1 Introduction

Atrial fibrillation is a highly prevalent arrhythmia which affects thirty-five million people worldwide and a significant risk factor for ischemic and hemorrhagic strokes [1, 2]. The current guidelines indicate patients with atrial fibrillation should take oral anticoagulants to reduce the risk of future ischemic strokes, unless contraindications exist [3, 4]. However, anticoagulant therapy also simultaneously increases the risk of bleeding such as intracerebral hemorrhage (ICH) and is associated with high mortality rates [5,6,7,8,9,10]. Although this risk of ICH in the setting of anticoagulation is well-documented in the literature, little is known regarding ethnic/racial and geographic disparities of this phenomenon. Therefore, we sought out to investigate ICH mortality trends in patients with atrial fibrillation with a focus on identifying disparities based on sex, race and ethnicity, and geographic regions in the United States.

2 Methods

This cross-sectional analysis utilized mortality data from the Centers for Disease Control and Prevention (CDC) Wide-ranging ONline Data for Epidemiologic Research database through death certificate queries, capturing > 99% of the mortality death rates in the US [11]. Data were collected from the years 1999 to 2020. Death certificates included causes of death in the form of International Classification of Diseases, 10th revision (ICD-10). We queried for all deaths with the underlying causes of death related to intracerebral hemorrhage (ICD-10: I61.0, I61.1, I61.2, I61.3, I61.4, I61.5, I61.6, I61.8, and I61.9). We included all decedents which that reported atrial fibrillation / flutter (ICD: I48) as the multiple causes of death, or contributors of death. In addition, we obtained overall death rates related to ICH without underlying atrial fibrillation as a reference group for comparison.

Death certificate information included demographic information such as age, sex, race and ethnicity, and geographic region. Race and ethnicity were documented by the funeral director as reported by an informant, typically the next of kin. Geographic region was included in death certificates as a component of the US Census Regions (i.e., Northeast, Midwest, South, and West). Based on the 2013 National Center for Health Statistics Urbanization Classification Scheme, we stratified all US counties into metropolitan and non-metropolitan regions. Metropolitan regions are described in Table 1. Nonmetropolitan counties are divided into two categories: micropolitan counties, which are part of a micropolitan statistical area, and noncore counties, which are not part of any micropolitan statistical area.

Table 1 2013 National center for health statistics urbanization classification scheme. Characterization and definition of metropolitan counties
Full size table

Crude death counts and population sizes for each corresponding year were obtained. We adjusted mortality data for age using the direct method, with the year 2000 as the standard US population. The age-adjusted mortality rates (AAMR) were compared among all subpopulations based on demographic factors. The lower 95% confidence intervals were calculated by multiplying the crude death rate with the lower 95% confidence limit factor for death rates, which is based on the Poisson variable of the number of deaths. Similarly, the upper 95% confidence interval was obtained by multiplying the crude death rate with the upper 95% confidence limit factor for death rates, also based on the Poisson variable of the number of deaths. To analyze mortality trends over the included 22-year period, we utilized log-linear regression models where temporal variation occurred by identifying inflection points in mortality trends (Joinpoint Regression) [12,13,14,15,16,17]. By use of the Monte-Carlo permutation test, we estimated annual percentage change (APC). Average-annual percentage change (AAPC) were calculated by taking weighted averages of the APC. Two-tailed t-test statistics were used to determine statistical significance in increasing or decreasing trends, with p-level set at < 0.05. Due to confidentiality constraints, yearly mortality data that included fewer than 10 or 20 death counts were suppressed or classified as unreliable, respectively. Therefore, we did not calculate the AAPC for the following populations: Hispanic, non-Hispanic, American Indian/Alaska Native, and Asian/Pacific Islander populations. We conducted a comparison between the AAMRs and AAPCs associated with ICH with and without underlying atrial fibrillation. To do this, we fitted linear regression lines to the annual changes in AAMR for both groups to ascertain the slopes. Following this, a test of parallelism was performed to establish whether the slopes of the two regression lines were statistically different, both cumulatively and across the various demographic subpopulations (i.e., sex, urbanization, and US census regions). Institutional Review Board approval was not mandated due to the use of publicly available and deidentified data. Data visualization and analysis was completed using Stata (Release 17.0; StataCorp LLC).

3 Results

A total of 21,349 deaths identified from ICH in the setting of atrial fibrillation/flutter were identified (Table 2). Yearly population sizes and death counts (crude and age-adjusted) are depicted in Table S1. Deaths counts increased from 729 individuals in 1999 to 1,266 individuals in 2020. Crude mortality increased from 0.26 [95% CI, 0.24–0.28] in 1999 to 0.38 [95% CI, 0.36–0.41] in 2020. The AAMR increased from 0.27 (95% CI, 0.25–0.29) in 1999 to 0.30 (95% CI, 0.29–0.32) in 2020 with an AAPC of + 0.5% [95% CI, 0.2–0.7, p = 0.001] (Fig. 1). No statistically significant inflection points were observed between 1999 and 2020, cumulatively and among subpopulations. The cumulative AAMR during our 22-year study period was 0.29 (95% CI, 0.28–0.29). Males (AAMR: 0.33 [95% CI, 0.33–0.34]) were impacted by higher mortality rates compared to females (AAMR: 0.26 [95% CI, 0.26–0.27]). The AAPC in males (AAPC: + 0.5% [95% CI, 0.1–0.8], p = 0.019) and females (AAPC: + 0.3% [95% CI, 0.0–0.7], p = 0.044) were similar (Fig. 2).

Table 2 Intracerebral hemorrhage mortality data in patients with atrial fibrillation. Cumulative and subpopulation characteristics including population sizes and death counts (crude and age-adjusted)
Full size table
Fig. 1
figure 1

ICH mortality in patients with atrial fibrillation. Yearly AAMR in the US from 1999 to 2020. Abbreviations: AAMR = age-adjusted mortality rate, ICH = intracerebral hemorrhage, US = United States

Full size image
Fig. 2
figure 2

ICH mortality in patients with atrial fibrillation. Yearly AAMR in the US from 1999 to 2020, stratified by sex. Abbreviations: AAMR = age-adjusted mortality rate, ICH = intracerebral hemorrhage, US = United States

Full size image

Non-Hispanic populations (AAMR: 0.29 [95% CI, 0.29–0.29]) had a higher AAMR compared to Hispanic populations (AAMR: 0.20 [95% CI, 0.18–0.21]). Asian populations (AAMR: 0.32 [95% CI, 0.30–0.34]) had the highest AAMR, followed by White (AAMR: 0.30 [95% CI, 0.29–0.30]), Black (AAMR: 0.15 [95% CI, 0.14–0.16]), and American Indian/Alaska Native populations (AAMR: 0.11 [95% CI, 0.08–0.14]). The AAPC among Black (AAPC: + 1.1% [95% CI, -0.1–2.4], p = 0.069) and White (AAPC: + 0.5% [95% CI, 0.2–0.8], p = 0.002) populations were similar (Fig. 3).

Fig. 3
figure 3

ICH mortality in patients with atrial fibrillation. Yearly AAMR in the US from 1999 to 2020, stratified by race. Abbreviations: AAMR = age-adjusted mortality rate, ICH = intracerebral hemorrhage, US = United States

Full size image

Western regions (AAMR: 0.35 [95% CI, 0.34–0.36]) had the highest AAMR, followed by Midwestern (AAMR: 0.31 [95% CI, 0.30–0.32]), Northeastern (AAMR: 0.28 [95% CI, 0.27–0.29]), and Southern (AAMR: 0.25 [95% CI, 0.25–0.26]) regions (Fig. 4). AAPC was highest among the Southern regions (AAPC: + 1.3% [95% CI, 0.8–1.8], p < 0.001), followed by Midwestern regions (AAPC: + 1.0% [95% CI, 0.6–1.4], p < 0.001), Northeastern (AAPC: -0.5% [95% CI, -1.1–0.1], p = 0.082) and Western (AAPC: + 0.3% [95% CI, -0.2–0.7], p = 0.299) regions. Non-metropolitan regions (AAMR: 0.30 [95% CI, 0.29–0.31]) had a similar AAMR compared to metropolitan regions (AAMR: 0.28 [95% CI, 0.28–0.29]). AAPC was higher in non-metropolitan (AAPC: + 1.9% [95% CI, 1.4–2.4], p < 0.001) regions compared to metropolitan (AAPC: + 0.3% [95% CI, -0.1–0.6], p = 0.111) regions (Fig. 5).

Fig. 4
figure 4

US map with age-adjusted ich mortality rates in patients with atrial fibrillation. State-level AAMR per 100,000 population related to ICH mortality in patients with atrial fibrillation. Abbreviations: AAMR = age-adjusted mortality rate, ICH = intracerebral hemorrhage, US = United States

Full size image
Fig. 5
figure 5

ICH mortality in patients with atrial fibrillation. Yearly AAMR in the US from 1999 to 2020, stratified by urbanization scheme. Abbreviations: AAMR = age-adjusted mortality rate, ICH = intracerebral hemorrhage, US = United States

Full size image

The mortality rate of ICH in the absence of atrial fibrillation consistently exhibited higher AAMRs compared to ICH mortality with underlying atrial fibrillation. Specifically, the overall AAMR for ICH without atrial fibrillation was 5.29 (95% CI, 5.27–5.30), whereas it was significantly lower at 0.29 (95% CI, 0.28–0.29) for ICH with underlying atrial fibrillation. Additionally, the AAPC indicated a decreasing trend in both cumulative and subpopulation AAMR changes associated with ICH without atrial fibrillation. In contrast, primarily opposite trends were observed for ICH with underlying atrial fibrillation (Table 3). The yearly changes in AAMR were found to be statistically different across the two groups, cumulatively and within subpopulations.

Table 3 ICH with and without underlying atrial fibrillation. Comparison table depicting the AAMR and AAPC in decedents related to ICH with underlying atrial fibrillation and without atrial fibrillation. Test of parallelism comparing yearly AAMR changes from 1999 to 2020 in ICH with and without atrial fibrillation
Full size table

4 Discussion

In this study, we aimed to investigate mortality trends and disparities related to ICH in patients with atrial fibrillation. Findings from our study indicate that annual ICH related AAMR in patients with atrial fibrillation increased steadily over the study period and a higher mortality rate was observed in males than in females. The highest mortality was found in Asian/Pacific Islander populations, followed by White, Black, and American Indian/Alaska Native populations. Hispanic individuals experienced lower mortality than their non-Hispanic counterparts. Geographically, Western US regions exhibited a highest mortality rate, followed by Midwestern, Northeastern, and Southern US regions in descending order. There was a more rapid increase in ICH related mortality rates in patients with atrial fibrillation within non-metropolitan regions compared to metropolitan regions even though there were no significant differences in the overall AAMR between these regions. Our results provide substantial insight regarding the current epidemiological understanding of mortality trends and disparities related to ICH in patients with atrial fibrillation.

Our analysis showed that the ICH mortality in patients with atrial fibrillation has steadily increased between 1999 and 2020; whereas ICH mortality in the absence of atrial fibrillation decreased over this same period. This could be explained by the increase in both the prevalence of atrial fibrillation and proportion of atrial fibrillation treated with anticoagulation [2, 18]. Earlier studies reported the overall mortality associated with ICH in patients with atrial fibrillation on coagulation to be between 49 to 57% [19, 20]. A study from Dijon Stroke Registry found that patients with atrial fibrillation had a higher ICH mortality at discharge than patients without atrial fibrillation (47.4% versus 30%) and adjusted premorbid anticoagulation use remained highly associated with mortality (Odds Ratio: 2.53) [18]. Although anticoagulation may not directly initiate a life-threatening bleed, it has been shown to prolong the hematoma formation after a bleeding focus develops [21]. Flibotte et al. supports this theory with their finding that warfarin did not determine the initial hematoma volume but was the only predictor of hematoma expansion (Odds Ratio: 6.2), which is associated with higher mortality [22].

Our data showed that there was a similar AAPC in mortality in both males and females during the study period, but males had a higher overall AAMR compared to females. This is consistent with previous reports identifying no sex differences in patients with ICH [23, 24]. Consistently, a previous analysis revealed that the age-adjusted mortality for ICH for adults above 25 years were lower for women compared to men with a risk ratio of 0.82 [23]. Furthermore, male sex was independently associated with a higher 90-day and 1-year mortality related to ICH [25]. This could be explained by female patients having atrial fibrillation were less likely to receive anticoagulant therapy than males even when their estimated stroke risk was higher [26]. In addition, higher rates of hematoma expansion were seen in males as compared to females even after adjustment for known predictors of hematoma growth [25].

Our findings related to racial and ethnic differences in ICH mortality in patients with atrial fibrillation revealed significant findings. Our study showed that Asian/Pacific Islander populations had the highest ICH mortality in patients with atrial fibrillation, followed by White, Black, and American Indian/Alaska Native populations. Shen et al. reported that warfarin predisposed patients with atrial fibrillation to higher rates of ICH across all races, among which Asians had the highest risk with four times the risk as Whites [27]. Furthermore, the highest age-specific rate of ICH deaths among Asian populations was 1.91 times greater at 45–64 years compared with the death rate in the corresponding age group of White populations [28]. These reports consistently indicate a higher ICH mortality rate in Asian patients with atrial fibrillation. In contrast to our findings, earlier studies reported a higher risk and mortality rate of ICH among Black populations compared with White populations [27,28,29]. However, patient ascertainment in these previous studies were conducted at least two decades ago when there was a low adoption rate for Direct Oral Anticoagulants (DOACs) among minority groups while warfarin was commonly used in these populations [30]. Compared to Vitamin K antagonists, DOACs are associated with lower risk of bleeding and all-cause mortality in patients with atrial fibrillation [31]. Another key finding from our study is a lower mortality rate among Hispanic individuals compared to non-Hispanic individuals, which is consistent with previous reports [28].

Despite the similarity in cumulative AAMR among non-metropolitan and metropolitan regions, our analyses revealed a higher AAPC within non-metropolitan regions. This disparity is likely to be multifactorial as the availability of neurocritical care services is disproportionately seen at higher rates in metropolitan regions [32, 33]. Close neurologic and blood pressure monitoring through specialized tools and procedures with low nurse-patient ratios are essential components of acute ICH management which are likely to contribute to the existing ICH mortality disparities [34]. US census region analyses also revealed the highest mortality rate in Western regions and the fastest increase in annual mortality rates in Southern regions. The rapid growth in the annual mortality in these Southern regions could be explained by the rising prevalence of cigarette smoking, obesity, and hypertension which are known risk factors for atrial fibrillation and prognostic factors for early neurologic deterioration and mortality in the acute phase of ICH [35,36,37,38,39].

Our study has significant implications. ICH accounts for approximately half of all stroke-related deaths and considering the absence of significant changes in the treatment of ICH in the previous few decades, mortality rates have not improved [40, 41]. Given the aging population in the US, the burden of atrial fibrillation is likely to increase and the risk of anticoagulant-related ICH is also expected to rise [41,42,43]. A previous study identified that patients with ICH and atrial fibrillation did worse in regard to vital and functional outcomes compared to patients with ICH without atrial fibrillation [18]. A thorough understanding of the implications of diagnosis and therapy for atrial fibrillation remains vital in identifying disparities in care. Therefore, our results provide substantial epidemiological insight in ICH mortality trends and outcomes related to US populations with atrial fibrillation. Our population-level analyses serve as a hypothesis-generating platform for further research in this area. Future analyses should incorporate and account for multiple aspects of socioeconomic determinants of health to identify areas of inequitable healthcare related to ICH in the setting of atrial fibrillation.

This study has limitations that should be considered while interpreting these findings. The analysis was undifferentiated between traumatic versus non-traumatic ICH. Furthermore, given the limitations of our utilized data repositories, we are unable to differentiate the cohorts of patients that were on anticoagulation versus those that were not. Nonetheless, the diagnosis of atrial fibrillation in the absence of anticoagulant therapy remains associated with a higher risk of ICH mortality [18]. Another limitation is the use of ICD-10 codes, which may lead to misclassification errors [44]. Lastly, our study is a cross-sectional design that cannot determine causality.

5 Conclusion

In this nationwide analysis of ICH mortality in atrial fibrillation, we observed an increasing mortality trend between 1999 and 2020, with evident disparities based on demographic factors including sex, race and ethnicity, and geographic location. Southern regions had the highest annual increase in ICH mortality rates compared to the other regions. These findings are hypothesis provoking and highlight the need to further investigate the potential causes and social determinants contributing to these disparities.

figure a