Hospitalization trends for thromboembolic events in atrial fibrillation: A 12-year national analysis

Abstract

Background

Direct oral anticoagulants and percutaneous left atrial appendage occlusion (LAAO) devices were approved for use in 2010 and 2015, respectively. It is unknown to what extent, if any, these new stroke preventive therapies have impacted hospitalizations for thromboembolic (TE) events.

Objectives

To evaluate temporal trends in AF-related hospitalizations for acute ischemic stroke (AIS), transient ischemic attack (TIA), and systemic embolism (SEE) in the U.S. from 2010 to 2021.

Methods

Using the National Inpatient Sample, we identified hospitalizations for TE events with comorbid AF. Data were grouped into two periods (2010–2015 and 2016–2021). Linear regression assessed trends in TE frequency. We also examined anticoagulation (AC) use and LAAO procedures among inpatients with AF.

Results

A total of 1,692,373 AF-related TE hospitalizations were identified: 798,413 (2010–2015) and 893,960 (2016–2021). The frequency of hospitalizations for any AF-TE event, as a fraction of total hospitalizations in patients with AF, declined from 3.69 % to 3.35 % (P < 0.001). AF-related AIS hospitalizations rose from 2.71 % to 2.89 % in 2010–2015 (P < 0.001) but declined from 3.02 % to 2.89 % in 2016–2021 (P < 0.001). TIA (0.75 % to 0.35 %) and SEE (0.22 % to 0.10 %) hospitalizations also decreased (P < 0.001). AC use increased from 21.2 % in 2010 to 42.4 % in 2021 (P < 0.001), while LAAO procedures rose sharply from 5129 in 2016 to 46,080 in 2021 (P < 0.001).

Conclusion

TE hospitalizations among inpatients with comorbid AF declined from 2010 to 2021, primarily driven by a decrease in TIAs and SEE. Acute ischemic stroke hospitalizations declined after 2016, coinciding with increased AC use and LAAO adoption.

1. Introduction

Atrial fibrillation (AF) is associated with a 4- to 5-fold increased risk of stroke and a 3-fold increased risk of extracranial systemic embolic events (SEE) [1,2]. It accounts for 25 % of cerebrovascular accidents in the United States (US) [2,3]. Although vitamin K antagonists (VKA) greatly reduce the risk of stroke in AF, compliance with anticoagulation (AC) therapy had been suboptimal, in part due to low adherence and potential or realized bleeding complications [4,5]. Following the introduction of direct oral anticoagulants (DOAC), first approved in 2010 by the US Food and Drug Administration (FDA), these medications have emerged as the primary therapeutic alternative to VKAs. In 2015, regulatory approval of the first transcatheter left atrial appendage occlusion (LAAO) device introduced an alternative nonpharmacologic stroke prevention therapy for patients who are unable to tolerate or have other justification to avoid long-term anticoagulation (AC) [6,7]. However, whether these milestones have impacted hospitalizations for AF-related thromboembolic (TE) events in the US remains unclear.

Previous National Inpatient Sample (NIS) database analyses had reported increasing prevalence of AF in acute ischemic stroke (AIS) and transient ischemic attacks (TIA) hospitalizations prior to 2010 and up to 2015 [[8], [9], [10]]. A more recent NIS study reported decreasing prevalence of comorbid AF in patients admitted for AIS from 2015 to 2020, specially in women >60. [11]. Recent data from outside the US has reported a decline in AF-related AIS from 2010 to 2020, paralleling an increase in the use of anticoagulation therapy [[12], [13], [14], [15], [16], [17]]. These studies, however, did not include data on TIAs and SEEs, nor have considered the total of all TE event hospitalizations in this population. Thus, using the NIS database, we evaluated whether the rate of primary hospitalizations for AIS, TIAs, and SEEs among inpatients with comorbid AF declined between 2010 and 2021.

2. Methods

We conducted a retrospective study using the NIS database, the largest inpatient care database in the US, which is derived from billing data submitted by hospitals to statewide data organizations as part of the Healthcare Cost and Utilization Project (HCUP) [18]. Each observation in the NIS represents an individual hospitalization with a primary diagnosis, up to 39 secondary diagnoses, and up to 25 procedure codes. To estimate the total number of hospitalized patients across the United States, a uniform sampling and weighting approach recommended by HCUP was used. The NIS has been used extensively to assess national trends and inpatient outcomes in patients with AF, among other diseases [11,[19], [20], [21], [22]].

IRB approval exemption was granted as a deidentified administrative database was used. We adhered strictly to the NIS survey methodology on data interpretation, research design, and data analysis as described by the HCUP. Guidelines according to the Strengthening the Reporting of Observational Studies in Epidemiology Statement were followed, along with methodologies for best practices using claims datasets [23].

To identify our sample, adult hospitalizations with International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes for AF were identified. Among those, hospitalizations with ICD codes of AIS, TIA, and SEE in the primary diagnosis position were then identified. ICD-9-CM codes were used from 2010 to the third quarter of 2015, and ICD-10 codes were used from the fourth quarter of 2015 to 2021. Records with diagnosis codes for rheumatic mitral valve stenosis or the presence of a mitral prosthetic valve were excluded from the study.

Data were divided into two temporal periods: the first included hospitalizations from 2010 to 2015, and the second included hospitalizations from 2016 to 2021. Baseline patient characteristics and clinically relevant comorbidities were described, using previously validated ICD-10 and ICD-9-CM codes, when available. Components of the CHA2DS2-VASc score were defined by the age of 65 to 74 years, age ≥ 75 years, a diagnosis of heart failure, hypertension, diabetes mellitus, previous ischemic stroke, female sex, and vascular disease (prior myocardial infarction, prior revascularization, peripheral arterial disease, or coronary artery disease). The ICD-10 and ICD-9-CM codes used to identify comorbidities are provided in Supplemental Table 1.

To identify TE events, we exclusively utilized the primary hospitalization diagnosis to increase the probability that the hospitalization was attributable to one of the adverse events identified in our study rather than a subsequent complication or a previously documented issue. A composite of all TE event hospitalizations was described, consisting of the sum of AIS, TIA, and SEE hospitalizations each year. SEE included all cases of presumed embolic acute limb ischemia or acute visceral ischemia. Hospitalizations were expressed in frequency, calculated by dividing the weighted number of primary hospitalizations for AIS, SEE, and/or TIAs with comorbid AF (numerator) by the total weighted number of hospitalizations with a diagnosis of AF in a secondary position (denominator).

To describe the utilization of AC use, changes in frequency of the ICD-9-CM code V58.69 and ICD-10 code Z79.01 were used. Yearly AC use rates were calculated by dividing the weighted number of inpatients with a diagnosis of AF and concurrent use of AC (numerator) by the total weighted number of patients with a diagnosis for AF (denominator) in the NIS. The total number of reported LAAO procedures per year were also described. Annual percent changes (APC) between years were described for each outcome.

Survey methodology was used for all analyses to account for the clustering and stratification of hospitalizations in the NIS as recommended by the Agency for Healthcare Research and Quality [25]. The trend weight files were merged onto the original NIS files by year and hospital identification number. For years before 2012, the trend weight was used to create national estimates for trend analysis. For 2012 and after, no trend weight was needed, and the regular discharge weight was used, consistent with the redesigned NIS trend analysis [26]. Cochran-Mantel-Haenszel trend test was used for categorical variables, and linear regression was used for continuous variables. Continuous variables were expressed as means and standard deviations. Categorical data were expressed as frequencies. Categorical variables were compared between cohorts using Pearson's Chi-Squared (χ2) test. All reported P values were two-sided, and the significance cut-off was set at 0.05. The Bonferroni adjustment was used to adjust for multiple testing between categorical χ² tests. Continuous variables were compared using the independent samples Student t-test if normally distributed, and Mann-Whitney U test for non-normally distributed continuous variables. All statistical analyses were performed using SPSS (IBM SPSS Statistics, Version 28.0, IBM Corporation, Armonk, NY).

3. Results

A total of 42,535,490 hospitalizations with secondary diagnosis codes for AF between 2010 and 2021 were identified. The prevalence of AF as a comorbidity among all hospitalizations in the NIS increased from 7.3 % in 2010 to 11.8 % in 2021 as shown in Fig. 1. Among these, 1,692,373 hospitalizations had primary diagnosis codes for TE events: 798,413 hospitalizations (47.2 %) from 2010 to 2015 (mean age 78.9 [standard deviation (SD) 10.7] years; 57.8 % female) and 893,960 hospitalizations (52.8 %) from 2016 to 2021 (mean age 77.5 [SD 10.9] years; 53.1 % female).

Fig. 1.

Changes in yearly prevalence of comorbid atrial fibrillation among all inpatients in the United States from 2010 to 2021.

Temporal changes in baseline characteristics of patients with AF hospitalized for TE events are listed in Table 1. Hospitalizations for AIS accounted for 75.9 % of all TE hospitalizations in 2010 to 2015 and 84.2 % in 2016 to 2021 (P < 0.001). A significant increase in the prevalence of stroke risk factors between periods was observed, such as hypertension (88.9 % vs. 83.8 %; P < 0.001), hyperlipidemia (61.3 % vs. 52.1 %; P < 0.001), diabetes (36.0 % vs. 31.7 %; P < 0.001), and tobacco use (32.6 % vs 21.1 %; P < 0.001). Inpatients with AF in 2016–2021 more frequently had a history of previous stroke or TIA (28.3 % vs 23.5 %; P < 0.001) compared to 2010–2015. The mean CHA2DS2-VASc score minimally increased from 4.4 in 2010–2015 to 4.5 in 2016–2021 (P < 0.001). The proportion of inpatients with a CHA2DS2-VASc of 2 to 4 decreased (48.9 % vs 51.4 %, P < 0.001) but the proportion of inpatients with a CHA2DS2-VASc ≥5 increased (47.6 % vs 45.2 %, P < 0.001) between periods.

Table 1.

Temporal changes in baseline characteristics in hospitalizations for AF-associated thromboembolism between in 2010 to 2015 and 2016 to 2021.

Variable	2010 to 2015, n = 798,413		2016 to 2021, n = 893,960		P-valuea
	N	(%)	N	(%)
Age, years ± SD	78.9 ± 10.7		77.5 ± 11.0		<0.001
18–44	5018	0.6	7335	0.8	<0.001
45–64	81,406	10.2	111,950	12.5
65–74	141,507	17.7	188,075	21.0
>75	570,456	71.5	586,585	65.6
Female	461,212	57.8	474,675	53.1	<0.001
White	600,721	80.6	672,070	77.4	<0.001
Black	66,863	9.0	91,470	10.5
Hispanic	41,553	5.6	57,090	6.6
Other races	36,244	4.9	47,830	5.5
Mean CHADSVASc Score ± SD	4.38 ± 1.56		4.45 ± 1.62		<0.001
CHADSVASc 0–1	27,172	3.4	31,745	3.6	<0.001
CHADSVASc 2–4	410,255	51.4	437,000	48.9
CHADSVASc 5–9	360,986	45.2	425,215	47.6
Hospitalization rates
Ischemic stroke	606,265	75.9	752,765	84.2	<0.001
Transient Ischemic Attacks	144,463	18.1	106,425	11.9	<0.001
Systemic Embolic Events	47,685	6.0	34,770	3.9	<0.001
Comorbidities
Coagulopathy	4526	0.6	22,925	2.6	<0.001
Hypertension	669,302	83.8	794,960	88.9	<0.001
Hyperlipidemia	415,606	52.1	546,200	61.3	<0.001
Diabetes	249,103	31.7	319,380	36.0	<0.001
Obesity	62,960	7.9	122,705	13.7	<0.001
Tobacco use	167,689	21.0	291,015	32.6	<0.001
Chronic lung disease	141,826	17.8	168,900	18.9	<0.001
Chronic liver disease	11,162	1.4	21,705	2.4	<0.001
CKD	152,193	19.1	214,785	24.0	<0.001
Anemia	98,462	12.3	138,080	15.6	<0.001
CAD	267,973	33.6	296,890	33.2	<0.001
Heart Failure	238,581	29.9	275,935	30.9	<0.001
Malignancy	25,633	3.2	45,930	5.1	<0.001
Peripheral vascular disease	94,643	11.9	94,210	10.5	< 0.001
Alcohol use	4157	0.5	28,085	3.1	<0.001
Previous history
History of TIA or Stroke	186,970	23.5	252,960	28.3	<0.001
History of MI	67,955	8.5	82,775	9.3	<0.001

MI = myocardial infarction, CKD = Chronic Kidney Disease, CAD = Coronary artery disease, HF = Heart Failure.

^aP value adjusted by Bonferroni correction to account for multiple testing. P-value for significance after correction < 0.002.

3.1. Thromboembolic events

The temporal changes in the frequency of AF-related TE hospitalizations are shown in Fig. 2. Changes in trends of TE hospitalizations between periods are shown in Fig. 3.

Fig. 2.

Temporal changes in the frequency of hospitalizations for acute ischemic stroke (B), transient ischemic attacks (C), systemic embolic events (C), and the composite of all thromboembolic events (A) among inpatients with atrial fibrillation from 2010 to 2021.

Fig. 3.

Comparison of temporal trends in hospitalizations for all atrial fibrillation-related thromboembolic events (A), acute ischemic stroke (B), transient ischemic attacks (C), and systemic embolic events (D) between periods (2010–2015 and 2016–2021) among inpatients with atrial fibrillation.

Hospitalizations for TE events decreased from 3.69 % in 2010 to 3.35 % in 2021 (APC −9.2 %, P < 0.001) (Fig. 2A). There was no significant temporal change in TE hospitalizations from 2010 to 2015 (3.69 % to 3.65 %, P = 0.608), while they significantly decreased from 2016 to 2021 (P < 0.001) (Fig. 3A).

There was a small increase in AIS from 2010 to 2021 (2.71 % to 2.89 %;APC +6.6 %). From 2010 to 2015, AIS hospitalizations increased (2.71 % to 2.91 %, P < 0.001) (Fig. 2B). However, from 2016 to 2021, AIS hospitalizations peaked in 2016 at 3.02 % and steadily decreased thereafter to 2.89 % by 2021 (P < 0.001), diverging from the upward trend seen in 2010–2015 (Fig. 3B).

Hospitalizations for TIA significantly decreased from 0.75 % in 2010 to 0.35 % in 2021 (APC −53.3 %) (Fig. 2C). Significant downward trends in AF-related TIA hospitalizations were observed in both periods, decreasing from 0.75 % to 0.54 % in 2010–2015 (P < 0.001) and from 0.52 % to 0.35 % in 2016–2021 (P < 0.001) (Fig. 3C).

Hospitalizations for AF-related SEE significantly decreased from 0.22 % in 2010 to 0.10 % in 2021 (APC −53.3 %) (Fig. 2D). No significant trend was observed in AF-related SEE hospitalizations from 2010 to 2015 (0.22 % to 0.20 %; P = 0.116). However, a significant downward trend was noted from 2016 to 2021 (0.18 % to 0.10 %; P < 0.001) (Fig. 3D).

3.2. Stroke prevention therapies

Temporal trends in AC use and percutaneous LAAO procedures are shown in Fig. 4.

Fig. 4.

Changes in frequency of anticoagulation use (A) and weighted number of left atrial appendage occlusion procedures (B) in patients with comorbid atrial fibrillation from 2010 to 2021.

Among all hospitalizations with a diagnosis of AF between 2010 and 2021, the proportion of patients on documented AC therapy increased from 21.2 % in 2010 to 42.4 % in 2021 (P_trend < 0.001) (Fig. 4A). The number of LAAO procedures on patients hospitalized with AF increased from 334 procedures in 2010 to 46,080 procedures in 2021 (P_trend < 0.001) (Fig. 4B).

4. Discussion

In this analysis of over a million hospitalizations in patients with AF-associated TE events, the key findings are: 1) There was a significant decline in AF-related hospitalizations for TE events between 2010 and 2021, particularly SEEs and TIAs. Hospitalizations for AF- associated AIS have steadily declined since 2016, despite an increase in the prevalence of comorbid AF, mean CHA2DS2-VASc score, and other stroke risk factors. Hospitalizations for TIA and SEE declined between 2010 and 2021. 2) From 2010 to 2021, there was a marked increase in the reported use of AC therapy among inpatients with AF, accompanied by an incipient rise in transcatheter LAAO procedures between 2016 and 2021.

Previous NIS analyses had reported an uptick in AIS hospitalizations in the general population between 2009 and 2018 and among patients with AF over 90 years of age from 2004 to 2015 [10,21]. A recent NIS study found that AF prevalence in AIS hospitalizations rose from 2010 to 2020, aligning with cohort studies from China and Denmark reporting increased AF-associated ischemic events into the early 2010s, possibly due to a true rise in incidence or improved detection [10,11,21,27,28]. Our study notes a small but significant increase in the frequency of AF-associated AIS hospitalizations from 2010 to 2015, opposite to the decreasing trend in TIA hospitalizations during the same period. This discordance may be partially explained by the 2009 redefinition of TIA, from a symptom-based diagnosis to one requiring absence of infarction on MRI. Thus, the increase in AIS during this period may reflect improved stroke detection through advanced imaging techniques or possibly a greater number of strokes occurring in patients with previously undiagnosed, and therefore untreated, AF [29,30]. These factors may have obscured the early impact of DOACs in reducing thromboembolic events. Nevertheless, we observed that hospitalizations for AIS in our inpatient AF population peaked in 2016 and showed a significant decreasing trend through 2021. This trend indicates that despite the rising prevalence of stroke risk factors, TE events, and AIS in particular, have constituted a decreasing proportion of the hospitalization burden among patients with a diagnosis of AF. This shift may be attributed to the increased use of AC noted in our study, consistent with previous reports from high-income countries that AF-related stroke rates are declining with increased availability of DOACs [[12], [13], [14], [15], [16], [17]].

The proportion of inpatients with AF on AC therapy doubled from 2010 to 2021, with 41.2 % of inpatients reported on AC by 2021. Although not indicative of the true rates of AC use among all patients with AF in the US, this certainly indicates a considerable rise in AC use in this population. This increase is supported by data from other high-income countries showing a rise in AC use among patients with AF after the launch of DOACs and their incorporation into guidelines as the first line AC therapy [[12], [13], [14], [15], [16], [17]]. Our findings notably highlight an increase in reported anticoagulation use beginning in 2016, coinciding with the period when DOAC prescriptions surpassed warfarin prescriptions for patients with AF in the US [31,32]. This timeframe also marked the onset of a downward trend in AIS hospitalizations. The growing use of DOACs during this time may be explained by several factors, including guideline endorsements, increased clinician familiarity, the introduction of reversal agents (e.g. idarucizumab in 2015 for dabigatran, andexanet alfa in 2018 for rivaroxaban/apixaban), better insurance coverage, and increased patient familiarity with DOACs and AF in general due to direct-to-consumer advertising [33]. The adoption of the CHA₂DS₂-VASc score over CHADS₂ in the 2014 AHA/ACC/HRS AF guidelines could have also played a role by expanding indications for anticoagulation [34].

While AC usage has grown, a notable portion of inpatients with AF were not on AC by 2021, suggesting there is still much work to be done in terms of optimizing stroke prevention strategies in patients with AF. Patients presenting with new onset AF and underutilization of the ICD code for AC use may have impacted the reported proportion of inpatients with AF on AC therapy in our data. Many patients also have a contraindication to AC, such as bleeding and fall risk. The utilization of LAAO in patients at high risk for stroke who are not suitable candidates for AC could potentially lead to a decrease in ischemic events in this population in the future. Given the retrospective nature of our data, it is difficult to determine the impact this rise may have had in reducing TE events in patients with AF by 2021. It is unlikely that LAAO significantly impacted the rate of TE events during 2016–2021 given the slow initial adaption of this procedure. However, the effects of LAAO are cumulative and, hopefully, its impact on TE events will be evident over the ensuing 5 years. Several randomized controlled trials are currently ongoing, looking to compare LAAO to DOACs in patients with AF without a contraindication for AC, such as the CHAMPION-AF (NCT 04394546) and the CATALYST (NCT 04226547) trials. These trials aim to show non-inferiority for stroke prevention but superiority for bleeding, which would make LAAO an attractive therapy over the lifelong need for AC.

5. Limitations

There are several limitations to this study. The data sourced from the NIS database utilizes ICD codes, which are used for billing purposes and to encapsulate clinical scenarios retrospectively. This reliance on “claims data” does not provide insights into clinical presentations. Moreover, the accuracy and consistency of coding within the NIS might suffer due to individual provider preferences and variations in clinical documentation. In this regard, the rates of anticoagulant use might be underrepresented if the prescriptions were not adequately documented, as previously described rates in the US are around 55–65 % [16]. Furthermore, it is uncertain how much data from patients with repeated hospitalizations influenced the analysis. The NIS may also omit certain confounding variables not recorded within its scope, such as patient preferences and the reasoning behind healthcare providers' decisions on AC prescriptions. Although all TE hospitalizations had a concurrent secondary code for AF, we are unable to say whether AF was causally related to these TE events. AF may coexist in patients with other TE event etiologies. Additionally, it is difficult to quantify the impact the COVID-19 pandemic may have had in our data. Patients may have avoided hospitals during the early phases of the pandemic, which could have led to underreporting or reduced hospitalization rates. COVID-19 infection itself may increase the risk of AF and TE events. Lastly, due to limitations of our administrative claims data, we are unable to differentiate if AF was present before hospitalizations or was detected after the TE event. It is unknown how many of the patients in this cohort had a new diagnosis of AF during the hospitalization. Acute AF in hospitalized patients with sepsis, pneumonia or other comorbidities has been well described, and could have played a role in our data (35).

6. Conclusion

The frequency of AF-associated TE event hospitalizations in the US has significantly decreased from 2010 to 2021, driven by a continuous decline in TIA and SEE hospitalizations. AIS hospitalizations declined from 2016 to 2021, despite a slightly older patient population with more risk factors for TE events. The use of oral AC and LAAO procedures has significantly risen, particularly since 2016. These are encouraging trends and validate the need to pursue guideline-directed stroke prevention strategies in patients with AF.

The following is the supplementary data related to this article.

Supplemental Table 1

ICD-10 diagnosis (CM) and procedure (PCS) codes.

CRediT authorship contribution statement

Moises A. Vasquez: Writing – original draft, Methodology, Investigation, Formal analysis, Conceptualization. Crystal Yan: Writing – review & editing, Methodology, Investigation, Conceptualization. Natacha Vargas: Writing – original draft, Investigation, Data curation. Samuel Vasquez: Writing – review & editing, Investigation, Conceptualization. Litsa Lambrakos: Writing – review & editing, Supervision, Project administration. Alex Velasquez: Writing – review & editing, Validation, Supervision. Jeffrey J. Goldberger: Writing – review & editing, Validation, Supervision. Raul D. Mitrani: Writing – review & editing, Validation, Supervision, Project administration, Methodology, Conceptualization.

Declaration of Generative AI and AI-assisted technologies in the writing process

Generative AI and AI-assisted technologies were NOT used in the preparation of this work.

Funding

This study was funded in part by unrestricted grants from the Palm Health Foundation and Marmot Donor Advised Fund.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The NIS database is available for purchase online and can be accessed through: https://hcup-us.ahrq.gov/nisoverview.jsp.