such as evolving practice patterns, the current
emphasis on controlling costs in the medical community, and the influence
of managed health care. The CCP may have contributed to declining lengths
of stay because quality improvement mechanisms such as critical pathways promoted
by the PROs during CCP feedback frequently have goals for both resource conservation
and quality improvement. We believe that CCP provides some evidence that quality
improvement is achievable in today's environment of cost control.
We should not be too complacent about the positive results. As compared
with current clinical trials for AMI that report short-term mortality results
of about 5% in selected patients, AMI in the older population remains a deadly
disease. Mortality rates are 18% at 30 days and 30% at 1 year, and more patients
die before they reach a hospital.16 We suspect
that there may still be room for improvement even in the pilot states and
that we have more lessons to learn from CCP and other sources about the optimal
care of older patients diagnosed as having AMI.
Because CCP pilot baseline results suggested that opportunities for
improving Medicare AMI care existed, HCFA proceeded with CCP activities in
the rest of the nation in 1995 and 1996. The CCP national effort has proceeded
from data collection (an 8-month sample of Medicare AMI discharges from all
hospitals in nonpilot states) through feedback from the PROs. We plan to resample
all states for 1997, although with a sample size that will produce reasonably
precise national statistics but not hospital-specific statistics. We have
learned much from the lessons of the CCP pilot, and are hopeful that the national
effort will produce results comparable to those reported herein.