American Journal of Epidemiology
ª The Author 2008. Published by the Johns Hopkins Bloomberg School of Public Health.
All rights reserved. For permissions, please e-mail: [email protected].
Vol. 168, No. 2
DOI: 10.1093/aje/kwn107
Advance Access publication May 8, 2008
Practice of Epidemiology
Impact of New Biomarkers of Myocardial Damage on Trends in Myocardial
Infarction Hospital Admission Rates from Population-based Administrative Data
F. M. Sanfilippo1, M. S. T. Hobbs1, M. W. Knuiman1, and J. Hung2
1
School of Population Health, University of Western Australia, Crawley, Western Australia, Australia.
School of Medicine and Pharmacology, Sir Charles Gairdner Hospital Unit, University of Western Australia, Nedlands, Western
Australia, Australia.
2
Received for publication August 7, 2007; accepted for publication March 31, 2008.
Use of troponin testing in the diagnosis of myocardial infarction substantially increases the number of cases
diagnosed as myocardial infarction among suspected cases in comparison with previous criteria. However, the
impact of troponin testing on rates reported in national statistics that use routinely collected hospital morbidity data
is uncertain. The authors developed Poisson regression models to estimate the effect of troponin testing on longterm trends in hospital admission rates in Perth, Western Australia, from 1980 to 2004. Troponin tests were used for
10.5% of patients with suspected myocardial infarction in 1996, rising rapidly to more than 90% of patients from
2001 onward. Fitted models that assumed a continuing linear decline estimated that 100% use of troponin testing in
cases of suspected myocardial infarction would lead to an apparent increase in hospital admission rates of 42%
(95% confidence interval (CI): 28, 56) in men and 21% (95% CI: 4, 41) in women as compared with rates that would
be expected if previous linear trends had continued. Smaller effects of 30% (95% CI: 14, 48) in men and 2% (95%
CI: 21, 20) in women were found in fitted models that assumed an underlying attenuating trend in the rates.
Similarly constructed logistic regression trend models found no significant effect of troponin testing on trends in
28-day case-fatality.
coronary disease; diagnosis; medical record linkage; mortality; myocardial infarction; troponin
Abbreviation: ICD, International Classification of Diseases.
Mortality from coronary heart disease in Australia has
been declining for nearly 40 years, but the disease remains
a major cause of death and disability (1). Monitoring of
trends in the incidence and prevalence of coronary heart
disease is therefore essential for guiding public health initiatives and planning clinical services.
At the population level, there is no practical method for
monitoring changes in the prevalence of coronary heart disease. Trends in acute coronary events (coronary heart disease
death or hospitalization for myocardial infarction) are widely
accepted as surrogate measures of incidence, but these measures should ideally be based on dedicated disease registers
that apply standardized objective diagnostic criteria over time
(2–4). However, such registers are costly, and despite the
participation of collaborating centers in Newcastle and Perth
in the World Health Organization MONICA Project (Monitoring of Trends and Determinants in Cardiovascular Disease;
1984–1993), there are no registers in Australia at present. In
their absence, routinely collected mortality data and hospital
morbidity data provide acceptable alternative sources of information for population surveillance of trends if they are
supported by periodic validation studies of disease coding
(3, 5–8), including previous studies using Western Australian
data (5, 8).
Apart from consistency of disease coding, the accuracy of
trends in hospital admissions for myocardial infarction may
Correspondence to Dr. Frank Sanfilippo, School of Population Health M431, University of Western Australia, 35 Stirling Highway, Crawley,
WA 6009, Australia (e-mail: frank.sanfi[email protected]).
225
Am J Epidemiol 2008;168:225–233
226 Sanfilippo et al.
be affected by many factors, including changes in admission
policies, changes in clinical terminology, and changes in the
sensitivity and specificity of diagnostic tests. Of particular
importance in the past 10 years has been the widespread adoption in clinical practice of immunoassays for troponins, which,
compared with previously available biomarkers of myocardial
infarction, are highly specific and sensitive tests for myocardial necrosis. In 2000, the European Society of Cardiology
and the American College of Cardiology recommended in
a joint statement that any elevation of troponin in the context
of symptoms of acute coronary heart disease should be considered diagnostic of myocardial infarction (9). The statement
acknowledged that this would lead to an increase in the diagnosis of mild cases of myocardial infarction and have major
implications for epidemiologic studies that had previously relied on traditional diagnostic biomarkers. This prediction has
been confirmed in several single-hospital studies, as summarized by Roger et al. (10), and in registry studies, with estimates of increased myocardial infarction in cases of acute
coronary heart disease ranging from 23 percent to 195 percent
(10, 11). Changes in case mix resulting in reduced case fatality
have also been reported, but with less consistency than increases in the number of cases of myocardial infarction
(10, 11). However, it is unclear what impact the adoption of
troponin testing has had on trends in myocardial infarction in
national statistics based on administrative data.
Our objectives in this study were to 1) describe long-term
trends in hospital admissions for myocardial infarction in
the Perth Statistical Division of Western Australia from
1980 to 2004 and 2) evaluate the hypotheses that the widespread use of troponin testing in Perth hospitals since 1998
was associated with a relative increase in admission rates for
myocardial infarction and a relative decline in 28-day case
fatality in 1998–2004 as compared with previous periods.
MATERIALS AND METHODS
Population-based linked administrative health data were
extracted from the Western Australian Data Linkage System
(12) for the period 1980–2004 to establish a person-linked
file of death records and hospital admissions for cardiovascular disease or diabetes mellitus (equivalent to International Classification of Diseases, Ninth Revision (ICD-9),
codes 390–459 and 250) (13). We identified 28-day episodes
of myocardial infarction based on hospital admissions with
a principal discharge diagnosis of myocardial infarction, including cases of myocardial infarction with onset in the hospital. That is, admissions for the same person were counted as
new events only if they occurred more than 28 days from the
date of a previous admission for myocardial infarction (4).
Fatal cases were those in which death from any cause occurred within 28 days of admission for myocardial infarction,
either before or after discharge. The data were used to calculate age-specific rates, age-standardized rates, and 28-day
case fatality among persons aged 35–79 years residing in
the Perth Statistical Division (population 1.46 million in
2004 (14)). The study was approved by the Human Research
Ethics Committee of the University of Western Australia and
the Confidentiality of Health Information Committee of the
Western Australian Department of Health.
TABLE 1. Proportion (%) of patients aged 35–79 years with
emergency hospital admissions who received troponin testing
in each year relative to all cardiac biomarker tests used in
emergency admissions of the same year (troponin, creatine
kinase, creatine kinase-MB), Perth Statistical Division, Western
Australia, 1980–2004
Sex and age group (years)
Year(s)
No. of
study
hospitals*
Males
<65
Males
65
Females
<65
Females
65
Total
0
0
0
0
0
1996
10.2
12.0
7.9
10.2
10.5
1
1997
15.3
17.0
10.2
12.5
14.4
5
1998
44.8
41.4
41.9
39.5
42.1
8
1999
65.2
60.9
61.8
58.6
61.8
8
2000
81.2
79.2
79.6
78.8
79.8
8
2001
92.9
91.8
93.1
92.7
92.6
8
2002
95.7
94.6
96.5
95.3
95.4
8
2003
94.8
94.5
97.0
96.0
95.3
8
2004
95.1
95.6
96.5
96.7
95.8
8
1980–1995
* Calculations were based on cardiac biomarker tests performed in
the eight major hospitals (three teaching and five private) in Perth,
Australia, in which cardiac admissions are investigated and treated.
The numbers indicate how many of these hospitals began using
troponin testing, starting from 1996 (initially there was gradual
uptake, followed by more rapid uptake beginning in 1999).
Diagnoses in the hospital morbidity data are recorded by
trained clinical coders in each hospital using ICD codes and
rules which are revised at various intervals. During 1980–
2004, there were three such revisions: the ICD-9 (13) was
used from 1980 to 1987, the International Classification of
Diseases, Ninth Revision, Clinical Modification (ICD-9CM) (15) was used from 1988 to June 30, 1999 (the Australian Version (16) of the ICD-9-CM was used from July 1,
1995, to June 30, 1999), and the International Classification
of Diseases, Tenth Revision, Australian Modification (ICD10-AM) (17) has been used since July 1, 1999. The following ICD codes were used to identify hospital admissions for
myocardial infarction: 410 (ICD-9 and ICD-9-CM) and I21
and I22 (ICD-10-AM).
Trends in the use of troponin testing
We determined trends in the use of troponin testing in
cases of suspected myocardial infarction through linkage
of laboratory test results for cardiac biomarkers (troponin,
creatine kinase, and creatine kinase-MB) to corresponding
hospital morbidity records of emergency hospital admissions (fatal and nonfatal) for any variant of coronary heart
disease or chest pain. Laboratory data were provided by
pathology laboratories servicing major public and private
hospitals. The level of use of troponin testing in each year
(regardless of test result) was calculated as the number of
patients aged 35–79 years having troponin tests divided by
the total number of patients aged 35–79 years in which any
cardiac biomarker test was ordered (table 1). These proportions were used in the Poisson regression trend models.
Am J Epidemiol 2008;168:225–233
Troponin Tests and Myocardial Infarction Admission Rates
Statistical methods
Age-specific rates were calculated as the number of hospital admissions for myocardial infarction in each year and
each 5-year age group (35–79 years) divided by the Perth
Statistical Division population of the same age group and
year obtained from the Australian Bureau of Statistics. Agestandardized rates were calculated by the direct method
using as the standard population the age distribution of the
Australian resident population of June 30, 2001 (the most
recent population census) (18). Case fatality was calculated
as the number of patients who died of any cause within 28
days of admission for myocardial infarction divided by the
number of admissions for myocardial infarction. For plotting, case fatality was age-standardized by the direct method
using 5-year age groups and the age distribution of the 28day myocardial infarction cohort.
Poisson log-linear regression models were used to estimate trends in rates with calendar year and to estimate the
effect of the introduction of troponin testing on rates. Troponin tests were introduced in Perth in 1996 but were not
adopted by all public and private hospitals admitting cases
of myocardial infarction until 1998, which we defined as the
beginning of the troponin era. To examine overall changes in
the slope of the trend over time, Poisson models were initially fitted separately for the three periods 1980–1988,
1989–1997, and 1998–2004. For each period, a Poisson linear model including 5-year age group (35–39, . . ., 75–79
years) and calendar year of admission (as a continuous variable) was fitted, and the rate ratio describing the relative
annual change was calculated from the beta coefficient for
calendar year. Similarly, logistic regression was used to estimate the relative annual change in case fatality for these
three periods.
To estimate the effect of troponin testing on rates of admission for myocardial infarction, we fitted Poisson regression models for the period 1989–2004. The earlier period
was excluded because the trends in rates showed a marked
increase in the rate of decline from 1989 onward. These
models included variables for age and calendar year of admission and a term reflecting the use of troponin testing.
Further models including a term for interaction between age
(35–64 years or 65–79 years) and the troponin variable were
fitted to evaluate whether the effect of troponin testing on
rates differed for the two age groups. Linear and quadratic
trend models were fitted. The linear trend model assumes
a linear trend in the rate (on a log scale) throughout the
period 1989–2004 with a troponin-test effect superimposed,
whereas the quadratic trend model assumes a curved (attenuating) trend in the rate (on a log scale) with a troponin-test
effect superimposed.
The variable representing the use of troponin testing took
two forms. The first was a binary variable indicating
whether or not troponin testing was used in a particular year.
This variable was 0 for years prior to 1998 (the year of full
uptake of troponin testing in Perth) and 1 for the years
1998–2004. The second was a continuous variable in which
individual years were given a value from 0 to 1, calculated
as the proportion of troponin tests used relative to all cardiac
biomarker tests as described above. This variable was 0 for
Am J Epidemiol 2008;168:225–233
227
1989–1995 and then increased from 0.10 in 1996 to 0.96 in
2004 (table 1). There was little variation in troponin test
proportions by age and sex, and therefore the overall proportions were used for all age and sex groups.
In a similar manner, we used logistic regression trend models to examine the effect of troponin testing on unadjusted
28-day case fatality. All p values reported are two-sided.
RESULTS
During the 25-year period 1980–2004, there were 36,028
hospital admissions for a 28-day episode of myocardial infarction as the principal discharge diagnosis among persons
aged 35–79 years in the Perth Statistical Division. In the 16year period 1989–2004, there were 22,896 admissions for
myocardial infarction (16,252 for men and 6,644 for
women), and 2,095 patients died within 28 days of admission (1,246 men and 849 women). Troponin tests were used
in 10.5 percent of patients with admissions for coronary
heart disease variants or chest pain in 1996, rising rapidly
to 42.1 percent in 1998 and 95.8 percent in 2004 (table 1).
Admission rates for myocardial infarction
Age-standardized rates of admission for myocardial infarction showed a declining trend over the period 1980–
2004, with declines of 3.0 percent per year in men and 3.3
percent per year in women. Rates declined more rapidly
during the period 1989–1997 than in 1980–1988, and there
was an upward shift in rates for the period 1998–2004 (the
troponin period) (figure 1). In men, myocardial infarction
admission rates declined by 5.9 percent per year in 1989–
1997 as compared with 2.5 percent per year in 1998–2004,
while in women the rates declined by 5.7 percent per year in
1989–1997 as compared with 1.7 percent per year in 1998–
2004. Figure 2 shows that the declines in rates of myocardial
infarction during the period 1980–2004 were similar in
younger and older men and women.
Rate ratios from the Poisson trend models incorporating
a troponin variable showed the estimated effect of the introduction of troponin testing on admission rates (table 2).
In men, the introduction of troponin testing produced estimated increases of 19 percent (linear model with binary
troponin variable) and 14 percent (quadratic model with
binary troponin variable) in the rate of admission for myocardial infarction. That is, rates in the troponin period were
19 percent (or 14 percent) higher than would have been
expected if troponin testing had not been introduced and
the linear (or quadratic) trend had continued during the
troponin period. Estimates from trend models with the continuous troponin variable reflect the gradual uptake of troponin testing, and the reported rate ratio shows the estimated
increase in the rate that is associated with 100 percent use of
troponin testing. The estimated increase was 42 percent
from the linear trend model and 30 percent from the quadratic trend model. When examined within age groups, the
estimated effect of troponin testing in men was slightly (but
not significantly) larger in the age group 35–64 years than in
the age group 65–79 years.
228 Sanfilippo et al.
1980–1988
1,000
Age-standardized rate
(per 100,000 person-years)
800
1989–1997
RR(slope) = 0.986
(95% CI: 0.978, 0.993)
1998–2004
RR(slope) = 0.941
(95% CI: 0.933, 0.948)
600
RR(slope) = 0.975
(95% CI: 0.964, 0.986)
400
Men
RR(slope) = 0.974
(95% CI: 0.962, 0.986)
RR(slope) = 0.943
(95% CI: 0.932, 0.955)
200
RR(slope) = 0.983
(95% CI: 0.965, 1.002)
Women
100
1980
1984
1988
1992
1996
2000
2004
Year
FIGURE 1. Age-standardized rates of hospital admission for myocardial infarction as the principal discharge diagnosis during 1980–2004 among
patients aged 35–79 years residing in the Perth Statistical Division, Western Australia. RR(slope) is the rate ratio for the annual change in
admission rates from a log-linear Poisson model with 5-year age group (35–39, 40–44, . . ., 75–79 years) and calendar year (as a continuous
variable from 1980 to 2004), and the 95% confidence interval (CI) is shown in parentheses. Rates were calculated for 28-day events from linked
administrative health data. The standard population was the final Australian estimated resident population on June 30, 2001 (18).
For women, a similar pattern of results was seen, with
a significantly greater effect of troponin testing in the age
group 35–64 years than in the age group 65–79 years; however, the effect sizes were smaller and not statistically significant, except for the age group 35–64 years. Overall, for
women the introduction of troponin testing produced
estimated increases of 9 percent (binary linear model) and
3 percent (binary quadratic model) in the rate of admission
for myocardial infarction. The corresponding effect sizes for
women aged 35–64 years were 22 percent and 16 percent,
respectively. Estimates of the effect size associated with 100
percent use of troponin testing (from respective continuous
Age-specific rate
(per 100,000 person-years)
10,000
Age (years)
Males ≥65
Females ≥65
Males <65
Females <65
1,000
100
10
1980
1984
1988
1992
1996
2000
2004
Year
FIGURE 2. Age-specific rates of hospital admission for myocardial infarction as the principal discharge diagnosis during 1980–2004 among
patients aged 35–79 years residing in the Perth Statistical Division, Western Australia. Rates were calculated for 28-day events from linked
administrative health data.
Am J Epidemiol 2008;168:225–233
Troponin Tests and Myocardial Infarction Admission Rates
229
TABLE 2. Estimated effects of use of troponin testing (Poisson regression trend models) on rates of
hospital admission for myocardial infarction during the period 1989–2004 among patients aged 35–79 years
residing in the Perth Statistical Division, Western Australia*
Sex
Male
Troponin
testing
modely
No. of
admissions
16,252
Binary
Continuous
Female
6,644
Binary
Continuous
Age group
(years)
Linear models
Quadratic models
RRz
95% CIz
p value§
RR
95% CI
p value§
All
1.19
1.12, 1.26
<0.0001
1.14
1.07, 1.22
<0.0001
65–79
1.15
1.07, 1.23
1.11
1.03, 1.19
35–64
1.22
1.14, 1.31
0.070
1.17
1.09, 1.26
0.073
All
1.42
1.28, 1.56
<0.0001
1.30
1.14, 1.48
0.0001
65–79
1.36
1.22, 1.52
1.25
1.09, 1.44
35–64
1.46
1.32, 1.63
0.058
1.34
1.17, 1.54
0.060
All
1.09
0.99, 1.20
0.081
1.03
0.93, 1.14
0.53
65–79
1.04
0.94, 1.15
0.99
0.89, 1.10
35–64
1.22
1.08, 1.38
0.003
1.16
1.02, 1.31
0.003
All
1.21
1.04, 1.41
0.016
0.98
0.79, 1.20
0.81
65–79
1.14
0.97, 1.33
0.92
0.75, 1.15
35–64
1.37
1.15, 1.64
1.11
0.88, 1.40
0.004
0.005
* Results were based on 28-day events from linked administrative health data.
y The binary model indicates whether troponin testing was used (1 for 1998–2004) or not used (0 for 1989–1997).
The continuous model used the proportion of patients who received troponin testing in a given year out of all patients
who had biomarker tests (troponin, creatine kinase, and creatine kinase-MB tests for men and women combined) in
the same year. Values used were those shown in table 1, expressed as fractions from 0 to 1. Period was
a continuous variable ranging from 1 (1989) to 16 (2004).
z RR, rate ratio; CI, confidence interval.
§ The p value for all ages was obtained from a two-sided Wald chi-squared test for RR ¼ 1; the p value for age
group (two-sided Wald chi-squared test) tested whether the RRs were equal for the two age groups.
models) were 21 percent and 2 percent overall and 37
percent and 11 percent for women aged 35–64 years.
Case fatality
During 1980–2004, age-standardized 28-day case fatality
decreased from 18.1 percent to 7.1 percent in women and
from 13.5 percent to 4.7 percent in men (figure 3). Table 3
shows the estimated effects of troponin use on 28-day case
fatality in men and women based on trend models for the
period 1989–2004. The overall estimated effects of troponin
testing in binary models were increases of 9–11 percent in
men and 6–8 percent in women, while continuous models
showed changes of 11–16 percent in men and 7 to 10
percent in women. However, none of these changes were
statistically significant. There were also no statistically significant changes when results were examined within age
groups. For men, the year 1989 had an atypically high case
fatality (figure 3), and when we omitted this year from the
models to examine the sensitivity of the results to this value,
we found that the odds ratios were only slightly smaller and
remained statistically nonsignificant.
DISCUSSION
During the period 1980–2004, rates of hospital admission
for myocardial infarction in Perth, Western Australia, declined by 3.0 percent per year in men and 3.3 percent per year
Am J Epidemiol 2008;168:225–233
in women. However, the rates of decline were not uniform,
with a modest decline in the first 9 years (1.4 percent decline
per year in men, 2.6 percent in women) being followed by
an accelerated decline in 1989–1997 (approximately 6 percent
per year for both men and women) before slowing to 2.5
percent per year in men and 1.7 percent in women in 1998–
2004. The departure in the last period from the linear decline
of the previous period of 1989–1997 is consistent with our
hypothesis that the uptake of troponin testing would be associated with a relative increase in admission rates for myocardial infarction. This does not exclude the possibility of real
attenuation of the rate of decline, although there is no evidence of slowing of the long-term decline in mortality from
ischemic heart disease in Australia to support this (19).
To estimate the effect of troponin testing on rates of myocardial infarction, we fitted both linear and quadratic Poisson regression trend models. The former assumed that
recent linear trends would have continued were it not for
troponin testing (possibly overestimating the effect of troponin testing), while the quadratic models captured the effect of troponin testing additional to real attenuation of the
downward trends (possibly underestimating the effect of
troponin testing). Both models were fitted using ‘‘troponin
period (1998–2004)’’ as a binary variable or alternatively
using a continuous variable reflecting the actual level of use
of troponin testing in emergency cardiac admissions in each
of the years 1996–2004. The binary troponin variable estimated the average effect over the period 1998–2004,
230 Sanfilippo et al.
1980–1988
Age-standardized 28-day case fatality (%)
25
1989–1997
1998–2004
OR(slope) = 1.004
(95% CI: 0.974, 1.034)
20
OR(slope) = 0.942
(95% CI: 0.909, 0.976)
OR(slope) = 0.924
(95% CI: 0.866, 0.985)
15
10
OR(slope) = 0.965
(95% CI: 0.941, 0.989)
Women
OR(slope) = 0.921
(95% CI: 0.895, 0.948)
5
Men
OR(slope) = 0.927
(95% CI: 0.880, 0.977)
0
1980
1984
1988
1992
2000
1996
2004
Year
FIGURE 3. Age-standardized 28-day case fatality (all causes) following myocardial infarction during 1980–2004 among patients aged 35–79
years residing in the Perth Statistical Division, Western Australia. OR(slope) is the odds ratio for annual change in case fatality from a logistic
regression model with 5-year age group (35–39, 40–44, . . ., 75–79 years) and calendar year (as a continuous variable from 1980 to 2004), and the
95% confidence interval (CI) is shown in parentheses. Case fatality was calculated for 28-day events from linked administrative health data.
whereas the continuous variable estimated the expected effect if troponin testing were used in all cases.
Estimates from the continuous linear model suggest that
in men admission rates for myocardial infarction would be
approximately 40 percent higher with complete use of troponin testing, whereas the continuous quadratic model was
associated with a 30 percent increase in rates. The effects
were not modified by age.
TABLE 3. Estimated effects of use of troponin testing (logistic regression trend models) on 28-day case
fatality (any cause) following hospital admission for myocardial infarction during the period 1989–2004
among patients aged 35–79 years residing in the Perth Statistical Division, Western Australia*
Sex
Male
No. of
deaths
1,246
Troponin
testing
modely
Binary
Continuous
Female
849
Binary
Continuous
Age group
(years)
Linear models
Quadratic models
ORz
95% CIz
p value§
OR
95% CI
p value§
All
1.11
0.87, 1.42
0.40
1.09
0.83, 1.42
0.53
65–79
1.17
0.90, 1.51
1.14
0.87, 1.51
35–64
0.96
0.69, 1.34
0.20
0.94
0.67, 1.33
0.20
All
1.16
0.79, 1.69
0.46
1.11
0.62, 1.98
0.72
65–79
1.21
0.82, 1.79
1.16
0.65, 2.09
35–64
0.99
0.62, 1.59
0.95
0.50, 1.81
All
1.06
0.78, 1.43
0.72
1.08
0.78, 1.50
65–79
1.04
0.77, 1.42
1.07
0.76, 1.49
35–64
1.14
0.70, 1.85
0.70
1.16
0.70, 1.92
0.70
All
0.93
0.58, 1.48
0.76
0.90
0.44, 1.84
0.78
65–79
0.92
0.57, 1.48
0.89
0.44, 1.83
35–64
0.99
0.51, 1.93
0.96
0.41, 2.27
0.28
0.79
0.28
0.65
0.79
* Results were based on 28-day events from linked administrative health data.
y The binary model indicates whether troponin testing was used (1 for 1998–2004) or not used (0 for 1989–1997).
The continuous model used the proportion of patients who received troponin testing in a given year out of all patients
who had biomarker tests (troponin, creatine kinase, and creatine kinase-MB tests for men and women combined) in
the same year. Values used were those shown in table 1, expressed as fractions from 0 to 1. Period was
a continuous variable ranging from 1 (1989) to 16 (2004).
z OR, odds ratio; CI, confidence interval.
§ The p value for all ages was obtained from a two-sided Wald chi-squared test for OR ¼ 1; the p value for age
group (two-sided Wald chi-squared test) tested whether the ORs were equal for the two age groups.
Am J Epidemiol 2008;168:225–233
Troponin Tests and Myocardial Infarction Admission Rates
In women, the estimated effects of troponin testing were
smaller, and there was a significant interaction with age. In
the continuous linear model, rates were estimated to be
nearly 40 percent higher (95 percent confidence interval:
15, 64) in women under age 65 years as compared with
14 percent higher (95 percent confidence interval: 3, 33)
in women aged 65–79 years. In the quadratic models, there
was no significant effect of troponin testing except for the
age group 35–64 years, which showed a 16 percent increase
(binary model). These findings, which contrast with those
of previous studies, are unexpected, especially in older
women, since troponin testing should identify more cases
of myocardial infarction among patients in whom the diagnosis of myocardial infarction is often difficult (11, 20).
Our study nevertheless suggests that troponin testing is associated with a substantial apparent increase in admission
rates for myocardial infarction, at least among men.
The relative increase in rates of myocardial infarction
during the troponin period in our study was substantially
smaller than the increases observed in two major populationbased registry studies in Finland (11) and the United States
(10), in which the introduction of troponin testing increased
the numbers of cases diagnosed as myocardial infarction by
83 percent and 74 percent, respectively. The major difference between these studies and ours was that the present
study, based on administrative data, related relative changes
in trends in myocardial infarction rates to levels of overall
use of troponin testing, whereas the other studies compared
troponin testing in diagnostic algorithms in place of traditional cardiac biomarkers. However, the US study (10) also
found that the increase in cases with dismissal diagnoses
of myocardial infarction (ICD-9 code 410) when meeting troponin-based criteria was substantially smaller (42
percent vs. 74 percent). The authors attributed this discrepancy to the failure of clinicians to fully adopt the
recommended criteria for the diagnosis of myocardial infarction incorporating troponin results, perhaps because
they were reluctant to label as myocardial infarction all
additional cases diagnosed from troponin testing (10). Anecdotal evidence suggests that this was also the case in
Perth, particularly in later years, in which there was progressive lowering of diagnostic thresholds for troponin
testing. If this is the case generally, a lag of several years
may occur before the new troponin-based criteria for myocardial infarction are fully expressed in administrative data
nationally.
Our second hypothesis, that the decline in 28-day case
fatality would accelerate due to an increase in the number of
less severe cases of myocardial infarction diagnosed because of troponin testing, was not confirmed. In women,
case fatality was further reduced by an apparent 7–10 percent, but it increased by 11–16 percent in men. However, in
neither instance were the changes statistically significant. It
is possible that our initial assumption that troponin testing
would lead to the diagnosis of less severe cases of myocardial infarction was incorrect, as several studies have shown
that cases of non-ST-elevation myocardial infarction with
low troponin levels, which account for the majority of cases
reclassified as myocardial infarction, are often associated
with poor outcomes (10, 11).
Am J Epidemiol 2008;168:225–233
231
In interpreting these inconclusive results, readers should
recognize that by 1998, 28-day case fatality for myocardial
infarction in Perth had already fallen by more than 50 percent in men and women since 1980. Hence, the numbers of
deaths in each year were relatively low, as reflected by the
wider confidence intervals in logistic regression models.
This improvement coincided with major and rapid changes
in proven medical treatment of myocardial infarction, particularly during the period 1984–1993, in Perth and in the
majority of the populations included in the MONICA Project (21–23). The benefits of these advances in treatment
may have been largely realized by the time troponin testing
was introduced.
Limitations of the study
During the period of introduction of troponin testing,
there was also progressive lowering of diagnostic troponin
thresholds for myocardial infarction as the precision of measurement improved. This may have led to the diagnosis of
more cases of myocardial infarction in later years, which
could have affected our results, but this could not be confirmed statistically.
The accuracy of long-term trends in rates of hospital admission for myocardial infarction based on administrative
data may be affected by factors other than changes in diagnostic tests. During the study period, there were three
major changes and one minor revision in the version of
the ICD used in Western Australian hospitals, as noted
above in Materials and Methods. These transitions did not
produce any obvious discontinuities in the coding of myocardial infarction, but they did affect the coding of unstable
angina pectoris from 1996 onward, with reciprocal declines
in cases coded as ‘‘other angina.’’ It is unlikely that this has
affected the coding of myocardial infarction.
The clinical terminology for acute coronary heart disease
has changed during the last decade. New insights into the
pathophysiology of myocardial infarction and unstable angina pectoris, together with common approaches to treatment, have led to the use of the broad clinical term ‘‘acute
coronary syndromes’’ (24). The extent to which these
changes in clinical nomenclature have influenced the coding
of myocardial infarction in hospitals is uncertain. Finally,
publicity advising persons with sudden chest pain to seek
immediate hospital care has undoubtedly increased the
numbers of people visiting emergency departments for chest
pain, possibly increasing admissions of persons with less
severe cases of myocardial infarction in many jurisdictions
(25). Nevertheless, previous studies in which the coding of
myocardial infarction in administrative data was validated
against myocardial infarction registers suggested that administrative data reflect the trends in myocardial infarction
reasonably well (5, 7, 8, 26).
The present study and the US study (10) suggest that the
validity of trends based on administrative data is now in
doubt and will remain so until the new diagnostic criteria
for myocardial infarction are fully accepted by clinicians
and there is stability in diagnostic thresholds for troponin
testing. Given the importance of monitoring rates of myocardial infarction for both public health and health-service
232 Sanfilippo et al.
purposes, we suggest that there is a need for further studies
to validate the coding of myocardial infarction in administrative data in different jurisdictions and to develop correction factors for adjusting rates of myocardial infarction and
other variants of coronary heart disease. Ideally, such studies should be based on a standardized international protocol
incorporating the new diagnostic criteria (27, 28).
ACKNOWLEDGMENTS
This research was funded by a National Health and
Medical Research Council of Australia project grant (grant
353671).
The authors thank the following people and institutions
for providing assistance during the study: Steve Ridout of
the School of Population Health, University of Western
Australia, for his work in merging the laboratory data with
records from the Linked Vascular File; Di Rosman and the
staff of the Health Information Linkage Branch, Western
Australian Department of Health, for extraction and linkage of
data that produced the research file (Linked Vascular File); Dr.
Chotoo Bhagat and Brett Richards of Pathwest Laboratory
Medicine for provision of laboratory data for teaching hospitals; and Frank Natalotto (St. John of God Pathology) and
Geraldine Ormonde (Western Diagnostic Pathology) for organizing the provision of laboratory data for private hospitals.
Conflict of interest: none declared.
REFERENCES
1. Mathur S. Epidemic of coronary heart disease and its treatment
in Australia. (Cardiovascular disease series no. 20). Canberra,
Australia: Australian Institute of Health and Welfare, 2002.
(AIHW catalogue no. CVD 21).
2. Salomaa V, Arstila M, Kaarsalo E, et al. Trends in the incidence of and mortality from coronary heart disease in Finland,
1983–1988. Am J Epidemiol 1992;136:1303–15.
3. Rosamond WD, Chambless LE, Folsom AR, et al. Trends in
the incidence of myocardial infarction and in mortality due to
coronary heart disease, 1987 to 1994. N Engl J Med 1998;339:
861–7.
4. Office of Cardiovascular Diseases, World Health Organization.
WHO MONICA Project—MONICA manual. Geneva, Switzerland: World Health Organization, 1999. (www.ktl.fi/
publications/monica/manual/).
5. Martin CA, Hobbs MS, Armstrong BK, et al. Trends in the
incidence of myocardial infarction in Western Australia between 1971 and 1982. Am J Epidemiol 1989;129:655–68.
6. Hammar N, Nerbrand C, Ahlmark G, et al. Identification of
cases of myocardial infarction: hospital discharge data and
mortality data compared to myocardial infarction community
registers. Int J Epidemiol 1991;20:114–20.
7. Mahonen M, Salomaa V, Brommels M, et al. The validity of
hospital discharge register data on coronary heart disease in
Finland. Eur J Epidemiol 1997;13:403–15.
8. Jamrozik K, Dobson AJ, Hobbs MS, et al. Monitoring the incidence of cardiovascular disease in Australia. (Cardiovascular
disease series no. 17). Canberra, Australia: Australian Institute
of Health and Welfare, 2001. (AIHW catalogue no. CVD 16).
9. Joint European Society of Cardiology/American College of
Cardiology Committee. Myocardial infarction redefined—
a consensus document of the Joint European Society of
Cardiology/American College of Cardiology Committee for
the Redefinition of Myocardial Infarction. Eur Heart J 2000;21:
1502–13.
10. Roger VL, Killian JM, Weston SA, et al. Redefinition of
myocardial infarction: prospective evaluation in the community. Circulation 2006;114:790–7.
11. Salomaa V, Koukkunen H, Ketonen M, et al. A new definition
for myocardial infarction: what difference does it make? Eur
Heart J 2005;26:1719–25.
12. Holman CD, Bass AJ, Rouse IL, et al. Population-based
linkage of health records in Western Australia: development of
a health services research linked database. Aust N Z J Public
Health 1999;23:453–9.
13. World Health Organization. International classification of
diseases: manual of the international statistical classification
of diseases, injuries, and causes of death. Ninth Revision.
Geneva, Switzerland: World Health Organization, 1977.
14. Australian Bureau of Statistics. Demography, Western
Australia, 2004 final. (ABS catalogue no. 3311.5.55.001).
Canberra, Australia: Australian Bureau of Statistics, 2006.
(www.abs.gov.au/ausstats/[email protected]/mf/3311.5.55.001).
15. Commission on Professional and Hospital Activities, National
Center for Health Statistics. International classification of
diseases, Ninth Revision, clinical modification: ICD-9-CM
annotated diseases. Ann Arbor, MI: Commission on Professional and Hospital Activities, 1986.
16. National Coding Centre, University of Sydney. The Australian
version of ICD-9-CM: tabular list of diseases. Sydney,
Australia: National Coding Centre, 1995.
17. National Centre for Classification in Health, University of
Sydney. ICD-10-AM tabular list of diseases. Sydney, Australia: National Centre for Classification in Health, 1999.
18. Australian Bureau of Statistics. Population by age and sex,
Australian states and territories, Jun 1997 to Jun 2002: 2001
census edition final. Canberra, Australia: Australian Bureau of
Statistics, 2003. (ABS catalogue no. 3201.0).
19. Australian Institute of Health and Welfare. Mortality over the
twentieth century in Australia: trends and patterns in major
causes of death. (Mortality surveillance series no. 4). Canberra, Australia: Australian Institute of Health and Welfare,
2006. (AIHW catalogue no. PHE73).
20. Meier MA, Al-Badr WH, Cooper JV, et al. The new definition
of myocardial infarction: diagnostic and prognostic implications in patients with acute coronary syndromes. Arch Intern
Med 2002;162:1585–9.
21. Thompson PL, Parsons RW, Jamrozik K, et al. Changing
patterns of medical treatment in acute myocardial infarction.
Observations from the Perth MONICA Project 1984–1990.
Med J Aust 1992;157:87–92.
22. McElduff P, Dobson A, Jamrozik K, et al. The WHO
MONICA Study in Australia, 1984–93: a summary of the
Newcastle and Perth MONICA Projects. (Cardiovascular
disease series no. 13). Canberra, Australia: Australian Institute
of Health and Welfare, 2000. (AIHW catalogue no. 11).
23. Tunstall-Pedoe H, Vanuzzo D, Hobbs M, et al. Estimation of
contribution of changes in coronary care to improving survival,
event rates, and coronary heart disease mortality across the WHO
MONICA Project populations. Lancet 2000;355:688–700.
24. Bertrand ME, Simoons ML, Fox KAA, et al. Management
of acute coronary syndromes in patients presenting without
persistent ST-segment elevation. Eur Heart J 2002;23:
1809–40.
Am J Epidemiol 2008;168:225–233
Troponin Tests and Myocardial Infarction Admission Rates
25. Clancy M. Chest pain units: evidence of their usefulness is
limited but encouraging. (Editorial). BMJ 2002;325:116–17.
26. Rosamond WD, Chambless LE, Sorlie PD, et al. Trends in the
sensitivity, positive predictive value, false-positive rate, and
comparability ratio of hospital discharge diagnosis codes for
acute myocardial infarction in four US communities, 1987–
2000. Am J Epidemiol 2004;160:1137–46.
27. Luepker RV, Apple FS, Christenson RH, et al. Case definitions
for acute coronary heart disease in epidemiology and clinical
Am J Epidemiol 2008;168:225–233
233
research studies: a statement from the AHA Council on Epidemiology and Prevention; AHA Statistics Committee; World
Heart Federation Council on Epidemiology and Prevention;
the European Society of Cardiology Working Group on Epidemiology and Prevention; Centers for Disease Control and
Prevention; and the National Heart, Lung, and Blood Institute.
Circulation 2003;108:2543–9.
28. Thygesen K, Alpert JS, White HD, et al. Universal definition
of myocardial infarction. Circulation 2007;116:2634–53.