European Heart Journal: Acute Cardiovascular Care http://acc.sagepub.com/
GRACE score predicts heart failure admission following acute coronary syndrome David A McAllister, Nynke Halbesma, Kathryn Carruthers, Martin Denvir and Keith A Fox European Heart Journal: Acute Cardiovascular Care published online 1 July 2014 DOI: 10.1177/2048872614542724 The online version of this article can be found at: http://acc.sagepub.com/content/early/2014/06/27/2048872614542724
Published by: European Society of Cardiology
ESC Working Group on Acute Cardiac Care
and http://www.sagepublications.com
Additional services and information for European Heart Journal: Acute Cardiovascular Care can be found at: Email Alerts: http://acc.sagepub.com/cgi/alerts Subscriptions: http://acc.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav
>> OnlineFirst Version of Record - Jul 1, 2014 What is This?
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
542724 research-article2014
ACC0010.1177/2048872614542724European Heart Journal: Acute Cardiovascular CareMcAllister et al.
EUROPEAN SOCIETY OF CARDIOLOGY ®
Original scientific paper
GRACE score predicts heart failure admission following acute coronary syndrome
European Heart Journal: Acute Cardiovascular Care 1–7 © The European Society of Cardiology 2014 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/2048872614542724 acc.sagepub.com
David A McAllister1, Nynke Halbesma1, Kathryn Carruthers2, Martin Denvir2 and Keith A Fox2
Abstract Background: Congestive heart failure (CHF) is a common and preventable complication of acute coronary syndrome (ACS). Nevertheless, ACS risk scores have not been shown to predict CHF risk. We investigated whether the atdischarge Global Registry of Acute Coronary Events (GRACE) score predicts heart failure admission following ACS. Methods and Results: Five-year mortality and hospitalization data were obtained for patients admitted with ACS from June 1999 to September 2009 to a single centre of the GRACE registry. CHF was defined as any admission assigned WHO International Classification of Diseases 10 diagnostic code I50. The hazard ratio (HR) for CHF according to GRACE score was estimated in Cox models adjusting for age, gender and the presence of CHF on index admission. Among 1,956 patients, CHF was recorded on index admission in 141 patients (7%), and 243 (12%) were admitted with CHF over 3.8 median years of follow-up. Compared to the lowest quintile, patients in the highest GRACE score quintile had more CHF admissions (116 vs 17) and a shorter time to first admission (1.2 vs 2.0 years, HR 9.87, 95% CI 5.93–16.43). Per standard deviation increment in GRACE score, the instantaneous risk was more than two-fold higher (HR 2.28; 95% CI 2.02–2.57), including after adjustment for CHF on index admission, age and gender (HR 2.49; 95% CI 2.06–3.02). The C-statistic for CHF admission at 1-year was 0.74 (95% CI 0.70–0.79). Conclusions: The GRACE score predicts CHF admission, and may therefore be used to target ACS patients at high risk of CHF with clinical monitoring and therapies. Keywords Heart failure, myocardial infarction, prognosis Received: 9 Jan 2014; accepted: 18 Jun 2014
Introduction There were 4.2 million emergency department attendances for chronic heart failure in the United States in 2010.1 In Europe approximately 5% of all hospital admissions are related, and it accounts for approximately 2% of all health care-related costs.2 Improved survival following myocardial infarction (MI) and an ageing population are expected to contribute to a 25% increase in the prevalence of heart failure by 2013.3 This expanding healthcare burden of heart failure poses a significant challenge for clinicians and health services. As the commonest cause of heart failure in developed nations, coronary heart disease represents a key opportunity to reduce the impact of heart failure, particularly in the
post-MI setting where heart failure is both common and associated with increased mortality,4,5 and where there is strong evidence that specific drug therapies reduce heart failure related events in the years following discharge from hospital.6–8
1Centre 2Centre
for Population Health Sciences, University of Edinburgh, UK for Cardiovascular Science, University of Edinburgh, UK
Corresponding author: David A McAllister, Centre for Population Health Sciences, Medical School, Teviot Place, London EH8 9AG, UK. Email: [email protected]
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
2
European Heart Journal: Acute Cardiovascular Care
International guidance recommends the use of a validated prognostic score among patients with acute coronary syndrome (ACS) in order to identify those most likely to benefit from more intensive management.9 No published models, however, have been shown to predict the risk of heart failure re-admission in patients with MI or ACS. Consequently, it is currently difficult to predict which patients are at highest risk of heart failure and are therefore most likely to benefit from interventions that are known to prevent CHF complications, such as ACE inhibitors, beta blockers and aldosterone receptor antagonists. Among a number of prognostic scores developed to predict outcomes following acute coronary syndrome,10–21 the Global registry of Acute Coronary Events Risk Score (GRACE Score) has emerged as one of the most widely used.22,23 Therefore, we examined whether the at-discharge GRACE Score could be used to predict CHF events following acute coronary syndrome to allow optimal targeting of clinical monitoring and CHF therapies in this high risk group of patients.
Methods Five-year mortality and hospitalization data were obtained for patients recruited at the Scottish centre of the GRACE registry from 28 June 1999 to 6 September 2009 via record linkage to routine healthcare data. A detailed description of the GRACE protocol has been published previously.24 Briefly, eligibility criteria for GRACE included aged 18 or older, admission for presumptive acute coronary syndrome and one or more of: ECG changes consistent with ACS, serial increases in serum biochemical markers of cardiac necrosis, and/or documentation of coronary artery disease. Patients transferred into or out of the registry hospital were enrolled regardless of the time spent at the transferring hospital. All patients consented to all study protocols and procedures and the study was reviewed by Lothian Research Ethics Committee (LREC/1999/4/61) (UK/EU equivalent to Institutional Review Boards). For patients transferred out of a registry hospital, data collection for the initial case report form ended with the transfer and indication of purpose transfer. Patients hospitalized for less than one day who died were enrolled, provided that the cause of death was confirmed to be due to acute coronary syndrome. Standard clinical definitions of diagnostic characteristics and outcomes during hospitalization were used. All cases were assigned to STEMI, NSTEMI, unstable angina, and other cardiac/non-cardiac diagnoses via independent review of clinical presentation, ECG findings and the results of serum biochemical markers of myocardial necrosis. Specifically, unstable angina was defined as ACS with normal biochemical markers of necrosis. Mortality and hospitalization data during follow-up were obtained for all patients via deterministic linkage to a
national database the Scottish Morbidity Record of hospitalization and death data (SMR01). In accordance with a decision by the relevant privacy body (NHS Information Services Division Privacy Advisory Committee), follow-up was for five years following discharge from hospital. Otherwise, events were censored on 30 January 2010. CHF admission as defined as any admission (not including the index admission) assigned the Tenth International Classification of Diagnosis (ICD-10) code I50 in diagnostic position 1 or 2 (of up to 6 available for coders), MI was defined as any admission (not including the index admission) assigned the code I21 or I22. Diagnostic coding accuracy for hospital discharges is high within Scotland, with 95% accuracy for the main condition, and over 90% accuracy for additional conditions.25 GRACE Score was calculated (without re-weighting) using the previously described ‘At Discharge’ GRACE tool (www.outcomes-umassmed.org/grace/acs_risk/acs_risk_ content.html) based on age, history of MI, history of CHF, pulse rate at presentation, systolic blood pressure at presentation, initial serum creatinine level, initial serum cardiac biomarker levels, ST-segment depression on presenting electrocardiogram, and percutaneous coronary intervention performed while in hospital.22 In total, 58 patients (2.8%) had missing characteristics required for the calculation with serum creatinine being most frequently missing (39 patients) followed by heart rate (30 patients). We assigned missing categorical variables to null and used the mean value for continuous variables. This prognostic model was originally developed to predict the risk of death within 6 months of discharge from hospital. This GRACE discharge score was chosen for practical reasons since it represents an important time-point in the assessment of a patient in the post-MI setting representing a key opportunity for careful assessment and optimization of discharge medications likely to influence future CHF outcomes.
Statistical analysis Summary statistics were obtained for patient characteristics by quintile of GRACE Score. Cause-specific hazard ratios (HR) for heart failure admission were estimated according to GRACE score in Cox regression models treating death as a competing cause, adjusting for age, gender and the presence of CHF on the index admission. Heterogeneity from type of event (STEMI versus other) was examined using an interaction term. The cumulative incidence (risk) of admission for CHF, CHF and/or death, recurrent MI, recurrent MI and/or death, and death at 1-year follow-up was estimated in logistic regression models. Patients were grouped into vigintiles (twentieths) according to GRACE score and the proportion experiencing CHF and MI at 1 year was plotted against the risk obtained from the logistic regression models. Departure from linearity was explored via polynomial terms.
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
3
McAllister et al. Table 1. Baseline characteristics total population.
Number GRACE score, median GRACE score, range Age (years), mean (SD) Male sex Smoking Past medical history Myocardial infarction CHF Angina Stroke/transient ischemic attack Peripheral artery disease Diabetes Hypertension Event type STEMI Non-STEMI Unstable angina Killip classification 1 2 3 4
1st quintile
2nd quintile
3rd quintile
4th quintile
5th quintile
392 2.9 1.5–3.23 50.2 (7.3) 76 74.2
391 3.5 3.23–3.78 59.4 (7.2) 68 72.1
391 4 3.78–4.31 66.1 (6.8) 65 71.4
391 4.6 4.31–4.9 71.9 (6.7) 63.7 67.4
391 5.4 4.9–8.01 78.1 (7.5) 58.8 60.9 56.3 32.2 67.5 16.1 12.8 16.6 51.9 25.6 43.2 21.0 51.7 40.1 7.4 0
15.6 0 37 4.6 4.1 10.5 33.4
22.5 0.8 45.8 5.6 4.6 15.1 39.6
33.6 3.3 51 8.7 6.9 14.4 47.7
39.1 5.9 52.4 10.5 11.3 15.9 46.8
31.9 20.7 34.2
35.5 22.3 33.8
34.0 25.1 32.2
28.4 34.0 31.2
90.3 9.2 0.5 0
75.2 22.8 1.8 0
70.1 27.1 2.3 0.5
68.5 27.6 3.8 0
Values are percentages except where indicated.
Area under the curve for a single model, and comparisons of area under the curve (C-statistic) were both calculated in the R-package pROC via bootstrapping.26 The categorical net reclassification index (NRI) was calculated using pre-specified cut-offs of 1%, 5%, 10% and 20% 1-year risk of CHF admission. The NRI, a measure of reclassification following the addition of a new predictor, is the percentage of patients assigned to a less appropriate risk category (cases placed in a lower risk category or controls placed in a higher risk category) minus the percentage of patients assigned to a more appropriate risk category (cases placed in a higher risk category or controls placed in a lower risk category).27 Bootstrap methods were also used to obtain 95% confidence intervals for the NRI. For a review of these standard methods, see Cook.28 We conducted sensitivity analyses following exclusion of patients who developed CHF during the index admission, and excluding events where MI was recorded alongside CHF. Following reviewer comments we also report the associations in patients with and without ST-elevation MI and according to severity of left ventricular ejection fraction (LVEF) impairment (none, mild, moderate and severe) and index heart failure admission, and determine whether GRACE score continues to predict heart failure admission after adjusting for LVEF. All analyses were carried out in R version 3.0.1 (R project for statistical computing, Vienna, Austria).
Results In total, 1,956 patients were admitted from 28 June 1999 to 6 September 2009. The mean age was 66 years, and 66% were male. In addition to a higher prevalence of older patients and those in Killip class 3 and 4, those with higher GRACE scores were more likely to be female, and had a higher prevalence of known coronary, cerebrovascular and peripheral vascular disease as well as hypertension and diabetes (Table 1). CHF was recorded on the index admission in 141 (7%) patients, and 243 (12%) were admitted with CHF over 3.8 median years of follow-up. Compared to the lowest quintile for GRACE score, patients in the highest GRACE score quintile had more CHF admissions (116 vs 17) and a lower time to first CHF admission (1.2 vs 2.0 years, HR 9.87, 95% CI 5.93–16.43). Per standard deviation increment in GRACE score, the instantaneous risk of admission with CHF was more than two-fold higher (HR 2.28; 95% CI 2.02–2.57). This association persisted following adjustment for easily recorded patient characteristics CHF on index admission, age and gender (HR 2.49; 95% CI 2.06– 3.02, Table 2). The risk of admission with CHF was 0.84-fold lower in patients with ST-elevation MI than those without, but the confidence interval was compatible with either a higher or lower risk (95% CI 0.62–1.14) and the p value was not statistically
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
4
European Heart Journal: Acute Cardiovascular Care
Table 2. Events and hazard ratios per quintile of GRACE score.
Heart failure admissions at, n 30 days 90 days 1 year 5 years Follow-up time (days), median Cox models (n = 1956, 243 events), HR (95% confidence interval) GRACE score GRACE score + age GRACE score + age + CHF on index admission
1st quintile
2nd quintile
3rd quintile
4th quintile
5th quintile
0 2 7 17 1800
3 3 8 19 1772
2 8 17 41 1699
6 11 24 50 1595
22 42 62 116 1020
1 (reference) 1 (reference) 1 (reference)
1.16 (0.60–2.23) 1.28 (0.65–2.50) 1.24 0.63–2.43)
2.57 (1.46–4.52) 3.04 (1.62–5.70) 2.96 (1.58–5.56)
3.22 (1.86–5.60) 4.04 (2.07–7.86) 3.95 (2.02–7.69)
9.87 (5.93–16.43) 13.16 (6.52–26.57) 11.97 (5.90–24.30)
25
Risk of event (%)
20
15
Outcome Heart failure at 12 months Myocardial infarction at 12 months Predicted death at 6 months
10
5
0 3
4 5 Grace Score (at discharge)
6
Figure 1. Risk of CHF admission, and readmission with MI, according to the at-discharge-GRACE score, at one-year following discharge.
Solid lines represent the predicted risk obtained from logistic regression models of each outcome on GRACE score. Dashed line indicates predicted mortality at 6 months post discharge according to GRACE score. Points represent the proportion within each vigintile according to GRACE score (n = 97–98 in each) who had the outcome of interest at 1 year (Y-axis) at the mean GRACE score for that vigintile (X-axis).
significant (p = 0.26) so this finding should be interpreted cautiously. The GRACE score was similarly associated with the risk of admission with CHF in patients with and without ST-elevation MI (HR 2.24 and HR 2.33, respectively, HR for interaction 1.03, 95% CI 0.78–1.39, p for interaction = 0.80). Discrimination for CHF admission within 1 year (C-statistic 0.74; 95% CI 0.70–0.79) and for CHF admission or death from any cause within 1 year (0.70; 95% CI 0.66–0.75) were slightly higher than the equivalent statistics for MI (both C-statistics 0.65; 95% CI 0.60–0.70). Adding GRACE score to age and CHF on index admission improved the C-statistic from 0.69 (95% CI 0.64–0.74) to
0.76 (95% CI 0.71–0.80) and this difference was statistically significant (p = 0.003). The percentage of patients within each vigintile of GRACE Score corresponded to that predicted by the model for both CHF and MI. The predicted 1-year risk of CHF admission was similar to the predicted 1-year risk of MI admission. Importantly, and unexpectedly, the 12-month risk of CHF closely parallels the risk of 6-month postdischarge death estimated using the original at-discharge GRACE score risk calculator (Figure 1). When GRACE score was added to age and CHF on index admission, the NRI (using pre-specified cut-offs of
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
5
McAllister et al. 1%, 5%, 10% and 20%) was 13.6% (95% CI −5.1–26.3) for cases, 11.8% (4.7%–18.5%) for controls and 25.3% (95% CI 11.8–42.2%, p < 0.001) overall. The number and percentage assigned to each category with and without the GRACE score is provided in the online supplement. In sensitivity analyses, similar results were obtained after excluding the 141 patients with CHF on the index admission, and after excluding the 17 events where MI was recorded alongside CHF as a cause of hospital admission. There was no evidence of departure from linearity in logistic regression models at 1 and/or 5 years of follow-up (p = 0.11 and p = 0.91, respectively). LVEF was recorded in a subset of 684 patients, of whom the number (%) with none, mild, moderate or severe LVEF impairment was 310 (46.8%), 172 (27.0%), 121 (20.5%) and 27 (5.7%) respectively. The proportion that developed heart failure admission within 1 year was 3.1%, 7.0%, 13.6% and 30.8% for none, mild, moderate and severe, respectively, and compared to those with normal LVEF, those with mild, moderate and severe LVEF had higher rates of admission with heart failure (HR 1.81, 3.12 and 6.20, respectively, p < 0.001). Nonetheless, the association between GRACE score and subsequent heart failure admission was not modified by adding LVEF grade (HR 2.01 per SD, 95% CI 1.66–2.43 versus HR 1.92 per SD, 95% CI 1.57–2.34).
Discussion Although developed for predicting recurrent MI and death rather than CHF admissions, the GRACE Score is a good predictor of CHF admissions, and improved prediction when added to a simple model comprising the presence of CHF on index admission and age. The absolute risk of CHF admission closely tracked the absolute risk of MI readmission across the range of GRACE scores. The proportion of patients who experienced a subsequent heart failure admission even among those with mild LVEF was 7%, more than half of that for people with moderate LVEF. Moreover, the GRACE score continued to predict heart failure admission even after adjusting for LVEF impairment. Consequently, GRACE score offers additional information over using LVEF alone for predicting subsequent risk of heart failure. The original GRACE at-discharge Score was designed to be readily usable at the bedside, and allows clinicians to calculate the risk of death at 6 months using a paper-based point system, an online calculator (http://www.outcomesumassmed.org/grace/acs_risk/acs_risk_content.html) and mobile apps. Since the 12-month risk of CHF closely parallels the 6-month risk of all-cause mortality, each of these GRACE tools can now be readily used to predict the CHF risk. The risk of heart failure readmission according to the GRACE score was similar for patients with and without
ST-elevation MI, and having an ST-elevation MI did not predict increased risk. Our findings are consistent with a recently published report indicating that among 21,516 patients with acute coronary syndrome who were discharged alive without heart failure, the proportion who developed incident heart failure (defined on outpatient clinic visits and/or hospitalization) was not related to type of MI (12.0% for ST-elevation MI, 12.8% for non-ST elevation MI and 11.4% for unstable angina).29 A large number of other prognostic scores have been developed to evaluate outcomes following acute coronary syndrome and/or ST elevation MI,10–21 and these have similar discrimination to GRACE score for all-cause mortality.30 Consequently, it is likely that a number of these will also predict CHF risk. Future studies comparing the performance of these tools with the GRACE score would be of interest, particularly for the most widely used alternative, TIMI score.14 Unfortunately, all variables required to calculate the TIMI score were not available in our GRACE register cohort. A new prediction model developed specifically for CHF admission following ACS may perform better than the GRACE score, which was originally developed for a different outcome. The GRACE score is recommended in current international guidance for management of ACS,9 it is in widespread use and it therefore represents a readily available tool to help flag patients at high-risk of CHF events. Therefore, a substantial improvement over the level of prediction offered by the GRACE score would be required to justify the additional costs and practical challenges of introducing a novel score.
Limitations We examined the performance of GRACE score for evaluating CHF in a single centre, rather than in a multi-centre international study. Over-fitting to our data is unlikely to be a problem as our analysis did not involve the development of a new model. Nonetheless, calibration of GRACE score for risk of CHF admission would be useful for implementing the GRACE score in other settings. We used routine data to estimate the risk of CHF rather than prospective data collection. However, previously published international findings suggest that CHF is reliably recorded in routine data systems,29 whereas local auditing of data quality within our centre has consistently demonstrated 95% accuracy for ICD-10 diagnostic coding.25 Moreover, in sensitivity analyses, we excluded admissions from CHF where MI was concurrently recorded and found similar associations as in the main analysis.
Conclusion In summary, the risk of admission to hospital with CHF is similar to the risk of readmission to hospital with MI among
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
6
European Heart Journal: Acute Cardiovascular Care
patients with ACS, and closely tracks the risk of death within 6 months as predicted using the GRACE Score risk calculator. This extends the utility of the GRACE score to CHF events and could provide added incentive for clinical teams to optimize pharmacological and device treatments at discharge and ensure close clinical monitoring for these patients in the ensuing weeks and months following discharge. Acknowledgements The authors would like to thank the Information and Services Division, NHS National Services Scotland and the British Heart Foundation.
Disclosures DM has no conflicts of interest. NH has no conflicts of interest. MD has no conflict of interest. KAF reports grants from Astra Zeneca, grants from Sanofi Aventis, during the conduct of the study; grants and personal fees from Astra Zeneca, grants and personal fees from Bayer/Janssen, grants and personal fees from Lilly, outside the submitted work.
Funding This research was supported by a peer reviewed grant awarded by the Chief Scientist Office, Scotland (KAAF, KC) and additional support from the British Heart Foundation (KAAF). The original data collection was supported by an unrestricted grant from Sanofi-Aventis, Paris, France, with additional support from Astra Zeneca. The current analysis was unfunded.
References 1. NAMCS and NHAMCS Web Tables, http://www.cdc.gov/ nchs/ahcd/web_tables.htm#2010 (accessed 10 June 2013). 2. Braunschweig F, Cowie MR and Auricchio A. What are the costs of heart failure? Europace 2011; 13: ii13–ii17. 3. Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics—2013 update: a report from the American heart association. Circulation 2013; 127: e6–e245. 4. Steg PG, Dabbous OH, Feldman LJ, et al. Determinants and prognostic impact of heart failure complicating acute coronary syndromes observations from the Global Registry of Acute Coronary Events (GRACE). Circulation 2004; 109: 494–499. 5. Ezekowitz JA, Kaul P, Bakal JA, et al. Declining in-hospital mortality and increasing heart failure incidence in elderly patients with first myocardial infarction. J Am Coll Cardiol 2009; 53: 13–20. 6. Pitt B, White H, Nicolau J, et al., and EPHESUS Investigators. Eplerenone reduces mortality 30 days after randomization following acute myocardial infarction in patients with left ventricular systolic dysfunction and heart failure. J Am Coll Cardiol 2005; 46: 425–431. 7. Dargie HJ. Effect of carvedilol on outcome after myocardial infarction in patients with left-ventricular dysfunction: the CAPRICORN randomised trial. Lancet 2001; 357: 1385–1390.
8. Hall AS, Murray GD and Ball SG. Follow-up study of patients randomly allocated ramipril or placebo for heart failure after acute myocardial infarction: AIRE Extension (AIREX) Study. Acute Infarction Ramipril Efficacy. Lancet 1997; 349: 1493–1497. 9. Hamm CW, Bassand J-P, Agewall S, et al., and ESC Committee for Practice Guidelines. ESC Guidelines for the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation: the task force for the management of acute coronary syndromes (ACS) in patients presenting without persistent ST-segment elevation of the European Society of Cardiology (ESC). Eur Heart J 2011; 32: 2999–3054. 10. Lee KL, Woodlief LH, Topol EJ, et al. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction. Results from an international trial of 41,021 patients. GUSTO-I Investigators. Circulation 1995; 91: 1659–1668. 11. Savonitto S, Ardissino D, Granger CB, et al. Prognostic value of the admission electrocardiogram in acute coronary syndromes. JAMA J Am Med Assoc 1999; 281: 707–713. 12. Krumholz HM, Chen J, Wang Y, et al. Comparing AMI mortality among hospitals in patients 65 years of age and older: evaluating methods of risk adjustment. Circulation 1999; 99: 2986–2992. 13. Jacobs DR Jr, Kroenke C, Crow R, et al. PREDICT: a simple risk score for clinical severity and long-term prognosis after hospitalization for acute myocardial infarction or unstable angina: the Minnesota heart survey. Circulation 1999; 100: 599–607. 14. Antman EM, Cohen M, Bernink PJ, et al. The TIMI risk score for unstable angina/non-ST elevation MI: a method for prognostication and therapeutic decision making. JAMA J Am Med Assoc 2000; 284: 835–842. 15. Lindahl B, Toss H, Siegbahn A, et al. Markers of myocardial damage and inflammation in relation to long-term mortality in unstable coronary artery disease. FRISC Study Group. Fragmin during instability in coronary artery disease. N Engl J Med 2000; 343: 1139–1147. 16. Morrow DA, Antman EM, Charlesworth A, et al. TIMI risk score for ST-elevation myocardial infarction: a convenient, bedside, clinical score for risk assessment at presentation: An intravenous nPA for treatment of infarcting myocardium early II trial substudy. Circulation 2000; 102: 2031–2037. 17. Morrow DA, Antman EM, Giugliano RP, et al. A simple risk index for rapid initial triage of patients with ST-elevation myocardial infarction: an InTIME II substudy. Lancet 2001; 358: 1571–1575. 18. Newby LK, Bhapkar MV, White HD, et al., and SYMPHONY and 2nd SYMPHONY Investigators. Predictors of 90-day outcome in patients stabilized after acute coronary syndromes. Eur Heart J 2003; 24: 172–181. 19. Kurz DJ, Bernstein A, Hunt K, et al. Simple point-of-care risk stratification in acute coronary syndromes: the AMIS model. Heart Br Card Soc 2009; 95: 662–668. 20. Mahaffey KW, Yang Q, Pieper KS, et al., and SYNERGY Trial Investigators. Prediction of one-year survival in highrisk patients with acute coronary syndromes: results from the SYNERGY trial. J Gen Intern Med 2008; 23: 310–316.
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
7
McAllister et al. 21. Poçi D, Hartford M, Karlsson T, et al. Role of the CHADS2 score in acute coronary syndromes: risk of subsequent death or stroke in patients with and without atrial fibrillation. Chest 2012; 141: 1431–1440. 22. Eagle KA, Lim MJ, Dabbous OH, et al., and GRACE Investigators. A validated prediction model for all forms of acute coronary syndrome: estimating the risk of 6-month postdischarge death in an international registry. JAMA J Am Med Assoc 2004; 291: 2727–2733. 23. Fox KAA, Dabbous OH, Goldberg RJ, et al. Prediction of risk of death and myocardial infarction in the six months after presentation with acute coronary syndrome: prospective multinational observational study (GRACE). BMJ 2006; 333: 1091. 24. Fox KA, Eagle KA, Gore JM, et al. The Global Registry of Acute Coronary Events, 1999 to 2009–GRACE. Heart 2010; 96: 1095–1101. 25. NHS ISD Scotland. ISD Services | Data Quality Assurance | ISD Scotland, May 2012. http://www.isdscotland.org/
26.
27.
28. 29.
30.
Products-and-Services/Data-Quality/ (accessed 23 June 2014). Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011; 12: 77. doi:10.1186/1471-210512-77 www.biomedcentral.com/1471-2105/12/77 Pencina MJ, D’Agostino RB, D’Agostino RB and Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008; 27: 157–172. Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 2007; 115: 928–935. Kaul P, Ezekowitz JA, Armstrong PW, et al. Incidence of heart failure and mortality after acute coronary syndromes. Am Heart J 2013; 165: 379–385.e2. Gale CP, Manda SOM, Weston CF, et al. Evaluation of risk scores for risk stratification of acute coronary syndromes in the Myocardial Infarction National Audit Project (MINAP) database. Heart 2009; 95: 221–227.
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
GRACE score predicts heart failure admission following acute coronary syndrome David A McAllister, Nynke Halbesma, Kathryn Carruthers, Martin Denvir and Keith A Fox European Heart Journal: Acute Cardiovascular Care published online 1 July 2014 DOI: 10.1177/2048872614542724 The online version of this article can be found at: http://acc.sagepub.com/content/early/2014/06/27/2048872614542724
Published by: European Society of Cardiology
ESC Working Group on Acute Cardiac Care
and http://www.sagepublications.com
Additional services and information for European Heart Journal: Acute Cardiovascular Care can be found at: Email Alerts: http://acc.sagepub.com/cgi/alerts Subscriptions: http://acc.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav
>> OnlineFirst Version of Record - Jul 1, 2014 What is This?
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
542724 research-article2014
ACC0010.1177/2048872614542724European Heart Journal: Acute Cardiovascular CareMcAllister et al.
EUROPEAN SOCIETY OF CARDIOLOGY ®
Original scientific paper
GRACE score predicts heart failure admission following acute coronary syndrome
European Heart Journal: Acute Cardiovascular Care 1–7 © The European Society of Cardiology 2014 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/2048872614542724 acc.sagepub.com
David A McAllister1, Nynke Halbesma1, Kathryn Carruthers2, Martin Denvir2 and Keith A Fox2
Abstract Background: Congestive heart failure (CHF) is a common and preventable complication of acute coronary syndrome (ACS). Nevertheless, ACS risk scores have not been shown to predict CHF risk. We investigated whether the atdischarge Global Registry of Acute Coronary Events (GRACE) score predicts heart failure admission following ACS. Methods and Results: Five-year mortality and hospitalization data were obtained for patients admitted with ACS from June 1999 to September 2009 to a single centre of the GRACE registry. CHF was defined as any admission assigned WHO International Classification of Diseases 10 diagnostic code I50. The hazard ratio (HR) for CHF according to GRACE score was estimated in Cox models adjusting for age, gender and the presence of CHF on index admission. Among 1,956 patients, CHF was recorded on index admission in 141 patients (7%), and 243 (12%) were admitted with CHF over 3.8 median years of follow-up. Compared to the lowest quintile, patients in the highest GRACE score quintile had more CHF admissions (116 vs 17) and a shorter time to first admission (1.2 vs 2.0 years, HR 9.87, 95% CI 5.93–16.43). Per standard deviation increment in GRACE score, the instantaneous risk was more than two-fold higher (HR 2.28; 95% CI 2.02–2.57), including after adjustment for CHF on index admission, age and gender (HR 2.49; 95% CI 2.06–3.02). The C-statistic for CHF admission at 1-year was 0.74 (95% CI 0.70–0.79). Conclusions: The GRACE score predicts CHF admission, and may therefore be used to target ACS patients at high risk of CHF with clinical monitoring and therapies. Keywords Heart failure, myocardial infarction, prognosis Received: 9 Jan 2014; accepted: 18 Jun 2014
Introduction There were 4.2 million emergency department attendances for chronic heart failure in the United States in 2010.1 In Europe approximately 5% of all hospital admissions are related, and it accounts for approximately 2% of all health care-related costs.2 Improved survival following myocardial infarction (MI) and an ageing population are expected to contribute to a 25% increase in the prevalence of heart failure by 2013.3 This expanding healthcare burden of heart failure poses a significant challenge for clinicians and health services. As the commonest cause of heart failure in developed nations, coronary heart disease represents a key opportunity to reduce the impact of heart failure, particularly in the
post-MI setting where heart failure is both common and associated with increased mortality,4,5 and where there is strong evidence that specific drug therapies reduce heart failure related events in the years following discharge from hospital.6–8
1Centre 2Centre
for Population Health Sciences, University of Edinburgh, UK for Cardiovascular Science, University of Edinburgh, UK
Corresponding author: David A McAllister, Centre for Population Health Sciences, Medical School, Teviot Place, London EH8 9AG, UK. Email: [email protected]
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
2
European Heart Journal: Acute Cardiovascular Care
International guidance recommends the use of a validated prognostic score among patients with acute coronary syndrome (ACS) in order to identify those most likely to benefit from more intensive management.9 No published models, however, have been shown to predict the risk of heart failure re-admission in patients with MI or ACS. Consequently, it is currently difficult to predict which patients are at highest risk of heart failure and are therefore most likely to benefit from interventions that are known to prevent CHF complications, such as ACE inhibitors, beta blockers and aldosterone receptor antagonists. Among a number of prognostic scores developed to predict outcomes following acute coronary syndrome,10–21 the Global registry of Acute Coronary Events Risk Score (GRACE Score) has emerged as one of the most widely used.22,23 Therefore, we examined whether the at-discharge GRACE Score could be used to predict CHF events following acute coronary syndrome to allow optimal targeting of clinical monitoring and CHF therapies in this high risk group of patients.
Methods Five-year mortality and hospitalization data were obtained for patients recruited at the Scottish centre of the GRACE registry from 28 June 1999 to 6 September 2009 via record linkage to routine healthcare data. A detailed description of the GRACE protocol has been published previously.24 Briefly, eligibility criteria for GRACE included aged 18 or older, admission for presumptive acute coronary syndrome and one or more of: ECG changes consistent with ACS, serial increases in serum biochemical markers of cardiac necrosis, and/or documentation of coronary artery disease. Patients transferred into or out of the registry hospital were enrolled regardless of the time spent at the transferring hospital. All patients consented to all study protocols and procedures and the study was reviewed by Lothian Research Ethics Committee (LREC/1999/4/61) (UK/EU equivalent to Institutional Review Boards). For patients transferred out of a registry hospital, data collection for the initial case report form ended with the transfer and indication of purpose transfer. Patients hospitalized for less than one day who died were enrolled, provided that the cause of death was confirmed to be due to acute coronary syndrome. Standard clinical definitions of diagnostic characteristics and outcomes during hospitalization were used. All cases were assigned to STEMI, NSTEMI, unstable angina, and other cardiac/non-cardiac diagnoses via independent review of clinical presentation, ECG findings and the results of serum biochemical markers of myocardial necrosis. Specifically, unstable angina was defined as ACS with normal biochemical markers of necrosis. Mortality and hospitalization data during follow-up were obtained for all patients via deterministic linkage to a
national database the Scottish Morbidity Record of hospitalization and death data (SMR01). In accordance with a decision by the relevant privacy body (NHS Information Services Division Privacy Advisory Committee), follow-up was for five years following discharge from hospital. Otherwise, events were censored on 30 January 2010. CHF admission as defined as any admission (not including the index admission) assigned the Tenth International Classification of Diagnosis (ICD-10) code I50 in diagnostic position 1 or 2 (of up to 6 available for coders), MI was defined as any admission (not including the index admission) assigned the code I21 or I22. Diagnostic coding accuracy for hospital discharges is high within Scotland, with 95% accuracy for the main condition, and over 90% accuracy for additional conditions.25 GRACE Score was calculated (without re-weighting) using the previously described ‘At Discharge’ GRACE tool (www.outcomes-umassmed.org/grace/acs_risk/acs_risk_ content.html) based on age, history of MI, history of CHF, pulse rate at presentation, systolic blood pressure at presentation, initial serum creatinine level, initial serum cardiac biomarker levels, ST-segment depression on presenting electrocardiogram, and percutaneous coronary intervention performed while in hospital.22 In total, 58 patients (2.8%) had missing characteristics required for the calculation with serum creatinine being most frequently missing (39 patients) followed by heart rate (30 patients). We assigned missing categorical variables to null and used the mean value for continuous variables. This prognostic model was originally developed to predict the risk of death within 6 months of discharge from hospital. This GRACE discharge score was chosen for practical reasons since it represents an important time-point in the assessment of a patient in the post-MI setting representing a key opportunity for careful assessment and optimization of discharge medications likely to influence future CHF outcomes.
Statistical analysis Summary statistics were obtained for patient characteristics by quintile of GRACE Score. Cause-specific hazard ratios (HR) for heart failure admission were estimated according to GRACE score in Cox regression models treating death as a competing cause, adjusting for age, gender and the presence of CHF on the index admission. Heterogeneity from type of event (STEMI versus other) was examined using an interaction term. The cumulative incidence (risk) of admission for CHF, CHF and/or death, recurrent MI, recurrent MI and/or death, and death at 1-year follow-up was estimated in logistic regression models. Patients were grouped into vigintiles (twentieths) according to GRACE score and the proportion experiencing CHF and MI at 1 year was plotted against the risk obtained from the logistic regression models. Departure from linearity was explored via polynomial terms.
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
3
McAllister et al. Table 1. Baseline characteristics total population.
Number GRACE score, median GRACE score, range Age (years), mean (SD) Male sex Smoking Past medical history Myocardial infarction CHF Angina Stroke/transient ischemic attack Peripheral artery disease Diabetes Hypertension Event type STEMI Non-STEMI Unstable angina Killip classification 1 2 3 4
1st quintile
2nd quintile
3rd quintile
4th quintile
5th quintile
392 2.9 1.5–3.23 50.2 (7.3) 76 74.2
391 3.5 3.23–3.78 59.4 (7.2) 68 72.1
391 4 3.78–4.31 66.1 (6.8) 65 71.4
391 4.6 4.31–4.9 71.9 (6.7) 63.7 67.4
391 5.4 4.9–8.01 78.1 (7.5) 58.8 60.9 56.3 32.2 67.5 16.1 12.8 16.6 51.9 25.6 43.2 21.0 51.7 40.1 7.4 0
15.6 0 37 4.6 4.1 10.5 33.4
22.5 0.8 45.8 5.6 4.6 15.1 39.6
33.6 3.3 51 8.7 6.9 14.4 47.7
39.1 5.9 52.4 10.5 11.3 15.9 46.8
31.9 20.7 34.2
35.5 22.3 33.8
34.0 25.1 32.2
28.4 34.0 31.2
90.3 9.2 0.5 0
75.2 22.8 1.8 0
70.1 27.1 2.3 0.5
68.5 27.6 3.8 0
Values are percentages except where indicated.
Area under the curve for a single model, and comparisons of area under the curve (C-statistic) were both calculated in the R-package pROC via bootstrapping.26 The categorical net reclassification index (NRI) was calculated using pre-specified cut-offs of 1%, 5%, 10% and 20% 1-year risk of CHF admission. The NRI, a measure of reclassification following the addition of a new predictor, is the percentage of patients assigned to a less appropriate risk category (cases placed in a lower risk category or controls placed in a higher risk category) minus the percentage of patients assigned to a more appropriate risk category (cases placed in a higher risk category or controls placed in a lower risk category).27 Bootstrap methods were also used to obtain 95% confidence intervals for the NRI. For a review of these standard methods, see Cook.28 We conducted sensitivity analyses following exclusion of patients who developed CHF during the index admission, and excluding events where MI was recorded alongside CHF. Following reviewer comments we also report the associations in patients with and without ST-elevation MI and according to severity of left ventricular ejection fraction (LVEF) impairment (none, mild, moderate and severe) and index heart failure admission, and determine whether GRACE score continues to predict heart failure admission after adjusting for LVEF. All analyses were carried out in R version 3.0.1 (R project for statistical computing, Vienna, Austria).
Results In total, 1,956 patients were admitted from 28 June 1999 to 6 September 2009. The mean age was 66 years, and 66% were male. In addition to a higher prevalence of older patients and those in Killip class 3 and 4, those with higher GRACE scores were more likely to be female, and had a higher prevalence of known coronary, cerebrovascular and peripheral vascular disease as well as hypertension and diabetes (Table 1). CHF was recorded on the index admission in 141 (7%) patients, and 243 (12%) were admitted with CHF over 3.8 median years of follow-up. Compared to the lowest quintile for GRACE score, patients in the highest GRACE score quintile had more CHF admissions (116 vs 17) and a lower time to first CHF admission (1.2 vs 2.0 years, HR 9.87, 95% CI 5.93–16.43). Per standard deviation increment in GRACE score, the instantaneous risk of admission with CHF was more than two-fold higher (HR 2.28; 95% CI 2.02–2.57). This association persisted following adjustment for easily recorded patient characteristics CHF on index admission, age and gender (HR 2.49; 95% CI 2.06– 3.02, Table 2). The risk of admission with CHF was 0.84-fold lower in patients with ST-elevation MI than those without, but the confidence interval was compatible with either a higher or lower risk (95% CI 0.62–1.14) and the p value was not statistically
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
4
European Heart Journal: Acute Cardiovascular Care
Table 2. Events and hazard ratios per quintile of GRACE score.
Heart failure admissions at, n 30 days 90 days 1 year 5 years Follow-up time (days), median Cox models (n = 1956, 243 events), HR (95% confidence interval) GRACE score GRACE score + age GRACE score + age + CHF on index admission
1st quintile
2nd quintile
3rd quintile
4th quintile
5th quintile
0 2 7 17 1800
3 3 8 19 1772
2 8 17 41 1699
6 11 24 50 1595
22 42 62 116 1020
1 (reference) 1 (reference) 1 (reference)
1.16 (0.60–2.23) 1.28 (0.65–2.50) 1.24 0.63–2.43)
2.57 (1.46–4.52) 3.04 (1.62–5.70) 2.96 (1.58–5.56)
3.22 (1.86–5.60) 4.04 (2.07–7.86) 3.95 (2.02–7.69)
9.87 (5.93–16.43) 13.16 (6.52–26.57) 11.97 (5.90–24.30)
25
Risk of event (%)
20
15
Outcome Heart failure at 12 months Myocardial infarction at 12 months Predicted death at 6 months
10
5
0 3
4 5 Grace Score (at discharge)
6
Figure 1. Risk of CHF admission, and readmission with MI, according to the at-discharge-GRACE score, at one-year following discharge.
Solid lines represent the predicted risk obtained from logistic regression models of each outcome on GRACE score. Dashed line indicates predicted mortality at 6 months post discharge according to GRACE score. Points represent the proportion within each vigintile according to GRACE score (n = 97–98 in each) who had the outcome of interest at 1 year (Y-axis) at the mean GRACE score for that vigintile (X-axis).
significant (p = 0.26) so this finding should be interpreted cautiously. The GRACE score was similarly associated with the risk of admission with CHF in patients with and without ST-elevation MI (HR 2.24 and HR 2.33, respectively, HR for interaction 1.03, 95% CI 0.78–1.39, p for interaction = 0.80). Discrimination for CHF admission within 1 year (C-statistic 0.74; 95% CI 0.70–0.79) and for CHF admission or death from any cause within 1 year (0.70; 95% CI 0.66–0.75) were slightly higher than the equivalent statistics for MI (both C-statistics 0.65; 95% CI 0.60–0.70). Adding GRACE score to age and CHF on index admission improved the C-statistic from 0.69 (95% CI 0.64–0.74) to
0.76 (95% CI 0.71–0.80) and this difference was statistically significant (p = 0.003). The percentage of patients within each vigintile of GRACE Score corresponded to that predicted by the model for both CHF and MI. The predicted 1-year risk of CHF admission was similar to the predicted 1-year risk of MI admission. Importantly, and unexpectedly, the 12-month risk of CHF closely parallels the risk of 6-month postdischarge death estimated using the original at-discharge GRACE score risk calculator (Figure 1). When GRACE score was added to age and CHF on index admission, the NRI (using pre-specified cut-offs of
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
5
McAllister et al. 1%, 5%, 10% and 20%) was 13.6% (95% CI −5.1–26.3) for cases, 11.8% (4.7%–18.5%) for controls and 25.3% (95% CI 11.8–42.2%, p < 0.001) overall. The number and percentage assigned to each category with and without the GRACE score is provided in the online supplement. In sensitivity analyses, similar results were obtained after excluding the 141 patients with CHF on the index admission, and after excluding the 17 events where MI was recorded alongside CHF as a cause of hospital admission. There was no evidence of departure from linearity in logistic regression models at 1 and/or 5 years of follow-up (p = 0.11 and p = 0.91, respectively). LVEF was recorded in a subset of 684 patients, of whom the number (%) with none, mild, moderate or severe LVEF impairment was 310 (46.8%), 172 (27.0%), 121 (20.5%) and 27 (5.7%) respectively. The proportion that developed heart failure admission within 1 year was 3.1%, 7.0%, 13.6% and 30.8% for none, mild, moderate and severe, respectively, and compared to those with normal LVEF, those with mild, moderate and severe LVEF had higher rates of admission with heart failure (HR 1.81, 3.12 and 6.20, respectively, p < 0.001). Nonetheless, the association between GRACE score and subsequent heart failure admission was not modified by adding LVEF grade (HR 2.01 per SD, 95% CI 1.66–2.43 versus HR 1.92 per SD, 95% CI 1.57–2.34).
Discussion Although developed for predicting recurrent MI and death rather than CHF admissions, the GRACE Score is a good predictor of CHF admissions, and improved prediction when added to a simple model comprising the presence of CHF on index admission and age. The absolute risk of CHF admission closely tracked the absolute risk of MI readmission across the range of GRACE scores. The proportion of patients who experienced a subsequent heart failure admission even among those with mild LVEF was 7%, more than half of that for people with moderate LVEF. Moreover, the GRACE score continued to predict heart failure admission even after adjusting for LVEF impairment. Consequently, GRACE score offers additional information over using LVEF alone for predicting subsequent risk of heart failure. The original GRACE at-discharge Score was designed to be readily usable at the bedside, and allows clinicians to calculate the risk of death at 6 months using a paper-based point system, an online calculator (http://www.outcomesumassmed.org/grace/acs_risk/acs_risk_content.html) and mobile apps. Since the 12-month risk of CHF closely parallels the 6-month risk of all-cause mortality, each of these GRACE tools can now be readily used to predict the CHF risk. The risk of heart failure readmission according to the GRACE score was similar for patients with and without
ST-elevation MI, and having an ST-elevation MI did not predict increased risk. Our findings are consistent with a recently published report indicating that among 21,516 patients with acute coronary syndrome who were discharged alive without heart failure, the proportion who developed incident heart failure (defined on outpatient clinic visits and/or hospitalization) was not related to type of MI (12.0% for ST-elevation MI, 12.8% for non-ST elevation MI and 11.4% for unstable angina).29 A large number of other prognostic scores have been developed to evaluate outcomes following acute coronary syndrome and/or ST elevation MI,10–21 and these have similar discrimination to GRACE score for all-cause mortality.30 Consequently, it is likely that a number of these will also predict CHF risk. Future studies comparing the performance of these tools with the GRACE score would be of interest, particularly for the most widely used alternative, TIMI score.14 Unfortunately, all variables required to calculate the TIMI score were not available in our GRACE register cohort. A new prediction model developed specifically for CHF admission following ACS may perform better than the GRACE score, which was originally developed for a different outcome. The GRACE score is recommended in current international guidance for management of ACS,9 it is in widespread use and it therefore represents a readily available tool to help flag patients at high-risk of CHF events. Therefore, a substantial improvement over the level of prediction offered by the GRACE score would be required to justify the additional costs and practical challenges of introducing a novel score.
Limitations We examined the performance of GRACE score for evaluating CHF in a single centre, rather than in a multi-centre international study. Over-fitting to our data is unlikely to be a problem as our analysis did not involve the development of a new model. Nonetheless, calibration of GRACE score for risk of CHF admission would be useful for implementing the GRACE score in other settings. We used routine data to estimate the risk of CHF rather than prospective data collection. However, previously published international findings suggest that CHF is reliably recorded in routine data systems,29 whereas local auditing of data quality within our centre has consistently demonstrated 95% accuracy for ICD-10 diagnostic coding.25 Moreover, in sensitivity analyses, we excluded admissions from CHF where MI was concurrently recorded and found similar associations as in the main analysis.
Conclusion In summary, the risk of admission to hospital with CHF is similar to the risk of readmission to hospital with MI among
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
6
European Heart Journal: Acute Cardiovascular Care
patients with ACS, and closely tracks the risk of death within 6 months as predicted using the GRACE Score risk calculator. This extends the utility of the GRACE score to CHF events and could provide added incentive for clinical teams to optimize pharmacological and device treatments at discharge and ensure close clinical monitoring for these patients in the ensuing weeks and months following discharge. Acknowledgements The authors would like to thank the Information and Services Division, NHS National Services Scotland and the British Heart Foundation.
Disclosures DM has no conflicts of interest. NH has no conflicts of interest. MD has no conflict of interest. KAF reports grants from Astra Zeneca, grants from Sanofi Aventis, during the conduct of the study; grants and personal fees from Astra Zeneca, grants and personal fees from Bayer/Janssen, grants and personal fees from Lilly, outside the submitted work.
Funding This research was supported by a peer reviewed grant awarded by the Chief Scientist Office, Scotland (KAAF, KC) and additional support from the British Heart Foundation (KAAF). The original data collection was supported by an unrestricted grant from Sanofi-Aventis, Paris, France, with additional support from Astra Zeneca. The current analysis was unfunded.
References 1. NAMCS and NHAMCS Web Tables, http://www.cdc.gov/ nchs/ahcd/web_tables.htm#2010 (accessed 10 June 2013). 2. Braunschweig F, Cowie MR and Auricchio A. What are the costs of heart failure? Europace 2011; 13: ii13–ii17. 3. Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics—2013 update: a report from the American heart association. Circulation 2013; 127: e6–e245. 4. Steg PG, Dabbous OH, Feldman LJ, et al. Determinants and prognostic impact of heart failure complicating acute coronary syndromes observations from the Global Registry of Acute Coronary Events (GRACE). Circulation 2004; 109: 494–499. 5. Ezekowitz JA, Kaul P, Bakal JA, et al. Declining in-hospital mortality and increasing heart failure incidence in elderly patients with first myocardial infarction. J Am Coll Cardiol 2009; 53: 13–20. 6. Pitt B, White H, Nicolau J, et al., and EPHESUS Investigators. Eplerenone reduces mortality 30 days after randomization following acute myocardial infarction in patients with left ventricular systolic dysfunction and heart failure. J Am Coll Cardiol 2005; 46: 425–431. 7. Dargie HJ. Effect of carvedilol on outcome after myocardial infarction in patients with left-ventricular dysfunction: the CAPRICORN randomised trial. Lancet 2001; 357: 1385–1390.
8. Hall AS, Murray GD and Ball SG. Follow-up study of patients randomly allocated ramipril or placebo for heart failure after acute myocardial infarction: AIRE Extension (AIREX) Study. Acute Infarction Ramipril Efficacy. Lancet 1997; 349: 1493–1497. 9. Hamm CW, Bassand J-P, Agewall S, et al., and ESC Committee for Practice Guidelines. ESC Guidelines for the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation: the task force for the management of acute coronary syndromes (ACS) in patients presenting without persistent ST-segment elevation of the European Society of Cardiology (ESC). Eur Heart J 2011; 32: 2999–3054. 10. Lee KL, Woodlief LH, Topol EJ, et al. Predictors of 30-day mortality in the era of reperfusion for acute myocardial infarction. Results from an international trial of 41,021 patients. GUSTO-I Investigators. Circulation 1995; 91: 1659–1668. 11. Savonitto S, Ardissino D, Granger CB, et al. Prognostic value of the admission electrocardiogram in acute coronary syndromes. JAMA J Am Med Assoc 1999; 281: 707–713. 12. Krumholz HM, Chen J, Wang Y, et al. Comparing AMI mortality among hospitals in patients 65 years of age and older: evaluating methods of risk adjustment. Circulation 1999; 99: 2986–2992. 13. Jacobs DR Jr, Kroenke C, Crow R, et al. PREDICT: a simple risk score for clinical severity and long-term prognosis after hospitalization for acute myocardial infarction or unstable angina: the Minnesota heart survey. Circulation 1999; 100: 599–607. 14. Antman EM, Cohen M, Bernink PJ, et al. The TIMI risk score for unstable angina/non-ST elevation MI: a method for prognostication and therapeutic decision making. JAMA J Am Med Assoc 2000; 284: 835–842. 15. Lindahl B, Toss H, Siegbahn A, et al. Markers of myocardial damage and inflammation in relation to long-term mortality in unstable coronary artery disease. FRISC Study Group. Fragmin during instability in coronary artery disease. N Engl J Med 2000; 343: 1139–1147. 16. Morrow DA, Antman EM, Charlesworth A, et al. TIMI risk score for ST-elevation myocardial infarction: a convenient, bedside, clinical score for risk assessment at presentation: An intravenous nPA for treatment of infarcting myocardium early II trial substudy. Circulation 2000; 102: 2031–2037. 17. Morrow DA, Antman EM, Giugliano RP, et al. A simple risk index for rapid initial triage of patients with ST-elevation myocardial infarction: an InTIME II substudy. Lancet 2001; 358: 1571–1575. 18. Newby LK, Bhapkar MV, White HD, et al., and SYMPHONY and 2nd SYMPHONY Investigators. Predictors of 90-day outcome in patients stabilized after acute coronary syndromes. Eur Heart J 2003; 24: 172–181. 19. Kurz DJ, Bernstein A, Hunt K, et al. Simple point-of-care risk stratification in acute coronary syndromes: the AMIS model. Heart Br Card Soc 2009; 95: 662–668. 20. Mahaffey KW, Yang Q, Pieper KS, et al., and SYNERGY Trial Investigators. Prediction of one-year survival in highrisk patients with acute coronary syndromes: results from the SYNERGY trial. J Gen Intern Med 2008; 23: 310–316.
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
7
McAllister et al. 21. Poçi D, Hartford M, Karlsson T, et al. Role of the CHADS2 score in acute coronary syndromes: risk of subsequent death or stroke in patients with and without atrial fibrillation. Chest 2012; 141: 1431–1440. 22. Eagle KA, Lim MJ, Dabbous OH, et al., and GRACE Investigators. A validated prediction model for all forms of acute coronary syndrome: estimating the risk of 6-month postdischarge death in an international registry. JAMA J Am Med Assoc 2004; 291: 2727–2733. 23. Fox KAA, Dabbous OH, Goldberg RJ, et al. Prediction of risk of death and myocardial infarction in the six months after presentation with acute coronary syndrome: prospective multinational observational study (GRACE). BMJ 2006; 333: 1091. 24. Fox KA, Eagle KA, Gore JM, et al. The Global Registry of Acute Coronary Events, 1999 to 2009–GRACE. Heart 2010; 96: 1095–1101. 25. NHS ISD Scotland. ISD Services | Data Quality Assurance | ISD Scotland, May 2012. http://www.isdscotland.org/
26.
27.
28. 29.
30.
Products-and-Services/Data-Quality/ (accessed 23 June 2014). Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 2011; 12: 77. doi:10.1186/1471-210512-77 www.biomedcentral.com/1471-2105/12/77 Pencina MJ, D’Agostino RB, D’Agostino RB and Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008; 27: 157–172. Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 2007; 115: 928–935. Kaul P, Ezekowitz JA, Armstrong PW, et al. Incidence of heart failure and mortality after acute coronary syndromes. Am Heart J 2013; 165: 379–385.e2. Gale CP, Manda SOM, Weston CF, et al. Evaluation of risk scores for risk stratification of acute coronary syndromes in the Myocardial Infarction National Audit Project (MINAP) database. Heart 2009; 95: 221–227.
Downloaded from acc.sagepub.com at KANSAS STATE UNIV LIBRARIES on July 12, 2014
