The Korean Acute Heart Failure (KorAHF) registry includes the data of 5,625 hospitalized patients with acute HF treated at 10 tertiary hospitals in Republic of Korea between March 2011 and February 2014.11 This registry included hospitalized patients who had signs or symptoms of HF and at least one objective sign of lung congestion or objective findings of left ventricular systolic dysfunction or structural heart disease. The only exclusion criterion of this registry was the withdrawal of consent.11 For this study, we only selected patients with HFpEF with 1 year of follow-up data. HFpEF was defined as a left ventricular ejection fraction ≥ 50%. AF was defined based on electrocardiographic confirmation of AF at admission or history of AF.

To assess adherence to the guidelines, we adopted and modified the QUality of Adherence to guideline recommendations for LIFe-saving treatment in heart failure surveY (QUALIFY) international survey.612 A global guideline adherence score, QUALIFY, was constructed based on physicians' adherence to guidelines for prescriptions of the following classes of medications: ACC/AHA/ESC guidelines recommended ACEIs, ARBs, β-blockers, MRAs, and ivabradine, for patients with HFrEF with normal sinus rhythm, unless contraindicated or not tolerated.461213 However, ivabradine was not included from this study because it was not available in South Korea during the study period. Additionally, anticoagulation therapy was included as a guideline-recommended therapy since anticoagulation is an established therapeutic strategy to reduce the risk of stroke in AF patients.1415 In addition to reducing the risk of stroke, anticoagulants markedly reduce the risk of death.16 The adherence was defined as the actual prescription of guideline recommended therapy that should have theoretically been prescribed for each of ACEIs or ARBs, β-blockers, MRAs, and an oral anticoagulant (if indicated). Only guideline-recommended β-blockers (including bisoprolol, carvedilol, sustained-release metoprolol succinate, and nebivolol) were recognized as adherence to guidelines. The non-adherence was defined as non-prescription of therapy in the absence of specified contraindications. Non-prescription of the recommended therapies because of specific contraindications or intolerance was also scored for adherence to guidelines.12 Additionally, non-prescription of anticoagulation for a CHA2DS2-VASc score < 2 was defined as adherence to guidelines. The level of adherence was defined as good for use of more than half of the indicated medications and poor for use of ≤ 50% of the indicated medications.61217

In this study, the outcomes were defined as all-cause mortality, rehospitalization due to aggravated HF, and a composite endpoint of mortality and HF rehospitalization at 60 day and 1 year. The attending physicians collected follow-up data. For the patients who were not directly followed by, the outcome data were ascertained via transtelephonic monitor or using the National Death Records database. All clinical events were monitored and verified by a Clinical Event Committee comprising independent experts on HF who did not participate in the patient enrollment part of this study.11

Data are presented as the mean ± standard deviation for continuous variables and as numbers and frequencies for categorical variables. For comparison between groups, χ² test was used for categorical variables and the unpaired Student t-test was used for continuous variables. To address the systematic differences between the two groups, we used inverse probability of treatment weighting (IPTW) analyses to reduce imbalances in measured confounders.18 The propensity score was estimated using a logistic regression model with the following covariates: sex; type of HF (acute decompensated HF or de novo HF); New York Heart Association functional class; history of hypertension, diabetes, ischemic heart disease, chronic obstructive pulmonary disease, chronic kidney disease, or cardiovascular accident; etiology of HF; right bundle branch block; left bundle branch block and corrected QT time on electrocardiogram; systolic blood pressure; diastolic blood pressure; serum sodium level; hemoglobin level; creatinine level; and left ventricular ejection fraction. The estimated propensity scores were used to calculate the inverse probability of the level of adherence for each patient. Weighting by the inverse probability of treatment results in an artificial population in which baseline covariates are independent of treatment status. Post-weighting balance in covariates between treatment groups was evaluated using the standardized difference approach. The balance between the treatment groups was assessed using weighted standardized mean differences (SMDs). An SMD value ≤ 20% was considered acceptable. The association between the degree of guideline adherence and clinical outcomes was analyzed using weighted Kaplan-Meier curves and Cox proportional hazards regression with a robust variance estimator to calculate weighted hazard ratios (wHRs) (95% confidence intervals [CIs]). Cox proportional hazards regression models were adjusted for sex; age; history of hypertension, diabetes mellitus, ischemic heart disease, and chronic obstructive pulmonary disease; type of HF (de novo vs. acute decompensated HF); New York Heart Association functional class; systolic blood pressure; heart rate; creatinine level; and left ventricular ejection fraction. In all cases, a P value ≤ 0.05 was considered statistically significant. Statistical analyses were performed using SPSS Version 23.0 (IBM Corp., Armonk, NY, USA).

The study protocol was approved by the ethics committee at each hospital and the Wonju Christian Hospital, Wonju College of Medicine, Yonsei University (Wonju, Korea; Approval No. CR311003), and was conducted according to the principles of the Declaration of Helsinki. All patients provided written informed consent for participation in the registry.

In the 687 patients without AF, the median age of the cohort was 74 (interquartile range [IQR]: 63–81) years, and 58.8% of these patients were women. Modified guideline adherence was good in 53.6% and poor in 46.4% of patients. In contrast to our expectation, patients with a poorer guideline adherence had favorable clinical characteristics (Table 1). Patients with poor guideline adherence had lower systolic blood pressure at admission, less prevalent comorbidities (hypertension, diabetes mellitus, and chronic kidney disease), different etiology of HF (less likely to have ischemic heart disease and more likely to have valvular heart disease), unique laboratory characteristics (higher levels of hemoglobin and lower levels of creatinine, B-type natriuretic peptide [BNP], and troponin), and favorable echocardiographic characteristics (smaller left ventricular diastolic and systolic volume and greater preserved ejection fraction). All baseline differences were balanced after IPTW except residual borderline differences in left ventricular end diastolic volume and serum creatinine. In the poor adherence group, the guideline adherence rates were 34.8%, 11.0%, and 14.7% for ACEIs or ARBs, β-blockers, and MRAs, respectively. The prescription rate of loop diuretics was also significantly lower in the poor adherence group than in the good adherence group (Supplementary Table 1).

In the 707 patients with AF, the median age of the cohort was 76 (IQR: 69–82) years, and 62.0% of these patients were women. Modified guideline adherence was good in 33.2% and poor in 66.8% of patients. In contrast to patients without AF, the baseline clinical characteristics of patients with AF were similar according to the degree of adherence, except for the level of BNP, corrected QT interval, and bundle branch block on electrocardiogram (Table 2). All baseline differences were balanced after IPTW. In the poor adherence group, the guideline adherence rates were 42.4%, 25.0%, 32.0%, and 39.2% for ACEIs or ARBs, β-blockers, MRAs, and warfarin, respectively. The calculated equivalent doses of each medication did not significantly differ according to adherence. The prescription rate of loop diuretics was also significantly lower in the poor adherence group than in the other two groups; however, the prescription rate of other medications including amiodarone and digoxin did not significantly differ (Supplementary Table 2).

Of the 687 patients without AF, 111 (16.2%) patients died, and 98 (14.3%) patients were re-hospitalized during the 1-year follow-up period. The incidences of clinical outcomes at 60 days and 1 year did not differ according to the degree of guideline adherence (Table 3). The IPTW adjusted Kaplan-Meier survival curves demonstrated better 60-day composite endpoint-free survival in the poor adherence group (Supplementary Fig. 1). After IPTW, good adherence was associated with poor 60-day composite endpoint-free survival in the multivariable Cox model (wHR, 1.74; 95% CI, 1.01–3.00; P = 0.045). Poor adherence was not associated with any risks for 1-year outcomes when compared with moderate or good adherence (Fig. 1, Supplementary Table 3).

Of the 707 patients with AF, 137 (19.4%) patients were re-hospitalized, and 130 (18.4%) patients died. The incidences of mortality and the composite endpoint during the follow-up period were higher in the poor adherence group than in the good adherence group (Table 4). The IPTW adjusted Kaplan–Meier survival curves demonstrated poor event-free survival among those with poor guideline adherence in terms of 60-day mortality and the composite endpoint. Additionally, the IPTW adjusted multivariate Cox regression analysis revealed that good adherence was significantly associated with the 60-day composite endpoint (wHR, 0.47; 95% CI, 0.27–0.79; P = 0.005) (Fig. 2, Supplementary Table 4). Good adherence was not associated with 1-year clinical outcomes. In the analysis excluding warfarin, good adherence was associated with 60-day re-hospitalization (wHR, 0.60; 95% CI, 0.37–0.98; P = 0.040), 1-year re-hospitalization (wHR, 0.67; 95% CI, 0.48–0.93; P = 0.018), and the composite endpoint (wHR, 0.77; 95% CI, 0.59–0.99; P = 0.041) (Fig. 3, Supplementary Table 5).