Trends in Atrial Fibrillation Incidence Rates Within an Integrated Health Care Delivery System, 2006 to 2018.

Importance: Atrial fibrillation (AF) is the most common cardiac arrhythmia, and multiple studies have reported increasing AF incidence rates over time, although the underlying explanations remain unclear.
Objectives: To estimate AF incidence rates from 2006 to 2018 in a community-based setting and to investigate possible explanations for increasing AF by evaluating the changing features of incident AF cases and the pool of patients at risk for AF over time. Design, Setting, and Participants: This cohort study included 500 684 patients who received primary care and other health care services for more than 2 years through a single integrated health care delivery network in Pennsylvania. Data collection was conducted from January 2003 to December 2018. The base study population had no documentation of AF in the electronic medical record for at least 2 years prior to baseline. Data analysis was conducted from May to December 2019. Main Outcomes and Measures: Incident AF cases were identified through diagnostic codes recorded at inpatient or outpatient encounters. Age- and sex-adjusted AF incidence rates were estimated by calendar year from 2006 to 2018 both overall and across subgroups, including according to diagnostic setting (inpatient vs outpatient) and priority (primary vs secondary diagnosis).
Results: Among 514 293 patients meeting criteria for the base study population, the mean (SD) age at baseline was 47 (18) years and 282 103 (54.9%) were women; 13 609 (2.6%) met AF diagnostic criteria on or prior to the baseline date and were excluded. Among 500 684 patients free of AF at baseline, standardized AF incidence rates from 2006 to 2018 increased from 4.74 (95% CI, 4.58-4.90) to 6.82 (95% CI, 6.65-7.00) cases per 1000 person-years, increasing significantly over time (P < .001). Incidence rates increased in all age and sex subgroups, although absolute rate increases were largest among those aged 85 years or older. The fraction of incident AF cases among individuals aged 85 years or older increased from 135 of 1075 (12.6%) in 2006 to 451 of 2427 (18.6%) in 2017. Patients with incident AF were more likely over time to have high body mass index (1351 of 3389 patients [39.9%] in 2006-2008 vs 4504 of 9214 [48.9%] in 2015-2018; P < .001), hypertension (2764 [81.6%] in 2006-2008 vs 7937 [86.1%] in 2015-2018; P < .001), and ischemic stroke (328 [9.7%] in 2006-2008 vs 1455 [15.8%] in 2015-2018; P < .001), but less likely to have coronary artery disease (1533 [45.2%] in 2006-2008 vs 3810 [41.4%] in 2015-2018; P < .001). Among 22 077 new cases of AF, 9146 (41.4%) were diagnosed as inpatients and 5731 (26.0%) as the primary diagnosis. Incidence rates of AF increased significantly in all diagnostic setting and priority pairings (eg, inpatient, primary: rate ratio, 1.07; 95% CI, 1.06-1.08; P < .001). Among patients at risk for AF, high BMI and hypertension increased over time (BMI: 71 433 of 198 245 [36.0%] in 2007 to 130 218 of 282 270 [46.1%] in 2017; hypertension: 79 977 [40.3%] in 2007 to 134 404 [47.6%] in 2017). Documentation of short-term ECG increased over time (23 297 of 207 349 [11.2%] in 2008 to 45 027 [16.0%] in 2017); however, long-term ECG monitoring showed no change (1871 [0.9%] in 2007 to 4036 [1.4%] in 2017). Conclusions and Relevance: In this community-based study, AF incidence rates increased significantly during the study period. Concurrent increases were observed in AF risk factors in the at-risk population and short-term ECG use.

Introduction

The community and clinical burdens of atrial fibrillation (AF) have been growing in recent years, with projections estimating sustained increases in the immediate future.[1,2,3,4,5,6] Multiple explanations have been posited for the AF increase, including an aging population, prolonged survival of individuals with cardiac conditions that predispose them to AF, and an increasing prevalence of AF risk factors, such as obesity and hypertension, in the at-risk populations.[3,5,7,8,9,10,11,12,13,14,15] Moreover, some authors have perceived increased attention being directed to AF in recent years—commonly attributed to the introduction of new anticoagulants for stroke prevention—that may be increasing the propensity to document AF clinically, especially as a secondary diagnosis.[14,16,17] Furthermore, prior studies have shown disproportionate increases in AF incidence by diagnostic setting (inpatient vs outpatient), although studies were discordant on the nature of these differences.[18,19] Lastly, increasing AF may be partially attributed to more frequent utilization of electrocardiography (ECG) in both routine clinical environments and through implanted or wearable devices with continuous AF-detecting capabilities, thereby increasing the opportunity to uncover AF.[3,7,9,10,14]

Prior studies have been unable to assess all these possible factors associated with the increasing AF burden. Accordingly, the goal of the current study was to provide a contemporary view of the evolving epidemiology of AF by calculating updated AF incidence rates through 2018 in a community-based population; describing the changing features of incident AF cases over time with respect to patient characteristics, diagnostic setting (inpatient vs outpatient), and diagnostic priority (primary vs secondary diagnosis); and characterizing the profile of patients at risk for AF over time, including the changing frequency of AF risk factors and use of short-term and long-term ECG monitoring. The primary hypothesis was that the prevalence of 1 or more of the aforementioned factors is increasing concurrently with rising AF incidence rates.

Methods

This cohort study extends previous work involving the patient population of the Geisinger Health System (Geisinger), an integrated health care delivery network with inpatient and outpatient facilities distributed over a 20 000 square-mile region in a largely rural part of central and northeast Pennsylvania.[1] The study was approved by the Geisinger institutional review board, which granted a waiver of patient consent because the study used retrospective data generated as part of usual clinical care, posing minimal risk to participants. The study adhered to the applicable sections of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

All study data were drawn from Geisinger’s electronic medical record (EMR) data repository; all health care services provided at Geisinger have been documented in the EMR since January 1, 2001. The base study population included all adult Geisinger patients receiving primary care and other health care services through Geisinger for at least 2 years between January 1, 2001, and December 31, 2018. Among 1 505 921 unique adult patients with at least 1 EMR-documented Geisinger encounter during the study interval, 514 293 (34.2%) met the base study population criteria. Among patients in the base study population, a baseline date was assigned corresponding to the first Geisinger office visit after inclusion criteria were met.

AF diagnostic criteria were operationalized according to the appropriate International Classification of Diseases, Ninth Revision (ICD-9) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes (ICD-9: 427.31; ICD-10: I48.x, excluding atrial flutter) documented on a patient’s problem list at 1 or more inpatient encounters or 2 or more outpatient encounters. Similar encounter-based diagnostic rules have been applied in prior electronic data studies with acceptable performance characteristics.[1,4,19,20,21,22] Outpatient encounters included any patient contact with Geisinger outside the emergency department or hospital. Study patients meeting AF diagnostic criteria on or before the baseline date were labeled as preexisting AF on the baseline date, while those meeting diagnostic criteria after baseline were considered newly diagnosed (ie, incident) AF.

Statistical Analysis

Among individuals without AF at baseline, person-years at risk for AF was defined by the time interval between the baseline date and the date of an AF diagnosis or the last EMR-documented encounter, up to December 31, 2018. AF incidence rates per 1000 person-years were calculated by calendar year from 2006 to 2018, with numerators defined by the number of incident AF cases diagnosed in the respective calendar years and denominators by the total person-years at risk within the corresponding calendar year. Incidence rates across age and sex strata were calculated using the following age groups: younger than 45, 45 to 54, 55 to 64, 65 to 74, 75 to 84, and 85 years and older. Standardized incidence rates adjusted for age and sex only were estimated, incorporating the age and sex distribution from the 2010 United States census.[23] Linear trends in incidence rates over time were evaluated by Poisson regression, and rate ratios (RRs) with 95% CIs were estimated. RRs are interpreted as the annual relative change in AF incidence rate over time. The offset in Poisson models was the natural log of person-years at risk by calendar year.

Characteristics of incident AF cases were reported by year of diagnosis to evaluate changes over time. All patient characteristics as of the diagnosis date, including demographic characteristics, vital signs, and medical history (diagnoses, procedures, and implanted devices), were assembled from EMR documentation on or before the diagnosis date. Medical history was determined through the appropriate ICD-9 or ICD-10 and Current Procedural Terminology codes. Missing data for vital signs were as follows: systolic blood pressure, less than 0.1%; diastolic blood pressure, less than 0.1%; heart rate, 0.2%; body mass index (BMI; calculated as weight in kilograms divided by height in meters squared), 3.6%. An AF diagnosis was considered inpatient when diagnostic criteria were first met in the hospital or emergency department setting, otherwise the diagnosis was considered outpatient. Furthermore, AF diagnoses were considered primary when an inpatient or outpatient encounter documented AF as the primary diagnosis for the encounter; all other diagnoses were considered secondary. Inpatient diagnoses refer to those made at discharge. AF diagnoses first identified from problem lists were considered outpatient, secondary diagnoses. Trends over time in continuous baseline characteristics were evaluated using general linear regression models and categorical characteristics using logistic regression.

To characterize the changing characteristics of the patient population at risk for AF, patients were included in the denominator of every calendar year during which they were active within Geisinger for the entire calendar year with no AF diagnosis or active within Geisinger and AF free on January 1 of a particular year but with a subsequent new AF diagnosis during that calendar year. Patients were considered active within Geisinger between their first and last EMR-documented Geisinger encounters. Patient characteristics were assembled per calendar year using EMR documentation from the current year or before (but no later than 30 days after the diagnosis date for new AF cases). In addition, documentation of short-term and long-term ECG monitoring (observed within the respective calendar year only) was noted. Short-term ECG refers to 12-lead ECG or rhythm strip done in outpatient settings, while long-term ECG refers to Holter monitors, implanted loop recorders, patient-activated event recorders, and other devices that can detect AF outside formal health care settings. Use of implanted therapeutic devices with AF-detecting capabilities, such as pacemakers (single-ventricular or biventricular) and implantable cardioverter defibrillators, were considered separately. Data analysis was conducted with SAS statistical software version 9.4 (SAS Institute). For all analyses, statistical significance was determined by a 2-tailed P < .05.

Results

Changes Over Time in Atrial Fibrillation Incidence Rates

Among 514 293 patients meeting criteria for the base study population, the mean (SD) age at baseline was 47 (18) years and 282 103 (54.9%) were women; 13 609 (2.6%) met AF diagnostic criteria on or prior to the baseline date and were excluded. Among 500 684 patients free of AF at baseline, the total cumulative person-years at risk from 2006 to 2018 was 3 342 034, a mean (SD) of 6.7 (4.8) years per person. New AF diagnoses were documented in 22 077 patients (4.4%), and the overall age-adjusted and sex-adjusted AF incidence rate was 5.53 (95% CI, 5.45-5.62) per 1000 person-years. Standardized AF incidence rates increased over time, demonstrating a relatively steady increase from 2006 to 2018 (RR, 1.03, 95% CI, 1.02-1.03) (Figure 1). Standardized incidence rates by year ranged from a minimum of 4.74 (95% CI, 4.58-4.90) per 1000 person-years in 2008 to 6.82 (95% CI, 6.65-7.00) per 1000 person-years in 2018, a 44% relative difference. In all age and sex subgroups, RR increases were statistically significant (eFigure and eTable 1 in the Supplement). However, absolute changes in incidence rates over time were largest among the oldest age group. The fraction of incident AF cases among individuals aged 85 years or older increased from 135 of 1075 (12.6%) in 2006 to 451 of 2427 (18.6%) in 2017; the absolute differences between the lowest and highest per-year incidence rates in men and women aged 85 years or older were 23.56 per 1000 person-years (60.43 per 1000 person-years in 2017 vs 36.87 per 1000 person-years in 2006) and 19.99 per 1000 person-years (49.74 per 1000 person-years in 2018 vs 29.75 per 1000 person-years in 2006), respectively, while in all other age groups, the analogous differences were less than 10.

Figure 1.

Standardized Atrial Fibrillation Incidence Rates From 2006 to 2018, Overall and Stratified by Diagnostic Setting

Changes Over Time in Characteristics of Incident Atrial Fibrillation Cases

The mean age of incident AF cases did not change significantly over time, yet shifts across age categories were observed such that the proportion of all AF cases among individuals aged 75 to 84 years old at diagnosis decreased, while the proportions of all AF cases among individuals aged 55 to 64, 65 to 74, and 85 years or older increased (Table 1; eTable 2 in the Supplement). The proportion of incident AF cases with hypertension, dyslipidemia, diabetes, ischemic stroke, and a BMI of at least 30 were 18 696 (84.7%), 17 795 (80.6%), 8095 (36.5%), 3125 (14.2%), and 10 196 (46.2)%, respectively, and the prevalence of each risk factor increased significantly over time (eg, hypertension: 2764 of 3389 patients [81.6%] in 2006-2008 vs 7937 of 9214 patients [86.1%] in 2015-2018; P < .001; BMI ≥30: 1351 [39.9%] in 2006-2008 vs 4504 [48.9%] in 2015-2018; P < .001; diabetes: 1143 [33.7%] in 2006-2008 vs 3521 [38.2%] in 2015-2018; P < .001; chronic lung disease 1107 [32.7%] vs 3400 [36.9%]; P < .001; ischemic stroke: 328 [9.7%] in 2006-2008 vs 1455 [15.8%] in 2015-2018; P < .001). In contrast, the proportion of AF cases with prior myocardial infarction remained stable over time (626 [18.5%] in 2006-2008 vs 1646 [17.9%] in 2015-2017; P = .75), while there was a decrease in the proportion with coronary artery disease (1533 [45.2%] in 2006-2008 vs 3810 [41.4%] in 2015-2018; P < .001), and prior coronary artery bypass surgery (520 [15.3%] in 2006-2008 vs 976 [10.6%] in 2015-2017; P < .001). There was no change in the proportion of new AF cases with an implanted pacemaker over time, although the proportion with an implantable cardioverter defibrillator increased.

Table 1.

Characteristics of Incident Atrial Fibrillation Cases by Year of Diagnosis

Characteristic	Patients, No. (%)					P value
	All (N = 22 077)	2006-2008 (n = 3389)	2009-2011 (n = 4224)	2012-2014 (n = 5250)	2015-2018 (n = 9214)
Inpatient	9146 (41.4)	1444 (42.6)	1536 (36.4)	2045 (39.0)	4121 (44.7)	<.001
Primary	2437 (11.0)	297 (8.8)	404 (9.6)	570 (10.9)	1166 (12.7)	<.001
Secondary	6709 (30.4)	1147 (33.8)	1132 (26.8)	1475 (28.1)	2955 (32.1)	.48
Outpatient	12 931 (58.6)	1945 (57.4)	2688 (63.6)	3205 (61.0)	5093 (55.3)	<.001
Primary	3294 (14.9)	523 (15.4)	596 (14.1)	780 (14.9)	1395 (15.1)	.83
Secondary	9637 (43.7)	1422 (42.0)	2092 (49.5)	2425 (46.2)	3698 (40.1)	<.001
Age group, y
<45	653 (3.0)	100 (3.0)	132 (3.1)	161 (3.1)	260 (2.8)	.16
45-54	1415 (6.4)	219 (6.5)	248 (5.9)	348 (6.6)	600 (6.5)	.48
55-64	3596 (16.3)	498 (14.7)	655 (15.5)	882 (16.8)	1561 (16.9)	<.001
65-74	5744 (26.0)	856 (25.3)	1036 (24.5)	1357 (25.8)	2495 (27.1)	.001
75-84	6831 (30.9)	1218 (35.9)	1443 (34.2)	1561 (29.7)	2609 (28.3)	<.001
≥85	3838 (17.4)	498 (14.7)	710 (16.8)	941 (17.9)	1689 (18.3)	<.001
Women	10 677 (48.4)	1583 (46.7)	2056 (48.7)	2556 (48.7)	4482 (48.6)	.22
Smoking status
Current	2417 (10.9)	321 (9.5)	451 (10.7)	570 (10.9)	1075 (11.7)	<.001
Former	9520 (43.1)	1413 (41.7)	1759 (41.6)	2359 (44.9)	3989 (43.3)
Never	10 140 (45.9)	1655 (48.8)	2014 (47.7)	2321 (44.2)	4150 (45.0)
BMI ≥30	10 196 (46.2)	1351 (39.9)	1908 (45.2)	2433 (46.3)	4504 (48.9)	<.001
Hypertension	18 696 (84.7)	2764 (81.6)	3532 (83.6)	4463 (85.0)	7937 (86.1)	<.001
Dyslipidemia	17 795 (80.6)	2450 (72.3)	3322 (78.6)	4294 (81.8)	7729 (83.9)	<.001
Heart failure	6429 (29.1)	1044 (30.8)	1237 (29.3)	1496 (28.5)	2652 (28.8)	.06
Diabetes	8059 (36.5)	1143 (33.7)	1539 (36.4)	1856 (35.4)	3521 (38.2)	<.001
Myocardial infarction	3974 (18.0)	626 (18.5)	731 (17.3)	971 (18.5)	1646 (17.9)	.75
Coronary artery disease	9541 (43.2)	1533 (45.2)	1869 (44.2)	2329 (44.4)	3810 (41.4)	<.001
Coronary artery bypass graft	2603 (11.8)	520 (15.3)	527 (12.5)	580 (11.0)	976 (10.6)	<.001
Percutaneous coronary intervention	2385 (10.8)	343 (10.1)	405 (9.6)	578 (11.0)	1059 (11.5)	<.001
Ischemic stroke	3125 (14.2)	328 (9.7)	548 (13.0)	794 (15.1)	1455 (15.8)	<.001
Venous thromboembolism	2191 (9.9)	258 (7.6)	363 (8.6)	527 (10.0)	1043 (11.3)	<.001
Pulmonary embolism	1280 (5.8)	152 (4.5)	218 (5.2)	280 (5.3)	630 (6.8)	<.001
Implanted pacemaker	1574 (7.1)	279 (8.2)	312 (7.4)	324 (6.2)	659 (7.2)	.17
Implantable cardioverter defibrillator	650 (2.9)	90 (2.7)	111 (2.6)	115 (2.2)	334 (3.6)	<.001

Abbreviation: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared).

Changes Over Time in Diagnostic Setting and Priority of Incident Atrial Fibrillation

Among patients with new AF, 12 931 (58.6%) were first documented as outpatients and 9146 (41.4%) as inpatients (Table 1; eTable 1 in the Supplement). AF was the primary diagnosis in 5731 (26.0%) of incident cases and a secondary diagnosis in 16 346 (74.0%). The ratio of secondary-to-primary diagnoses was similar in both inpatient (6709 [73.4%] vs 2437 [26.6%]) and outpatient (9637 [74.5%] vs 3294 [25.5%]) settings. AF incidence rates increased significantly over time in each diagnostic setting and priority pairing (inpatient, primary: RR, 1.07; 95% CI, 1.06-1.08, P < .001; outpatient, primary: RR, 1.03; 95% CI, 1.02-1.04, P < .001; inpatient, secondary: RR, 1.03; 95% CI, 1.02-1.03; P < .001; outpatient, secondary: RR, 1.02; 95% CI, 1.01-1.02; P < .001) (Figure 1, Figure 2, and Figure 3; eTable 3 in the Supplement). However, over time, the ratio of inpatient-to-outpatient diagnoses increased, such that the fraction of incident AF as a primary, inpatient diagnosis increased, while the fraction as secondary, outpatient diagnoses decreased (Table 1 and Figure 2; eTable 1 in the Supplement). No discernable differences in the ratio of inpatient-to-outpatient diagnoses were observed across age and gender subgroups (eTable 4 in the Supplement), but the ratio of primary-to-secondary diagnoses did show a pattern across subgroups in that AF was more likely a primary diagnosis among younger than older age groups (eTable 5 in the Supplement).

Figure 2.

Standardized Atrial Fibrillation Incidence Rates From 2006 to 2018, Overall and Stratified by Diagnostic Priority

Figure 3.

Standardized Atrial Fibrillation Incidence Rates From 2006 to 2018, Overall and Stratified by Diagnostic Setting and Priority

Changes Over Time in Characteristics of Patients At Risk for Atrial Fibrillation

The number of patients active within Geisinger and at risk for AF at the beginning of each calendar year from 2006 to 2017 ranged from 187 735 to 282 270 out of a total pool of 500 684 unique patients (Table 2; eTable 6 in the Supplement). The mean (SD) age of the at-risk population was in the low 50s (2007: 51 [17] years; 2017: 53 [18] years), approximately 60% were women (2007: 114 251 of 198 245 [57.6%]; 2017: 162 601 of 282 270 [57.6%]), and approximately 20% reported currently smoking (2007: 41 774 [21.1%]; 2017: 56 377 [20.0%]). Many AF risk factors increased in the at-risk population over time, most notably high BMI (71 433 [36.0%] in 2007 to 130 218 [46.1%] in 2017), hypertension (79 977 [40.3%] in 2007 to 134 404 [47.6%] in 2017), diabetes (26 346 [13.3%] in 2007 to 47 089 [16.7%] in 2017), and ischemic stroke (2788 [1.4%] in 2007 to 10 527 [3.7%]). Documentation of short-term ECG increased over time from a low of 23 297 of 207 349 at-risk patients (11.2%) in 2008 to a high of 45 027 (16.0%) in 2017. Use of long-term ECG monitoring showed no clear pattern over time (1871 [0.9%] in 2007 to 4036 [1.4%] in 2017).

Table 2.

Characteristics of the Patient Population At Risk of Atrial Fibrillation by Calendar Year

Characteristic	Patients, No. (%)
	2007 (n = 198 245)	2009 (n = 217 752)	2011 (n = 237 464)	2013 (n = 256 452)	2015 (n = 272 260)	2017 (n = 282 270)
Age group, y
<45	76 485 (38.6)	81 648 (37.5)	86 432 (36.4)	91 354 (35.6)	92 262 (33.9)	90 429 (32.0)
45-54	42 563 (21.5)	47 485 (21.8)	51 555 (21.7)	54 618 (21.3)	57 184 (21.0)	58 055 (20.6)
55-64	34 979 (17.6)	40 185 (18.4)	46 098 (19.4)	50 781 (19.8)	56 208 (20.6)	60 442 (21.4)
65-74	22 851 (11.5)	25 540 (11.7)	28 584 (12.0)	33 342 (13.0)	38 225 (14.0)	43 417 (15.4)
75-84	16 609 (8.4)	17 369 (8.0)	18 228 (7.7)	18 830 (7.3)	20 090 (7.4)	21 315 (7.6)
≥85	4758 (2.4)	5525 (2.5)	6567 (2.8)	7527 (2.9)	8291 (3.0)	8612 (3.0)
Women	114 251 (57.6)	125 438 (57.6)	137 086 (57.7)	148 638 (58.0)	155 940 (57.3)	162 601 (57.6)
Smoking status
Current	41 774 (21.1)	46 741 (21.5)	50 987 (21.5)	54 292 (21.2)	56 049 (20.6)	56 377 (20.0)
Former	45 450 (22.9)	52 499 (24.1)	59 478 (25.0)	66 937 (26.1)	75 122 (27.6)	80 457 (28.5)
Never	111 021 (56.0)	118 512 (54.4)	126 999 (53.5)	135 223 (52.7)	141 089 (51.8)	145 436 (51.5)
BMI ≥30	71 433 (36.0)	86 543 (39.7)	97 323 (41.0)	107 594 (42.0)	120 519 (44.3)	130 218 (46.1)
Hypertension	79 977 (40.3)	90 730 (41.7)	101 569 (42.8)	111 886 (43.6)	124 557 (45.8)	134 404 (47.6)
Dyslipidemia	91 098 (46.0)	107 503 (49.4)	121 877 (51.3)	134 497 (52.4)	148 350 (54.5)	157 208 (55.7)
Heart failure	5938 (3.0)	6495 (3.0)	6832 (2.9)	7433 (2.9)	8518 (3.1)	9510 (3.4)
Diabetes	26 346 (13.3)	30 210 (13.9)	33 859 (14.3)	37 476 (14.6)	42 883 (15.8)	47 089 (16.7)
Myocardial infarction	5520 (2.8)	6344 (2.9)	7077 (3.0)	7997 (3.1)	9592 (3.5)	11 132 (3.9)
Coronary artery disease	17 921 (9.0)	19 499 (9.0)	21 637 (9.1)	23 905 (9.3)	26 866 (9.9)	28 621 (10.1)
Coronary artery bypass graft	3582 (1.8)	3767 (1.7)	3863 (1.6)	4061 (1.6)	4595 (1.7)	4964 (1.8)
Percutaneous coronary intervention	3708 (1.9)	4701 (2.2)	5647 (2.4)	6579 (2.6)	7984 (2.9)	8989 (3.2)
Ischemic stroke	2788 (1.4)	4930 (2.3)	5942 (2.5)	7143 (2.8)	8722 (3.2)	10 527 (3.7)
Venous thromboembolism	3092 (1.6)	3910 (1.8)	4860 (2.0)	6130 (2.4)	7814 (2.9)	9889 (3.5)
Pulmonary embolism	1507 (0.8)	1953 (0.9)	2461 (1.0)	3078 (1.2)	3922 (1.4)	4720 (1.7)
Implanted pacemaker	1310 (0.7)	1530 (0.7)	1776 (0.8)	1898 (0.7)	2196 (0.8)	3911 (1.4)
Implantable cardioverter defibrillator	380 (0.2)	497 (0.2)	604 (0.2)	744 (0.3)	934 (0.3)	2657 (0.9)
Short-term ECG monitoring	22 324 (11.3)	24 941 (11.4)	32 596 (13.7)	34 648 (13.5)	41 081 (15.1)	45 027 (16.0)
Long-term ECG monitoring	1871 (0.9)	2327 (1.1)	2371 (1.0)	2432 (1.0)	2899 (1.1)	4036 (1.4)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); ECG, electrocardiographic.

Discussion

The current study, undertaken within a community-based setting among individuals receiving health care services through a single integrated health care delivery network, found increasing rates of new AF diagnoses from 2006 to 2018. AF incidence rates increased significantly in all age and sex subgroups, and across all diagnostic setting and priority pairings. Although the mean age of incident AF cases showed no significant change over time, the fraction of new cases among individuals aged 85 years or older increased; furthermore, absolute incidence rate increases over time were by far largest among this age group. Several comorbidities increased in prevalence among incident AF cases over time (most notably high BMI and hypertension), in parallel with their increasing frequency in the at-risk population. Notably, the proportion of new AF cases with heart failure or ischemic heart disease, common AF antecedents, either decreased or showed no change over time. Documentation of short-term ECG in patients at risk of AF increased over time, but long-term ECG monitoring showed no clear pattern.

Multiple studies have reported increasing AF incidence and prevalence rates over time, yet others have reported more stable rates.[1,2,4,6,9,18,19,24,25,26,27,28,29] Our results provide a contemporary view of the evolving AF burden, showing increasing AF incidence rates in a community-based setting through the end of 2018. The current study extends our previous work by adding 2 years of follow-up and evaluating several potential explanations for the observed increase.[1] A population dynamics phenomenon has been offered as a partial explanation for an increased AF burden, ie, more people are surviving to older ages and thus chronic conditions (such as AF) inevitably increase in parallel. Although increasingly older populations will increase the absolute number of individuals with AF, increasing age-adjusted rates suggest that overall increases are more than a function of an aging population. Notably, the mean age of incident AF cases did not change over time in our study, although the percentage of new AF cases in the most extreme age group (ie, ≥85 years) increased from 12.6% to 18.6% during the study period. This observation is consistent with prior studies showing modest increases in age over time in newly diagnosed AF.[2,18,19,30] Although AF incidence rates increased significantly in all age groups, absolute rate changes were 2-fold larger in the oldest age group compared with other age groups. Thus, increasing AF rates in the oldest age group made the largest contribution to the overall observed trend.

Consistent with prior studies, newly diagnosed AF cases had increasing prevalence over time of several comorbidities including high BMI, hypertension, diabetes, and chronic lung disease.[2,9,19,28] These AF risk factors also demonstrated increasing prevalence over time in the at-risk population, supporting the contention that an increased risk factor burden in the at-risk population is associated with the observed increase in incidence rates. Importantly, incident AF cases were not more likely over time to have a history of heart failure, coronary artery disease, myocardial infarction, or coronary bypass surgery, refuting the hypothesis that prolonged survival of individuals with cardiac disorders predisposing them to AF partially explains the AF increase. Similar observations have been reported in prior studies, yet the mechanistic explanations remain unclear.[2,9,19,28] Our study also observed increases over time in prior thrombotic events, including ischemic stroke. Although speculative, a history of these events at the time of diagnosis suggests subclinical AF may have existed prior to the current diagnosis in some cases.[31,32]

Two prior studies that evaluated changing AF incidence rates by diagnostic setting (inpatient vs outpatient) and priority (primary vs secondary diagnosis) showed discordant results and served as important motivators for the current study.[18,19] In the United Kingdom, an overall increasing AF incidence rate was explained predominantly by increases in AF as a secondary diagnosis among hospitalized patients aged 80 years or older.[18] Incidence rates for AF as a primary hospital or primary care (outpatient) diagnosis showed no change over time. In contrast, Piccini et al[19] reported that new AF diagnoses in outpatient settings among a Medicare population were increasing at a faster rate than in inpatient settings. Our findings provide little resolution to this matter, finding that AF incidence rates increased significantly in all diagnostic setting and priority pairings, with some evidence suggesting larger increases in the inpatient, primary group relative to others. Thus, our results do not support the speculative hypothesis that incident AF as a secondary diagnosis is increasing more rapidly than as a primary diagnosis. Of note, the percentage of new AF diagnoses first documented as inpatient in our study (41.4%) was lower than previously observed (49%-66%).[18,19,20,22,33] This may be attributable to methodologic differences, particularly our extended outpatient definition. Lastly, with respect to the balance of AF documented as a primary vs secondary diagnosis, our results are largely similar to prior work showing that 74% of new AF cases were diagnosed as secondary, differing little between inpatient (73.4%) and outpatient (74.5%) settings.[5,18,27,34]

As early-stage AF is frequently sporadic and asymptomatic, a significant fraction of extant AF is believed to be latent yet detectable if sought.[7,35] Indeed, a rational explanation for our results is that in more recent years, a subset of AF is being identified (and diagnosed) that was historically unappreciated (and undiagnosed) within usual clinical channels. Supporting this contention, we observed increased utilization of short-term ECG in the at-risk population over time coupled with increasing AF incidence rates in all subgroups examined. Furthermore, a 50-year AF incidence study from the Framingham population[24] showed decreasing AF incidence rates when restricting diagnoses to study-specific ECGs, but increasing AF incidence when diagnoses from outside medical records were included. These collective results may reflect an enhanced capacity of US health care systems to uncover AF—not unexpected given the numerous technological means by which AF can be detected.[36,37,38,39] Importantly, the increased diagnostic yield from many modern devices may come predominantly from more sporadic forms of AF, and even these less persistent forms are associated with an increased risk of thromboembolic events.[36,40]

Limitations

Some limitations of this study should be noted. First, study patients were patrons of a single health care system. Thus, the attributes of the at-risk population unlikely represent completely those of the underlying local community as suggested by the greater proportion of women (approximately 60%) and individuals who currently smoke (approximately 20%). Healthier individuals in the community who never or seldom use formal health care services would not be included in our denominator (ie, the base study population), which may partially explain our higher incidence rates relative to comparable studies.[28] Furthermore, patients meeting study criteria but with new AF diagnoses made at non-Geisinger facilities may have their diagnosis missed, delayed, and/or misclassified according to diagnostic setting or priority. Finally, as our AF case definition relied exclusively on diagnostic codes, our results may be susceptible to bias from changes in administrative coding practices over time.

Conclusions

This community-based study found increasing AF incidence rates over time. Concurrently, an increasing burden of AF risk factors in the at-risk population and more frequent short-term ECG use were observed. In contrast, over time, AF was not more frequently documented as a secondary diagnosis, nor were new AF cases more likely to have preexisting cardiac conditions.