New York Heart Association Class vs Patient-Reported Outcomes in Heart Failure (2024)

Key Points

Question In routine US clinical practice, how does the New York Heart Association (NYHA) class compare with patient-reported outcomes for serial assessment of health status among patients with heart failure with reduced ejection fraction?

Findings During outpatient follow-up of 12 months in this cohort study, 75% of 2872 patients had a clinically meaningful change in Kansas City Cardiomyopathy Questionnaire Overall Summary Score (KCCQ-OS) of 5 or more points, whereas 35% had a change of 1 class or more in NYHA class. Improvement in KCCQ-OS of 5 or more points was independently associated with decreased mortality, whereas improvement in NYHA class was not.

Meaning Compared with the clinician-assigned NYHA class, the patient-reported KCCQ-OS is more likely to detect meaningful change in health status over time, and changes in KCCQ-OS may have more prognostic value than changes in NYHA class.

Importance It is unclear how New York Heart Association (NYHA) functional class compares with patient-reported outcomes among patients with heart failure (HF) in contemporary US clinical practice.

Objective To characterize longitudinal changes and concordance between NYHA class and the Kansas City Cardiomyopathy Questionnaire Overall Summary Score (KCCQ-OS), and their associations with clinical outcomes.

Design, Setting, and Participants This cohort study included 2872 US outpatients with chronic HF with reduced ejection fraction across 145 practices enrolled in the CHAMP-HF registry between December 2015 and October 2017. All patients had complete NYHA class and KCCQ-OS data at baseline and 12 months. Longitudinal changes and correlations between the 2 measure were examined. Multivariable models landmarked at 12 months evaluated associations between improvement in NYHA and KCCQ-OS from baseline to 12 months with clinical outcomes occurring from months 12 through 24. Statistical analyses were performed from March to August 2020.

Exposure Change in health status, as defined by 12-month change in NYHA class or KCCQ-OS.

Main Outcomes and Measures All-cause mortality, HF hospitalization, and mortality or HF hospitalization.

Results In total, 2872 patients were included in this analysis (median [interquartile range] age, 68 [59-75] years; 872 [30.4%] were women; and 2156 [75.1%] were of White race). At baseline, 312 patients (10.9%) were NYHA class I, 1710 patients (59.5%) were class II, 804 patients (28.0%) were class III, and 46 patients (1.6%) were class IV. For KCCQ-OS, 1131 patients (39.4%) scored 75 to 100 (best health status), 967 patients (33.7%) scored 50 to 74, 612 patients (21.3%) scored 25 to 49, and 162 patients (5.6%) scored 0 to 24 (worst health status). At 12 months, 1002 patients (34.9%) had a change in NYHA class (599 [20.9%] with improvement; 403 [14.0%] with worsening) and 2158 patients (75.1%) had a change of 5 or more points in KCCQ-OS (1388 [48.3%] with improvement; 770 [26.8%] with worsening). The most common trajectory for NYHA class was no change (1870 [65.1%]), and the most common trajectory for KCCQ-OS was an improvement of at least 10 points (1047 [36.5%]). After adjustment, improvement in NYHA class was not associated with subsequent clinical outcomes, whereas an improvement of 5 or more points in KCCQ-OS was independently associated with decreased mortality (hazard ratio, 0.59; 95% CI, 0.44-0.80; P < .001) and mortality or HF hospitalization (hazard ratio, 0.73; 95% CI, 0.59-0.89; P = .002).

Conclusions and Relevance Findings of this cohort study suggest that, in contemporary US clinical practice, compared with NYHA class, KCCQ-OS is more sensitive to clinically meaningful changes in health status over time. Changes in KCCQ-OS may have more prognostic value than changes in NYHA class.

The New York Heart Association (NYHA) functional classification system has been a cornerstone nomenclature for quantifying the health status of patients with heart failure (HF) for almost a century, and remains foundational to eligibility criteria for contemporary HF trials and application of clinical guidelines.¹ However, despite widespread use, the NYHA functional classification may carry considerable limitations.^2-4 For example, it assesses patients’ functional class through the lens of clinicians, who may vary in their rigor of acquiring this information, resulting in marked variability across clinicians.⁴

Since the initial development of the NYHA classification, patient-reported outcomes (PROs) have evolved substantially.² Given the increasing availability of validated HF-specific PRO instruments, such as the Kansas City Cardiomyopathy Questionnaire (KCCQ), the appropriate role and relative importance of clinician-assessed NYHA class, as compared with directly assessing health status from patients in clinical practice, is unclear. The CHAMP-HF (Change the Management of Patients With Heart Failure) registry represents a novel opportunity to compare the role and clinical value of NYHA class and PROs in contemporary US clinical practice. This cohort study aimed to (1) assess longitudinal changes and level of agreement between NYHA functional class and PROs for quality of life and symptoms, (2) identify factors associated with discordance between NYHA class and PROs, and (3) compare the prognostic value associated with changes in HF status as defined by changes in NYHA class vs KCCQ.

The design of the CHAMP-HF registry has been previously described.⁵ In brief, CHAMP-HF was a prospective, observational, noninterventional study that enrolled adult outpatients with HF with reduced ejection fraction (HFrEF) in the US between December 2015 and October 2017. Eligible patients had received a diagnosis of chronic HF, had a left ventricular ejection fraction of 40% or lower on most recent imaging within 12 months of enrollment, and were receiving at least 1 oral medication for HF at study enrollment. The registry was conducted in accordance with the Declaration of Helsinki⁶ and with institutional review board or ethics committee approval at all sites. All patients provided written informed consent that was obtained in a manner consistent with the Declaration of Helsinki. No one received compensation or was offered any incentive for participating in this study.

The NYHA class consists of a 4-tier schema of class I through class IV, with class IV denoting worst functional status (ie, HF symptoms at rest).⁷ Four PROs were evaluated in the current study as follows: (1) KCCQ Overall Summary Score (KCCQ-OS), a 12-item form composed of 4 domains (ie, physical limitation, symptom frequency, quality of life score, and social limitation) with the overall score and each individual domain scored 100 to 0, with 100 indicating the best health status and 0 indicating the worst; (2) EuroQoL 5-dimensions (EQ-5D) utility index, which reflects the degree of patient difficulty with mobility, self-care, usual activity, pain or discomfort, and anxiety or depression, with a maximal score of 1 designating perfect health, 0 equivalent to death, and lower than 0 worse than death; (3) EQ-5D visual analog scale (VAS), a global assessment of health status ranging from 100 to 0, with 0 indicating worst health status; and (4) number of HF symptoms, that is, the number of symptoms reported by patients among the following 5 options: dyspnea at rest, dyspnea with exertion, orthopnea, edema, and exercise intolerance. In addition, 3 KCCQ subscores were evaluated: (1) KCCQ Clinical Summary Score, which is equal to the mean of the scores from the physical limitation and symptom frequency domains; (2) KCCQ Physical Limitation Score, that is, the score from the physical limitation domain of the KCCQ-OS; and (3) KCCQ Symptom Frequency Score, the score from the symptom frequency domain of the KCCQ-OS.

There are no established ranges for KCCQ or other PRO scores that map to each of the 4 NYHA classes. Thus, discordance among the NYHA class, KCCQ-OS, and EQ-5D scores was determined at baseline and 12 months using a prespecified framework. To define discordance, each scale was organized into 4 levels from “best” to “worst,” as outlined in eFigure 1 in the Supplement. Prespecified cut points for the KCCQ-OS and EQ-5D scores were chosen for consistency with prior publications and ease of communication, and reflected the investigators’ best judgment.^8,9 For each pair of measures, concordance was defined as each scale registering the same level. Mild discordance was defined when the pair of measures differed by 1 level, and moderate or severe discordance was defined as measures differing by 2 to 3 levels. Furthermore, recognizing potential bias based on cut points selected in categorical analyses of discordance, complementary analyses of correlation between continuous NYHA and PROs were performed, as described in the Statistical Analysis.

Clinical outcomes included all-cause mortality, HF hospitalization, and the composite of all-cause mortality or HF hospitalization. Outcomes were determined by review of the medical record.

Patient characteristics were compared by degree of discordance between NYHA and KCCQ-OS at baseline, with patients assigned to 1 of 3 groups (concordance, mild discordance [differed by 1 level], and moderate to severe discordance [differed by 2-3 levels]). Recognizing that discordance can occur in 2 directions (eg, NYHA worse than KCCQ-OS or NYHA better than KCCQ-OS), for each pair of measures, we determined the proportions of patients differing in 1, 2, or 3 levels in both directions.

Correlations between NYHA class, KCCQ scores, EQ-5D scores, and number of symptoms were assessed at baseline and 12 months. Specifically, correlations between NYHA class and KCCQ-OS were calculated, and correlations were calculated for each of these core measures with remaining PROs. Spearman rank correlation, a nonparametric test appropriate for both ordinal data (NYHA class, number of symptoms) and continuous but nonnormal data (KCCQ scores, EQ-5D scores), was used.

Changes in NYHA class and KCCQ-OS scores between baseline and 12 months were summarized in categorical and continuous analyses. For NYHA class, change categories included worsening of 2 or more classes, worsening of 1 class, no change, improvement of 1 class, and improvement of 2 or more classes. Recognizing that a change of 5 or more points in KCCQ-OS is considered clinically significant, change categories for KCCQ-OS included worsening of 10 or more points, worsening of 5 to 9 points, no significant change (<5-point change), improvement of 5 to 9 points, and improvement of 10 or more points.¹⁰ For each change category of NYHA class and KCCQ-OS, median (interquartile range) 12-month change in other measures were calculated.

Multivariable models were constructed to determine whether certain patient characteristics were associated with greater likelihood of discordance between clinician-assessed and patient-reported assessments. These analyses included the NYHA class, KCCQ-OS, and EQ-5D VAS. For each pair of measures, any level of discordance (ie, ≥1 level) was considered a binary outcome, with concordance considered a nonevent. Logistic regression models were used, with generalized estimating equations to account for clustering within sites. Separate models were constructed for baseline and 12-month time points. Values for candidate variables were taken from baseline for baseline models and taken from 12 months (or most recent available) for the 12-month models. Model selection was based on backward elimination. Variables with a P ≥ .05 were removed, with the highest P value removed first, and subsequent assessment continuing with remaining variables. Model selection continued until all remaining variables were P < .05.

Unadjusted and adjusted Cox proportional hazards models were used to separately evaluate associations between improvements in NYHA class and KCCQ-OS and clinical outcomes. Models were landmarked at 12 months so that change in each measure between baseline and 12 months could be associated with clinical outcomes occurring between months 12 through 24. Models were adjusted for 20 prespecified covariates, including age, sex, race/ethnicity, total household income, body mass index (BMI), systolic blood pressure, heart rate, ejection fraction, glomerular filtration rate, atrial fibrillation, coronary artery disease, diabetes, chronic obstructive pulmonary disease, HF hospitalization in the prior 12 months, cardiac resynchronization therapy, implantable cardioverter-defibrillator, angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker, sacubitril/valsartan, evidence-based β-blocker, and mineralocorticoid receptor antagonist. Time-updated covariate values collected at 12 months (or most recent available) were used. Robust sandwich variance estimators were used to account for clustering within site. Hazard ratios and P values were calculated. Statistical analyses were performed from March to August 2020 using SAS, version 9.4 (SAS Institute Inc). A 2-tailed P < .05 was considered statistically significant.

From the total sample of 5009 patients, the present analysis included 2872 patients with complete NYHA and KCCQ-OS data at baseline and 12 months (eFigure 2 in the Supplement). Overall, median age was 68 years (interquartile range [IQR], 59-75 years), 872 (30.4%) were women, 2156 (75.1%) were of White race, and the median EF was 30% (IQR, 23%-35%) (eTable 1 in the Supplement).

At baseline, 312 patients (10.9%) were NYHA class I, 1710 patients (59.5%) were NYHA class II, 804 patients were NYHA class III (28.0%), and 46 patients (1.6%) were NYHA class IV. For KCCQ-OS, 1131 patients (39.4%) scored 75 to 100, 967 patients (33.7%) scored 50 to 74, 612 patients (21.3%) scored 25 to 49, and 162 patients (5.6%) scored 0 to 24 (eTables 2 and 3 in the Supplement).

At baseline, 1085 patients (37.8%) showed concordance between NYHA class and KCCQ-OS, whereas 1494 patients (52.0%) had mild discordance by 1 level and 293 patients (10.2%) had moderate to severe discordance by 2 to 3 levels (Table 1). Patient characteristics were generally similar across discordant groups, with few exceptions.

For NYHA class, KCCQ-OS, and other PROs, the proportions of patients with varying directionality and severity of discordance are given in eTables 4 and 5 in the Supplement. At baseline, among patients with discordance between NYHA class and KCCQ-OS, most had worse NYHA class (1222 of 1787 [68.4%]) rather than worse KCCQ-OS (565 of 1787 [31.6%]). Discordance with the EQ-5D index and EQ-5D VAS was more common for NYHA class (EQ-5D index, 2160 [75.2%] and EQ-5D VAS, 1854 [64.6%]) than for KCCQ-OS (EQ-5D index, 1500 [52.2%] and EQ-5D VAS, 1364 [47.5%]). For both NYHA class and KCCQ-OS, discordance with EQ-5D scales was mostly owing to better EQ-5D status. Discordance with the number of symptoms was comparable for NYHA class (1850 [64.4%]) and KCCQ-OS (2042 [71.1%]), and in both cases was more frequently due to worse number of symptoms. For all measures, the degree and direction of discordance was generally similar at both baseline and 12 months.

In assessment of correlation coefficients between NYHA class, KCCQ-OS, and other PROs, all measures showed statistically significant correlations with each other at baseline and 12 months (eTable 6 in the Supplement). The magnitude of correlation between NYHA class and KCCQ-OS at baseline was modest (ρ = 0.33; P < .001). Likewise, correlations for the NYHA class with other KCCQ subscales and PROs were modest (ρ ranging from 0.22 to 0.35 at baseline), whereas correlations between KCCQ-OS and other PROs were stronger (ρ ranging from 0.35 to 0.66 at baseline).

At baseline, older patients had a higher likelihood of NYHA class being worse than KCCQ-OS (eTable 7 in the Supplement). By contrast, being a woman or of Hispanic ethnicity and having higher BMI, chronic obstructive pulmonary disease, and coronary artery disease were each associated with a lower likelihood of worse NYHA class than KCCQ-OS. At 12 months, age and Hispanic ethnicity continued to be associated with the likelihood of NYHA class being worse than KCCQ-OS, with addition of higher left ventricular ejection fraction associated with lower likelihood of NYHA class–KCCQ-OS category discordance.

In terms of KCCQ-OS being worse than NYHA class, lower household income and higher heart rate up to 65 beats per minute were associated with higher risk of discordance at baseline, whereas older age up to 75 years was associated with lower risk of discordance. At 12 months, lower household income, chronic obstructive pulmonary disease, diabetes, and being a woman were associated with higher risk of discordance, and no factors were associated with lower risk of discordance. Factors associated with discordance between NYHA class and EQ-5D VAS and discordance between KCCQ-OS and EQ-5D VAS are presented in eTables 8 and 9 in the Supplement.

In comparing status at baseline and 12 months, 1002 patients (34.9%) had a change in NYHA class, whereas 2158 (75.1%) had a meaningful change in KCCQ-OS (Figure 1). For NYHA class, 599 patients (20.9%) had any improvement, and 403 (14.0%) had any worsening, with the large majority of changes representing a 1-class change. For KCCQ-OS, 1388 patients (48.3) had at least a 5-point improvement, and 770 (26.8%) had at least a 5-point worsening. Overall, the most common scenario for NYHA class trajectory was no change (1870 [65.1%]), and the most common scenario for KCCQ-OS trajectory was at least a 10-point improvement (1047 [36.5%]). Patterns of longitudinal change in NYHA class and KCCQ-OS were similar among patients enrolled from sites with cardiologist investigators vs noncardiologist investigators (ie, family medicine, internal medicine, and other) (eFigure 3 in the Supplement).

Of patients with no change in NYHA class, the median change in KCCQ-OS was 3 (IQR, −6 to 16) (Table 2). For patients who worsened by 2 or more NYHA classes, the median change in KCCQ-OS was 2 points (IQR, −24 to 9 points), and the median change in EQ-5D VAS was −4 points (IQR, −11 to 8 points). For patients who improved by 2 or more NYHA classes, the median change in KCCQ-OS was 11 points (IQR, 0 to 31 points), and the median change in EQ-5D VAS was 0 points (IQR, −9 to 11 points).

For all categories of KCCQ-OS change ranging from worsening of 10 or more points to improvement of 10 or more points, the median change in NYHA class was 0 (IQR, 0-0 for all categories, except 0-1 for improvement of 10 or more) (Table 3). Changes in KCCQ subscores were consistent with changes in KCCQ-OS. Changes in EQ-5D VAS scores ranged from a median of −9 points (IQR, −20 to 2 points) for patients with worsening of 10 or more points in KCCQ-OS to 6 points (IQR, −3 to 20 points) for patients with improvement of 10 or more points in KCCQ-OS.

In landmark analysis at the 12-month time point, there were 206 (7.2%) deaths, 231 (8.0%) HF hospitalizations, and 398 (13.9%) deaths or HF hospitalizations between months 12 and 24. In unadjusted and adjusted analyses, there were no significant associations between improvement in NYHA class and subsequent all-cause death, HF hospitalization, or the composite of death or HF hospitalization during 1 year of follow-up (Figure 2). By contrast, after adjustment, improvement in KCCQ-OS of 5 points or more was associated with decreased risk of all-cause mortality (hazard ratio, 0.59; 95% CI, 0.44-0.80; P < .001) and the composite of all-cause death or HF hospitalization (hazard ratio, 0.73; 95% CI, 0.59-0.89; P = .002). There was no significant association between improvement in KCCQ-OS and HF hospitalization.

In this large contemporary US registry of outpatients with HFrEF, there was discordance in health status between the clinician-assigned NYHA class and the patient-reported KCCQ-OS. This discordance was moderate to severe in more than 10% of patients and was mostly due to assigned NYHA classes being disproportionately worse than KCCQ-OS. In multivariate analysis, multiple patient factors were independently associated with higher risk of discordance, including age, sex, household income, and comorbidities. During 12 months of follow-up, the majority of patients had meaningful changes in KCCQ-OS, yet most patients had no change in NYHA class. Improvement in NYHA class was not associated with subsequent clinical outcomes, whereas improvement in KCCQ-OS was independently associated with lower risks of death and the composite of death and HF hospitalization, but not with HF hospitalization alone.

To our knowledge, we present the first comprehensive direct comparison of NYHA class and PROs in contemporary US clinical practice. Prior large studies have assessed NYHA class and KCCQ in clinical trial cohorts, but data from large clinical practice-based patient populations have not been available.^11-13 These data from the CHAMP-HF registry suggest relative advantages of measuring health status via serial KCCQ assessments, as compared with NYHA class, in the longitudinal care of patients with HFrEF. Although numerous reliable clinical risk markers for HF already exist, the sheer magnitude of the 41% lower risk of mortality independently associated with improvement in KCCQ-OS in the present study is notable. The relative strength of this association is further magnified by the lack of a corresponding association between clinical outcomes and NYHA class. A similarly clear distinction between KCCQ-OS and NYHA class was observed when comparing changes over time. Although cross-sectional analyses showed statistically significant correlations between the 2 measures, more than twice as many patients had meaningful changes in KCCQ-OS compared with NYHA class during 12 months, and meaningful changes in KCCQ were not reliably reflected in the NYHA class. Even among patients with improvement or worsening of 10 points or more in KCCQ-OS, there was generally no corresponding change in NYHA class. Although the exact reasons for such dissociations are unclear, limited change in clinician assessment of NYHA class would be consistent with other examples of clinician inertia in ambulatory HF care.¹⁴ In contrast to NYHA class, KCCQ-OS was not only more sensitive for detecting change in health status, but such patient-reported changes were further validated by strong prognostic value.

To date, the uptake of PRO measures within HF care has been slow and varied, even though they have been recommended as measures of health care quality.^15-17 Barriers have been proposed, including concerns regarding logistics of data collection, results interpretation, and perceived utility and value of PROs by clinicians.¹⁵ This may stand in contrast with NYHA class, which can conceivably be assigned in seconds by a clinician familiar with the patient. Nonetheless, routine collection of PROs within the HF clinic has proven feasible at some centers, with mean PRO assessment times of less than 7 minutes and computerized scoring within the electronic health record.¹⁸ Emphasizing clinical education, automated data collection and scoring, and easily interpretable presentation may be critical elements for more widespread adoption of PRO measures.¹⁵

Aside from prognostic implications, these data highlight the potential for clinician and patient perspectives to differ regarding health status, with discordance occurring in both directions (ie, clinician more favorable than patient and vice versa). Such dissociations were more likely in certain demographic subsets based on age, sex, race/ethnicity, and household income, raising the question of clinician bias in assigning the NYHA class and the potential contribution to disparities in care.¹⁹ Likewise, these findings could reflect challenges in patient-clinician communication in the care of particular patient subsets. Certain comorbidities may also increase risk for discordance. For example, it may be difficult for clinicians to reliably distinguish pulmonary vs HF contributions to health status among patients with HF and chronic obstructive pulmonary disease, and the NYHA class assessment may be limited to the perceived HF symptoms. Nonetheless, given the weight NYHA class carries in application of guideline recommendations and patient eligibility for clinical trials, any potential for bias in clinician assignment of NYHA class and dissociation from PROs is highly relevant. By comparison, despite the clear association with clinical outcomes observed here, little to no guidance is available for using PROs to inform HF management decisions. To best ensure patient-centered care, future studies may consider novel KCCQ-based eligibility criteria.

Limitations of this study should be acknowledged. First, there is no clear range of KCCQ scores that matches the 4 NYHA classes, in part because of the variability in assigning NYHA class. Thus, we selected, a priori, 4 equally sized 25-point ranges of the PROs, which are consistent with prior literature and may facilitate ease of interpretation.⁹ Moreover, recognizing potential bias with arbitrary cut points, we assessed correlation between continuous NYHA and PROs scales in complementary analyses. Second, although study sites included a broad set of HF, cardiology, and primary care practices, these data reflect sites and patients who elected to participate in the registry, and thus may not be generalizable to patients receiving care at all US HF clinics. Although exact concordance between NYHA class and PROs may not be expected, the magnitude of difference found here is noteworthy and supports the conclusion that clinicians and patients rate patient health status differently. Third, this analysis should be interpreted in the context of including patients with complete NYHA class and PRO data at baseline and 12 months, thus excluding patients with interval death or with missing data. Survival bias may have contributed to NYHA class and KCCQ-OS generally showing more improvement than worsening over time.

In this contemporary US outpatient HFrEF registry, as compared with NYHA class, KCCQ-OS was more likely to detect meaningful changes in health status during 12 months of follow-up, and even patients with large changes in KCCQ-OS generally had no change in NYHA class. Improvement in KCCQ-OS was independently associated with lower risk of subsequent mortality and the composite of mortality and HF hospitalization, whereas improvement in NYHA class was not associated with clinical outcomes.

Accepted for Publication: December 29, 2020.

Published Online: March 24, 2021. doi:10.1001/jamacardio.2021.0372

Corresponding Author: Gregg C. Fonarow, MD, Ahmanson-UCLA Cardiomyopathy Center, University of California Los Angeles, 10833 LeConte Ave, Room 47-123 CHS, Los Angeles, CA 90095 (gfonarow@mednet.ucla.edu).

Author Contributions: Drs Greene and Fonarow had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Greene, Butler, Spertus, Albert, Patterson, Williams, Fonarow.

Acquisition, analysis, or interpretation of data: Greene, Butler, Spertus, Hellkamp, Vaduganathan, DeVore, Duffy, Patterson, Thomas, Hernandez, Fonarow.

Drafting of the manuscript: Greene, Butler, Hellkamp.

Critical revision of the manuscript for important intellectual content: Greene, Butler, Spertus, Vaduganathan, DeVore, Albert, Duffy, Patterson, Thomas, Williams, Hernandez, Fonarow.

Statistical analysis: Hellkamp, DeVore.

Obtained funding: Hernandez, Fonarow.

Administrative, technical, or material support: Hernandez, Fonarow.

Supervision: Greene, Butler, Patterson, Thomas, Fonarow.

Conflict of Interest Disclosures: Dr Greene reported receiving research support from the American Heart Association; Amgen; AstraZeneca; Bristol Myers Squibb; Merck & Co; the National Heart, Lung, and Blood Institute (NHLBI); and Novartis and receiving personal fees from Amgen, Cytokinetics, and Merck & Co. Dr Butler reported receiving personal fees from Abbott, Adrenomed, Amgen, Applied Therapeutics, Arena Pharma, Array, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Cardior, CVRx, Eli Lilly, G3 Pharma, Imbria, Impulse Dynamics, Innolife, Janssen, LivaNova, Luitpold, Medtronic, Merck & Co, Novartis, Novo Nordisk, Relypsa, Sequana Medical, V-Wave Limited, and Vifor outside the submitted work. Dr Spertus reported receiving personal fees from Novartis during the conduct of the study; receiving personal fees from Amgen, AstraZeneca, Bayer, Blue Cross Blue Shield of Kansas City, Janssen, Merck & Co, Myokardia, and United Healthcare outside the submitted work; receiving royalties for copyright of SAQ, Kansas City Cardiomyopathy Questionnaire, and PAQ; and having equity in Health Outcomes Sciences. Ms Hellkamp reported receiving grants from Novartis during the conduct of the study. Dr Vaduganathan reported receiving research grant support or personal fees from American Regent, Amgen, AstraZeneca, Bayer AG, Baxter Healthcare, Boehringer Ingelheim, Cytokinetics, and Relypsa; and participating on committees for studies sponsored by Galmed, the National Institutes of Health, and Novartis. Dr DeVore reported receiving grants from Novartis and personal fees from Novartis during the conduct of the study; receiving grants from American Heart Association, Amgen, AstraZeneca, Bayer, Intra-Cellular Therapies, American Regent Inc, the NHLBI, Novartis, and PCORI; receiving personal fees from Amgen, AstraZeneca, Bayer, CareDx, InnaMed, LivaNova, Mardil Medical, Procyrion, scPharmaceuticals, Story Health, and Zoll Consulting; and receiving nonfinancial support from Abbott outside the submitted work. Dr Albert reported receiving personal fees from Novartis during the conduct of the study; receiving personal fees from Amgen and AstraZeneca outside the submitted work; and receiving nonfinancial support from Boston Scientific. Dr Duffy reported being an employee of Novartis Pharmaceutical Corporation with minimal stock options outside the submitted work. Dr Patterson reported receiving personal fees from Novartis during the conduct of the study; receiving personal fees from Alnylam outside the submitted work; and receiving research funding from Amgen, Boehringer Ingelheim, Bristol Myers Squibb, Merck, Novartis, and Otsuka. Dr Hernandez reported receiving grants from American Regent, AstraZeneca, Boehringer Ingelheim, Merck & Co, Novartis, and Verily during the conduct of the study; receiving research support from AstraZeneca, GlaxoSmithKline, Luitpold, Merck & Co, and Novartis; and receiving personal fees from AstraZeneca, Bayer, Boehringer Ingelheim, Boston Scientific, Cytokinetics, Merck & Co, Myokardia, Novartis, and Sanofi outside the submitted work. Dr Fonarow reported receiving grants from the National Institutes of Health; receiving personal fees from Novartis during the conduct of the study; and receiving personal fees from Abbott, Amgen, AstraZeneca, Bayer, CHF Solutions, Edwards, Janssen, Medtronic, Merck & Co, and Novartis outside the submitted work. No other disclosures were reported.

Funding/Support: This analysis and the CHAMP-HF registry were funded by Novartis Pharmaceuticals Corporation.

Role of the Funder/Sponsor: The sponsor participated in the design, conduct, and management of the study; and the preparation, review, and approval of this manuscript. The sponsor had no role in the collection and analysis of data, or the interpretation of the data. The sponsor did not have the right to prevent submission or publication of the data.

Disclaimer: Dr Thomas is the Assistant Editor for Statistics of JAMA Cardiology, Dr Hernandez is an Associate Editor for JAMA Cardiology, and Dr Fonarow is the Associate Editor of Health Care Quality and Guidelines for JAMA Cardiology, but they were not involved in any of the decisions regarding review of the manuscript or its acceptance.