1. International comparison of outcomes among survivors after acute myocardial infarction: real-world evidence from electronic health and administrative records
Professor Harry Hemingway
on behalf of the APOLLO investigators
European Society of Cardiology Congress 2014
Registry Hot Line Session: Atrial fibrillation and myocardial infarction
ORIGINAL ABSTRACT
Background
International comparisons have shown differences in acute and short-term outcomes in myocardial infarction (MI) patients, but many remain at elevated risk beyond this period. Multiple registries derived from electronic health records offer an innovative approach to understanding which factors have greatest impact on the prognosis of high risk patients and how this varies across different healthcare systems, with the potential to inform quality improvement initiatives.
Objective
To compare the risk of atherothrombotic events and hospitalized bleeding in patients with stable CAD across the USA, Sweden, England, and France.
Patient population and methods
We analysed linked electronic health records and disease registries (England, Sweden) and administrative data (US, France) according to common disease definitions based on admission ICD-10 codes and common analysis protocol. Eligible patients were those who had survived for 1 year following hospitalization for MI (‘post-MI survivors’) in [N=77,976 (Sweden), 53,909 (US; ≥65 years old patients), 7,238 (England), and 1,757 (France). Outcome predictors were identified and used in multivariate CoxPH models to estimate adjusted risks.
Outcomes
Individual and composites of MI, stroke, CVD death, all-cause mortality, and hospitalized bleeding.
Results
Compared to Swedish and English patients (mean age ~70), French patients were younger (mean age 66) and healthier, and US patients older (minimum age 65; mean age 79) with more comorbidities. Outcome predictors (prevalence range) were age (54-100% aged ≥65), male gender (51-68%), no PCI (38-58%) or CABG (83-94%) performed within 1 year post-MI, diabetes (21-35%), history of >1 MI (12-19%), stroke (7-10%), heart failure (21-45%), peripheral arterial disease (1-10%), renal disease (3-7%), prior hospitalization for bleeding (4-17%), atrial fibrillation (14-28%), COPD (8-28%), and cancer (7-12%), all of which had similar strong positive associations with the outcomes in each country (multivariate hazard ratios ranging from 1.5-3).
By year 3 the unadjusted (Kaplan-Meier) cumulative risk of MI, stroke, or death was highest for US patients (36.2% [95% CI, ]), followed by Sweden (26.9% [ ]), England (24.1% [ ]), and France (17.3% [ ]). After multiple adjustments these event rates were similar (US, 19.7% [ ]; Sweden, 20.0% [ ]; England, 19.1% [ ]; France, 16.7% [ ]).
The observed 3-year cumulative risk of hospitalized bleeding was >2-fold higher for patients in Medicare (7.1% [ ]), compared to patients in England (3.6% [ ]), Sweden (2.5% [ ]), and France (2.2% [ ]). These differences reduced but persisted after risk adjustments (3-year adjusted risk of bleeding; US, 4.7% [ ]; England, 3.3% [ ]; Sweden, 2.1% [ ]; France, 2.4% [ ]).
Conclusion
Risk of CV events remained high across the 3 years and across the 4 countries studied. The risk factors for predicting post-MI survivors with increased risk of subsequent cardiovascular and bleeding events are similar across different countries. Differences in observed risks of cardiovascular events or death among post-MI survivors in different countries are partially explained by differences in the distribution of risk factors and revascularization treatments, but differences in the risk of bleeding remain despite adjustments. This study highlights the importance of comorbidity management for this high risk population and exemplifies the power and generaliseability of prognosis research that utilizes the clinical records of different countries.

2. Authors
Eleni Rapsomaniki1, Magnus Janzon2, David J. Cohen3, Tomas Jernberg4, Nicholas Moore5, Marcus Thuresson6, Erru Yang7, Patrick Blin5, Saga Johansson8, Harry Hemingway1
1Farr Institute of Health Informatics Research, University College London, UK
2Linkoping University, Sweden
3Saint Luke’s Mid America Heart Institute, Kansas City, USA
4Karolinska University Hospital, Sweden
5Department of Pharmacoepidemiology, University of Bordeaux, France
6Statisticon AB, Uppsala, Sweden
7Health Economics & Epidemiology, Evidera, Lexington, USA
8Observational Research Center, AstraZeneca R&D, Mölndal, Sweden

3. Conflicts
AstraZeneca funded the APOLLO Programme alongside the PEGASUS-TIMI 54 study which is aimed at determining the clinical efficacy and safety of long-term dual antiplatelet therapy with ticagrelor plus aspirin for the prevention of secondary cardiovascular events in patients with a recent myocardial infarction and additional atherothrombotic risk factors

4. Motivation Importance Uncertainty Novel opportunity MI survivors
International comparisons
Uncertainty
Unselected populations
Long-term follow-up
Non-fatal and fatal
Benefits and harms
Novel opportunity
Electronic health and administrative records

5. Objective
To compare atherothrombotic events, death and bleeding risks in 1-year post-MI survivors across Sweden, the USA, England and France over 3 years of follow-up

6. Methods: electronic health records and administrative data sources in APOLLO Programme
Countries
National
registries
Nationwide
Longitudinal data
Hospital discharge data linked to prescribed data register and death registry
Sweden
Medicare
Age >65 years
Demographics and health insurance claims
Linked to death registry
USA
CPRD, MINAP, HES
Four linked datasets
Longitudinal data
Primary and secondary care, and disease registry and death registry
England
EGB,
PMSI
Sample of national healthcare insurance data
Hospital discharge data linked to death registry
France
Record sources
Details
The global real world evidence program comprises 5 studies conducted in different geographical regions reflecting different health care systems
CPRD, Clinical Practice Research Datalink; EGB, Echantillon Généraliste des Bénéficiaires; HES, Hospital Episodes Statistics; MINAP, Myocardial Ischaemia National Audit Project; PMSI, Programme de médicalisation des systèmes d'information

7. Methods: study population, disease definitions and statistics
Patients entered the study 1 year after the most recent discharge for MI (study period: 2002–2011)
Disease definitions were harmonised using ICD9/ICD10 diagnostic codes
Data from each country were analysed using a common protocol
Cox models were utilised to estimate adjusted risks and relative risks, using Sweden as reference
ICD, International Classification of Diseases

8. Age- and sex-standardised prevalence of comorbidities and secondary prevention treatments
Baseline comorbidities (%)
Treatments prescribed at 1 year post-MI (%)
Revascularisation (%)
Hypertension
History of heart failure
Diabetes
History of atrial fibrillation
History of >1 MI
History of cancer
History of stroke
History of COPD
History of hospitalised bleeding
History of renal disease
History of PAD
Sweden, n=77 798
USA, n=53 909
England, n=7238
France, n=1764 20 40 60 80
100
Percentage
PCI
CABG
Statins
b-blockers
ACEI/ARBs
Aspirin
ADP-receptor blockers
Dual antiplatelet
Vitamin K antagonists
54.9%
48.1%
38.7%
61.6%
14.1%
19.7%
11.4%
7.1%
9.8%
8.4%
8.9%
54.4%
41.0%
22.9%
65.0%
49.7%
26.4%
80.4%
86.7%
76.1%
73.5%
68.5%
81.3%
71.3%
80.1%
71.4%
80.0%
79.4%
82.3%
63.9%
79.0%
69.1%
68.7%
29.0%
40.7%
24.2%
29.6%
25.4%
38.5%
23.2%
28.0%
22.0%
20.8%
16.9%
14.4%
11.7%
12.2%
13.5%
0.4%
7.2%
10.5%
4.2%
7.8%
11.2%
5.4%
13.0%
7.9%
8.0%
15.2%
11.3%
6.4%
8.3%
3.7%
10.8%
28.6%
11.9%
9.1%
16.1%
7.5%
3.8%
ACEI, angiotensin-converting enzyme inhibitor; ADP, adenosine diphosphate; ARB, angiotensin receptor blocker; CABG, coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; MI, myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention

9. Prognostic validity: adjusted HRs of all-cause death
Sweden
USA
England
France
Overall mean
CABG, no vs yes
Age per 10 years
PCI, no vs yes
History of renal disease
History of heart failure
0.5 1 2 3 4 5 6 HR
Number of
events
% in study
15 233
22 498
659
222
12.6
16.9
10.4
5.6
55.8
42.9
41.8
65.0
5.0
3.4
7.0
6.9
27.7
45.0
21.0
23.2
95% CI
1.67
2.21
2.00
1.75
1.92
(1.58–1.78)
(2.12–2.31)
(1.42–2.83)
(0.95–3.22)
(1.54–2.40)
2.15
1.79
1.69
1.85
(2.11–2.20)
(1.75–1.82)
(1.55–1.84)
(1.57–2.04)
(1.63–2.11)
1.90
1.71
1.87
1.81
(1.83–1.98)
(1.65–1.76)
(1.50–2.32)
(1.39–2.46)
(1.66–1.98)
1.56
1.76
1.73
(1.72–1.92)
(1.50–1.62)
(1.57–2.34)
(1.22–2.54)
(1.53–1.96)
1.68
1.72
1.45
1.70
(1.63–1.74)
(1.67–1.77)
(1.32–1.85)
(1.07–1.06)
(1.66–1.74)
CABG, coronary artery bypass graft; CI, confidence interval; HR, hazard ratio; PCI, percutaneous coronary intervention

10. 3-year cumulative absolute risks
All-cause death
MI/stroke/all-cause death
Follow-up (years)
Observed risk (%)
Adjusted risk (%) 50 40 30 20 10
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Sweden 20.1 (19.7–20.4)
USA 30.2 (29.8–30.7)
England 13.7 (12.6–14.8)
France 14.3 (12.5–16.1)
Sweden 26.9 (26.5–27.2)
USA 36.2 (35.7–36.6)
England 24.1 (22.7–25.5)
France 17.9 (16.0–19.8) 15 5 25
Sweden 11.2 (10.9–11.5)
USA 12.8 (12.3–13.4)
England 8.7 (6.9–10.5)
France 12.4 (10.1–14.7)
Sweden 19.8 (19.4–20.2)
USA 18.2 (17.6–18.9)
England 21.3 (18.2–24.2)
France 16.7 (14.3–19.2)
Shaded areas correspond to 95% confidence intervals
MI, myocardial infarction

11. Relative risks vs SwedenRR
All-cause death
MI/stroke/ all-cause death
1.55
95% CI
1.46
1.11
1.17
0.77
1.04
0.90
0.69
1.14
1.09
(1.50–1.61)
(1.39–1.53)
(1.05–1.18)
(1.11–1.24)
(0.66–0.90)
(0.86–1.26)
(0.85–1.28)
(0.73–1.12)
(0.56–0.84)
(0.93–1.41)
(0.88–1.39)
(0.86–1.37)
1.33
0.83
0.88
0.89
1.12
0.60
0.82
0.78
(1.29–1.37)
(1.05–1.13)
(0.80–0.87)
(0.84–0.92)
(0.79–1.00)
(0.98–1.28)
(0.98–1.30)
(0.90–1.20)
(0.50–0.72)
(0.70–0.99)
(0.68–0.98)
(0.64–0.94)
0.75 1
1.5
Unadjusted (KM)
Age and sex
+ comorbiditiesa
+ PCI/CABG
USA
England
France
aComorbidities adjusted for history of >1 MI, hypertension, renal disease, heart failure, PAD, stroke, atrial fibrillation, hospitalised bleeding, cancer and COPD. French group adjusted only for age, sex, and year of index MI. All models were additionally adjusted for year of index MI
CABG, coronary artery bypass graft; CI, confidence interval; COPD, chronic obstructive pulmonary disease; KM, Kaplan–Meier; MI, myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention; RR, relative risk

12. 3-year cumulative risks of hospitalised bleeding events
Adjusted risk (%)
Follow-up (years)
Observed risk (%)
RR (vs Sweden) RR
0.0 6
0.5
1.0
1.5
2.0
2.5
3.0 5 4 3 2 1
Sweden 2.5 (2.3–2.6)
USA 5.3 (5.1–5.5)
England 3.6 (2.9–4.3)
France 2.2 (1.4–3.0)
Sweden 2.0 (1.9–2.1)
USA 3.6 (3.2–4.0)
England 4.9 (2.7–7.0)
France 2.2 (1.5–3.4)
95% CI
Unadjusted (KM)
Age and sex
+ comorbiditiesa
+ PCI/CABG
USA
2.12
2.14
1.64
1.69
(1.92–2.34)
(1.92–2.38)
(1.45–1.84)
(1.51–1.90)
England
1.35
2.07
1.94
(0.91–2.01)
(1.40–3.07)
(1.28–2.93)
France
0.75
0.99
1.06
(0.42–1.33)
(0.58–1.69)
(0.62–1.81)
Shaded areas correspond to 95% CIs
aComorbidities adjusted for history of >1 MI, hypertension, renal disease, heart failure, PAD, stroke, atrial fibrillation, hospitalised bleeding, cancer and COPD. French group adjusted only for age, sex, and year of index MI. All models additionally adjusted for year of index MI
CABG, coronary artery bypass graft; CI, confidence interval; COPD, chronic obstructive pulmonary disease; KM, Kaplan–Meier; MI, myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention; RR, relative risk

13. Limitations Medication information Data on cause-specific mortality
Lacking for USA
Data on cause-specific mortality
Lacking for USA, France
Socioeconomic data
Lacking for Sweden, USA, France
Patient age
USA >65 years only

14. Main results summary
Among year post-MI survivors drawn from unselected electronic health and administrative records populations:
High-risk state (>3% annual all-cause death risk1,2)
About half of deaths are non-cardiovascular
Compared with European populations, US patients had
higher (age and sex-standardised) prevalence of comorbidities
higher adjusted all-cause mortality
Risk of further MI, stroke or death remained high (about 1 in 5) across the 3 years and across the 4 countries studied, with fairly constant annual risks
Difference in risk of hospitalised bleeding in the USA and England vs Sweden remained substantial
1Montalescot G et al. Eur Heart J 2013;34:2949–3003
2Fihn SD et al. J Am Coll Cardiol 2012;60:e44–e164

15. Clinical implications
Guidelines1,2
Definition of ‘high risk’ needs to be considered
Policy
Impetus to improve quality of healthcare systems
Primary care
Need for a generalist approach
Evidence for regulators and clinicians
New interventions and generalisability of trial results
National electronic health record resources
Quality, scope and comparability
1Montalescot G et al. Eur Heart J 2013;34:2949–3003
2Fihn SD et al. J Am Coll Cardiol 2012;60:e44–e164

16. Acknowledgements
Editorial support was provided by Oxford PharmaGenesis™ Ltd