Author + information
- Received September 17, 2015
- Revision received November 4, 2015
- Accepted November 23, 2015
- Published online February 23, 2016.
- Rashmee U. Shah, MD, MSa,∗ (, )
- James A. de Lemos, MDb,
- Tracy Y. Wang, MD, MHS, MScc,
- Anita Y. Chen, MSc,
- Laine Thomas, PhDc,
- Nadia R. Sutton, MD, MPHd,
- James C. Fang, MDa,
- Benjamin M. Scirica, MD, MPHe,
- Timothy D. Henry, MDf and
- Christopher B. Granger, MDc
- aUniversity of Utah, Division of Cardiovascular Medicine, Salt Lake City, Utah
- bDivision of Cardiology, University of Texas Southwestern Medical Center, Dallas, Texas
- cDuke Clinical Research Institute, Durham, North Carolina
- dUniversity of Michigan Medical Center, Ann Arbor, Michigan
- eCardiovascular Division, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts
- fCedars-Sinai Heart Institute, Division of Cardiology, Los Angeles, California
- ↵∗Reprint requests and correspondence:
Dr. Rashmee Shah, University of Utah, 30 North 1900 East, Room 4A100, Salt Lake City, Utah 84132.
Background Many patients with acute myocardial infarction (AMI) and cardiogenic shock survive hospitalization; little is known about their subsequent prognosis.
Objectives This study sought to evaluate the associations between cardiogenic shock and post-discharge mortality and all-cause hospitalization among hospital survivors.
Methods We included patients ≥65 years of age with AMI from the ACTION Registry–GWTG (Acute Coronary Treatment and Intervention Outcomes Network Registry–Get With The Guidelines) who survived hospitalization and linked these patients with Medicare claims data. We used proportional hazards models to test the association between cardiogenic shock and outcomes, adjusting for patient and hospital characteristics. Hazard ratios (HRs) are reported for early (1 to 60 days) and late (61 to 365 days) post-discharge time periods.
Results Among 112,668 AMI survivors, 5% had cardiogenic shock during hospitalization. The rate of death was significantly higher among patients with cardiogenic shock at 60 days (9.6% vs. 5.5%) and 1 year (22.4% vs. 16.7%). After accounting for baseline characteristics, the risk of death remained higher for cardiogenic shock patients in the first 60 days after discharge (adjusted HR: 1.62; 95% confidence interval [CI]: 1.46 to 1.80), but was similar to nonshock patients thereafter (adjusted HR: 1.08 for days 61 to 365; 95% CI: 1.00 to 1.18). The rate of all-cause hospitalization or death was significantly higher among shock patients at 60 days (33.9% vs. 24.9%) and 1 year (59.1% vs. 52.3%). After adjustment, the risk of this outcome was also clustered in the first 60 days (adjusted HR: 1.28; 95% CI: 1.21 to 1.35) and was similar thereafter (adjusted HR: 0.95 for days 61 to 365; 95% CI: 0.89 to 1.01).
Conclusions Hospital survivors of AMI who had cardiogenic shock have a higher risk of death and/or hospitalization during the first year after discharge. The risk is time-dependent and is clustered in the early post-discharge period, after which the prognosis is similar in patients with and without cardiogenic shock.
Historically, cardiogenic shock in the setting of acute myocardial infarction (AMI) has had a dismal prognosis. In GUSTO-I (Global Utilization of Streptokinase and Tissue Plasminogen Activator for Occluded Coronary Arteries), only 44% of patients survived the hospitalization (1). The SHOCK (SHould we emergently revascularize Occluded Coronaries for cardiogenic shocK) registry had similar results: only 40% of patients survived hospitalization (2). In recent decades, as revascularization emerged as the standard treatment for AMI, notable gains were made in short-term mortality. The NRMI (National Registry of Myocardial Infarction) demonstrated that, by 2005, over one-half of AMI cardiogenic shock patients survived to hospital discharge (3). In fact, a recent publication from the NCDR (National Cardiovascular Data Registry) ACTION Registry–GWTG (Acute Coronary Treatment and Intervention Outcomes Network Registry–Get With The Guidelines) showed that two-thirds of patients with cardiogenic shock survive to hospital discharge (4).
In light of these advances in early survival, improvement in post-hospital outcomes is gaining increasing importance for patients with AMI and cardiogenic shock. As more patients survive the primary shock hospitalization, we need a better understanding of their long-term prognoses. The effects of acute illness extend beyond the short term—patients face consequences well beyond the first 30 days. The purpose of this study was to provide prognostic information about patients with cardiogenic shock who survive the initial hospitalization. We used data from ACTION Registry–GWTG linked with Centers for Medicare & Medicaid Services (CMS) claims data to evaluate the association between cardiogenic shock during hospitalization for AMI and follow-up outcomes.
The NCDR ACTION Registry–GWTG is a voluntary registry for hospitals designed to improve treatment and outcomes of AMI patients (5). We linked AMI patients enrolled in ACTION Registry–GWTG from January 1, 2007, to December 31, 2012, to CMS administrative claims data using 5 indirect identifiers (date of birth, sex, hospital identifier, date of admission, and date of discharge) (6). The starting population for these analyses was patients who survived the index hospitalization and had Medicare Part A and B fee-for-service insurance for at least 12 consecutive months prior to the index admission. The sequential exclusions for cohort development are displayed in Figure 1. Among patients who died during the index admission, 49.4% had cardiogenic shock (see Online Table 1 for patient characteristics). After exclusions, the analysis population for all-cause mortality consisted of 112,668 AMI survivors (81% of the starting population) from 677 sites. For the secondary outcomes, including hospitalizations, patients who were ineligible for Medicare Part A and B fee-for-service insurance during the index hospitalization (n = 107) were further excluded, which left an analysis sample of 112,561 patients.
Exposure and outcome variables
Cardiogenic shock was defined as: 1) a sustained (>30 min) episode of systolic blood pressure <90 mm Hg; 2) cardiac index <2.2 l/min/m2 determined to be secondary to cardiac dysfunction; and/or 3) the requirement for parenteral inotropic or vasopressor agents or mechanical support (e.g., intra-aortic balloon pump, extracorporeal circulation, or ventricular assist devices) to maintain blood pressure and cardiac index above those specified levels. Transient episodes of hypotension reversed with intravenous fluid or atropine did not constitute cardiogenic shock (7).
The primary outcome of interest was all-cause mortality. Secondary outcomes were composite endpoints: 1) all-cause mortality or all-cause hospitalization to an acute care facility; and 2) all-cause mortality or heart failure–specific hospitalization. Heart failure readmissions were classified with the following primary International Classification of Diseases, Ninth Revision (ICD-9) discharge diagnoses: 428, 402.01, 402.11, 402.91, 398.91, 404.x1, or 404.x3. We used previously published methodology to exclude planned, elective hospitalizations (8), such as staged revascularization procedures. Readmissions for percutaneous coronary intervention (PCI) or coronary artery bypass graft (CABG) within 60 days of hospital discharge were considered planned hospitalizations unless a diagnosis code consistent with an acute event was also present, including ICD-9 codes for heart failure (428.xx, 425.x, 415.0, 398.91, 402.01, 402.11, 402.91, 404.x1, 404.x3), AMI (410.x1), unstable angina (411.1), arrhythmia (426.xx, 427.xx, 785.0, 785.1, 99.61, 99.62, 99.69), and/or cardiac arrest (427.5, 668.1x, 997.1, V12.53, 99.60).
Baseline clinical characteristics, in-hospital procedures and complications, discharge medications, and hospital characteristics were compared between patients with and without cardiogenic shock using Wilcoxon rank-sum tests and Pearson’s chi-square tests. Baseline characteristics and in-hospital variable definitions can be found on the NCDR ACTION Registry–GWTG website (9).
The unadjusted cumulative event rates were estimated by Kaplan-Meier methods and displayed separately for patients with and without cardiogenic shock. The log-rank test was used to assess whether the differences between the curves were statistically significant at p < 0.05. Kaplan-Meier curves are presented separately for the early post-discharge period (within 60 days) and the late post-discharge period (after 60 days). This corresponds to time epochs that were important for hazards modeling and allows us to visualize potential differences in the early and later post-discharge period.
The unadjusted and adjusted associations between cardiogenic shock and follow-up outcomes were evaluated in Cox proportional hazards regression models. We observed that the proportional hazards assumption did not hold for cardiogenic shock, but the relationship could be well captured by fitting a separate hazard ratio (HR) for each of 2 time intervals: the early (1 to 60 days) and late (61 to 365 days) post-discharge periods. As the HR is a conditional parameter, the HR for the late post-discharge period is only representative of patients who remained event-free until 60 days. We considered other approaches to model the changing hazard, but the current approach was favored for its simplicity, clinically meaningful interpretation, and good fit to the data. Other cutpoints were explored, but were less consistent across outcomes.
Potential model variables were adapted from prior long-term mortality risk models (10) and supplemented with variables considered to be clinically important. We adjusted for patient characteristics (baseline demographics and comorbid conditions), hospital characteristics (region, bed size, teaching status, and interhospital transfer status), in-hospital interventions and events (diagnostic catheterization, PCI, CABG, stroke, peak serum creatinine, peak troponin, lowest hemoglobin, and bleeding), and discharge medications. Robust standard errors were used to account for patient clustering within hospitals. We also performed stratified analyses to evaluate the effect of AMI type on the association between cardiogenic shock and outcomes. Finally, we created 2 models (60 days and 1 year) to identify the predictors of post-discharge mortality among cardiogenic shock patients only. These results were similar to the model including all AMI patients (results not shown).
Missing data were <1.5% for all variables, except documented left ventricular ejection fraction (7.1% missing). Missing continuous covariates were imputed to the AMI-type and sex-specific median of the nonmissing values. Missing categorical variables were imputed to the most frequent nonmissing category. All analyses were performed using SAS version 9.3 (SAS Institute, Cary, North Carolina), and p values <0.05 were considered statistically significant.
Among 112,668 AMI survivors, 4.9% (n = 5,555) experienced cardiogenic shock during the index hospitalization. Cardiogenic shock patients were younger, and had a similar sex distribution compared with noncardiogenic shock patients, and were less likely to have had a prior MI, PCI, or CABG (Table 1). Patients with shock were also more likely to have a reduced left ventricular ejection fraction and to receive diagnostic catheterization and revascularization (Table 2).
By 1 year, 22.4% of patients with cardiogenic shock had died compared with 16.7% of those without shock. In the early post-discharge period (1 to 60 days), shock patients were more likely than nonshock patients to die (Kaplan-Meier rate at 60 days: 9.6% [95% confidence interval (CI): 8.9% to 10.4%] vs. 5.5% [95% CI: 5.4% to 5.6%]) (Central Illustration, panel A). The early-term association of cardiogenic shock with mortality persisted after multivariable adjustment; the hazard of death was 1.62 (95% CI: 1.46 to 1.80) for shock versus nonshock patients. Among patients who survived the first 60 days, more shock patients died subsequently than nonshock patients (Kaplan-Meier rate at 365 days 14.1% [95% CI: 13.1% to 15.2%] vs. 11.8% [95% CI: 11.6% to 12.0%]) (Central Illustration, panel B). This late-term association between shock and death was attenuated in the multivariable model, with an adjusted HR of 1.08 (95% CI: 1.00 to 1.18) (Table 3).
By 1 year, 59.1% of AMI patients with cardiogenic shock were hospitalized for any reason or died compared with 52.3% of those without shock. The probability of this composite outcome was higher for shock patients in the early post-discharge period (Kaplan-Meier rate at 60 days: 33.9% [95% CI: 32.6% to 35.1%] vs. 24.9% [95% CI: 24.7% to 25.2%]) (Central Illustration, panel C) and the late post-discharge period (Kaplan-Meier rate at 365 days: 38.2% [95% CI: 36.5% to 39.9%] vs. 36.4% [95% CI: 36.1% to 36.8%]) (Central Illustration, panel D). After accounting for patient baseline and hospital characteristics, the hazard of all-cause hospitalization or death was higher for shock compared with nonshock patients in the first 60 days (adjusted HR: 1.28; 95% CI: 1.21 to 1.35) and similar thereafter (adjusted HR for 61 to 365 days: 0.95; 95% CI: 0.89 to 1.01) (Table 3).
Heart failure–specific readmissions accounted for 23.8% of all readmissions. The rate of heart failure–specific hospitalization or death within 1 year post-discharge was higher among patients with versus without cardiogenic shock: 32.5% versus 23.6%. In the early post-discharge period, 16.6% (95% CI: 15.6% to 17.6%) of shock patients, compared with 9.8% (95% CI: 9.6% to 10.0%) of nonshock patients were hospitalized for heart failure or died. In the late post-discharge period, 19.1% (95% CI: 17.9% to 20.4%) of shock patients and 15.3% (95% CI: 15.1% to 15.6%) of nonshock patients were hospitalized for heart failure or died. The adjusted hazard of heart failure hospitalization or death was higher for shock compared with nonshock patients in the first 60 days (adjusted HR: 1.48; 95% CI: 1.37 to 1.61). Between 61 and 365 days, the increased hazard associated with cardiogenic shock was attenuated, but still present (adjusted HR: 1.10; 95% CI: 1.02 to 1.18) (Table 3).
Other factors associated with 1-year mortality in the multivariable model were older age, ejection fraction ≤40%, discharge to a skilled nursing facility, number of hospitalizations during the year prior to AMI admission, and peak serum creatinine (Online Table 2). Factors associated with the secondary outcomes were similar (Online Tables 3 and 4), except prior HF diagnosis was notably associated with the combined outcome of mortality or HF hospitalization.
In AMI type-stratified analyses, non–ST-segment elevation myocardial infarction (NSTEMI) patients fared worse than ST-segment elevation myocardial infarction (STEMI) patients, with higher unadjusted event rates in the early and late post-discharge periods. In the adjusted analyses, however, similar patterns emerged, regardless of AMI type: the risks of mortality and mortality and/or all-cause hospitalization were clustered in the early post-discharge period (Table 4).
Many patients with AMI and cardiogenic shock survive to hospital discharge, and long-term outcomes are increasingly important to patients, payers, and health care delivery systems. This study of older AMI survivors with and without cardiogenic shock includes several important findings. More than one-half of AMI survivors were hospitalized or died during the year after the index hospitalization, with higher rates among patients whose AMI was complicated by cardiogenic shock. The risk of poor outcomes for cardiogenic shock patients was greater than for noncardiogenic shock patients during the early post-discharge period. Beyond this vulnerable time period, however, outcomes were similar between shock and nonshock patients. This observation suggests that: 1) tailored interventions are needed to address the vulnerable, immediate post-hospital period; and 2) cardiogenic shock is not the primary driver of adverse outcomes in the later post-hospital period.
Revascularization rates for AMI have increased rapidly in the past decade, which has contributed to significant reductions in short-term morbidity and mortality (11,12). Thus far, much of the published data focuses on these short-term gains. In contrast to the extensive in-hospital outcome data (3,4), few studies report post-hospital prognosis among survivors of cardiogenic shock. Two prior studies of AMI patients found similar, diminishing effects of cardiogenic shock on outcomes over time. The CRUSADE (Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation) registry, including only NSTEMI patients, showed that cardiogenic shock had no bearing on outcomes after survival to 6 months (13). Whereas our analysis was limited to 1 year of follow-up, FAST MI (French registry of Acute STEMI and NSTEMI) evaluated outcomes through 5 years and found that cardiogenic shock had no effect on death for patients who survived the first year post-discharge (14). Although different methods and patient populations can explain variable risk time frames, the underlying concept is the same: the highest risk for cardiogenic shock patients who survive hospitalization is clustered in the early post-discharge period.
AMI cardiogenic shock patients are vulnerable during the early post-discharge period for several reasons. Perhaps cardiogenic shock patients present to the hospital late after the index event, when reperfusion is less effective or may not be offered (15). These patients may survive to hospital discharge, but fare worse in the follow-up. Hospitalized patients experience sleep disturbances, deconditioning, and malnutrition, all of which adversely affect post-discharge prognosis (16). Cardiogenic shock patients are sicker and have longer hospitalizations compared with noncardiogenic shock patients (1), and so may be more susceptible to these factors. In addition, cardiogenic shock patients may face challenges in transitions of care or be less likely to receive evidence-based medical treatment; these are potential opportunities to improve outcomes during the high-risk period. Alternatively, future studies could focus on identifying the sickest patients with the worst prognosis during the early, post-hospital period. This information could be used to tailor interventions, such as palliative care and hospice, to match expected patient outcomes. Our findings are hypothesis-generating, and future studies should identify specific reasons for poor outcomes soon after discharge. Some of these reasons may be modifiable, meaning that targeted interventions could improve survival during the vulnerable early post-discharge period.
The unadjusted 1-year rate of all-cause hospitalization or death was high (>50%) in this cohort of AMI survivors (with and without shock), a reflection of the older age of the population. Less than one-fourth of all-cause hospitalizations were heart failure–related in our study. A recent report from Medicare demonstrated declining heart failure hospitalizations after AMI (17); in combination with the heart failure–specific rehospitalization rate seen here, these observations suggest a shift toward non–heart failure determinants of post-hospital outcomes. As acute outcomes improve for cardiac conditions, together with widespread use of post-AMI medical therapy, the effect of noncardiac disease may be increasing (18). A report from MIDAS (Myocardial Infarction Data Acquisition System), a data collection system for AMI hospitalizations in New Jersey, showed a steady increase in noncardiovascular causes of death between 1986 and 2008 (19). Medicare patients hospitalized with pneumonia, heart failure, or AMI have frequent readmissions within 30 days of discharge, but the readmission diagnosis is different from the index diagnosis in two-thirds of cases (20). Therefore, as we refine and improve upon in-hospital treatments for AMI, we may need to increase the focus on noncardiac conditions. Future investigations should specifically identify the effect of noncardiac conditions on long-term outcomes in the AMI population.
NSTEMI patients with cardiogenic shock experienced higher event rates (mortality alone and mortality and/or all-cause hospitalization) in the post-discharge period, compared with STEMI patients. However, AMI type did not appreciably modify the association between cardiogenic shock and outcomes in the adjusted analyses. In other words, after accounting for variables, including age and comorbid conditions, NSTEMI and STEMI patients with cardiogenic shock have similar, increased relative risks of poor outcomes clustered in the early post-discharge period.
Our dataset is not detailed enough to delineate the mechanisms underlying the time-dependent attenuation of hazards associated with cardiogenic shock. In addition, hospital participation in ACTION Registry–GWTG is voluntary and may be biased toward high-performing hospitals. ACTION Registry–GWTG does not include information about mechanical support (intra-aortic balloon pump and temporary ventricular assist devices), although the mortality and morbidity benefit of these devices is unclear at this time (21). This investigation is limited to Medicare patients and may not be applicable to the younger, non-Medicare population. In addition, ICD-9 coding accuracy varies widely (22), and given Medicare reimbursement penalties, hospitals may have an incentive to code non–heart failure diagnoses. These factors may affect the accuracy of heart failure–specific analyses.
We were unable to account for cardiac arrest during the index admission, as this data was not collected in ACTION Registry–GWTG until 2011. In sensitivity analyses, we limited the analyses to years during which cardiac arrest data was collected (40.7% [n = 45,876] of the analysis population). Exclusion of patients with cardiac arrest (3.2%; n = 1,462) or missing arrest data (<1%; n = 287) did not appreciably change the results; cardiogenic shock patients experienced increased death and death or all-cause hospitalization in the early, but not in the late post-discharge period. Finally, although we attempted to address the association between cardiogenic shock and post-hospital outcomes by adjusting for a broad range of patient-level clinical factors and treatments, the possibility of confounding by unmeasured covariates remains.
In this study, we leveraged ongoing data collection systems (ACTION Registry–GWTG and CMS claims data) to assess patterns of mortality and morbidity for survivors of AMI with and without cardiogenic shock. We found an association between cardiogenic shock and outcomes during the early post-hospital period, but no association among patients who survive the early post-hospital period. In other words, cardiogenic shock had a diminishing association with death and death or all-cause hospitalization over time. This finding should not undermine the high rates of in-hospital death for cardiogenic shock patients—many patients die in the hospital (4). Still, hospital survivors of AMI complicated by cardiogenic shock are vulnerable to adverse outcomes during the immediate post-discharge period. Future investigations should identify specific modifiable and nonmodifiable reasons for this pattern. This information would be valuable for tailoring interventions to improve early survival and for identifying the sickest patients who may be better served with palliative care or hospice.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Survivors of AMI complicated by cardiogenic shock remain at increased risk of death and repeated hospitalization during the early period after hospital discharge.
TRANSLATIONAL OUTLOOK: Future studies should identify the causes of adverse events early after hospital discharge and identify modifiable targets for intervention.
The ACTION Registry–GWTG (Acute Coronary Treatment and Intervention Outcomes Network Registry–Get With The Guidelines) is an initiative of the American College of Cardiology and the American Heart Association, with partnering support from the Society of Cardiovascular Patient Care and the American College of Emergency Physicians. The American College of Cardiology Foundation’s National Cardiovascular Data Registry (NCDR) supported this research. The views expressed in this paper represent those of the authors and do not necessarily represent the official views of the NCDR or its associated professional societies. Dr. Shah owns stock in Gilead Sciences. Dr. de Lemos has received grant support from Abbott Diagnostics; and has received consulting fees for membership on a data and safety monitoring board from St. Jude Medical. Dr. Wang has received research grants through Duke Clinical Research Institute from AstraZeneca, Boston Scientific, Bristol-Myers Squibb, Daiichi-Sankyo, Eli Lilly, Gilead Sciences, GlaxoSmithKline, and Regeneron Pharmaceuticals; and has received consulting fees or honoraria from AstraZeneca, Eli Lilly, and Premier, Inc. Dr. Scirica has received research grants via the TIMI Study and Brigham and Women’s Hospital from AstraZeneca, Bristol-Myers Squibb, Daiichi-Sankyo, GlaxoSmithKline, Johnson & Johnson, Bayer Healthcare, Gilead, Eisai, and Merck; and has received consulting fees from Arena, AstraZeneca, BioagenIdec, Boehringer Ingelheim, Boston Clinical Research Institute, Bristol-Myers Squibb, Covance, Eisai, Elsevier Practice Update Cardiology, Forest Pharmaceuticals, GE Healthcare, Gilead, GlaxoSmithKline, Lexicon, Merck, St. Jude Medical, and the University of Calgary. Dr. Granger has received research support, salary support, and/or consulting fees from Boehringer Ingelheim, Bristol-Myers Squibb, GlaxoSmithKline, Hoffman-La Roche, Medtronic Foundation, Merck & Co., Pfizer, Sanofi, Takeda, The Medicines Company, Daiichi-Sankyo, Janssen Pharmaceuticals, Bayer, and Armethon. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- acute myocardial infarction
- coronary artery bypass graft
- confidence interval
- Center for Medicare & Medicaid Services
- non-ST-segment elevation myocardial infarction
- percutaneous coronary intervention
- ST-segment elevation myocardial infarction
- Received September 17, 2015.
- Revision received November 4, 2015.
- Accepted November 23, 2015.
- American College of Cardiology Foundation
- Holmes D.R. Jr..,
- Bates E.R.,
- Kleiman N.S.,
- et al.,
- for the GUSTO-I Investigators
- Hochman J.S.,
- Buller C.E.,
- Sleeper L.A.,
- et al.
- Anderson M.L.,
- Peterson E.D.,
- Peng A.S.,
- et al.
- Peterson E.D.,
- Roe M.T.,
- Rumsfeld J.S.,
- et al.
- Anderson H.V.,
- Weintraub W.S.,
- Radford M.J.,
- et al.
- Krumholz H.M.,
- Lin Z.,
- Drye E.E.,
- et al.
- ↵American College of Cardiology. NCDR Data Collection. Available at: https://www.ncdr.com/webncdr/action/home/datacollection. Accessed December 8, 2015.
- Shah R.U.,
- Henry T.D.,
- Rutten-Ramos S.,
- et al.
- Roe M.T.,
- Messenger J.C.,
- Weintraub W.S.,
- et al.
- Aissaoui N.,
- Puymirat E.,
- Simon T.,
- et al.
- Schömig A.,
- Ndrepepa G.,
- Kastrati A.
- Chen J.,
- Hsieh A.,
- Dharmarajan K.,
- et al.
- Roger V.L.
- Kostis W.J.,
- Deng Y.,
- Pantazopoulos J.S.,
- et al.,
- for the Myocardial Infarction Data Acquisition System Study Group
- Naidu S.S.