Outcomes and Resource Utilization Associated With Readmissions After Atrial Fibrillation Hospitalizations

Abstract

Background

Atrial fibrillation is the most common arrhythmia worldwide. Data regarding 30‐day readmission rates after discharge for atrial fibrillation remain poorly reported.

Methods and Results

The Nationwide Readmission Database (2010–2014) was queried using the International Classification of Diseases, Ninth Revision (ICD‐9) codes to identify study population. Incidence, etiologies of 30‐day readmission and predictors of 30‐day readmissions, and cost of care were analyzed. Among 1 723 378 patients who survived to discharge, 249 343 (14.4%) patients were readmitted within 30 days. Compared with the readmitted group, the nonreadmitted group had higher utilization of electrical cardioversion and catheter ablation. Atrial fibrillation was the most common cause of readmission (24.1%). Median time to 30‐day readmission was 13 days. Advancing age, female sex, and longer stay during index hospitalization predicted higher 30‐day readmissions, whereas private insurance, electrical cardioversion, catheter ablation, higher income, and elective admissions correlated with lower 30‐day readmission. Comorbidities such as heart failure, neurological disorder, chronic obstructive pulmonary disease, diabetes mellitus, chronic kidney disease, chronic liver failure, coagulopathy, anemia, peripheral vascular disease, and electrolyte disturbance, correlated with increased 30‐day readmissions and cost burden. Trend analysis showed a progressive decline in 30‐day readmission rates from 14.7% in 2010 to 14.3% in 2014 (P trend, <0.001).

Conclusions

Approximately 1 in 7 patients were readmitted within 30 days of discharge, with symptomatic atrial fibrillation being the most common cause. We identified a predictive model for increased risk of readmissions and treatment expense. Electrical cardioversion during index admission was associated with a significant reduction in 30‐day readmissions and service charges. The 30‐day readmissions correlated with a substantial rise in the cost of care.

Clinical Perspective

What Is New?

• Atrial fibrillation readmission rate was noted to be ≈14%.

• There is a slow, but progressive, decline in readmission rate over recent years.

• Atrial fibrillation readmissions were associated with a significant rise in cost of care, and novel techniques, such as electrical cardioversion and catheter ablation, predicted lower 30‐day readmission risk.

What Are Clinical Implications?

• Atrial fibrillation hospitalizations remain a high source of readmission and burden but it is possible to risk stratify patients based on clinical characteristics.

• We encourage increased utilization of novel techniques, such as electrical cardioversion and catheter ablation, to prevent readmissions.

Introduction

Atrial fibrillation (AF) is the most common arrhythmia worldwide and the estimated global age‐adjusted prevalence was 0.5% in 2010, representing nearly 33.5 million individuals.1 AF affects nearly 2.3 million Americans, and prevalence is expected to double by the year 2050.2, 3 Prevalence is likely underestimated given that a large proportion of asymptomatic individuals and those having transient symptoms remain undiagnosed. It is recognized as a global public health problem because of its significant burden of morbidity and mortality resulting from embolic stroke, congestive heart failure, and acute coronary syndrome.4 The incremental cost related to AF in the United States is estimated at $6 to 26 billion per year and much of it is related to inpatient hospitalization, with nearly 460 000 hospitalizations having a primary discharge diagnosis of AF every year.5, 6 Although recent advances in treatment of AF improved the overall symptom burden and quality of life, readmission rates continue to escalate and have been one of the major sources of AF‐related financial strains on healthcare economies throughout the world.7 Previous studies examined the readmission rates of AF; however, they had a short duration8 or included only selected patient populations.7 In this context, we sought to identify the etiologies, predictors, and trends of 30‐day readmissions after an index hospitalization for AF and factors associated with high cost of care over a 5‐year period using the largest, all‐payer readmission database available in the United States.

Methods

Data Source

The study was derived from the Healthcare Cost and Utilization Project's (HCUP) National Readmission Database (NRD) of 2010–2014, sponsored by the Agency for Healthcare Research and Quality. The NRD is one of the largest publicly available all‐payer inpatient care databases in the United States, which includes data on ≈15 million discharges in year 2010–2014, estimating roughly 35 million discharges from 22 states with reliable, verified linkage numbers. The NRD represents 49.3% of total US hospitalizations. Patients were tracked during the same year using the variable “NRD_visitlink,” and time between 2 admissions was calculated by subtracting the variable “NRD_DaysToEvent.” Time to readmission was calculated by subtracting length of stay (LOS) of index admissions to time between 2 admissions. Sampling weights provided by the sponsor were used to produce national estimates.9 The design of the NRD is available online.10 The data that support the findings of this study are available from the corresponding author upon reasonable request. The NRD is a de‐identified database and was deemed exempt from ethical review at institutional review boards of University of Arizona (Phoenix, AZ), University of South Florida (Tampa, FL), St. Francis Medical Center (Hartford, CT), Michigan State University (East Lansing, MI), Guthrie Robert Packer Hospital (Sayre, PA), Icahn School of Medicine (Mount Sinai, NY), and Mayo Clinic (Rochester, MN).

Data Selection

We queried the NRD using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9‐CM) diagnosis codes for AF (427.31) in the primary diagnosis field to extract the study population. Patients aged <18 years and with missing data for age and sex were excluded. We also excluded index admissions during the month of December because we did not have 30‐day follow‐up data for that month. We identified in total 1 723 378 index admissions with a primary discharge diagnosis of AF (Figure S1). A similar methodology for data extraction from the NRD has been used and validated in previously published studies.11 Patients who were readmitted to any hospital within 30 days (n=249 343) within the same calendar year were further evaluated.

Outcomes

The primary outcome of our study was 30‐day readmissions. Secondary analysis was performed to explore predictors of 30‐day readmission and cost of care associated with AF hospitalizations. Causes of readmission were identified by using ICD‐9‐CM codes in the primary diagnosis field during readmission observation. We identified 2523 different ICD‐9‐CM diagnosis codes and combined the ones with similar diagnoses to make clinically important groups (Tables S1 and S2). Additional analysis was performed to explore trends in mortality during index hospitalization and rehospitalizations (Figure S2).

Definition of Variables

NRD variables were used to identify patients’ demographic characteristics, including age, sex, hospital characteristics (bed size and teaching status), and patient‐specific characteristics, including median household income category for a patient's ZIP code, primary payer, admission type, admission day, and discharge disposition.12 Comorbidities such as hypertension, diabetes mellitus, chronic obstructive pulmonary disease, peripheral vascular disease, neurological disorder, hypothyroidism, coexisting malignancy, coagulopathy, valvular heart disease, anemia, and fluid and electrolyte disturbance were identified by variables provided in the NRD, which uses ICD‐9‐CM diagnoses and the diagnosis‐related group in effect on the discharge date. Electrical cardioversion was identified by the appropriate ICD‐9‐CM procedural code in the primary procedural field. Other comorbidities were identified by ICD‐9 codes in a secondary diagnosis field, which included heart failure, chronic kidney disease, chronic liver failure, previous coronary artery bypass graft, previous myocardial infarction, previous stroke/transient ischemic attack, etc (Table S3).

The Deyo modification of the Charlson comorbidity index, which contains 17 comorbid conditions with differential weights, was utilized to define severity of comorbid conditions. This score ranges from 0 to 33, with higher scores corresponding to larger burden of comorbid diseases, (Table S4).13 Annual hospital volume of procedures was calculated by using a unique hospital identification provided by the NRD. We also evaluated LOS provided by the NRD. Cost of index hospitalization was calculated by merging cost to charge ratio provided by the HCUP to the main data set and after adjusting for inflation.14

Statistical Analysis

SAS software (version 9.4; SAS Institute Inc, Cary, NC) was implemented for analyses. The Wilcoxon rank‐sum test was performed to test differences in continuous variables between the readmitted group and nonreadmitted group, because data were not normally distributed. The chi‐square test of independence was used for testing the difference between the 2 groups with respect to categorical variables. Multivariate predictors of 30‐day readmission were calculated using a hierarchical logistic regression model, and predictors of cost of care during index hospitalization were explored using a linear regression model. The hierarchical logistic regression model considers the effect of nesting (ie, patient‐level effects nested with hospital‐level effects in our study). In this model, a unique hospital identification number was incorporated as a random effect, creating a 2‐level model. This methodology has been recommended by the HCUP and has been used widely in previous studies.9, 15 In multivariate analysis, clinically significant variables, such as age, sex, primary payer, median household income, admission type and day, certain hospital characteristics (bed capacity and teaching status), as well as comorbidities, were incorporated. We only considered comorbidities with a statistically significant difference in readmission using the univariate method (congestive heart failure, neurological disorder, chronic lung disease, diabetes mellitus, chronic renal failure, hypertension, hypothyroidism, chronic liver disease, coagulopathy, valvular heart disease, anemia, peripheral vascular disease, and fluid and electrolyte disturbance). The same independent variables were included to create linear regression model to assess predictors of cost of care. For multivariable analyses for cost of care, data were log transformed to achieve a normal distribution. For trend analysis of categorical variables, such as 30‐day readmission and in‐hospital mortality, the modified chi‐squared test of trend for proportions (Cochrane–Armitage test) was used. For a continuous variable, such as cost of care, simple linear regression was used to obtain P values for yearly trends. All readmissions were identified using “NRD_visitlink,” and the hospital identifier “Hosp_NRD” was used further to track readmissions to same versus different hospitals.

Results

Baseline Characteristics

Our study included 1 723 378 patients who had an index hospitalization with AF as the primary discharge diagnosis between 2010 and 2014. Among these, 249 343 (14.5%) patients were readmitted within 30 days after discharge (Table 1. Compared with the nonreadmitted group, the majority of readmitted patients were females (56.2% versus 50.89%; P<0.001), aged >65 years (76.7% versus 67.77%; P<0.001), and covered through Medicare/Medicaid (82.4% versus 70.43%; P<0.001). Readmitted patients had a higher baseline burden of comorbidities represented by the Charlson comorbidity index ≥2 compared with those without a readmission. Hypertension, heart failure, and diabetes mellitus were the most common comorbidities among hospitalized patients, with higher prevalence among the readmitted group. Lower utilization of electrical cardioversion (6.96% versus 8.8%; P<0.001) and catheter ablation (0.44% versus 0.57%; P<0.001) were observed among the readmitted group. Patients who required readmission had a significantly higher frequency of unplanned hospitalizations (89.96% versus 86.3%; P<0.001), weekend index admissions (20.82% versus 20.12%; P<0.001), and disposition to nursing facilities (16.8% versus 9.77%; P<0.001) compared with patients who were not readmitted. Readmitted patients had higher median (interquartile range) LOS and cost of hospitalization during the index admission compared with those without a readmission. Thirty‐day readmission rates among different subgroups are presented in Table S5.

Table 1.

Baseline Characteristics of Patients With 30‐Day Readmission vs No Readmission With Atrial Fibrillation

	Readmission		Overall	P Value
	No	Yes
Index population	1 474 035	249 343 (14.47%)	1 723 378
Patient‐level variables
Age, y (%)				<0.001
18 to 49	8.27	4.26	7.69
50 to 64	23.97	19.03	23.25
65 to 79	39.13	39.78	39.22
≥80	28.64	36.93	29.84
Sex (%)				<0.001
Male	49.11	43.82	48.34
Female	50.89	56.18	51.66
Primary payer, %				<0.001
Medicare/Medicaid	70.43	82.41	72.16
Private including HMO	23.6	13.6	22.15
Self‐pay/no charge/other	5.76	3.82	5.48
Missing	0.21	0.18	0.21
Median household income category for patient's ZIP code, %a				<0.001
0 to 25th percentile	27.05	30.09	27.49
26 to 50th percentile	26.12	26.08	26.12
51 to 75th percentile	24.28	23.51	24.17
76 to 100th percentile	22.54	20.32	22.22
Deyo/Charlson Score, %b				<0.001
0 to 1	67.64	49.03	64.95
2 to 3	24.26	33.97	25.66
≥3	8.10	17.00	9.39
Comorbidities, %
Congestive heart failurec	27.55	40.73	29.45	<0.001
Neurological disorder or paralysisd	6.43	8.87	6.79	<0.001
Chronic lung diseased	20.25	29.51	21.59	<0.001
Diabetes mellitusd	24.34	30.82	25.28	<0.001
Chronic renal failured	12.28	21.20	13.57	<0.001
Hypertensiond	68.94	72.06	69.39	<0.001
Hypothyroidismd	15.98	17.84	16.25	<0.001
Hematological or oncological malignancyd	3.61	7.47	4.16	<0.001
Chronic liver diseased	1.54	2.45	1.67	<0.001
Coagulopathyd	3.13	4.70	3.36	<0.001
Valvular heart diseased	0.23	0.43	0.26	<0.001
Deficiency anemiad	11.38	19.47	12.55	<0.001
Previous MIc	6.73	8.50	6.99	<0.001
Previous CABGc	7.11	9.26	7.42	<0.001
Previous strokec	1.75	2.44	1.85	<0.001
Peripheral vascular diseased	6.38	9.51	6.83	<0.001
Fluid and electrolyte disturbancesd	19.62	26.35	20.60	<0.001
Electrical cardioversionc	8.80	6.96	8.54	<0.001
Pacemaker/implantable cardioverter defibrillatorc	1.17	1.13	1.17	0.066
Catheter ablationc	0.57	0.44	0.55	<0.0001
Hospital characteristics
Hospital bed size, %e				0.040
Small	13.45	13.33	13.43
Medium	23.71	23.92	23.74
Large	62.84	62.75	62.83
Hospital teaching status, %f				0.203
Nonteaching	53.91	54.05	53.93
Teaching	46.09	45.95	46.07
Admission type, %				<0.001
Nonelective	86.30	89.96	86.83
Elective	13.70	10.04	13.17
Admission day, %				<0.001
Weekdays	79.88	79.18	79.78
Weekend	20.12	20.82	20.22
Disposition, %				<0.001
Home	89.45	81.85	88.35
Facility/others	9.77	16.8	10.79

CABG indicates coronary artery bypass graft; HMO, Health Maintenance Organization; MI, myocardial infarction.

^aRepresents a quartile classification of the estimated median household income of residents in the patient's ZIP code, derived from ZIP code demographic data obtained from Claritas. The quartiles are identified by values of 1 to 4, indicating the poorest to wealthiest populations. Because these estimates are updated annually, the value ranges vary by year. https://www.hcup-us.ahrq.gov/db/vars/zipinc_qrtl/nrdnote.jsp.

^bCharlson/Deyo comorbidity index was calculated as per Deyo classification.

^cOther primary diagnosis: derived from appropriate International Classification of Diseases, Ninth Revision, Clinical Modification code mentioned in Table S2.

^dVariables are Agency for Healthcare Research and Quality comorbidity measures.

^eThe bed‐size cut‐off points divided into small, medium, and large have been done so that approximately one‐third of the hospitals in a given region, location, and teaching status combination would fall within each bed‐size category. https://www.hcup-us.ahrq.gov/db/vars/hosp_bedsize/nrdnote.jsp.

^fA hospital is considered to be a teaching hospital if it has an American Medical Association–approved residency program, is a member of the Council of Teaching Hospitals, or has a ratio of full‐time equivalent interns and residents to beds of 0.25 or higher. https://www.hcup-us.ahrq.gov/db/vars/hosp_ur_teach/nrdnote.jsp.

Etiologies of 30‐Day Readmission

Cardiac conditions were the most common causes (52.6%) of 30‐day readmission. Among cardiac conditions, AF was the most common (24.1%) followed by heart failure (12%) and ischemic heart disease (3%). The common noncardiac causes included pulmonary conditions (9.7%), infections (5.6%), and bleeding complications (4.1%; Figure 1.

Figure 1.

Etiologies of 30‐day readmissions after atrial fibrillation. GI indicates gastrointestinal.

Predictors of 30‐Day Readmission

Results of multivariable hierarchical logistic regression analysis for predictors of 30‐day readmission are presented in Table 2. Results showed that advancing age, female sex, and coexisting comorbid conditions like heart failure, neurological disorders, chronic lung disease, diabetes mellitus, chronic kidney disease, chronic liver disease, coagulopathy, anemia, peripheral vascular disease, and fluid/electrolyte disturbances were predictive of readmission within 30 days of discharge after an index admission for AF, whereas private insurance (compared with Medicare/Medicaid), higher median household income, elective cardioversion, and catheter ablation during the index admission and elective admissions were associated with lower risk of readmission within 30 days.

Table 2.

Multivariate Predictors of 30‐Day Readmission After Atrial Fibrillation

	Odds Ratio	LL	UL	P Value
Variable
Age	1.007	1.006	1.007	<0.001
Sex
Male	Referent	Referent	Referent
Female	1.06	1.05	1.08	<0.001
Primary payer
Medicare/Medicaid	Referent	Referent	Referent
Private including HMO	0.71	0.70	0.73	<0.001
Self‐pay/no charge/other	0.77	0.74	0.80	<0.001
Median household income category for patient's ZIP codea
0 to 25th percentile	Referent	Referent	Referent
26 to 50th percentile	0.94	0.92	0.95	<0.001
51 to 75th percentile	0.92	0.90	0.94	<0.001
76 to 100th percentile	0.89	0.87	0.91	<0.001
Comorbidities
Congestive heart failureb	1.38	1.36	1.40	<0.001
Neurological disorder or paralysisc	1.15	1.13	1.18	<0.001
Chronic lung diseasec	1.38	1.36	1.41	<0.001
Diabetes mellitusc	1.20	1.18	1.22	<0.001
Chronic renal failurec	1.33	1.31	1.36	<0.001
Hypertensionc	1.00	0.98	1.01	0.776
Hypothyroidismc	1.00	0.98	1.02	0.801
Chronic liver diseasec	1.43	1.37	1.50	<0.001
Coagulopathyc	1.14	1.10	1.18	<0.001
Valvular heart diseasec	0.90	0.80	1.02	0.088
Deficiency anemiac	1.35	1.32	1.37	<0.001
Peripheral vascular diseasec	1.16	1.14	1.19	<0.001
Fluid and electrolyte disturbancesc	1.16	1.14	1.18	<0.001
Electrical cardioversionb	0.89	0.87	0.92	<0.001
Catheter ablationb	0.90	0.81	0.99	0.031
Hospital bed sized
Small	Referent	Referent	Referent
Medium	1.01	0.98	1.04	0.425
Large	1.00	0.98	1.03	0.768
Hospital teaching statuse
Nonteaching	Referent	Referent	Referent
Teaching	1.04	1.02	1.05	<0.001
Admission type
Nonelective	Referent	Referent	Referent
Elective	0.83	0.81	0.85	<0.001
Admission day
Weekdays	Referent	Referent	Referent
Weekend	1.01	0.99	1.03	0.182
Length of stay	1.03	1.03	1.03	<0.001
C‐Index	0.658

HMO indicates Health Maintenance Organization; LL, Lower limit; UL, Upper limit.

^aRepresents a quartile classification of the estimated median household income of residents in the patient's ZIP code, derived from ZIP code demographic data obtained from Claritas. The quartiles are identified by values of 1 to 4, indicating the poorest to wealthiest populations. Because these estimates are updated annually, the value ranges vary by year. https://www.hcup-us.ahrq.gov/db/vars/zipinc_qrtl/nrdnote.jsp.

^bOther primary diagnosis: derived from appropriate International Classification of Diseases, Ninth Revision, Clinical Modification code mentioned in Table S2.

^cVariables are Agency for Healthcare Research and Quality comorbidity measures.

^dThe bed‐size cut‐off points divided into small, medium, and large have been done so that approximately one‐third of the hospitals in a given region, location, and teaching status combination would fall within each bed‐size category. https://www.hcup-us.ahrq.gov/db/vars/hosp_bedsize/nrdnote.jsp.

^eA hospital is considered to be a teaching hospital if it has an American Medical Association–approved residency program, is a member of the Council of Teaching Hospitals, or has a ratio of full‐time equivalent interns and residents to beds of 0.25 or higher. https://www.hcup-us.ahrq.gov/db/vars/hosp_ur_teach/nrdnote.jsp.

Predictors of Cost of Hospitalization for Index Admission

Results of multivariable hierarchical linear regression analysis for predictors of cost of hospitalization during index admission are presented in Table 3. Readmissions were associated with a 3% increase in cost of care (β coefficient, 0.030; P<0.001). Compared with the 18 to 49 years age group, the 50 to 64 years and 65 to 79 years age groups had a higher cost of care, but the age group ≥80 years was associated with reduced (β coefficient, −0.024; P<0.001) resource utilization. Higher burden of comorbidities was predictive of higher cost of hospitalization during the index admission for AF. Factors associated with the highest increase in cost of care were utilization of catheter ablation during the index admission (β coefficient, 0.974; P<0.001), heart failure (β coefficient, 0.134; P<0.001), coagulopathy (β coefficient, 0.087; P<0.001), and fluid/electrolyte disturbances (β coefficient, 0.085; P<0.001). Other factors associated with increased resource utilization were higher income, admission to teaching centers, elective admission, and longer LOS during the index hospitalization. Factors associated with reduced cost burden were female sex (β coefficient, −0.048; P<0.001), electrical cardioversion during the index admission (β coefficient, −0.116; P<0.001), and admission during the weekend (β coefficient, −0.008; P<0.001).

Table 3.

Multivariate Predictors of Cost of Care Associated With Atrial Fibrillation Hospitalizations

	Cost of Hospitalization			P Value
	β‐Coefficient	LL	UL
Intercept	8.21	8.19	8.22	<0.001
Age, y
18 to 49	Referent	Referent	Referent
50 to 64	0.05	0.04	0.05	<0.001
65 to 79	0.03	0.02	0.03	<0.001
≥80	−0.02	−0.03	−0.02	<0.001
30‐day readmission	0.03	0.03	0.03	<0.001
Sex
Male	Referent	Referent	Referent
Female	−0.05	−0.05	−0.05	<0.001
Primary payer
Medicare/Medicaid	Referent	Referent	Referent
Private including HMO	0.000	−0.004	0.005	0.987
Self‐pay/no charge/other	0.000	−0.007	0.007	0.957
Median household income category for patient's ZIP codea
0 to 25th percentile	Referent	Referent	Referent
26 to 50th percentile	0.008	0.003	0.012	0.001
51 to 75th percentile	0.007	0.002	0.011	0.005
76 to 100th percentile	0.006	0.001	0.011	0.027
Comorbidities
Congestive heart failureb	0.134	0.130	0.137	<0.001
Neurological disorder or paralysisc	0.038	0.033	0.044	<0.001
Chronic lung diseasec	0.074	0.071	0.078	<0.001
Diabetes mellitusc	0.043	0.040	0.046	<0.001
Chronic renal failurec	0.012	0.008	0.016	<0.001
Hypertensionc	0.004	0.001	0.007	0.011
Hypothyroidismc	0.000	−0.004	0.004	0.954
Chronic liver diseasec	0.047	0.036	0.058	<0.001
Coagulopathyc	0.087	0.079	0.094	<0.001
Valvular heart diseasec	0.023	−0.005	0.052	0.111
Deficiency anemiac	0.073	0.069	0.077	<0.001
Peripheral vascular diseasec	0.045	0.039	0.050	<0.001
Fluid and electrolyte disturbancesc	0.085	0.082	0.089	<0.001
Electrical cardioversionb	−0.116	−0.121	−0.111	<0.001
Catheter ablationb	0.974	0.956	0.993	<0.001
Hospital bed sized
Small	Referent	Referent	Referent
Medium	−0.025	−0.044	−0.005	0.014
Large	0.011	−0.008	0.029	0.260
Hospital teaching statuse
Nonteaching	Referent	Referent	Referent
Teaching	0.067	0.050	0.085	<0.001
Admission type
Nonelective	Referent	Referent	Referent
Elective	0.306	0.301	0.310	<0.001
Admission day
Weekdays	Referent	Referent	Referent
Weekend	−0.006	−0.010	−0.003	<0.001
Length of stay	0.106	0.106	0.107	<0.001

HMO indicates Health Maintenance Organization; LL, Lower limit; UL, Upper limit.

^aRepresents a quartile classification of the estimated median household income of residents in the patient's ZIP code, derived from ZIP code demographic data obtained from Claritas. The quartiles are identified by values of 1 to 4, indicating the poorest to wealthiest populations. Because these estimates are updated annually, the value ranges vary by year. https://www.hcup-us.ahrq.gov/db/vars/zipinc_qrtl/nrdnote.jsp.

^bOther primary diagnosis: derived from appropriate International Classification of Diseases, Ninth Revision, Clinical Modification code mentioned in Table S2.

^cVariables are Agency for Healthcare Research and Quality comorbidity measures.

^dThe bed‐size cut‐off points divided into small, medium, and large have been done so that approximately one‐third of the hospitals in a given region, location, and teaching status combination would fall within each bed‐size category. https://www.hcup-us.ahrq.gov/db/vars/hosp_bedsize/nrdnote.jsp.

^eA hospital is considered to be a teaching hospital if it has an American Medical Association–approved residency program, is a member of the Council of Teaching Hospitals, or has a ratio of full‐time equivalent interns and residents to beds of 0.25 or higher. https://www.hcup-us.ahrq.gov/db/vars/hosp_ur_teach/nrdnote.jsp.

Trends in 30‐Day Readmission and Cost of Hospitalization

Thirty‐day readmission pattern after discharge from index hospitalization is shown in Figure 2. Half of the readmissions occurred in the first 13 days of discharge. We noted a decrease in 30‐day readmission rate during the study period (14.75% in 2010 and 14.33% in 2014; P trend, <0.001). Median (interquartile range) cost of hospitalization decreased from USD 5740 (3549–10 025) in 2008 to USD 5660 (3529–9744) in 2014 (P trend, <0.001).

Figure 2.

A, Yearly trends in 30‐day readmission following atrial fibrillation–related hospitalization. B, Trends of readmission per day postdischarge from index hospitalization with atrial fibrillation.

Same Versus Different Hospital Readmission Outcomes

Outcomes related to same versus different hospital readmission after AF index hospitalization are shown in Figure 3. Readmission in hospitals different to index hospitalization was noted in 21.6% of patients. We noted higher in‐hospital mortality (5.75% versus 4.91%; P<0.001), LOS (6.2 versus 5.4 days; P<0.001), and cost of care (USD 8831 versus 7179; P<0.001) in patients admitted to different hospitals compared with the same hospitals.

Figure 3.

Outcomes related to same vs different hospital admission after index atrial fibrillation hospitalization. LOS indicates length of stay; USD, US dollars.

Discussion

Our study highlights several important findings related to 30‐day readmissions after index hospitalization for AF. First, ≈1 in 7 patients were readmitted within 30 days of discharge, and symptomatic AF was the most common cause of readmission. Second, advancing age, female sex, and certain coexisting comorbid conditions correlated with increased risk of 30‐day readmission. Third, there was a small, but progressive, decline in readmission rates and cost of care from 2010 to 2014. Fourth, readmissions were associated with a significant rise in cost of care for hospitalizations related to AF. Last, electrical cardioversion during the index hospitalization correlated independently with reduced 30‐day rehospitalizations and cost burden whereas catheter ablation was associated with reduced 30‐day readmissions, but a rise in cost of care, during the index hospitalization.

AF is the most common arrhythmia that is contributing to a substantial increase in morbidity and mortality with huge implications on healthcare economies throughout the world.16 In this context, numerous recent studies have focused on multiple aspects of AF, including trends in hospitalization,1, 6 treatment patterns,17, 18 mortality rates, and readmission rates.7, 8 Given that the burden of AF is known to increase with advancing age, readmission rates have been a topic of great interest. An estimated 2.6 million US seniors (ie, 1 in 5 Medicare beneficiaries) are readmitted within 30 days after hospital discharge, resulting in an annual financial burden of $26 billion on the US economy.19 Previously, Freeman et al studied the temporal trends of AF readmission rates among Medicare beneficiaries.7 They reported an overall 30‐day readmission rate of 15%, with a progressive decline by 1% per year from 1999 to 2013. In another study, Johnson et al reported a 30‐day readmission rate of 18% utilizing the MarketScan Hospital Drug Database.20 The readmission rate of 14.5% observed during our study period is similar to these earlier studies. However, the larger study sample and the utilization of an all‐payer source database with universal representation of overall AF population are some of the unique features of our study.

In our study, cardiac conditions were the most common causes of subsequent readmission within 30 days after the index hospitalization for AF. Among the cardiac conditions, AF was the most common etiology. Jencks et al reported that patients were more likely to get readmitted for the same medical condition they had been admitted and treated for during the index hospitalization.19 So, it is not uncommon for patients with a discharge diagnosis of AF to get readmitted for the same problem. We also found that 87% of these patients with a discharge diagnosis of AF were readmitted once, whereas 11.3% and 1.3% were readmitted twice and thrice, respectively (Figure 4). In fact, many of the patients with AF have a chronic underlying disease that is often unrecognized, and the multiple coexisting comorbidities could trigger AF requiring subsequent hospital admissions. For example, Scardi et al studied the characteristics of first detected lone AF at a single center in Italy, and found that 55% of the patients had paroxysmal AF whereas the rest had chronic AF at the time of diagnosis.21 Furthermore, Lubitz et al evaluated the patterns of AF in Framingham study participants and reported that it is uncommon to find AF without recurrence in a community during a follow‐up period of 1 year.22 They found that only 6% of their study population had not had AF during a follow‐up of 1 year. The long‐term recurrence rates of AF have been estimated at 22% to 49% within the first 3 years of detection23, 24, 25 and nearly 60% during a follow‐up of 5 years.26 The wide variation in recurrence rates can be attributed to the heterogeneity of the study samples and variations in the frequency of rhythm assessment in these studies. A previous study by Munir et al reported heart failure as the most common cause for 30‐day readmissions.8 However, their study period was limited to only 1 year, which resulted in a smaller sample size. In contrast, we included 1.7 million of AF‐related hospitalization records spanning over a period of 5 years.

Figure 4.

Frequency of 30‐day readmissions.

In our previous study, we analyzed the utilization of CHADS2 and CHA₂DS₂‐VASc scores as surrogate markers for predicting mortality and 30‐day readmission risk.27 We found that chances of getting readmitted within 30 days increased with an increment in the risk profile of the patient. In the current study, we extend our previous work by studying other predictors of 30‐day readmission and cost of care. Advancing age was associated with increased risk of 30‐day readmissions. This can be explained by the increasing burden of comorbidities and limited hemodynamic reserve in the elderly patient population resulting in a poor prognosis. Females were at higher odds of readmission compared with males after AF‐related hospitalization. This can be attributed to the sex disparity in the treatment utilization as reported previously.28, 29, 30 This provides an opportunity to focus attention on underlying biological and sociocultural mechanisms responsible for sex‐specific differences and also identifies barriers in the delivery of effective treatment for AF.

Another important finding that merits discussion is the progressive decline in 30‐day readmission rates during our study period. This observation is comparable with the previous study done by Freeman et al, who also reported a progressive decline in 30‐day readmission rates of AF.7 It may be related to the remarkable advances made in the treatment of AF. Particularly, electrical cardioversion and catheter ablation in AF have improved symptom burden and quality of life and also reduced the incidence of stroke and mortality.31, 32 There has been an increase in the utilization of antiarrhythmic drugs, changes in prescription pattern of anticoagulants with increased use of novel oral anticoagulants, and improvements in cardiac resynchronization therapy and permanent pacemakers.17, 33, 34 All of these practices might have contributed to the declining rates of rehospitalization.

To our knowledge, this is the first study to explore the cost burden of AF‐related readmissions from a national database. We noted a declining cost of care during our study period, which largely results from reduction in readmissions, attributable to increasing utilization and universal availability of effective treatment strategies.35 Through our analysis, we noted a 3% increase in cost with readmission. An interesting finding of our study was the lower cost of care in the elderly population (age ≥80 years), female patients, and weekend admissions. Underutilization of AF treatment modalities among these subgroups, as reported previously,36, 37, 38, 39 appears to be the principle reason behind lower cost of care noted in these patients. This study provides vital data and prediction modeling to identify a high‐risk AF population that can be further targeted to implement customized transitional care and reduce preventable readmissions.

We noted improved outcomes when electrical cardioversion and catheter ablation were performed for AF during the index hospitalization. We believe that an early rhythm control strategy with these tools is underutilized. Our findings call for more‐frequent implementation of electrical cardioversion and catheter ablation, especially in subpopulations with worse outcomes (ie, females, elderly patients).

Disparity in outcomes was also noted when patients were readmitted in hospitals other than the hospital of index hospitalization. Our findings are consistent with those noted by Lahewala et al.40 Familiarity with the patient, smooth posthospitalization, and avoidance of unnecessary investigations appear to be driving factors for improved outcomes with the same hospital readmissions.

Finally, our results support the need for development of readmission prediction models that can identify patients at increased risk. Future prospective studies should investigate the preventability of 30‐day readmissions after the index hospitalization for AF.

Limitations

The findings of our study should be viewed in the context of the following limitations. The use of an administrative database like the NRD is associated with risk of errors attributable to variations in coding practices. Only data related to readmission within 30 days after an index hospitalization are available. Information related to longitudinal follow‐up of these patients are not available. Information related to race, ethnicity, and hospital regions are not reported in the NRD. The NRD does not include causes of death. Individual operator‐ and procedure‐level data are not available. This study exclusively involves a US patient population; thus, results may not be applicable to other countries. The NRD does not include deaths out of the hospital or in the emergency room. Other factors that can affect a patient's prognosis, such as drugs, are not included in the database as well. The large sample size included in our study can partially offset some of these limitations, and the real‐world clinical experience can add to the current existing literature.

Conclusions

In this large, nationally representative study, ≈1 in 7 patients were readmitted within 30 days after the index hospitalization for AF. Cardiac conditions were the most common causes for readmission, and AF was the most common cardiac culprit. Our study demonstrated declining readmission rates and cost of care during the study period, possibly attributable to wide availability and utilization of effective treatment options. We propose a predictive model to identify the patients at high risk of 30‐day readmission and high cost of care.

Disclosures

None.

Supporting information

Table S1. Description of Etiologies for 30‐Day Readmission

Table S2. Description of ICD‐9‐CM Codes for Etiologies of 30‐Day Readmission

Table S3. International Classification of Disease, Ninth Edition, Clinical Modification (ICD‐9‐CM) for Comorbidities

Table S4. Deyo's Modification of Charlson Comorbidity Index

Table S5. Subgroup Analysis of 30‐Day Readmissions

Figure S1. Overflow diagram depicting patient selection algorithm.

Figure S2. Yearly trend of in‐hospital mortality following atrial fibrillation hospitalization.

(J Am Heart Assoc. 2019;8:e013026 DOI: 10.1161/JAHA.119.013026.)