Seasonal variations in admissions for atrial fibrillation or atrial flutter in the Northeast and the Midwest regions of the United States

Abstract

Background

Previous studies conflict on whether seasonal variability exists in atrial fibrillation (AF) admissions, and contemporary studies are lacking.

Methods

We identified admissions for AF or atrial flutter in the Midwest and Northeast regions of the US from the National Inpatient Database for 2016 to 2020, grouped them into the four seasons (spring, summer, fall, winter), and compared the number of admissions. Subgroup analyses were performed stratified to sex, age, race, AF alone, and geographical regions.

Results

A total of 955,320 admissions for AF or atrial flutter occurred. The number of admissions was highest during winter (243,990, 25.5% of the total), followed by fall (239,250, 25.0% of the total), summer (236,910, 24.8% of the total), and spring (235,170, 24.6% of the total). The differences were statistically significant (P < 0.001). An increasing trend in the number of admissions was observed from March to February of the next year (P trend <0.001). Admissions were most common in the winter and least common in the spring in subgroups of both sexes, age ≥65 years, Whites, non-Whites, AF alone, Northeast region, and Midwest region.

Conclusion

Contemporary analysis of a national database demonstrates seasonal variability in the number of admissions for AF, with a slight increase observed during the winter.

Atrial fibrillation (AF) is the most prevalent clinically important arrhythmia, with an estimated global prevalence of 33.5 million patients.¹ In the US, its prevalence is 1% to 2%, afflicting 3 to 6 million people, and it is estimated to reach up to 16 million people by 2050.^2,3 The incidence of AF has also been steadily increasing for over two decades, resulting in a higher number of hospitalizations and a higher healthcare burden in the US.^4–6 Amidst the growing epidemic of AF, some studies have considered whether the incidence of hospitalization for AF demonstrates seasonal variation.^7–9 A previous study that relied on studies predominantly published before 2010 concluded that paroxysmal AF varied by season, with winter peaks and summer troughs.⁸ However, a more recent study utilizing registry data from 2014 to 2018 concluded that no seasonal variation was observed in AF admissions.⁹ Given the conflicting results and the absence of a contemporary study on a national level in an era of climate change,¹⁰ we sought to reexamine seasonal variation in hospital admissions for AF or atrial flutter in the Northeast and Midwest regions of the US.

METHODS

We retrospectively reviewed the National Inpatient Sample (NIS), the largest inpatient healthcare database maintained by the Healthcare Cost and Utilization Project and the Agency for Healthcare Research and Quality.¹¹ The database samples about a fifth of all hospitals in 49 participating states in the US, which is used to estimate about 35 million admissions annually. Each admission contains information on age, sex, race, primary and secondary diagnoses, procedures, costs, and outcomes, which can be used to study in-hospital outcomes and trends over time. State, hospital, and patient identifiers are removed from the database to ensure patient confidentiality. Because the database is strictly deidentified, our study was exempt from the purview of our institutional review board. The NIS databases are openly available on the public website of the Healthcare Cost and Utilization Project.¹¹

Study population and covariates

From 2016 to 2020, in the NIS, we identified all hospitalizations for the primary diagnosis of AF or atrial flutter. We decided to include atrial flutter in our main analysis, although we excluded it in one of our subgroup analyses because of its close interrelationship and frequent coexistence with AF.¹² We then excluded the South and West regions of the US to capture regions that experience distinct seasons, notably colder winters.¹³ We also excluded patients under 18 years of age and entries that had missing data on demographics, hospital characteristics, primary payer, median income, day and month of hospitalization, admission type, and in-hospital outcomes. The remaining cohort was stratified to spring (March to May), summer (June to August), fall (September to November), and winter (December to February), as verified by a previous study, by using the variable AMONTH provided in the database.¹⁴

From the selected cohort, we extracted data on demographics (sex, age, race), hospital characteristics (region, bed size, urban location), primary payer, median income, day of hospitalization (weekday, weekend), and admission type (elective, nonelective). To characterize our cohort, we identified multiple comorbidities (smoking, hypertension, diabetes mellitus, hyperlipidemia, obesity, heart failure, chronic ischemic disease, valvular heart disease, peripheral artery disease, previous percutaneous coronary intervention, previous coronary artery bypass graft, previous stroke, previous pacemaker, chronic obstructive pulmonary disease, pulmonary hypertension, chronic kidney disease, end-stage renal disease, liver cirrhosis, history of malignancy, deficiency anemia, malnutrition, and dementia) from all the admissions. Admissions were also classified according to clinical presentation, with AF or atrial flutter. All the primary diagnoses and comorbidities used in our cohort were determined using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM). The individual codes can be found in Supplementary Table 1.

Study outcomes

Our primary outcome of interest was the number of admissions for AF or atrial flutter in each of the four seasons. Secondary outcomes included in-hospital mortality, length of hospital stay, and total hospital cost, which are present in the original database. The total hospital cost can be derived by multiplying the total hospital charge with the cost-to-charge ratios in the separate cost-to-charge ratio files.¹⁵

Statistical analysis

We applied survey analysis based on weights of hospital-level discharge by multiplying given weights (DISCWT) to each data entry to generate national estimates. We employed the chi-square test for categorical covariates and analysis of variance for continuous covariates to compare baseline characteristics among the four seasons. We used descriptive statistics whereby the relative number of admissions in each season was compared using its proportion of the total admissions, as done by previous verified studies.^9,14 We additionally assessed the trend in the number of admissions from spring to winter (March to February of the next year) using simple linear regression and also using linear regression adjusted for hospital characteristics. For secondary outcomes, multivariable logistic and linear regression models were utilized to produce adjusted odds ratios (aOR) and adjusted mean differences with respective 95% confidence intervals (CI). As done by previous studies, comparisons were made with spring as the reference.¹⁴ The models were adjusted using demographic information, comorbidities, hospital characteristics, and median income after a consensus from team discussions concluded that they can clinically impact in-hospital outcomes.

Multicollinearity among the covariates was ruled out using variance inflation factor, tolerance level, and eigensystem analysis of covariance. Multiple subgroup analyses were conducted to compare seasonal variations in AF or atrial flutter admissions according to sex (males, females), age (older adults defined as age ≥65 years), race (Whites, non-Whites), clinical presentation (AF, atrial flutter), and geographical region (Northeast, Midwest). All tests were two-sided, and P values <0.017 were considered significant to account for Bonferroni correction associated with three comparisons (spring versus summer, fall, and winter, respectively). All statistical analyses were conducted using SAS, version 9.4 (SAS Institute, Cary, NC).

RESULTS

A total of 955,320 admissions for AF or atrial flutter occurred in the Midwest and Northeast regions of the United States from 2016 to 2020 (Figure 1). The majority of the admissions (87.4%) were due to AF. The number of admissions was lowest during spring (235,170, 24.6% of total), followed by summer (236,910, 24.8% of total), fall (239,250, 25.0% of total), and winter (243,990, 25.5% of total) (Figure 2). These were statistically significant (P < 0.001). Summer, fall, and winter had 1740 (0.7%), 4080 (1.7%), and 8820 (3.8%) more admissions for AF or atrial flutter than spring. When stratified by month, a positive trend was seen from March to February (unadjusted P trend <0.001, adjusted P trend <0.001) (Figure 2). Small statistical differences were observed in sex, age, and a few comorbidities, but most characteristics showed no difference among the four seasons (Table 1). Similarly, no statistical difference was seen in hospital region, bed size, urban location, primary payer, median income, and day of hospitalization.

Figure 1.

Flow diagram illustrating the process for selecting the cohort of admissions for atrial fibrillation or atrial flutter in the Midwest and Northeast.

Figure 2.

Admissions for atrial fibrillation or atrial flutter by season and month. (a) The number of admissions for atrial fibrillation or atrial flutter in each of the four seasons. Please note that the ordinate begins with 230,000 and not 0. (b) The trend of admissions over 12 months of the year, beginning from March (beginning of spring) and ending in February (end of winter). Spring, summer, fall, and winter are color-coded in green, yellow, red, and blue, respectively. The dotted line represents the line of best fit, which shows a positive trend.

Table 1.

Baseline characteristics of hospitalizations for atrial fibrillation or atrial flutter

Variable	Spring	Summer	Fall	Winter	P value
Number of hospitalizations	235,170	236,910	239,250	243,990	<0.001
Clinical presentation (%)
Atrial fibrillation	87.5	87.0	87.5	87.8	0.005
Atrial flutter	12.5	13.0	12.5	12.2	0.005
Male sex (%)	50.4	51.6	51.0	50.6	0.002
Age, mean (SD)	70.7 (13.0)	70.6 (12.8)	70.8 (12.8)	71.0 (13.0)	<0.001
Race (%)					0.17
White	86.6	86.0	86.3	86.2
Black	7.0	7.3	7.0	7.1
Hispanic	3.3	3.4	3.4	3.3
Asian	0.9	1.0	1.0	1.0
American Indian/Alaska Native	0.3	0.2	0.2	0.2
Other	2.0	2.2	2.1	2.1
Comorbidities (%)
Smoking	41.6	41.8	41.8	41.1	0.09
Hypertension	41.8	41.2	39.1	40.8	<0.001
Diabetes mellitus	28.4	28.6	28.4	28.7	0.62
Hyperlipidemia	53.6	54.2	55.0	54.0	<0.001
Obesity	23.5	23.8	23.8	23.4	0.43
Heart failure	40.6	40.5	40.7	40.6	0.93
Chronic ischemic heart disease	4.6	4.7	4.6	4.2	0.002
Valvular heart disease	11.8	11.7	11.8	12.1	0.30
Peripheral artery disease	4.8	4.8	4.8	4.9	0.90
Previous PCI	1.1	1.0	1.1	1.1	0.82
Previous CABG	7.7	7.6	7.6	7.6	0.94
Previous stroke	11.1	11.4	11.5	11.4	0.32
Previous pacemaker	5.3	5.5	5.7	5.3	0.02
COPD	20.0	19.2	19.4	20.0	0.001
Pulmonary hypertension	8.1	8.1	7.9	8.3	0.12
Chronic kidney disease	19.1	19.4	19.2	19.7	0.11
End-stage renal disease	2.3	2.4	2.3	2.5	0.09
Liver cirrhosis	1.5	1.5	1.5	1.4	0.43
History of malignancy	13.4	13.5	13.7	14.1	0.006
Deficiency anemia	4.1	4.2	4.2	4.0	0.35
Malnutrition	2.6	2.7	2.9	2.8	0.15
Dementia	6.0	5.8	5.9	6.2	0.02
Hospital characteristics (%)
Hospital region					0.20
Northwest	45.5	46.0	45.8	46.2
Midwest	54.5	54.0	54.2	53.8
South	0	0	0	0
West	0	0	0	0
Hospital bed size					0.067
Small	24.1	23.5	23.6	24.0
Medium	27.9	28.0	28.1	28.3
Large	48.1	48.6	48.3	47.7
Urban location					<0.001
Rural	11.9	11.0	11.0	10.9
Urban nonteaching	17.9	17.5	17.6	17.8
Urban teaching	70.2	71.5	71.4	71.4
Primary payer (%)					0.19
Medicare	68.7	69.2	69.4	69.1
Medicaid	6.6	6.5	6.3	6.3
Private insurance	21.6	21.2	21.0	21.6
Self-pay	1.4	1.4	1.6	1.5
No charge	0.1	0.1	0.1	0.1
Others	1.6	1.6	1.6	1.5
Median income (%)					0.11
Quartile 1	20.6	20.6	20.0	20.5
Quartile 2	27.3	26.8	26.9	27.0
Quartile 3	27.1	27.2	27.4	27.4
Quartile 4	25.0	25.4	25.7	25.2
Day of hospitalization (%)					0.14
Weekday	81.0	81.3	80.9	80.7
Weekend	19.0	18.7	19.1	19.3
Admission type (%)					<0.001
Elective	13.3	15.0	15.6	13.8
Nonelective	86.7	85.1	84.4	86.2

CABG indicates coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; PCI, percutaneous coronary intervention; SD, standard deviation.

In-hospital mortality during spring, summer, fall, and winter were 0.8%, 0.7%, 0.7%, and 0.8%, respectively. With spring as the reference, no difference in in-hospital mortality was observed among admissions occurring in summer (aOR 0.89, 95% CI 0.77–1.03, P = 0.39), fall (aOR 0.87, 95% CI 0.75–1.01, P = 0.19), and winter (aOR 0.95, 95% CI 0.83–1.10, P = 0.56) (Table 2). The average lengths of hospital stay during spring, summer, fall, and winter were 3.4, 3.3, 3.3, and 3.4 days, respectively. The adjusted mean difference was small but statistically significant. Average total hospital costs during spring, summer, fall, and winter were $10,219, $10,564, $10,822, and $10,663, respectively. Compared with admissions in spring, those in summer, fall, and winter were associated with significantly higher total hospital costs (P < 0.001).

Table 2.

Comparison of outcomes among difference seasons

Outcome	Spring	Summera	Falla	Wintera
In-hospital mortality (%)	Reference	0.89 (0.77–1.03)	0.87 (0.75–1.01)	0.95 (0.83–1.10)
		P = 0.39	P = 0.19	P = 0.56
Length of stay (days ± SD)	Reference	0.07 (0.03–0.11)b	0.06 (0.02-0.10)b	0.07 (0.03–0.11)b
		P = 0.001	P = 0.005	P < 0.001
Total hospital cost ($ ± SD)	Reference	307 (153–462)b	535 (381–690)b	394 (244–544)b
		P < 0.001	P < 0.001	P < 0.001

^aAdjusted for demographics, comorbidities, hospital characteristics, and median income.

^bAdjusted mean difference with 95% confidence interval.

On subgroup analyses, more admissions for AF or atrial flutter occurred during winter (P < 0.001) than in any other season (Figure 3). This was consistent in both men and women, Whites and non-Whites, and older adults (Supplementary Table 2). Of note, 4.7% more admissions occurred in winter compared with spring in older adults (P < 0.001). When admissions were stratified to AF and atrial flutter separately, winter admissions were most frequent in the former (25.6%) but not in the latter (24.9%) (Figure 4). For atrial flutter, admissions were numerically more common during spring (25.6%), but the difference was not statistically significant (P = 0.14). Winter remained the most common season in the Northeast region (25.7%) and the Midwest region (25.4%) when analyzed separately (Supplementary Figure 1).

Figure 3.

Subgroup analyses for different demographic characteristics showing the number of admissions for atrial fibrillation or atrial flutter in each season stratified by (a) sex, (b) age ≥65 years, (c) White race, and (d) non-White race. Please note that the ordinate does not start from 0. Spring, summer, fall, and winter have been color-coded in green, yellow, red, and blue, respectively.

Figure 4.

Subgroup analyses according to rhythm showing the number of admissions for (a) atrial fibrillation and (b) atrial flutter. Please note that the ordinate does not start from 0. Spring, summer, fall, and winter have been color-coded in green, yellow, red, and blue, respectively.

DISCUSSION

Our study suggests a small seasonality phenomenon in contemporary AF admissions in the Midwest and Northeast regions of the US. Among the four seasons, winter had the greatest percentage of AF admissions, whereas spring occupied the lowest percentage, irrespective of sex, older age, and race (dichotomized into Whites and non-Whites). A trend of increasing admissions for AF or atrial flutter was observed from March to February despite recent temperature changes due to global warming.

Seasonality in cardiovascular disease patterns have been described in the literature for nearly a century since its first reported description in 1926.¹⁶ A considerable proportion of the literature describes associations between seasonal variation and major adverse cardiovascular events due to thrombotic events such as myocardial infarction (MI),^17,18 stroke,^18–20 and pulmonary embolism.²¹ Extreme temperature changes have been reported to increase cardiovascular mortality and morbidity, especially after sudden changes in weather such as “heat waves” and “cold snaps,” mostly in patients 65 years or older.²² Heat waves have shown a more extensive, immediate effect on cardiovascular admissions, morbidity, and mortality (3.4% increase per °C raised). In contrast, cold snaps elicit a lagged peak in admissions estimated within 11 days after an event and less dramatic effects in mortality (1.66% increase per °C raised) and morbidity.^22,23 However, a study that evaluated the association between cold weather and MI showed a cumulative 5% increase in mortality for each day of extreme cold exposure prior to MI, defined as a decrease of 13.9°C and 7.6°C during winter and nonwinter seasons, respectively.¹⁷ As a result, colder temperatures have been implicated to contribute significantly more to season-related mortality when compared with heat, according to large worldwide reports such as the World Health Organization Multinational Monitoring of Trends and Determinants in Cardiovascular Disease and multiple meta-analyses.^7,18,22,23

Previous studies conducted mostly in the European continent have shown conflicting evidence on a winter seasonality phenomenon.^{9,19,20,24–30} In addition, there is limited US-based evidence regarding AF and seasonality. The most extensive analysis in the US was conducted through the GWTG-AFib registry, a nationwide, optional initiative from the American Heart Association/American College of Cardiology for AF and atrial flutter monitoring with both epidemiologic and quality metric monitoring purposes.³¹ The study showed a higher percentage of admissions in fall months but no statistically significant cyclic variation across seasons on peak-to-trough ratio analysis, conflicting with the findings in our study.⁹ Methodologically, three notable differences could account for this discrepancy: total surveyed population, geographic population studied with associated climate data, and possible sampling bias within the registry. Our sample size was >15 times larger and covered areas with more prominent temperature variations within the US, as reported by the Environmental Protection Agency³² and Federal Emergency Management Agency National Risk Index for cold waves.³³ In addition, the GWTG-AFib registry may have sampled different populations since it was devised as an optional quality monitoring program from the American Heart Association, voluntarily recruiting centers interested in standard quality metrics for AF and regularly monitoring for goals and milestones.^31,34,35 As a result, the participating institutions may have had unique clinical profiles with variable access to resources and, thus, may include different clusters of patients. Our findings were more consistent with other previous studies, which suggest that colder temperatures, especially presenting as temperature variations, are associated with more AF admissions.^7,23

Hypothesized mechanisms exploring the seasonality phenomena are uncertain and diverse. Most mechanisms proposed have been described in thrombotic rather than arrhythmic events, with population-based studies correlating hypercoagulability and colder climates.^7,36 Previous studies have demonstrated a strong bidirectional relationship between AF and thrombotic events such as MI and stroke.^19,37 The Reasons for Geographic and Racial Differences in Stroke (REGARDS) study revealed a higher incidence of AF in the presence of strokes and MI^38,39 and vice versa, a higher incidence of coronary heart disease and cardiovascular events in patients with AF.⁴⁰ Risk factors for the pathogenesis of AF also intertwine with those for other cardiovascular diseases, such as coronary heart disease. These risk factors include aging, hypertension, diabetes, obesity, smoking, obstructive sleep apnea, left ventricular hypertrophy, heart failure, and MI, regardless of hemodynamically significant ischemic cardiomyopathy.⁴¹ Moreover, the seasonality in incidence, admissions, morbidity, and mortality found in this and previous studies matches that of thrombotic events. Altogether, these similarities raise suspicion about shared pathophysiological mechanisms as an unexplored cause of these matching features.

Other potential explicatory mechanisms relate to the increase in respiratory tract infections, including influenza, during winter and physiological responses to cold exposure. A national study demonstrated that influenza infection increased the risk of AF by 18%, with a systemic inflammatory response and increase in sympathetic tone implicated in the pathogenesis.⁴² Meanwhile, cold temperatures induce sympathetically mediated peripheral vasoconstriction and shivering to counteract core temperature decline through decreased convective heat loss and contraction thermogenesis.^7,43 These mechanisms can result in acutely elevated blood pressures leading to increased afterload, intra-atrial pressures, and atrial arrhythmia, which can be even worse in abrupt temperature drops.^44,45 Acclimation to cold exposure can protect against these temperature drops but is blunted in older patients, who, with other cardiovascular risk factors, have a higher degree of susceptibility to AF.⁴⁶ In our analysis, older adults had a greater relative difference in the number of AF hospitalizations in spring versus winter.

Finally, behavioral changes during the winter season may also contribute to seasonal fluctuations in AF admissions. The Seasonal Variation of Blood Cholesterol Levels study (SEASONS) and a recent systematic literature review support a seasonal variation with a significant decrease in physical activity and increased sedentary behaviors in the overall population regardless of the country, culture, or preexistent comorbidities during colder months.⁴⁷ Studies have also reported significant weight gains during winter, especially after the holidays.^48,49 Depression and seasonal affective disorders are more prevalent during the shorter days of winter, which may impact medication compliance and even increase the incidence of AF.^50,51 Sedentary lifestyle, weight gain, and depression are closely associated with AF, and we hypothesize that they may be contributing to the increased admissions during the winter season.^52–55

Our study was conducted using the NIS database, which is an administrative database that lacks granularity, such as information on medications, type of AF or atrial flutter, and electrocardiographic data. The retrospective nature of our study cannot be used to confirm causal relationships but should rather be used to make inferences and hypotheses. The NIS database only displays the months and not the days, so specific dates for the beginning of seasons could not be set. However, previous studies on seasonal variation determined using months have been validated.¹⁴ We did not have access to the climate data of each of the inpatient entries.

In conclusion, our findings demonstrate minimal seasonal variability in the number of admissions for AF or atrial flutter in the Midwest and Northeast regions of the US. Admissions were most common during winter and least common during spring, regardless of sex, race, and older age. No differences in in-hospital mortality and minimal differences in length of hospital stay and total hospital cost were observed across the seasons, with spring as the reference.

DISCLOSURE STATEMENT

The authors report no funding or conflicts of interest.