Heart failure with preserved ejection fraction (HFpEF) is characterized by a left ventricular ejection fraction (LVEF) of 50% or greater, with evidence of spontaneous or provoked increased LV filling pressures (LVFP), as evidenced by elevated brain natriuretic peptides or hemodynamic measurements [1][2]. HFpEF is a heterogeneous clinical syndrome; pathophysiology differs from heart failure with reduced ejection fraction (HFrEF, EF ≤ 40%) [1][2]. Obesity and diabetes are important risk factors and predictors of HFpEF [3]. High blood pressure is a significant risk factor for the development of HFpEF in all populations; obesity is a particularly pronounced risk factor for patients who identify as African American or Hispanic [4]. Other risk factors for the development of HFpEF include: increased age, female sex, type 2 diabetes, sleep apnea, hypertension, pulmonary hypertension, chronic obstructive pulmonary disease, iron deficiency, coronary artery disease, atrial fibrillation, and dysrhythmias, where the exact pathogenic role of each risk factor is poorly understood [5]. The lifetime risk of HFpEF approximates 19.3%, exceeding the approximate 11.4% lifetime risk of HFrEF, as per data from the second 25-year epoch of the Framingham Heart Study (1990-2014) [6]. The incidence and prevalence of HFpEF are increasing overall, partially due to improved diagnostic capabilities and more uniform diagnostic parameters [1]. Tsao et al. reported an increase of 53% in HFpEF diagnoses among participants in the Framingham Heart Study and Cardiovascular Health Study between 2000 and 2009 [7]. Chang et al. reported an 8.2% increase in HFpEF diagnoses among black women compared to a 5.9% increase for white women, and demonstrated a rise in HFpEF diagnoses for black men with 5.7% and white men with 6.3% [8]. In contrast to this, data from Olmstead County in Minnesota demonstrated a decrease of 28% in the incidence of HFpEF diagnoses between 2000 and 2010 [9]. There are also geographical differences in the worldwide incidence and prevalence of HFpEF. Shiga et al. reported a prevalence of 43% in 1245 patients with HF hospitalized in Japan between 2013-2014 [10], while Wang et al. evaluated 5236 patients in Australia and reported a prevalence of 37.4% [11]. The prevalence of HFpEF was only 16% in the European Society of Cardiology long-term registry and the Asian Sudden Cardiac Death in Heart Failure Registry [12].

A constellation of changes that occur at the cellular, tissue, and organ levels that lead to the pathogenesis of HFpEF. These include structural remodelling of the LV, reduced LV reserve, abnormal hemodynamics, and secondary organ dysfunction [13]. Obesity causes a chronic systemic inflammatory state that induces hemodynamic derangements and hormonal abnormalities, produces subclinical changes in the structure and function of the LV, even in the presence of normal LVEF [14][15]. The most significant hemodynamic finding in HFpEF is elevated LVFP, resulting from LV diastolic dysfunction secondary to incomplete myocardial relaxation, increased ventricular stiffness with subsequent decreased ventricular distensibility, and a shift in the Frank-Starling mechanism [13]. These elevated filling pressures cause remodelling of the left atrium, which is further exacerbated by atrial fibrillation, increasing the risk of pulmonary hypertension, which then induces right ventricular dysfunction, ultimately leading to a spectrum of signs and symptoms seen in HFpEF [16][17].

Although the underlying mechanisms are unknown, sudden cardiac death (SCD) is the most common mode of death in HFpEF [18]. Ventricular arrhythmias (VAs) are commonly reported in HFpEF; however, their burden and mechanisms have not been established [19]. Various randomized controlled trials (RCTs) and prospective registries have investigated the mode of death in HFpEF, where cardiovascular (CV) deaths were responsible for 60% of all deaths in HFpEF, and 30% of all deaths were from non-CV causes [20]. There was a lesser burden of CV deaths and reduced sudden deaths in HFpEF compared to HFrEF [20]. Adabag et al. developed a multivariate Cox regression model, which was based on age, male gender, history of diabetes mellitus and myocardial infarction, left bundle branch block on ECG, and the natural logarithm of N-terminal pro brain natriuretic peptide (NT-pro BNP), and identified patients with HFpEF who had a risk of sudden death over 5 years [21]. Vaduganathan et al. used the TOPCAT trial data to investigate predictors of sudden death in HFpEF; male sex and insulin-treated diabetes mellitus independently predicted risk of sudden death in HFpEF [22]. There were 282 patients with HFpEF in a study by Woolcott et al., who investigated the prevalence of shockable rhythms (ventricular tachycardia/VT or ventricular fibrillation/VF at the time of cardiac arrest), 27% had shockable rhythms, asystole in 43.3% of cases, and pulseless electrical activity in 28.7%. These findings suggested that VAs are the major cause of death in HFrEF, while VAs might play a less significant role in sudden death in HFpEF patients [23]. Several studies investigated the burden of VA in HFpEF. Gutierrez et al. performed ambulatory ECG monitoring or pacemaker interrogations, and non-sustained VT (NSVT) was found in 32.5%, which was associated with a 3.4-fold increased risk of death in HFpEF [24]. Cho et al. found NSVT in 37% of patients with HFpEF, sustained VT in 1/110 HFpEF patients, and covariate-adjusted logistic regression showed HFpEF was associated with increased risk of NSVT [25]. Veldhuisen et al. found that 18% of HFpEF patients had NSVT, and 1/113 developed sustained VT [26]. The mechanisms underlying VAs in HFpEF include the following: 1) Reduced conduction velocity, primarily due to ventricular hypertrophy, which affects the heart’s electrical conduction; 2) Delayed repolarization occurs due to down-regulation of potassium currents in HFpEF hearts, creating a potential for early afterdepolarizations that can trigger arrhythmias, 3) Calcium leakage and altered excitation-contraction coupling in HFpEF can lead to delayed afterdepolarizations, and 4) Increased ventricular fibrosis due to systemic inflammation provides substrates for functional re-entry and further contributes to arrhythmias. Hypertension and the resulting ventricular hypertrophy decrease conduction velocity in HFpEF hearts through a heterogeneous distribution of connexin 43. Collectively, these factors contribute to the development of VAs in patients with HFpEF [20].

It is crucial to understand the predictors of VAs in HFpEF, given that these patients are older, have a high burden of metabolic and systemic comorbidities, and are typically excluded from device trials. It will enable physicians to identify independent risk factors for better prognostication, arrhythmia monitoring, and timely electrophysiologic evaluation in this underrecognized population. Despite increasing recognition of HFpEF, predictors of VAs in this population remain poorly defined, particularly in large real-world datasets. To our knowledge, few large nationwide studies have systematically evaluated predictors of VAs specifically among hospitalized patients with HFpEF. Therefore, we analyzed a large nationally representative cohort from the National Inpatient Sample (NIS) to determine the prevalence of VAs among hospitalized patients with HFpEF, identify independent demographic, cardiometabolic, renal, and electrolyte predictors, and compare inpatient outcomes with patients hospitalized with HFrEF.

We conducted a retrospective cross-sectional study using the NIS database from 2016 through 2018, developed as part of the Healthcare Cost and Utilization Project (HCUP). The NIS represents approximately 20% of all hospital discharges in the United States and incorporates a stratified sampling design that allows for the generation of nationally representative estimates through the application of discharge weights. As the dataset is de-identified and publicly available, institutional review board approval was not required.

We identified adult hospitalizations (≥18 years) with a diagnosis of heart failure using International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes. Patients were categorized into two groups:

Heart failure with preserved ejection fraction (HFpEF): I50.3, I50.30, I50.31, I50.32, I50.33.

Heart failure with reduced ejection fraction (HFrEF): I50.2, I50.20, I50.21, I50.22, I50.23.

Hospitalizations with combined systolic and diastolic heart failure (I50.4, I50.40, I50.41, I50.42, I50.43) were excluded to avoid overlap between phenotypes. Because the NIS is a hospitalization-level database, each admission was treated as an independent observation.

VAs were identified using ICD-10-CM codes in any diagnosis field and included: Ventricular tachycardia (I47.2), Ventricular fibrillation (I49.01), Ventricular flutter (I49.02).

Heart failure diagnoses (HFpEF and HFrEF), VAs, comorbidities, and electrolyte abnormalities were each identified using ICD-10-CM codes present in any diagnosis field of the discharge record, without restriction to the principal (primary) diagnosis position. Consequently, the HFpEF and HFrEF cohorts include both hospitalizations in which heart failure was the primary reason for admission and hospitalizations in which heart failure was a co-existing condition, and VAs were similarly captured regardless of whether they precipitated the admission or developed during hospitalization. This any-position approach maximizes the sensitivity of case ascertainment but does not distinguish the clinical context in which each diagnosis arose; this should be considered when interpreting the composition of the study population and the generalizability of our findings to admissions specifically triggered by decompensated HFpEF or by a VA.

Because HFpEF and HFrEF were defined using administrative ICD-10-CM diagnosis codes rather than direct clinical, echocardiographic, or invasive hemodynamic data, these diastolic and systolic heart-failure codes served as a proxy for the underlying clinical phenotype. The NIS does not report left ventricular ejection fraction, natriuretic peptide levels, or hemodynamic filling pressures; therefore, the codes used to define HFpEF cannot confirm a preserved ejection fraction or verify that patients met contemporary hemodynamic diagnostic criteria for HFpEF, and some misclassification relative to echocardiographically confirmed HFpEF is possible.

The primary outcome was the occurrence of VAs among patients with HFpEF. Secondary outcomes included: In-hospital mortality, Length of hospital stay (LOS), Total hospitalization charges, and discharge disposition.

Additionally, comparative analyses between HFpEF and HFrEF cohorts were performed to evaluate differences in VA prevalence and in-hospital mortality.

Baseline variables included demographic and clinical characteristics known to influence heart failure outcomes and arrhythmogenesis. Demographics included Age, Sex, and race/ethnicity. Clinical Comorbidities included: Hypertension, Diabetes mellitus, Obesity, Hypothyroidism, Chronic kidney disease, Chronic obstructive pulmonary disease, Cirrhosis, and Obstructive sleep apnea. Electrolyte Abnormalities included: Hyperkalemia and hypokalemia, Hypermagnesemia and hypomagnesemia, Hypercalcemia and hypocalcemia. Complete ICD-10-CM code definitions for every comorbidity and electrolyte abnormality used to construct the study cohort are provided in Supplementary AppendixTable S1 to allow independent reproduction of the cohort definitions.

All analyses accounted for the complex survey design of the NIS using discharge-level weights, stratification, and clustering variables to generate nationally representative estimates. Categorical variables were expressed as weighted percentages and compared using the chi-square test, while continuous variables were expressed as survey-weighted means with standard errors and compared using the Student’s t-test. To identify independent predictors of VAs in patients with HFpEF, we performed multivariable logistic regression analysis, adjusting for demographic variables, comorbidities, and electrolyte abnormalities. Results were reported as adjusted odds ratios (aORs) with 95% confidence intervals (CIs). A two-sided p-value < 0.05 was considered statistically significant. All analyses were performed using Stata version 17.0 (StataCorp, College Station, TX). Because the NIS captures diagnoses present at any point during a single hospitalization without timestamps for individual conditions, the database cannot establish whether electrolyte abnormalities, acute kidney injury, or VAs occurred before, during, or after the index admission; accordingly, these variables are best interpreted as factors concurrently present during the same hospitalization rather than as clinical predictors with an established temporal or causal relationship to ventricular arrhythmias.

A total of 1,289,705 adult heart failure hospitalizations were identified from the NIS between 2016 and 2018. The mean age was 72.5 ± 13.1 years, and 52.1% were women. Most patients were White (71.0%), followed by Black (16.8%) and Hispanic (7.5%) individuals. HFpEF accounted for 66.1% of hospitalizations, whereas HFrEF accounted for 33.9%. Baseline characteristics are summarized in Table 1; Table 2 highlights the in-hospital outcomes; and Table 3 shows VAs and in-hospital outcomes by HF type.

Table 1.Baseline characteristics of hospitalizations with HFpEF and HFrEF.

Table 2.In-hospital outcomes.

Table 3.Ventricular arrhythmias and in-hospital outcomes by HF phenotype.

Overall, VAs occurred in 3.44% of all heart failure hospitalizations. However, the prevalence differed markedly according to heart failure phenotype, occurring significantly less frequently in patients with HFpEF than in those with HFrEF (1.89% vs. 6.48%, p < 0.001). Ventricular tachycardia was the predominant arrhythmia subtype (3.23%), whereas ventricular fibrillation (0.35%) and ventricular flutter (<0.01%) were uncommon. Overall in-hospital mortality was 3.9% and was significantly lower among patients with HFpEF than HFrEF (3.64% vs. 4.46%, p < 0.001). The mean hospital length of stay was 6.1 ± 6.9 days, with mean hospitalization charges of $61,781.

On multivariable logistic regression (Table 4), male sex emerged as the strongest demographic predictor of VAs (aOR 1.71, 95% CI 1.66 - 1.77; p < 0.001). Increasing age and Black race were also independently associated with higher odds of VAs. Among comorbid conditions, diabetes mellitus (OR ~1.16, p < 0.001), obesity (OR ~1.06, p = 0.002), hypothyroidism (OR ~1.17, p < 0.001), acute kidney injury (OR ~1.12, p < 0.001), chronic kidney disease (OR ~1.03, p = 0.042), chronic obstructive pulmonary disease (OR ~1.19, p < 0.001), and alcohol use (OR ~1.12, p = 0.012) were independently associated with higher odds of VAs. Hypertension was not significantly associated with VAs (p = 0.41). Figure 1 provides a forest plot representation of independent predictors of VAs in HFpEF.

Table 4.Multivariable logistic regression for predictors of ventricular arrhythmias.

Among electrolyte abnormalities, hypomagnesemia demonstrated the strongest association with VAs (aOR 1.37, 95% CI 1.29 - 1.46; p < 0.001), followed by hypokalemia and hypocalcemia. Hyperkalemia was not independently associated with VAs. Overall, although VA were substantially less common in HFpEF than in HFrEF, their occurrence in HFpEF was consistently associated with distinct demographic, cardiometabolic, renal, and electrolyte risk factors, suggesting a multifactorial arrhythmogenic substrate in this population.

In this nationwide analysis of hospitalized patients with HFpEF, VAs occurred in approximately 3.4% of hospitalizations and were independently associated with a distinct profile of demographic, cardiometabolic, renal, and electrolyte risk factors. Although VAs were substantially less frequent than in HFrEF, their occurrence identifies a clinically important subgroup with potentially increased risk of adverse outcomes. Male sex, Black race, diabetes mellitus, obesity, hypothyroidism, chronic kidney disease, acute kidney injury, chronic obstructive pulmonary disease, alcohol use, liver cirrhosis, and electrolyte abnormalities independently predicted VAs, highlighting the multifactorial nature of arrhythmogenesis in HFpEF.

Figure 1.Forest plot of independent predictors of ventricular arrhythmias in HFpEF.

Gutierrez et al. demonstrated a high prevalence of non-sustained ventricular tachycardia (NSVT) in HFpEF and showed an associated increase in mortality risk [24]. Similarly, Cho et al. reported increased ventricular tachycardia burden and QTc prolongation among HFpEF patients [25]. VIP-HF study demonstrated frequent NSVT episodes during follow-up [26]. Unlike prior studies that primarily evaluated ambulatory NSVT or sudden cardiac death in relatively small cohorts, our nationally representative analysis demonstrates that hospitalized patients with HFpEF exhibit distinct demographic, metabolic, renal, and electrolyte predictors of ventricular arrhythmias in real-world clinical practice.

A major finding of our study is the strong association between VAs and cardiometabolic comorbidities, including diabetes mellitus, obesity, and hypothyroidism. These associations are biologically plausible. Diabetes mellitus promotes myocardial fibrosis, autonomic dysfunction, oxidative stress, and ion-channel remodeling, all of which increase ventricular electrical instability. Obesity contributes to systemic inflammation, epicardial adipose tissue expansion, and adverse ventricular remodeling, whereas hypothyroidism may prolong ventricular repolarization and increase susceptibility to VAs through alterations in ion-channel function. Collectively, these metabolic abnormalities likely amplify the arrhythmogenic substrate already present in HFpEF. Additional associations were observed with COPD, alcohol use, acute kidney injury, and chronic kidney disease, suggesting that systemic illness and metabolic stress may further contribute to arrhythmogenesis in HFpEF. Ito et al. reviewed the risk factors for fatal VA events, including SCD (composite event) in 2423 HFpEF patients from the CHART -2 study, where the mean age was around 69 years, and 35% females, and decline in EF and NSVT were found to be independent risk factors in multivariate analysis [27]. Lavu et al. found that PVCs and NSVT during the first 48 hours of admission did not predict long-term mortality. [28] In another study, Ito et al. found a new onset of anemia associated with increased risk of VAs and SCD in HFpEF. [29]

We also observed important demographic associations. Male sex was independently associated with increased arrhythmic risk, consistent with prior VA literature. Black race was also independently associated with higher odds of VAs. The NIS does not capture genetic data, socioeconomic status, granular measures of healthcare access, or detailed indices of clinical severity, and this administrative dataset therefore cannot adequately evaluate these factors as explanations for the observed association. We caution against attributing this finding to genetic susceptibility; more plausible contributors include unmeasured differences in comorbidity burden or disease severity, socioeconomic factors, and disparities in access to cardiovascular care, and residual confounding cannot be excluded. This finding warrants further investigation using data sources capable of directly assessing these social and clinical determinants.

Electrolyte abnormalities, particularly hypokalemia, hypomagnesemia, and hypocalcemia, were significantly associated with VAs. These findings are biologically plausible given the established role of electrolyte disturbances in promoting triggered activity, delayed afterdepolarizations, and ventricular electrical instability. Hypomagnesemia demonstrated the strongest electrolyte association with VAs. Magnesium plays a critical role in myocardial membrane stability, potassium homeostasis, and suppression of triggered activity. Deficiency promotes early and delayed afterdepolarizations, facilitating ventricular tachyarrhythmias. Similarly, hypokalemia and hypocalcemia contribute to delayed repolarization and increased ventricular electrical heterogeneity, providing biologically plausible mechanisms for the observed associations.

Similarly, AKI was associated with increased arrhythmic risk, likely reflecting the contribution of acute physiologic stress and metabolic derangements to arrhythmogenesis. In contrast, the lack of association observed with hypertension and hyperkalemia may reflect residual confounding, treatment effects, or limitations inherent to administrative coding datasets. Consistent with the adjusted regression analysis, OSA was independently, though modestly, associated with higher odds of VAs in HFpEF (aOR 1.06, 95% CI 1.01 - 1.11). This modest association may reflect intermittent hypoxia, sympathetic activation, and autonomic dysfunction associated with untreated or undertreated sleep-disordered breathing, although residual confounding by comorbidity burden cannot be excluded. Further studies are needed to clarify the relationship between OSA and arrhythmogenesis in HFpEF.

The increasing prevalence of HFpEF highlights the importance of improved arrhythmic risk stratification in this population. Unlike HFrEF, current guideline-directed recommendations for implantable cardioverter-defibrillator therapy are largely based on reduced ejection fraction thresholds. Identification of these readily available clinical risk factors may facilitate earlier telemetry, ambulatory rhythm monitoring, timely electrophysiology referral, and optimization of modifiable metabolic and electrolyte abnormalities. As disease-modifying therapies for HFpEF continue to evolve, improved recognition of patients at increased arrhythmic risk may become increasingly important for individualized management strategies.

Our study has several strengths, including the use of a large nationally representative cohort and the evaluation of multiple demographic and clinical variables associated with VAs in HFpEF. However, several limitations should be acknowledged. The study relies on administrative ICD-10 coding and is therefore subject to coding inaccuracies and potential misclassification. The NIS lacks detailed clinical information, including medication use, imaging findings, arrhythmia burden, ventricular tachycardia morphology, and outpatient follow-up data. Temporal relationships and causality cannot be established, and residual confounding remains possible despite multivariable adjustment. In particular, because the NIS does not provide timestamps for individual diagnoses within a hospitalization, we cannot determine whether electrolyte abnormalities, acute kidney injury, or ventricular arrhythmias occurred before, during, or after the index event; these variables should therefore be interpreted as concurrent associations identified during the same hospitalization rather than as clinical predictors with demonstrated temporal precedence. Additionally, the NIS does not distinguish sustained from non-sustained ventricular arrhythmias and lacks information regarding antiarrhythmic therapy, implantable cardioverter-defibrillator use, catheter ablation, or sudden cardiac death occurring after discharge.

In conclusion, ventricular arrhythmias, although less common in HFpEF than in HFrEF, are associated with a distinct constellation of demographic, cardiometabolic, renal, and electrolyte risk factors. Recognition of these readily identifiable predictors may improve risk stratification, rhythm surveillance, and early electrophysiology evaluation in hospitalized patients with HFpEF. Prospective studies are warranted to validate these findings and determine whether targeted modification of these risk factors reduces ventricular arrhythmia burden and improves clinical outcomes.

Table S1.ICD-10-CM code definitions used for cohort, comorbidity, and electrolyte-abnormality ascertainment.