Author + information
- Received February 21, 2018
- Accepted March 27, 2018
- Published online August 27, 2018.
- Ayman Samman-Tahhan, MDa,
- Jeffrey S. Hedley, MD, MScb,
- Andrew A. McCue, MDa,
- Jonathan B. Bjork, MDc,
- Vasiliki V. Georgiopoulou, MD, MPH, PhDa,
- Alanna A. Morris, MD, MSca,
- Javed Butler, MD, MPH, MBAd and
- Andreas P. Kalogeropoulos, MD, MPH, PhDd,∗ ()
- aDivision of Cardiology, Department of Medicine, Emory University School of Medicine, Atlanta, Georgia
- bDepartment of Cardiology, Cleveland Clinic, Cleveland, Ohio
- cDepartment of Medicine, University of Minnesota Medical School, Minneapolis, Minnesota
- dDivision of Cardiology, Department of Medicine, Stony Brook University School of Medicine, Stony Brook, New York
- ↵∗Address for correspondence:
Dr. Andreas Kalogeropoulos, Stony Brook University Medical Center, Stony Brook University, Health Sciences Center, T-16, Room 080, 101 Nicolls Road, Stony Brook, New York 11794-8167.
Objectives This study sought to evaluate INTERMACS (Interagency Registry for Mechanically Assisted Circulatory Support) profiles for prognostic use among ambulatory non–inotrope-dependent patients with heart failure with reduced ejection fraction (HFrEF).
Background Data for INTERMACS profiles and prognoses in ambulatory patients with HFrEF are limited.
Methods We evaluated 3-year outcomes in 969 non–inotrope-dependent outpatients with HFrEF (EF: ≤40%) not previously receiving advanced HF therapies. Patients meeting an INTERMACS profile at baseline were classified as profile 7 (n = 348 [34.7%]); 146 patients (14.5%) were classified profile 6; and 52 patients (5.2%) were classified profile 4 to 5. Remaining patients were classified “stable Stage C” (n = 423 [42.1%]).
Results Three-year mortality rate was 10.0% among stable Stage C patients compared with 21.8% among INTERMACS profile 7 (hazard ratio [HR] vs. Stage C: 2.45; 95% confidence interval [CI]: 1.64 to 3.66), 26.0% among profile 6 (HR: 3.93; 95% CI: 1.64 to 3.66), and 43.8% among profile 4 to 5 (HR: 6.35; 95% CI: 3.51 to 11.5) patients. Hospitalization rates for HF were 4-fold higher among INTERMACS profile 7 (38 per 100 patient-years; rate ratio [RR] vs. Stage C: 3.88; 95% CI: 2.70 to 5.35), 6-fold higher among profile 6 patients (54 per 100 patient-years; RR: 5.69; 95% CI: 3.72 to 8.71), and 10-fold higher among profile 4 to 5 patients (69 per 100 patient-years; RR: 9.96; 95% CI: 5.15 to 19.3) than stable Stage C patients (11 per 100 patient-years). All-cause hospitalization rates had similar trends. INTERMACS profiles offered better prognostic separation than NYHA functional classifications.
Conclusions INTERMACS profiles strongly predict subsequent mortality and hospitalization burden in non–inotrope-dependent outpatients with HFrEF. These simple profiles could therefore facilitate and promote advanced HF awareness among clinicians and planning for advanced HF therapies.
Although risk scores have been proposed as decision-making aids for patients with advanced heart failure (HF) (1), acceptance in clinical practice has been limited (1), especially for patients who are potential candidates for long-term mechanical circulatory support (MCS). This is partially because HF progression is a complex process, rendering predictions imprecise at the individual level (1), but also because these scores have been developed mainly with heart transplant candidates in mind (2–4). The INTERMACS (Interagency Registry for Mechanically Assisted Circulatory Support) clinical profiles (5), which are based on clinical descriptors, were developed to better characterize MCS candidates (5). These profiles have strong prognostic properties for patients with HF, considered for MCS or heart transplantation (6–9). However, data for the usefulness of INTERMACS profiles for the broader population of ambulatory patients with HF and reduced ejection fraction (HFrEF) are limited (10). Also, no data exist for predicted hospitalization burden according to INTERMACS profiles, which could be an important driver of clinical decisions.
The objective of this study was to determine the association between baseline INTERMACS profiles and subsequent mortality and all-cause and HF-related hospitalization rates among non–inotrope-dependent ambulatory patients with HFrEF over a 3-year follow-up period. In addition, because INTERMACS profiles map to certain New York Heart Association (NYHA) functional classes, we compared the prognostic value of NYHA functional classification with that of INTERMACS profiles for the outcomes of interest.
We reviewed the records of consecutive adult (≥18 years of age) outpatients who received care for HF between January 1, 2012, and March 31, 2012, in Emory Healthcare by cardiologists, including HF specialists. Patients were identified using International Classification of Diseases-9th Revision-Clinical Modification (ICD-9-CM) codes 402.X1, 404.X1, 404.X3, and 428.XX. The inception time frame was selected to allow for ≥3 years of follow-up. Medical records were reviewed for symptoms, signs, and treatment of HF; documentation of left ventricular ejection fraction (LVEF); and special causes of HF. We verified the diagnosis of HF based on physician documentation of symptoms, signs, and guideline-based treatment for HF. We defined HFrEF as current LVEF ≤40% (11). We excluded patients with: 1) specific cardiomyopathies (e.g., hypertrophic, stress-induced, infiltrative, restrictive, and chemotherapy-induced; 2) complex congenital heart disease; 3) primary right-sided disease (e.g., right ventricular cardiomyopathy, Group I pulmonary arterial hypertension); and 4) primary valvular disease. Among 1,053 patients who fulfilled those criteria, information to reliably adjudicate INTERMACS profile was not available in 13 patients. In addition, 33 patients had previously received advanced HF therapies (27 MCS and 6 patients receiving palliative inotropes), and 35 patients were inotrope-dependent. We also excluded 3 patients who had had no follow-up visits. The final analysis cohort consisted of 969 patients (Figure 1). The Emory Institutional Review Board approved the study.
Baseline characteristics and outcomes
Baseline demographics, anthropometric measurements, vital signs, and laboratory data were collected through the Clinical Data Warehouse from the index clinic visit. Race was self-reported. For laboratory values and medications, we extracted the last information available up to the index visit date. The presence of chronic conditions was adjudicated according to the latest (2014) algorithm proposed by the Chronic Conditions Data Warehouse, used to analyze Medicare data (12).
INTERMACS profiles classification
INTERMACS is a North American registry established in 2005 for patients who are receiving MCS for advanced HF (5). Within INTERMACS profiles, patients are divided by their signs and symptoms into 7 clinical profiles (Table 1). This division further refines classification of patients with NYHA functional class III to IV symptoms and provides a detailed description of disease severity (5). For the current study, 2 investigators (A.S.T. and J.S.H.), blinded to outcomes data, were independently assigned an INTERMACS profile at baseline to each patient after a detailed review of medical records. For profile 7, in addition to symptomatic limitation with more than mild exertion, we required the absence of recent decompensation (within 6 months, in accordance to European Heart Failure Association criteria for advanced HF) (13). For profile 6, we required symptomatic limitation with minimal activity (as well as activities of daily living) but without overt fluid overload or symptoms at rest. To distinguish “stable Stage C” from profile 7, we sought further guidance from the INTERMACS manual of operations (14) (Online Appendix). Because signs and symptoms for patients with HF are documented using a structured approach according to institutional policy at Emory University, the information required to distinguish among the various levels of symptomatic limitations is available for most patients. We considered INTERMACS profiles 4 and 5 as a single group for analysis purposes as these profiles are uncommon in the ambulatory setting, they share important characteristics, and often patients “shuttle” between profiles 4 and 5. Patients whose HF severity did not meet any INTERMACS profile were classified stable Stage C. If there was discrepancy between reviewers for profile assignment, a one-third investigator (A.P.K.) adjudicated the final patient profile.
Assessment and classification of outcomes
Outcome data (death and hospitalizations) were collected through the Emory Clinical Data Warehouse database, with a minimum of 3 years of follow-up for each living patient who continued to receive care at Emory Healthcare. For patients who were alive until the last encounter but did not continue to receive care in Emory Healthcare throughout the study period, the last encounter was considered the last date of follow-up for analysis purposes. We classified hospitalizations based on the primary ICD-9-CM codes, using Clinical Classification software (2015 version: Healthcare Cost and Utilization project, Agency for Healthcare Research and Quality) (15). Hospitalizations with primary ICD-9-CM diagnosis mapping to Clinical Classifications Software category 7 (“Diseases of the circulatory system”) were classified as cardiovascular. Hospitalizations associated with primary codes 402.X1, 404.X1, 404.X3, and 428.XX were classified specifically as HF-related.
We evaluated: 1) all-cause mortality and the composite endpoint of death, left ventricular assist device (LVAD) implantation, or heart transplantation (time to first event); 2) hospitalization rates (all-cause and HF-related) over the entire 3-year follow-up period; and 3) days spent in hospital on an annual basis (all-cause and HF-related).
We used Cox proportional hazards models to estimate hazard ratios for mortality and the composite of death, LVAD implantation, or heart transplantation. For mortality estimation, LVAD and heart transplantation recipients were censored as alive at the time of event. We used the Schoenfeld residuals to evaluate the proportionality assumption. For hospitalization rates and days spent in hospital, we considered only the hospitalizations occurring before an LVAD implantation or heart transplantation event and the corresponding time at risk (i.e., time after these events were censored). We used negative binomial regression models to estimate rate ratios (RRs) for admission rates across clinical profiles. For days spent in hospital, we estimated RR by using zero-inflated negative binomial regression models (because of the cluster of observations at zero), as the Bayesian information criterion indicated that these models offered a better fit. In all adjusted models, we entered age, sex, race, and presence of ischemic heart disease, as INTERMACS profiles inherently stratify for other risk factors and the purpose of the present study was to examine whether INTERMACS profiles could be a useful prognostic tool for these patients. In exploratory analyses, we examined whether the association of clinical profiles with outcomes was modified by sex, race, and ischemic heart disease. In addition, we compared the incremental value of INTERMACS classification with the simpler NYHA functional class for prognostic classification by comparing model performance of INTERMCAS-based versus NYHA functional class-based models. We opted to avoid subclassification of NYHA functional class III into IIIA and IIIB, as there is no clear definition or consensus in published reports (16), and therefore we treated NYHA functional class III as a single class. We used the C statistic (range 0.5 to 1.0) as a measurement of model discrimination and Royston’s D (range 0 to ∞) as a measurement of prognostic separation (17), where improvements in D by >0.1 are considered meaningful (18). We examined agreement between raters for clinical profile classification by using Cohen’s κ statistic weighted for distance between categories (i.e., more discordant classifications between raters were assigned more weight). To examine for differential event ascertainment across clinical profiles (e.g., less sick patients could be receiving more local care with outside hospitalizations not captured in our system), we performed sensitivity analyses for event rates and outpatient visits based on residential ZIP codes. We used STATA 14.2 software (StataCorp LP, College Station, Texas) for all analyses.
Distribution of clinical profiles and characteristics
Among the 969 consecutive patients who fulfilled all study criteria, 423 (43.7%) did not meet any INTERMACS profile description and were therefore classified as stable Stage C patients. For the remaining patients, HF severity was classified as INTERMACS profile 7 in 348 patients (35.9%), profile 6 in 146 patients (15.1%), and profile 4 to 5 in 52 patients (5.4%). Baseline clinical characteristics of the patients are summarized in Table 2. Among HF patients in stable Stage C, 1.0% of patients had been previously evaluated for advanced HF therapies (LVAD or heart transplantation); in comparison, 4.6%, 11.2%, and 10.6% of patients with INTERMACS profiles 7, 6, and 4 to 5, respectively, had previously received a similar evaluation.
Inter-rater agreement for clinical profiles
The two raters agreed exactly on the clinical profile in 55.4% of cases; disagreed with adjacent classifications in 29.7% of cases; and disagreed with nonadjacent classifications in 14.9% of cases. Weighted agreement (i.e., more discordant profile classifications were assigned more weight) between the 2 primary raters on clinical profile assignment was 79.8%, and Cohen's weighted κ coefficient was 0.43 for the entire patient population, indicating moderate agreement.
Clinical profiles and mortality
After a median of 3.0 years (range 1.8 to 3.2 years), there were 186 endpoint events, including 159 deaths (without previous LVAD or heart transplantation), 19 LVAD implantations, and 8 heart transplantations. Mortality rates (with LVAD and heart transplant recipients censored as alive at the time of event) were 6.2%, 11.6%, and 18.1% at 1, 2, and 3 years, respectively. The corresponding event rates for the composite of death, LVAD, and heart transplantations were 6.9%, 13.2%, and 20.9%, respectively.
The 3-year rate of LVAD implantation was 1.5% for stable Stage C patients and 3.1%, 4.5%, and 8.5% for patients with INTERMACS profiles 7, 6, and 4 to 5, respectively. The 3-year rate of heart transplantation was 0.0% for stable Stage C patients and 0.4%, 3.2%, and 5.3% for patients with INTERMACS profiles 7, 6, and 4 to 5, respectively.
Table 3 summarizes the 3-year mortality rates (with LVAD and heart transplant recipients censored as alive at the time of event) and 3-year composite endpoints (death, LVAD, or heart transplantation) according to baseline clinical profile. As expected, relative risks increased sharply across worsening clinical profiles with early separation of Kaplan-Meier curves (Figures 2A and 2B). For mortality, INTERMACS profiles had a C statistic of 0.705 (vs. 0.670 for NYHA functional class) and a Royston D of 1.20 (vs. 0.96 for NYHA functional class) for prognostic separation across groups. For the composite endpoint INTERMACS profiles had a C statistic of 0.689 (vs. 0.640 for NYHA functional class) and a Royston D of 1.09 for prognostic separation across groups (vs. 0.79 for NYHA functional), indicating that INTERMACS classification was a meaningful refinement compared to NYHA functional classification.
One-year mortality rates, which is commonly used as a criterion for advanced HF therapy referral, were 2.5%, 5.1%, 15.1%, and 19.5% among patients with baseline stable Stage C and INTERMACS profiles 7, 6, and profile 4 to 5, respectively. Corresponding 1-year composite endpoint rates were 2.5%, 6.0%, 15.8%, and 25.6%, respectively.
Clinical profiles and hospitalizations
There were 1,556 all-cause hospitalizations; HF was the primary reason for 677 (43.5%) of these hospitalizations. The overall hospitalization rate was 65.9 admissions (95% confidence interval [CI]: 58.6 to 73.3) per 100 patient-years, and the HF hospitalization rate was 28.7 admissions (95% CI: 24.7 to 32.7) per 100 patient-years. Patients spent an average of 6.4 days (95% CI: 5.3 to 7.5) per year in the hospital; 2.5 of these days (95% CI: 2.1 to 2.9) were for HF. These numbers do not take into account any admissions and in-hospital days after LVAD implantation or heart transplantation.
Rates of hospitalizations increased sharply across worsening clinical profiles (Table 4), both for all-cause and for HF-related hospitalizations, with an approximately 5-fold increase for all-cause and a 10-fold increase for HF hospitalizations across profiles. Of note, HF accounted for fewer hospitalizations in Stage C patients (26.9% of total) compared with patients meeting an INTERMACS profile (50.3% of total). Patients with worse clinical profiles spent more total and HF-related days in the hospital (Online Table 1), ranging from 4.3 days per year (0.8 days for HF admissions) among stable Stage C patients to 16.4 days per year for INTERMACS profile 4 to 5 patients (6.0 days for HF). Patients with stable Stage C HF spent relatively fewer days (18.6%) in hospital for HF admissions, whereas patients with an INTERMACS profile spent 47.6% in hospital days for HF.
African-American patients had mortality and composite endpoint rates similar to those of Caucasian patients across clinical profiles. However, African-American patients had 52% higher all-cause hospitalization rates than their Caucasian counterparts in adjusted models (RR: 1.52; 95% CI: 1.20 to 1.93; p = 0.001), and this effect was consistent across clinical profiles (p = 0.37 for interaction of race with clinical profile). Similarly, there was a consistently higher rate of HF hospitalizations in African Americans (RR: 1.90; 95% CI: 1.37 to 2.63; p < 0.001) without modification across clinical profiles (p = 0.38 for interaction).
Patients with ischemic heart disease had, on average, a 34% higher all-cause hospitalization rate than their nonischemic counterparts (RR: 1.34; 95% CI: 1.06 to 1.69; p = 0.015); this effect was driven mainly by a relatively higher hospitalization rate in ischemic patients with Stage C clinical profile at baseline (p = 0.013 for interaction of ischemic heart disease with clinical profile). Rates of HF hospitalizations for ischemic patients did not differ from those for nonischemic patients (RR: 1.10; 95% CI: 0.80 to 1.52; p = 0.56) and were not significantly different across clinical profiles.
Finally, mortality and composite endpoint rates for men did not differ from those for women, and sex did not modify the association between clinical profiles and these outcomes. Sex-based results across clinical profiles were similar for all-cause and HF-related hospitalizations.
Mortality and composite of death, LVAD, or transplantations for metropolitan Atlanta area residents (56.6% of patients) were similar to those for residents of more distant areas (43.4% of patients). Outpatient visits were slightly more frequent among metropolitan Atlanta residents, but the gradient did not differ across clinical profiles. All-cause hospitalizations were nonsignificantly more frequent among metropolitan Atlanta residents than among residents of more distant areas, but hospitalization rates for HF were similar, and there was no interaction between residence and clinical profile either for all-cause or for HF hospitalization rates. Full details are provided in the Online Appendix and in Online Table 2.
In this single-center, retrospective study, we observed a significant risk gradient for mortality and hospitalization rates across INTERMACS-based clinical profiles in non–inotrope-dependent outpatients with HFrEF. Compared with stable Stage C HF patients, mortality at 3 years was 2.5-fold higher among INTERMACS profile 7 patients, 4-fold higher among INTERMACS profile 6 patients, and 6-fold higher among profile 4 to 5 patients. Also, hospitalization rates for HF were 4-fold higher among INTERMACS profile 7, 6-fold higher among profile 6, and 10-fold higher among profile 4 to 5 patients than among stable Stage C patients. We also observed that African-American patients had consistently higher hospitalization rates across all clinical profiles. These findings suggest that a simple clinical classification system such as INTERMACS profiles: 1) carries significant prognostic information for non–inotrope-dependent outpatients with HFrEF; 2) can facilitate advanced HF awareness among health care providers and appropriate patient referral; and 3) can aid in decision-making about advanced HF therapies for these patients.
Data for mortality according to INTERMACS profiles among outpatients not scheduled for MCS or heart transplantation evaluation are limited. In the prospective, observational, nonrandomized ROADMAP (Risk Assessment and Comparative Effectiveness of Left Ventricular Assist Device and Medical Management) study (19), the intention-to-treat 1-year mortality among patients who were assigned to the optimal medical management arm (62 patients with INTERMACS profile 4 to 5; 34 patients with profile 6; and 2 patients with profile 7) was 18%, which compares directly to the 1-year mortality of profile 4 to 5 (19.5%) and profile 6 (15.1%) patients in our study. In contrast, 1-year mortality for the same baseline profiles was considerably higher among patients who participated in the MedaMACS (Medical Arm of Mechanically Assisted Circulatory Support) screening pilot study, as these patients were selected on the basis of several additional high-risk criteria (10). Of note, INTERMACS profiles still offered significant prognostic separation in this high-risk population (10). Our study extends these previous findings to a large, unselected population of outpatients with HFrEF receiving care in a tertiary center. Of note, compared to NYHA functional classes, INTERMACS profiles offered better discrimination and prognostic separation for the outcomes of interest in head-to-head comparison in our study. Taken together, these data highlight the prognostic value of INTERMACS clinical profiles for a broad spectrum of patients with HFrEF and the potential for wider clinical application of these profiles.
Several quantitative risk stratification systems have been developed for patients with advanced HF (20,21). The performance of these prediction models has been variable (20,21), depending on the characteristics of the target population (22), the outcomes of interest (4), and the natural shift in HF therapy and outcomes over time (20). Regardless of the predictive accuracy of risk scores, the acceptance of these scores in practice has been minimal, partially because of complexity (i.e., additional tests needed to calculate the score) and partially because of limited clinical implications of such risk stratification at this point, besides decision support for heart transplant candidacy or MCS. Therefore, use of HF risk stratification schemes is limited to referral centers, usually in the context of advanced HF programs. Consequently, a practical clinical classification system for patients with HFrEF would be more conducive to assessment of disease severity and early recognition of transition to advanced HF in broader clinical settings. Late recognition of transition to advanced HF has been identified as an important gap in the management of HF, as late referral to HF programs limits therapeutic options (1). Although most non–inotrope-dependent patients with HFrEF who meet an INTERMACS profile would not be candidates for MCS with current standards, because of less severe HF or inadequate social support, it would still be important to refer these patients to advanced HF programs for further management. The boundary between Stages C and D is in constant evolution as new therapies and more complex risk scoring systems are developed, and therefore seeking expert advice early in the process of progression to Stage D would be reasonable. Preliminary data suggest that promoting advanced HF awareness among health care providers through, for example, a clinical decision support system could lead to improved patient outcomes (23). From this perspective, a practical risk stratification system would facilitate early recognition of advanced disease. Our findings suggest that INTERMACS profiles could play this role.
First, this was a retrospective study from a single, academic referral center, and therefore, the findings may not be generalizable to community outpatients with HFrEF. In addition, because of the retrospective nature of the study, the main exposure (clinical profiles) might have been misclassified. However, use of recommended therapies in our cohort was comparable to or higher than that reported in recent registries and clinical trials (24,25). Second, INTERMACS profiles, in contrast to quantitative risk scores, are based on clinical information, and, therefore, assignment of profiles can be subjective. In a study of 212 health care providers (26), assignment of INTERMACS profiles for 5 clinical scenarios varied widely among respondents. With the caveat that our study was retrospective, concordance among physicians for clinical profile assignment was modest. Third, the number of patients with worse clinical profiles was relatively small, and, therefore, our estimates for mortality and hospitalization rates for these patients are less precise. Fourth, although mortality data are incorporated consistently into Emory Clinical Data Warehouse data, data for outside hospitalizations are not, and, therefore, our estimates for hospitalizations and days spent in hospital are probably underestimates. Our sensitivity analyses based on residential ZIP codes suggest that, although we have probably missed some all-cause hospitalizations due to outside admissions, this effect was less prominent for HF-related hospitalizations and most likely equally distributed across clinical profiles. Also, the classification of hospitalizations was based on administrative data and not on individual medical record review, and, therefore, HF hospitalizations might have been misclassified. Finally, although the Emory Clinical Data Warehouse database is linked to multiple sources for mortality and is regularly updated, data for mode of death were not available for all patients in our study, and, therefore, we could not provide cause-specific mortality data.
A simple clinical classification system using INTERMACS profiles to identify and characterize advanced HF among non–inotrope-dependent outpatients with HFrEF provided significant information for subsequent mortality and hospitalization rates in this single-center study. Prospective multicenter data providing robust estimates of anticipated survival, hospitalization rates, and health-related quality of life could aid in decision-making about advanced HF therapies for these patients, especially patients with intermediate clinical profiles.
COMPETENCY IN MEDICAL KNOWLEDGE: INTERMACS profiles are strongly predictive of 3-year mortality and hospitalization burden in a broader population of non-inotrope-dependent outpatients with heart failure and reduced ejection fraction. Compared with patients in stable Stage C heart failure, INTERMACS profiles 7, 6, and 4 to 5 patients have a 2.5-fold, 4-fold, and 6-fold higher mortality, respectively. Also, hospitalization rates escalate across worsening clinical profiles. These simple clinical profiles can facilitate advanced heart failure awareness and appropriate patient referral and also can help with shared decision-making.
TRANSLATIONAL OUTLOOK 1: Data from a single referral center may not be representative of outcomes among heart failure patients corresponding to these clinical profiles in broader practice settings. Multicenter registries and cohort studies would be able to provide more definitive estimates of anticipated mortality and hospitalization according to INTERMACS-based clinical profiles to help with decision-making.
TRANSLATIONAL OUTLOOK 2: Preliminary data suggest that timely referral of patients transitioning to Stage D heart failure can improve outcomes. A potential strategy to facilitate early referral would be to incorporate clinical profiles as part of standard assessment of heart failure patients in the electronic medical record, with appropriate prompts for further action. Obviously, the benefits of that strategy would have to be tested in a pragmatic clinical trial.
Dr. Butler has consulted for Amgen, Array, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squib, CVRx, G3 Pharmacautical, Innolife, Janssen, Luitpold, Medtronic, Merck, Novartis, Relypsa, StealthPeptide, SC Pharma, Vifor, and ZS Pharma. All other authors have reported that they have no relationships with industry relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- heart failure
- heart failure with reduced ejection fraction
- International Classification of Diseases-9th Revision-Clinical Modification
- Interagency Registry for Mechanically Assisted Circulatory Support
- left ventricular assist devices
- left ventricular ejection fraction
- mechanical circulatory support
- New York Heart Association
- rate ratio
- Received February 21, 2018.
- Accepted March 27, 2018.
- 2018 American College of Cardiology Foundation
- Gorodeski E.Z.,
- Chu E.C.,
- Chow C.H.,
- Levy W.C.,
- Hsich E.,
- Starling R.C.
- Lanfear D.E.,
- Levy W.C.,
- Stehlik J.,
- et al.
- Barge-Caballero E.,
- Segovia-Cubero J.,
- Almenar-Bonet L.,
- et al.
- Stewart G.C.,
- Kittleson M.M.,
- Patel P.C.,
- et al.
- Writing Committee M,
- Yancy C.W.,
- Jessup M.,
- et al.
- ↵Centers for Medicare and Medicaid Services. Chronic conditions data warehouse: condition categories. Available at: www.ccwdata.org/web/guest/condition-categories. Accessed October 3, 2016.
- ↵UAB School of Medicine. Intermacs Documents. Available at: https://www.uab.edu/medicine/intermacs/intermacs-documents. Accessed April 9, 2018
- ↵Healthcare cost and utilization project (HCUP). HCUP clinical classifications software (CCS) for ICD-9-CM. Available at: http://hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed October 3, 2016.
- Austin P.C.,
- Pencinca M.J.,
- Steyerberg E.W.
- Estep J.D.,
- Starling R.C.,
- Horstmanshof D.A.,
- et al.
- Alba A.C.,
- Agoritsas T.,
- Jankowski M.,
- et al.
- Freitas P.,
- Aguiar C.,
- Ferreira A.,
- Tralhao A.,
- Ventosa A.,
- Mendes M.
- Kalogeropoulos A.P.,
- Georgiopoulou V.V.,
- Giamouzis G.,
- et al.
- Evans R.S.,
- Kfoury A.G.,
- Horne B.D.,
- et al.
- Chioncel O.,
- Lainscak M.,
- Seferovic P.M.,
- et al.
- Cowger J.,
- Shah P.,
- Stulak J.,
- et al.