Predictors and Trends of 30-day Readmissions in Patients With Acute Decompensated Heart Failure With Preserved Ejection Fraction: Insight From the National Readmission Database

Abstract

Background and Objectives

Hospital readmissions serve as a significant negative prognostic indicator and have a considerable impact on healthcare utilization among individuals diagnosed with heart failure with preserved ejection fraction (HFpEF). For our study, we aimed to elucidate predictors and trends of HFpEF readmissions within a 30-day period.

Methods

The Healthcare Cost and Utilization Project National Readmission Database (NRD) was queried between 2016–2020 to study the 30-day all-cause hospital readmission rate, predictors, duration of hospital stay, and the overall cost of hospitalization. Multivariate/univariate logistic and linear regression analysis were used to analyze the outcomes and adjust for possible confounders.

Results

A total of 3,831,156 index hospitalizations for acute decompensated HFpEF were identified between the years 2016–2020, of which 673,844 (18.4%) patients were readmitted within 30 days. The 30-day all-cause readmissions increased significantly from 17.4% to 19.9% (p<0.001) in the 5-year trend analysis. The most common cardiovascular cause for readmission was hypertensive heart disease with chronic kidney disease stage 1–4 (13.2%). Independent predictors associated with increased rate of readmissions were patients that left against medical advice (adjusted odds ratio [aOR], 2.06; 95% confidence interval [CI], 1.99–2.14; p<0.001), cirrhosis (aOR, 1.33; 95% CI, 1.30–1.36; p<0.001), and chronic obstructive pulmonary disease (aOR, 1.27; 95% CI, 1.25–1.29; p<0.001).

Conclusions

Nearly 1 in 5 patients with acute decompensated HFpEF were readmitted within 30 days (2016–2020), with readmissions rising over time. Identifying at-risk patients is crucial to reducing readmissions and costs.

INTRODUCTION

Over the past few decades, heart failure (HF) with preserved ejection fraction (HFpEF) prevalence has increased and has become an emerging epidemic.¹⁾ In the United States, around 3 million individuals have HFpEF, defined as having clinical signs of HF with an ejection fraction of 50% or higher.²,³⁾ HFpEF now constitutes more than half of all HF cases and is anticipated to become more prevalent than HF with reduced ejection fraction (HFrEF) in the near future.⁴⁾ This expected increase is linked to rising rates of common risk factors such as obesity and diabetes, an aging population, and increased diagnostic awareness of HFpEF.¹,⁵⁾

As the ratio of newly diagnosed HF being in those with HFpEF has risen over time, similarly, the ratio of HF patients admitted to hospitals with HFpEF has increased. Furthermore, in patients hospitalized with HFpEF, about 20% are readmitted within 30 days of hospital discharge, and approximately 50% are readmitted within one year.⁶⁾ These high rates of readmission lead to increased morbidity and mortality, significant use of healthcare resources, and a decline in the quality of life for patients with HFpEF.¹,³,⁵,⁶⁾ Consequently, there is growing attention and research focused on reducing readmissions, particularly within the first month post-discharge.

There is a paucity of studies examining HFpEF hospital readmissions on a large, nationwide scale. Previous studies have examined single centers, single payers (i.e., Medicare), or single years.⁷,⁸,⁹,¹⁰⁾ Additionally, there is a scarcity of data examining the trends of HFpEF-specific readmissions to confirm the predicted increase in hospitalizations and readmissions.

This study aimed to determine the predictors and temporal trends of acute decompensated HFpEF between 2016–2020 using a large, real-world, nationally representative database. We hypothesized that hypertension, diabetes, atrial fibrillation (AF), and obesity would be highly associated with readmissions. Additionally, that 30-day readmissions would decline over the 5 years studied.

METHODS

Research design

This retrospective cohort study utilized data from the Healthcare Cost and Utilization Project National Readmission Database (NRD), sourced from the Agency for Healthcare Research and Quality.¹¹⁾ We used the 5-year period spanning 2016 to 2020, containing discharge data from approximately 90 million unweighted discharges. The NRD is one of the largest publicly accessible databases in the United States, accounting for approximately two-thirds of all hospitalizations from various payer sources. Data within the NRD covers the period from January 1st to December 31st each year. Patients can be tracked within a year using a unique linkage number to monitor readmissions. Institutional Board Review approval was exempted for this study because the NRD database contains de-identified patient information.

Participants

Discharge and procedure diagnoses were collected for each patient using the International Classification of Diseases, 10th Revision Clinical Modification and Procedural Coding System (ICD-10-CM/PCS). Diagnoses within this study are categorized into 2 main types: principal diagnosis, denoting the primary reason for hospitalization, and secondary diagnosis, encompassing any additional discharge diagnoses beyond the principal one. Patients with a diagnosis of diastolic (congestive) HF, acute diastolic HF or acute-on-chronic diastolic HF were included (ICD-10-CM codes I50.3, I50.31, I50.33, respectively). Patients under 18 years old or transferred to another acute care hospital were excluded from the study. Because patient identifiers cannot be linked across years in the NRD, patients who had an index hospitalization within the month of December were excluded. Patients that died during index admission were excluded in readmission analysis. Patients were stratified into 30-day readmissions based on time to readmission. Time to readmission was calculated by subtracting the length of stay of index admission from the time between 2 admissions. Patients with at least one unplanned readmission within 30 days were included, whereas elective readmissions within 30 days were excluded. The causes of readmission were classified as all causes versus cardiovascular-specific causes based on ICD-10-CM codes.

Outcomes

The primary focus of this study centered on evaluating the independent predictors for readmission and the underlying causes of readmissions as delineated by ICD-10 codes. All cause and cardiovascular-specific reasons for readmission were studied based on patients admitted for HFpEF exacerbation on their index admission. Secondary outcomes encompassed an analysis of readmission mortality rates, mean length of hospital stay, and mean cumulative hospital charges. Furthermore, the study sought to identify trends in all-cause readmissions, all-cause mortality, length of stay, and total hospital charges over the 5 years studied.

Data collection

We performed our analysis using the Healthcare Cost and Utilization Project STATA survey data analysis packages. These packages incorporate NRD-specific variables, including hospital identifiers, stratum, and discharge weights, to account for clustering and large survey-weighted data analysis. Stratified weighted data were analyzed using the “svy” command to obtain nationwide estimates. By using “svy,” our data analysis software adjusts for the design effects of clustering and stratification and applies appropriate weights to each observation. The NRD includes variables on patient demographics, including age, sex, median household income, and primary payer. It also contains hospital-specific variables, including bed size, teaching status, and location. The severity of comorbidities was determined by the Elixhauser Comorbidity Index, a validated measurement tool for statistical use in survey-weighted samples.

Data analysis

Continuous variables were compared using the Student’s t-test, whereas categorical variables were compared using the Pearson χ² test. Multivariate logistic and linear regression analysis was performed with readmission as the independent variable and the readmission event as the dependent variable. A univariate analysis was used to identify variables associated with readmissions related to cardiovascular causes following acute decompensated HFpEF hospitalizations to obtain independent predictors of 30-day all-cause readmission. We included those variables having a p value <0.2 in the final multivariable regression analysis. Subsequently, we ran a multivariable logistic regression analysis to identify independent predictors of readmissions with p values <0.05 set as the threshold for statistical significance. By including these covariates in the model, better adjustment for confounders can be made, reducing interaction effects and ensuring that the observed relationships between the primary variables of interest and the outcome are not confounded by these covariates. A p value <0.05 was considered significant for all the analyses. All analyses were performed using STATA, version 18 (StataCorp, College Station, TX, USA).

RESULTS

Between 2016 and 2020, there were 3,831,156 recorded hospital admissions for acute decompensated HFpEF. Of these, 3,648,022 (95.2%) patients were discharged alive. Within 30 days of discharge, 673,844 patients (18.4%) were readmitted to the hospital. The in-hospital mortality rate during readmissions was significantly higher than that of the index admission (6.9% vs. 4.7%, p<0.001). Tables 1 and 2 show the baseline characteristics of the readmitted patients following their initial hospital stay. The majority of patients in both the non-readmitted and readmitted groups were over the age of 65 (78.3% and 75.7%, respectively) and female (59.6% and 59.2%, respectively). Nearly all the patients in both groups had 3 or more comorbidities (98.7% and 99.2%, respectively), according to the Elixhauser Comorbidity Index. There was a greater prevalence of chronic kidney disease (CKD, 53.7% vs 46.1%, p<0.001), anemia (47.0% vs. 39.9%, p<0.001), diabetes (52.4% vs. 47.3%, p<0.001), chronic obstructive pulmonary disease (COPD, 44.5% vs. 38.9%, p<0.001), and coronary artery disease (46.1% vs. 42.7%, p<0.001) in the readmitted group versus those not readmitted.

Table 1. Comparison of baseline characteristics during index admission between those with and without readmission within 30 days after acute decompensated heart failure with preserved ejection fraction.

Characteristics		Index admission without readmission (n=3,157,312)	Index admission with readmission (n=673,844)	p value
Mean age (years)		74.4	73.2	<0.001
Sex				<0.001
	Female	1,881,758 (59.6)	398,916 (59.2)
	Male	1,275,554 (40.4)	274,928 (40.8)
Age category (years)				<0.001
	18–45	75,775 (2.4)	18,193 (2.7)
	45–64	609,361 (19.3)	145,550 (21.6)
	>65	2,472,175 (78.3)	510,099 (75.7)
Total comorbidities (Elixhauser Comorbidity)				<0.001
	0	0	0
	1	6,315 (0.2)	674 (0.1)
	2	3,157 (0.1)	4,717 (0.7)
	3	3,116,267 (98.7)	668,453 (99.2)
Insurance status				<0.001
	Medicare	2,604,782 (82.5)	558,617 (82.9)
	Medicaid	211,540 (6.7)	56,603 (8.4)
	Private	296,787 (9.4)	51,212 (7.6)
	Self-pay	44,202 (1.4)	7,412 (1.1)
Median household income by ZIP code (USD)				<0.001
	$1–$45,999	921,935 (29.2)	213,609 (31.7)
	$46,000–$58,999	893,519 (28.3)	190,024 (28.2)
	$59,000–$78,999	767,227 (24.3)	156,332 (23.2)
	>$79,000	574,631 (18.2)	113,879 (16.9)
Hospital bed size				<0.001
	Small	615,676 (19.5)	127,357 (18.9)
	Medium	906,149 (28.7)	192,719 (28.6)
	Large	1,635,488 (51.8)	353,768 (52.5)
Hospital teaching status				<0.001
	Teaching	2,096,455 (66.4)	453,497 (67.3)
	Non-teaching	1,060,857 (33.6)	220,347 (32.7)
Hospital setting				<0.001
	Urban	3,024,705 (95.8)	648,912 (96.3)
	Rural	132,607 (4.2)	24,932 (3.7)
Discharge to skilled nursing rehab facility		1,755,465 (55.6)	396,220 (58.8)	<0.001
Patient left against medical advice*		25,258 (0.8)	10,108 (1.5)	<0.001
Mean length of stay (days)		7.1	6.4	<0.001
Mean total charges (USD)		73,613.94	66,929.44	<0.001
Total charges (USD)		28.2e+10	4.51e+10	-
Length of stay category (days)
	<3	18.2	15.3	<0.001
	3–6	39.2	36.7	<0.001
	6–9	19.9	21.0	<0.001
	>10	18.7	22.7	<0.001

Values are presented as number (%). Bold terms indicate statistical significance.

^*Patient left against medical advice defines any patient that chooses to be discharged from the hospital prior to being deemed medical ready.

Table 2. Likelihood of readmission within 30 days after acute decompensated heart failure with preserved ejection fraction based on comorbidities.

Comorbidities	Index admission without readmission (n=3,157,312)	Index admission with readmission (n=673,844)	Adjusted odds ratio	95% CI	p value
Hypertension	2,859,893 (90.58)	615,085 (91.28)	1.09	1.07–1.10	<0.001
Dyslipidemia	1,729,260 (54.77)	368,795 (54.73)	0.99	0.99–1.00	0.743
Atherosclerotic conditions*	1,578,025 (49.98)	358,283 (53.17)	1.14	1.13–1.15	<0.001
Diabetes	1,494,356 (47.33)	352,892 (52.37)	1.22	1.21–1.23	<0.001
Chronic kidney disease	1,455,837 (46.11)	361,652 (53.67)	1.35	1.34–1.37	<0.001
Coronary artery disease	1,349,119 (42.73)	310,373 (46.06)	1.14	1.13–1.15	<0.001
Anemia	1,258,820 (39.87)	316,774 (47.01)	1.34	1.32–1.35	<0.001
Smoker	1,254,716 (39.74)	276,883 (41.09)	1.06	1.05–1.07	<0.001
COPD	1,220,301 (38.65)	299,861 (44.50)	1.27	1.26–1.28	<0.001
Obesity	1,012,550 (32.07)	215,293 (31.95)	0.99	0.98–1.00	0.225
Hypothyroidism	642,513 (20.35)	136,386 (20.24)	0.99	0.98–1.00	0.177
Prior PCI	330,571 (10.47)	75,538 (11.21)	1.08	1.07–1.09	<0.001
Prior CABG	317,310 (10.05)	70,147 (10.41)	1.04	1.03–1.05	<0.001
Implantable cardioverter-defibrillator	305,944 (9.69)	64,217 (9.53)	0.98	0.97–0.99	0.008
Cancer	271,213 (8.59)	68,799 (10.21)	1.21	1.19–1.22	<0.001
Peripheral arterial disease	265,530 (8.41)	63,072 (9.36)	1.12	1.11–1.14	<0.001
Prior stroke or TIA	211,540 (6.70)	46,967 (6.97)	1.04	1.03–1.06	<0.001
Malnutrition	208,698 (6.61)	50,808 (7.54)	1.15	1.13–1.17	<0.001
Prior valve replacement	161,970 (5.13)	36,185 (5.37)	1.05	1.03–1.07	<0.001
Hemodialysis	130,081 (4.12)	42,183 (6.26)	1.56	1.53–1.58	<0.001
Cirrhosis	79,564 (2.52)	23,989 (3.56)	1.42	1.39–1.45	<0.001
Carotid atherosclerotic disease	56,832 (1.80)	12,870 (1.91)	1.06	1.03–1.09	<0.001
Aortic atherosclerotic disease	55,253 (1.75)	11,321 (1.68)	0.96	0.93–0.99	0.012
Renal arterial disease	17,997 (0.57)	4,784 (0.71)	1.25	1.19–1.31	<0.001

Values are presented as number (%). Bold terms indicate statistical significance.

CI = confidence interval; COPD = chronic obstructive pulmonary disease; PCI = percutaneous coronary intervention; CABG = coronary artery bypass graft; TIA = transient ischemic attack; CAD = coronary artery disease; CVA = cerebrovascular accident; PAD = peripheral arterial disease.

^*Composite variable on results (combining 5 different conditions: CAD, CVA, renal, aortic and PAD).

The mean length of stay during the index admission was 7.1 days, compared to the 6.4 days observed for readmitted patients (p<0.001). On further analysis, patients with shorter length of stay on index admission were less likely to be readmitted. Conversely, patients with longer index length of stay were more likely to be readmitted. Readmission rates were higher among patients discharged to skilled nursing rehabilitation facilities (58.8% vs. 55.6%, p<0.001) and among those who left against medical advice (1.5 vs. 0.8, p<0.001) after their index admission. The total hospital charges associated with readmissions totaled over 45 billion USD.

Socioeconomic analysis demonstrated that patients in the lowest income bracket made up a higher proportion of patients in the readmission group compared to the group not readmitted (31.7% vs. 29.2%, p<0.001). Medicare was the most common form of insurance for index admissions and readmissions. However, there was a higher proportion of Medicaid recipients in the readmitted group compared to the index group (8.4% vs. 6.7%, p<0.001) and a lower proportion of patients with private insurance in the readmitted group compared to the index admission group (7.6% vs. 9.4%, p<0.001).

Table 3 describes the significant independent predictors for 30-day readmissions. Significant factors associated with higher odds of readmission included leaving against medical advice (adjusted odds ratio [aOR], 2.06; p<0.001; 95% confidence interval [CI], 1.99–2.14), cirrhosis (aOR, 1.33; p<0.001; 95% CI, 1.30–1.36) COPD (aOR, 1.27; p<0.001; 95% CI, 1.25–1.29), cancer (aOR, 1.19; p<0.001; 95% CI, 1.17–1.21) and anemia (aOR, 1.19; p<0.001; 95% CI, 1.18–1.20). Length of stay was also a significant predictor of readmission, with different categories reflecting varying risks: shorter stays (category a and b) were associated with lower odds, while longer stays (category d) increased the likelihood of readmission. Patients from higher income quartiles showed a decreased likelihood of readmission (aOR, 0.96; p<0.001; 95% CI, 0.95–0.97).

Table 3. Multivariate analysis of predictors for readmission based on baseline characteristics of patients readmitted vs. not readmitted within 30 days after an index hospitalization for heart failure with preserved ejection fraction.

Baseline characteristics	Adjusted odds ratio	p value	95% CI
Female	1.02	<0.001	1.01–1.03
Age category	0.80	<0.001	0.79–0.81
Income quartile	0.96	<0.001	0.95–0.97
Hospital bed size	1.00	0.387	0.99–1.01
Teaching hospital	1.00	0.430	0.99–1.01
Insurance type	0.89	<0.001	0.90–0.91
Length of stay <3 days	0.81	<0.001	0.79–0.83
Length of stay 3–6 days	0.88	<0.001	0.86–0.90
Length of stay 7–10 days	0.95	<0.001	0.93–0.97
Length of stay >10 days	1.04	<0.001	1.02–1.07
Elixhauser Comorbidity Index	1.07	0.011	1.02–1.22
Urban location of the hospital	1.06	<0.001	1.04–1.08
Patient left against medical advice*	2.06	<0.001	1.99–2.14
Discharge to a skilled nursing facility	1.09	<0.001	1.08–1.10
Anemia	1.19	<0.001	1.18–1.20
Diabetes	1.13	<0.001	1.12–1.14
Hypertension	0.98	0.020	0.97–1.00
Smoker	1.00	0.367	0.99–1.02
Malnutrition	1.07	<0.001	1.05–1.09
CAD	1.12	<0.001	1.10–1.14
COPD	1.27	<0.001	1.25–1.29
Cancer	1.19	<0.001	1.17–1.21
Chronic kidney disease	1.00	0.924	0.92–1.08
Peripheral arterial disease	1.08	<0.001	1.06–1.10
Cirrhosis	1.33	<0.001	1.30–1.36
Prior stroke or TIA	1.01	0.453	0.98–1.05
Renal arterial disease	1.11	<0.001	1.05–1.17
Aortic atherosclerotic disease	0.96	0.017	0.93–0.99
Atherosclerosis†	0.99	0.621	0.97–1.02
Carotid atherosclerotic disease	0.99	0.809	0.95–1.04
Prior CABG	0.97	<0.001	0.95–0.98
Implantable cardioverter-defibrillator	1.01	0.263	0.99–1.02
Prior PCI	1.01	0.301	0.99–1.02
Prior valve replacement	1.07	<0.001	1.05–1.09

Bold terms indicate statistical significance.

CI = confidence interval; CAD = coronary artery disease; COPD = chronic obstructive pulmonary disease; TIA = transient ischemic attack; CABG = coronary artery bypass graft; PCI = percutaneous coronary intervention; CVA = cerebrovascular accident; PAD = peripheral arterial disease.

^*Patient left against medical advice defines any patient that chooses to be discharged from the hospital prior to being deemed medical ready.

^†Composite variable on results (combining 5 different conditions: CAD, CVA, renal, aortic and PAD).

Table 4 outlines the 25 most common causes of readmission. The most common cardiac causes were hypertensive heart disease, HF, and CKD. Other notable cardiac causes included non-ST elevation myocardial infarction at 1.5%, paroxysmal AF at 1.2%, and nonrheumatic aortic valve stenosis at 1.1%. Sepsis was the leading non-cardiac cause of readmission, accounting for 7.3% of cases followed by acute kidney injury at 4.0%, and acute COPD exacerbation at 2.8%.

Table 4. Percentage of 25 most common cardiac and noncardiac causes of 30-day readmission in patients initially admitted for acute decompensated heart failure with preserved ejection fraction.

Causes		%
Cardiac causes
	Hypertensive heart disease with heart failure and CKD stage 1–4	13.2
	Hypertensive heart disease with heart failure	7.1
	Acute on chronic diastolic (congestive) heart failure	3.7
	Hypertensive heart disease with heart failure and CKD stage 5 or end-stage renal disease	3.0
	NSTEMI	1.5
	Paroxysmal atrial fibrillation	1.2
	Nonrheumatic aortic (valve) stenosis	1.1
	Atrial fibrillation (unspecified)	0.9
	Acute on chronic combined systolic and diastolic heart failure	0.5
	Noncardiac causes
Noncardiac causes
	Sepsis	7.3
	Acute kidney failure	4.0
	Acute COPD exacerbation	2.8
	Acute and chronic respiratory failure with hypoxia	2.7
	Pneumonia	1.9
	Acute respiratory failure with hypoxia	1.4
	Pneumonitis due to inhalation of food and vomit	1.1
	Acute and chronic respiratory failure with hypercapnia	1.1
	Urinary tract infection	0.9
	Gastrointestinal hemorrhage	0.8
	COVID-19	0.5
	COPD with (acute) lower respiratory infection	0.5
	Melena	0.5
	Cellulitis of left lower limb	0.5
	Pleural effusion	0.5
	Sepsis due to Escherichia coli	0.5

CKD = chronic kidney disease; NSTEMI = non-ST elevation myocardial infarction; COPD = chronic obstructive pulmonary disease; COVID-19 = coronavirus disease 2019.

Figure 1 show the trend analysis for secondary outcomes. The readmission rates exhibited a general upward trend from 2016 to 2019, beginning at 17.4% in 2016 and peaking at 22.4% in 2019. In 2020, there was a noticeable decrease in the readmission rate to 19.9%. The length of stay for index admissions and readmissions had minor variations between 2016–2010. However, both had significant trends toward increased length of stay. For index admissions, there was a consistent year-over-year increase in average total charges from $64,989.71 in 2016 to $83,134.20 in 2020. Comparably, average readmissions charges started slightly higher than index admissions in 2016 at $66,827.73 and followed an increasing trend, reaching $85,743.75 in 2020. The readmissions tended to have higher total charges than index admissions, particularly notable in the last 2 years, 2019 and 2020, with a significant increase in charges for readmitted patients (Tables 5 and 6).

Figure 1. Non-linear trend analysis. (A) All-cause 30-day readmission rate following an index admission for acute decompensated heart failure with preserved ejection fraction between 2016–2020. (B) Mean total charges per patient during index admission for acute decompensated heart failure with preserved ejection fraction versus 30-day readmission between 2016–2020. (C) Mean length of stay during index admission for acute decompensated heart failure with preserved ejection fraction versus 30-day readmission between 2016–2020.

Table 5. Average total charges per patient during index admission for acute decompensated heart failure with preserved ejection fraction vs. 30-day readmission between 2016–2020.

Year	Index (USD)	95% confidence interval (USD)	Readmitted (USD)	95% confidence interval (USD)
2016	64,989.71	62,556.90–67,422.52	66,827.73	64,292.26–69,363.20
2017	68,758.65	66,295.33–71,221.96	67,905.65	65,518.61–70,292.69
2018	72,387.90	69,705.03–75,070.78	72,482.25	69,747.42–75,217.08
2019	75,381.97	72,646.21–78,117.73	80,674.42	77,569.70–83,779.15
2020	83,134.20	80,105.38–86,163.02	85,743.75	82,413.88–89,073.62

Table 6. Average length of stay for during index admission for acute decompensated heart failure with preserved ejection fraction vs. 30-day readmission between 2016–2020.

Year	Index (days)	95% confidence interval (days)	Readmitted (days)	95% confidence interval (days)
2016	6.97	6.89–7.06	7.44	7.33–7.55
2017	6.96	6.88–7.04	7.21	7.11–7.30
2018	6.93	6.85–7.01	7.24	7.15–7.33
2019	6.89	6.82–6.98	7.61	7.51–7.72
2020	7.28	7.20–7.36	7.75	7.65–7.86

DISCUSSION

This large retrospective analysis has revealed multiple findings that highlight the impact of HFpEF readmissions in the United States. First, approximately one out of 5 patients had early readmissions after discharge for acute decompensated HFpEF. Furthermore, if readmitted within 30 days after discharge, there was a significantly higher mortality rate than the index admission. Second, we found a worrisome increase in the trend of readmission rates in our study that confirms prior projections. Third, the predictors for readmission included complex care comorbidities such as cirrhosis and COPD, providing key insight into identifying those at the highest risk of readmission.

As described in the Methods section, the NRD is a publicly available nationwide database that links readmissions with index hospitalizations and their respective ICD-10-related data. This feature allows 3 cardinal advantages of the NRD for this project: 1) longitudinal tracking of unplanned hospitalizations, 2) determining the cause for readmissions, and 3) patient quantification rather than event quantification. Lastly, the embedded data in the NRD encompasses 170 million discharges irrespective of payer. These features power our findings and allow appropriate representation across geographically dispersed areas.

Our study showed that approximately 18% of patients experience early readmissions after discharge for acute decompensated HFpEF, with those readmitted within 30 days facing a significantly higher mortality rate than at the index admission. While readmission rates reported in the literature vary widely, from 21% to 71%, most of these studies have either analyzed single years or were greatly underpowered.⁸,⁹,¹²⁾ Other large studies have shown similar rates of readmission rates as our study, of approximately 20%.¹³⁾

The in-hospital mortality rate during readmissions was significantly higher than that of the index admission (6.9% vs 4.7%, p<0.0001). Previous studies examining the mortality rate have shown a similar relationship.¹⁴⁾ Our study’s increased mortality rate during readmissions was likely due to the complex comorbidities associated with readmission. Repeated hospitalizations are common in patients with HFpEF. Distinguishing the cause of these readmissions is crucial for providing appropriate care and improving patient outcomes.

The most common cardiovascular cause of readmission was hypertensive heart disease with CKD stage 1–4. Following hypertensive heart disease with CKD and HFpEF, combined paroxysmal and “unspecified” AF comprised a top cardiovascular cause of readmission (2.1%). Similar studies have identified similar causes and predictors for readmission.⁹,¹²,¹⁵,¹⁶⁾ It is well known that AF is associated with an incident diagnosis of HFpEF and is incorporated in the validated H₂FPEF scoring system.¹⁷⁾ More recently, conflicting evidence on the effect of AF on readmission rates has emerged. In a small cohort post hoc analysis of the ROPA-DOP trial, AF was an independent predictor of 30-day hospital readmission (OR, 4.17 [1.23, 13.73]; p=0.02).¹⁸⁾ In a post-hoc analysis, AF initially appeared to increase the risk of all-cause 30-day readmissions (37.5% vs. 17.5%; p=0.036) but it was not statistically significant after propensity score adjustment for clinical covariates.

Cirrhosis, COPD, and renal insufficiency were among the most significant independent predictors of readmission. Previous studies using NRD data have shown similar predictors for readmission.⁹,¹²,¹⁵,¹⁶⁾ These conditions are typically associated with complex care needs and comorbidities, underscoring the importance of effective management and follow-up for these patients to reduce readmission rates. Conversely, these conditions are common non-cardiac mimics of HFpEF. Therefore, it is possible these are conditions linked to HFpEF readmission or the misclassification of noncardiovascular comorbidities that exhibit clinical signs of congestion or dyspnea resembling a HFpEF exacerbation.²⁾ Utilizing ICD-10 codes to identify HFpEF patients and comorbid conditions may also lead to coding errors and misdiagnosis of HFpEF. Overall, there is a need for better diagnostic tools and more precise treatment strategies to distinguish between HFpEF exacerbations and symptoms caused by other comorbidities. This distinction is crucial for improving the management and outcomes of patients with HFpEF.

Our study found that patients in higher income brackets and with private insurance were less likely to be readmitted for a HFpEF exacerbation. These findings highlight the importance of socioeconomic status and importance of health equity. In the “2021 update to the 2017 ACC expert consensus decision pathway for optimization of heart failure treatment,” guidance was set forth on cost reduction measures for low-income individuals.¹⁹⁾ Strategies to reduce patient cost of care such as prescribing lower-cost medications of similar efficacy, reducing travel cost with using a single pharmacy, and incorporating social workers and pharmacists to help identify copay assistance programs were recommended. Although this guidance was recommended for individuals with HFrEF, the same can be applied for patients with HFpEF. Strategies to reduce costs for patients leads to improved medication adherence, which leads to reduction of readmissions for HFpEF. Ultimately, adding resources to resource limited areas and improving access to clinicians with the knowledge to treat HFpEF will lead to improved outcomes.

There was a significant overall increase in the incidence of HFpEF readmissions over the period studied. Readmission rates increased from 2016 to 2019, with a peak in 2019 at 22.4% and a notable decrease in 2020 to 19.9%. Other studies analyzing trends showed mixed results. In a study examining trends from 2004–2011, Cui et al.²⁰⁾ found no significant trend and a stable rate of HFpEF all-cause 30 day readmissions with an average of 21.6%. A study from the Veterans Affairs Health Care System revealed a downward trend from 2007 to 2017 in the risk-aOR for 30-day hospital readmissions among patients with HF and an left ventricular ejection fraction of 40% or higher.²¹⁾ Using the NRD, Khan et al.²²⁾ investigated all-cause readmissions for HF from 2010 to 2017 and identified a rising trend in readmission rates following an initial hospitalization for HF. In our study, the increasing trend could be associated with various factors, including the evolution of HFpEF diagnostic criteria, an increased prevalence of comorbid conditions, or changing healthcare system dynamics. The decline in 2020 may have been influenced by the global coronavirus disease 2019 pandemic, which could have led to changes in hospital admission practices or patient reluctance to seek care due to fear of infection. A trend analysis from 2021–2023 will likely clarify the direction of the trend. Future investigations may show declines in readmission rates with improved recognition, diagnosis, and increased use of sodium-glucose cotransporter-2 inhibitors.

There are several limitations to our study. First, the study relies on observational data, which inherently limits the ability to establish causation due to potential unmeasured confounders. Second, the NRD only allows tracking patient data for one year. This limitation restricts the longitudinal analysis of patient outcomes and patterns beyond this timeframe. Moreover, the same patients may be recorded in the database in multiple years. Due to anonymity and privacy protocols, these records cannot be definitively linked, potentially leading to data redundancy or overestimating unique patient counts. Third, the database collects data exclusively on in-hospital mortality, excluding post-discharge deaths. This collection could under-represent the true mortality rates and outcomes after hospitalization. Fourth, variables such as laboratory values, medications, medication adherence, follow up appointment adherence, imaging results, discharge to homeless shelter, discharge to home, and discharge to rehab facility are not included in the NRD database, preventing the analysis of health and demographic data that would otherwise add important readmission predictors. Lastly, utilizing ICD codes for analyzing administrative data is subject to potential error. These errors could arise from misclassification, incorrect coding, or varying coding practices across different hospitals or regions.

Our study emphasizes the significance of recognizing patients at risk for HFpEF readmission and customizing interventions that could decrease hospital readmissions and related expenses. Future studies should explore the patterns and determinants of readmission for HFpEF patients further.