Full document - International Hospital Federation

Innovation in clinical specialities: cardiovascular disease
Table 1: Major risk factors a in coronary heart disease mortality
Age groups (years)
Risk factor 25–44 45–54 55–64 65–74 75–84
Systolic blood pressure 7
Men (log hazard ratio per 1 mmHg) -0.036 -0.035 -0.032 -0.027 -0.021
Women (log hazard ratio per 1 mmHg)
-0.046 -0.046 -0.035 -0.032 -0.026
Total cholesterol 6
Men & women: mortality reduction per 1
mmol/l (38.6 mg/dl) change in total cholesterol 0.900 0.650 0.450 0.333 0.317
Log coefficient -1.2942 -0.8238 -0.5245 -0.3719 -0.3512
Body mass index (BMI) 8
Age-specific relative risks per 1 kg/m 2 1.04 1.03 1.02 1.01 1.01
c
Men & women: log coefficients
0.0363 0.0297 0.0165 0.0132 0.0099
a Used as input for the IMPACT model for the United States of America
Table 2: Relative risks a of smoking, diabetes and physical inactivity for coronary heart disease mortality
Men
Women
55 years >55 years ≤65 years > 65 years
Smoking 3·33 (2·80–3·95) b 2·52 (2·15–2·96) 4·49 (3·11–6·47) 2·14 (1·35–3·39)
Exercise 1·02 (0·83–1·25) b 0·79 (0·66–0·96) 0·74 (0·49–1·10) 0·75 (0·46–1·22)
Diabetes 2·66 (2·04–3·46) b 1·93 (1·58–2·37) 3·53 (2·49–5·01) 2·59 (1·78–3·78)
Source: INTERHEART study 27
a Used as input for the IMPACT model for the United States of America
b
Odds ratios for relative effect of ri sk factors with 99% confidence intervals
A variety of CHD policy models have been developed to
estimate the relative contributions and hence the population
impact of medical and public-health interventions. Good models
are able to integrate and simultaneously consider large amounts of
data on patient numbers, treatments and population risk factor
trends. 9–11 The CHD Policy Model developed in the USA was used
successfully to examined trends in that country between 1980 and
1990 9 and later demonstrated the potential advantages of a
population-based approach to prevention, 12 consistent with
European studies. 13-15
Capewell et al. have subsequently developed, refined and
applied the IMPACT model in a variety of populations. 10,16-19
Approximately 44% of the substantial CHD mortality decline in the
USA between 1980 and 2000 was attributable to changes in
major risk factors and 47% to specific cardiological treatments, 10
similar to findings in other industrialized countries. 16,17,19,20 Three
earlier analyses suggested that further modest reductions in major
risk factors could halve CHD deaths in the United Kingdom of
Great Britain and Northern Ireland. 15,21–23 To determine whether
similar gains may be possible for the population of the USA, or
whether they would be negated by recent dramatic rises in obesity
and diabetes, we used the previously calibrated IMPACT model 10
for the USA to estimate the number of CHD deaths that could
potentially be prevented or postponed in 2010, compared with the
number in 2000. The model was applied to three contrasting
scenarios, the most optimistic of which assumed that the
prevalence of risk factors in the entire population reaches the
levels already reported in its “low-risk stratum”. 24,25
Methods
The IMPACT model aims to explain changes in CHD mortality
rates in a population. It quantifies the contribution from temporal
trends in the major population
risk factors (smoking, high
systolic blood pressure,
elevated total cholesterol,
obesity, diabetes and physical
inactivity) and from medical
and surgical treatments given
to CHD patients. 10,17,26 The
model employs regression
coefficients produced by large
meta-analyses and cohort
studies. 6-8 Each coefficient
quantifies the independent (log linear) relationship between the
absolute change in a specific cardiovascular risk factor, such as
high systolic blood pressure or total cholesterol, and the
consequent change in population mortality rates from CHD (Table
1). 6-8 For each risk factor, the subsequent reduction in deaths in
2010 from base year 2000 could then be estimated as the product
of three variables:
Number fewer deaths = number of CHD deaths observed in
2000 × the risk factor reduction × the specific regression
coefficient exponentiated. 10,22,23
All the coefficients and relative risk values were obtained from
multivariate logistic regression analyses and were assumed to be
independent, having been adjusted for potential confounding from
the other major risk factors. The total deaths prevented could
therefore be summed. Independent regression coefficients were
not available for smoking, diabetes and physical inactivity. An
alternative methodology was therefore used, which involved
population-attributable risk fractions 9,21,22 calculated for the
INTERHEART study, a large, international multivariate analysis that
included data from the USA27 (Table 2).
The original 1980–2000 IMPACT model 10 was extended to 2010
using United States Census Bureau population projections and
mortality data for men and women aged 25–84 years. The number
of CHD deaths (ICD-10: I20-I25) expected in 2010 was then
calculated under three contrasting risk factor scenarios.
Risk factor scenarios
The first scenario assumed that recent risk factor trends would
continue to 2010. Using data from the Third National Health and
Nutrition Examination Survey (NHANES) 1988–1994, NHANES
1999– 2002, and the Behavioral Risk Factor Surveillance System
Hospital and Healthcare Innovation Book 2009/2010 69