Post-operative atrial fibrillation and risk of heart failure hospitalization 

Abstract

Aims

Post-operative atrial fibrillation (POAF) is associated with stroke and mortality. It is unknown if POAF is associated with subsequent heart failure (HF) hospitalization. This study aims to examine the association between POAF and incident HF hospitalization among patients undergoing cardiac and non-cardiac surgeries.

Methods and results

A retrospective cohort study was conducted using all-payer administrative claims data that included all non-federal emergency department visits and acute care hospitalizations across 11 states in the USA. The study population included adults aged at least 18 years hospitalized for surgery without a prior diagnosis of HF. Cox proportional hazards regression models were used to examine the association between POAF and incident HF hospitalization after making adjustment for socio-demographics and comorbid conditions. Among 76 536 patients who underwent cardiac surgery, 14 365 (18.8%) developed incident POAF. In an adjusted Cox model, POAF was associated with incident HF hospitalization [hazard ratio (HR) 1.33; 95% confidence interval (CI) 1.25–1.41]. In a sensitivity analysis excluding HF within 1 year of surgery, POAF remained associated with incident HF hospitalization (HR 1.15; 95% CI 1.01–1.31). Among 2 929 854 patients who underwent non-cardiac surgery, 23 763 (0.8%) developed incident POAF. In an adjusted Cox model, POAF was again associated with incident HF hospitalization (HR 2.02; 95% CI 1.94–2.10), including in a sensitivity analysis excluding HF within 1 year of surgery (HR 1.49; 95% CI 1.38–1.61).

Conclusions

Post-operative atrial fibrillation is associated with incident HF hospitalization among patients without prior history of HF undergoing both cardiac and non-cardiac surgeries. These findings reinforce the adverse prognostic impact of POAF and suggest that POAF may be a marker for identifying patients with subclinical HF and those at elevated risk for HF.

Structured Graphical Abstract

Associations between post-operative atrial fibrillation and incident heart failure hospitalization were observed following cardiac and non-cardiac surgeries. AF, atrial fibrillation.

Atrial fibrillation, Heart failure

See the editorial comment for this article ‘Post-operative AF and heart failure hospitalizations: what remains hidden in patients undergoing surgery’, by Melissa E. Middeldorp and Christine M. Albert, https://doi.org/10.1093/eurheartj/ehac335.

Introduction

Post-operative atrial fibrillation (POAF) occurs in up to 40% of patients undergoing cardiac surgery¹ and 2% of patients undergoing non-cardiac surgery.² In clinical practice, POAF is often viewed as a relatively benign event triggered by the stress and subsequent catecholamine release of surgery.² Emerging evidence suggests that POAF is associated with long-term sequelae such as stroke and all-cause mortality,³ but its clinical significance remains incompletely understood.

Recent data reveal that atrial fibrillation (AF) is associated with a myriad of pathologic abnormalities of the left atrium, including chamber dilatation, fibrosis, impaired calcium handling, electrical remodelling, and reduced function.⁴ Accordingly, contemporary viewpoints focus on AF as a mechanistic process that is closely intertwined with the pathophysiology of heart failure (HF), whereby AF begets HF and HF begets AF.⁵

Despite these shared pathophysiologic mechanisms, limited data are available regarding the association between POAF and future HF.⁵ With the present study, we aimed to better understand an important potential long-term consequence of POAF—incident HF hospitalization. An association between POAF and incident HF hospitalization would have important implications regarding the clinical relevance of POAF and potentially inform future studies and treatment paradigms for these common conditions.

Methods

Data source and study population

We performed a retrospective cohort study using all-payer claims data on all discharges from all non-federal emergency departments (ED) and acute care hospitals in Arkansas, Florida, Georgia, Iowa, Maryland, Massachusetts, Nebraska, New York, Utah, Vermont, and Wisconsin. These states have a combined population of ∼80 million residents, comprising 25% of the total US population.⁶ Data were available from calendar year 2016 for all 11 states; 2017 for all states except New York; and 2018 for Florida, Iowa, Maryland, Nebraska, Vermont, and Wisconsin. We followed the Reporting of Studies Conducted Using Observational Routinely-Collected Data guidelines.⁷ The Weill Cornell Medicine Institutional Review Board approved this analysis of deidentified data and waived the requirement for informed consent.

We included all adults ≥18 years old with a hospitalization for surgery, defined using surgery-related Medicare Severity-Diagnosis Related Group (MS-DRG) codes. For patients with multiple surgical hospitalizations captured in our data, we included the first hospitalization. We excluded patients with any documented ED or hospital discharge diagnosis of HF either before or during the index surgical hospitalization (Figure 1). Heart failure hospitalization was ascertained using previously validated ICD-10 codes that have >90% positive and negative predictive values.⁸ ICD-10 codes used for HF are given in Supplementary material online, Table S1. We excluded patients who did not survive the index surgical hospitalization, because we were interested in the long-term risk of HF after POAF.

Figure 1

Flow diagram of patients included in an analysis of post-operative atrial fibrillation and heart failure.

Measurements

Our primary exposure variable was AF status during the index surgical hospitalization, classified as POAF, previously diagnosed AF, and no AF. Post-operative atrial fibrillation was defined as a discharge diagnosis of AF from the index surgery hospitalization that was coded as not present on admission, along with the absence of any previously documented AF during a prior ED visit or hospitalization. Patients with an ED or hospital discharge diagnosis of AF at a prior encounter, or AF coded as present on admission at the time of the index surgical hospitalization, were coded as having previously diagnosed AF. All other patients were classified as having no AF. Atrial fibrillation was ascertained using previously validated ICD-10 codes with excellent sensitivity, specificity, and positive and negative predictive values.⁹ ICD-10 codes used for AF are given in Supplementary material online, Table S1. Present-on-admission codes in administrative data have been previously validated and used in other studies examining POAF.^10,11 For descriptive purposes, AF was subclassified as paroxysmal, persistent, or chronic AF, as defined by specific subgroups of ICD-10 codes (see Supplementary material online, Table S1).

Our primary outcome variable was incident HF hospitalization, ascertained using previously validated ICD-10 codes,⁸ after discharge from the index surgical hospitalization.

Covariates to control for potential confounders of the association between POAF and incident HF included socio-demographics (age, sex, race, and insurance status) and comorbid conditions (hypertension, diabetes, coronary disease, valvular disease, and body mass index). We used ICD-10 codes related to body mass index and obesity to categorize patients into three categories of body mass index: <30, 30–40, and ≥40 kg/m². For descriptive purposes, the type of surgery was classified according to the Major Diagnostic Categories (see Supplementary material online, Table S2).

Statistical analysis

We calculated means with standard deviations and made comparisons using the t-test for continuous variables, and calculated counts and proportions with 95% confidence intervals (CI) and made comparisons using the χ² test for categorical variables. We constructed Kaplan–Meier curves to calculate the cumulative incidence of HF hospitalization over the study period and used the log-rank test to compare cumulative rates among patients with POAF, previously diagnosed AF, and no AF. We calculated incident HF hospitalization rates per 1000 person-years for each group. We censored patients at the time of death or at the end of the available follow-up period for each state; all-cause mortality data were available from subsequent ED visits and hospitalizations. We used Cox proportional hazards regression models to examine the association between POAF and incident HF hospitalization after adjustment for socio-demographics and comorbid conditions. We examined log–log plots and confirmed that the proportional hazards assumptions were met. Given low proportions of missing data for demographics and primary diagnosis and procedure codes (<0.1%), we conducted a complete case analysis.

To ensure the robustness of our findings, we also conducted several sensitivity analyses. First, to address the potential that subclinical HF was present at the time of POAF, we excluded HF hospitalizations occurring earlier than 1 year after the index surgical hospitalization. Second, we counted only primary hospital discharge diagnoses of HF as our outcome. Third, we used an alternative definition of HF (see Supplementary material online, Table S1). Fourth, we additionally adjusted our primary model for Charlson comorbid conditions.¹² Fifth, we examined the association between POAF and HF hospitalization after adjustment for the CHA₂DS₂-VASc score instead of our primary model covariates. Sixth, we performed competing risk regression in a random 5% sample of the cohort to account for the competing risk of death. Seventh, in the non-cardiac surgery group, we additionally adjusted for the category of surgery.¹³ Eighth, in the cardiac surgery group, we limited our analysis to elective surgeries to avoid potential confounding by cardiac procedures related to AF itself.

We conducted the aforementioned analyses separately for patients who underwent cardiac surgery vs. those who underwent non-cardiac surgery. Cardiac surgery was identified based on MS-DRG codes listed in Supplementary material online, Table S2. We chose a priori to examine these strata separately because of fundamental differences in patient characteristics, POAF mechanisms, and anticipated variations in illness trajectory and prognosis among patients undergoing each surgery type.

We performed analyses using Stata, version 15 (StataCorp, College Station, TX, USA). A two-sided P-value of <0.05 was considered statistically significant.

Results

The study sample includes data from 3 006 390 patients from 11 different states in the USA with no prior history of HF undergoing surgery. The most common type of surgery involved the musculoskeletal/connective tissue system (35.8%). The mean age of the cohort was 57.0 ± 18.6 years, and 59.9% were women. Patients had follow-up for a median of 1.7 years (interquartile range 1.0–2.2).

At baseline, 38 128 had incident POAF (1.3%) and 201 101 (6.7%) had previously diagnosed AF. Patients with POAF were older than those without AF and slightly younger than those with previously diagnosed AF (Table 1). Those with POAF or previously diagnosed AF more frequently had vascular risk factors (hypertension, diabetes, ischaemic heart disease, and valvular heart disease) compared with those without AF. The cumulative rate of death at 3 years was 4.00% (95% CI 3.75–4.27%) in those with POAF, 4.64% (95% CI 4.51–4.76%) in those with previously diagnosed AF, and 1.56% (95% CI 1.53–1.58%) in those without AF (P < 0.001 for the log-rank test).

Table 1

Baseline characteristics of patients undergoing surgery, stratified by the occurrence of post-operative atrial fibrillation

Characteristic	Post-operative AF (N = 38 128)	Previously diagnosed AF (N = 201 101)	No AF (N = 2 767 161)
Age, mean (SD), years	70.7 (11.5)	73.8 (11.7)	55.6 (18.4)
Female sex	15 203 (39.9)	90 073 (44.8)	1 695 057 (61.3)
Racea
White	32 067 (84.9)	172 378 (86.3)	1 960 048 (71.4)
Black	2 375 (6.3)	11 921 (6.0)	390 208 (14.2)
Hispanic	1961 (5.2)	10 416 (5.2)	274 097 (10.0)
Asian	519 (1.4)	1599 (0.8)	44 351 (1.6)
Other	836 (2.2)	3431 (1.7)	75 476 (2.8)
Payment sourceb
Medicare	26 850 (70.4)	159 073 (79.1)	1 092 597 (39.5)
Medicaid	1631 (4.3)	5859 (2.9)	399 957 (14.5)
Private insurance	7896 (20.7)	30 037 (14.9)	1 054 072 (38.1)
Self-pay	537 (1.4)	1710 (0.9)	91 918 (3.3)
No charge	105 (0.3)	389 (0.2)	19 332 (0.7)
Other	1099 (2.9)	4001 (2.0)	108 055 (3.9)
Vascular risk factors
Hypertension	29 528 (77.4)	159 108 (79.1)	1 304 861 (47.2)
Diabetes	10 898 (28.6)	56 228 (28.0)	503 235 (18.2)
Body mass index
<30 kg/m2	29 663 (77.8)	161 534 (80.3)	2 205 140 (79.7)
30–40 kg/m2	6291 (16.5)	27 895 (13.9)	359 198 (13.0)
>40 kg/m2	2174 (5.7)	11 672 (5.8)	202 823 (7.3)
Ischaemic heart disease	6350 (16.7)	21 740 (10.8)	155 447 (5.6)
Valvular heart disease	8797 (23.1)	30 476 (15.2)	82 987 (3.0)
CHA2DS2-VASc score, mean (SD)	3.1 (1.5)	3.2 (1.4)	2.0 (1.4)
Cardiac surgeryc	14 365 (37.6)	12 481 (6.2)	49 690 (1.8)
Surgery typed
Musculoskeletal system and connective tissue	5957 (15.6)	69 309 (34.5)	1 002 157 (36.2)
Circulatory system	17 996 (47.2)	60 682 (30.2)	269 528 (9.7)
Digestive system	4523 (11.9)	19 989 (9.9)	308 464 (11.2)
Pregnancy, childbirth, and the puerperium	67 (0.2)	319 (0.2)	419 535 (15.2)
Endocrine, nutritional, and metabolic	257 (0.7)	2599 (1.3)	78 945 (2.9)
Nervous system	1577 (4.1)	12 551 (6.2)	125 088 (4.5)
Female reproductive system	165 (0.4)	1065 (0.5)	73 661 (2.7)
Respiratory system	3045 (8.0)	7769 (3.9)	71 328 (2.6)
Kidney and urinary tract	556 (1.5)	4413 (2.2)	59 806 (2.2)
Male reproductive system	100 (0.3)	1347 (0.7)	25 571 (0.9)
Other	3885 (10.2)	21 058 (10.5)	333 078 (12.0)
Type of AFe
Paroxysmal AF	14 932 (39.2)	67 996 (33.8)	—
Persistent AF	637 (1.7)	5531 (2.8)	—
Chronic AF	439 (1.2)	30 022 (14.9)	—
Unspecified AF	20 240 (53.1)	78 549 (39.1)	—
Atrial flutter	1880 (4.9)	19 003 (9.5)	—

Data are presented as number (%) unless otherwise specified. AF, atrial fibrillation; ICD-10, International Classification of Diseases, 10th revision; MS-DRG, Medicare Severity-Adjusted Diagnosis Related Group; SD, standard deviation.

a

Self-reported by patients or their surrogates. Numbers do not sum to group totals because of missing data in <1% of patients.

b

Numbers do not sum to group totals because of missing data in <0.1% of patients.

c

Defined using MS-DRG codes.

d

Defined based on Major Diagnostic Categories.

e

Defined based on ICD-10 codes.

Cardiac surgery cohort

Among 76 536 patients who underwent cardiac surgery, 14 365 (18.8%) developed incident POAF and 12 481 (16.3%) had a prior diagnosis of AF (Table 2). The HF hospitalization rate per 1000 person-years was 66 (95% CI 62–69) for patients with POAF, 98 (95% CI 94–103) for previously diagnosed AF, and 44 (95% CI 43–46) for no AF. The cumulative risk of HF hospitalization among patients with POAF was lower than patients with previously diagnosed AF but higher than those without AF (Figure 2). In a Cox model, we found associations with HF hospitalization for both POAF [hazard ratio (HR) 1.33; 95% CI 1.25–1.41] and previously diagnosed AF (HR 1.91; 95% CI 1.80–2.02). When compared with patients with previously diagnosed AF, those with POAF had a lower hazard for HF hospitalization (HR 0.70; 95% CI 0.66–0.75).

Figure 2

Cumulative rates of heart failure hospitalization after cardiac surgery, stratified by status of atrial fibrillation. Cumulative rates differed significantly among groups (P < 0.001 by log-rank test).

Table 2

Baseline characteristics of patients undergoing cardiac surgery, stratified by the occurrence of post-operative atrial fibrillation

Characteristic	Post-operative AF (N = 14 365)	Previously diagnosed AF (N = 12 481)	No AF (N = 49 690)
Age, mean (SD), years	68.6 (9.6)	69.8 (10.1)	63.2 (12.1)
Female sex	3770 (26.2)	3712 (29.7)	14 940 (30.1)
Racea
White	12 429 (87.6)	10 810 (87.5)	38 951 (79.3)
Black	554 (3.9)	496 (4.0)	4076 (8.3)
Hispanic	677 (4.8)	624 (5.1)	3508 (7.1)
Asian	199 (1.4)	144 (1.2)	956 (2.0)
Other	332 (2.3)	278 (2.3)	1631 (3.3)
Payment sourceb
Medicare	9488 (66.1)	8762 (70.2)	24 243 (48.8)
Medicaid	511 (3.6)	384 (3.1)	3497 (7.0)
Private insurance	3769 (26.2)	2949 (23.6)	18 713 (37.7)
Self-pay	192 (1.3)	106 (0.9)	1273 (2.6)
No charge	46 (0.3)	35 (0.3)	325 (0.7)
Other	355 (2.5)	245 (2.0)	1623 (3.3)
Vascular risk factors
Hypertension	12 116 (84.3)	10 252 (82.1)	39 389 (79.3)
Diabetes	4819 (33.6)	3933 (31.5)	17 589 (35.4)
Body mass index
<30 kg/m2	10 548 (73.4)	9453 (75.7)	37 529 (75.5)
30–40 kg/m2	3015 (21.0)	2358 (18.9)	9802 (19.7)
>40 kg/m2	802 (5.6)	670 (5.4)	2359 (4.8)
Ischaemic heart disease	3566 (24.8)	2569 (20.6)	13 668 (27.5)
Valvular heart disease	6164 (42.9)	6436 (51.6)	15 443 (31.1)
CHA2DS2-VASc score, mean (SD)	3.1 (1.4)	3.0 (1.4)	2.7 (1.4)
Type of AFc
Paroxysmal AF	5559 (38.7)	5393 (43.2)	—
Persistent AF	220 (1.5)	749 (6.0)	—
Chronic AF	37 (0.3)	1822 (14.6)	—
Unspecified AF	8154 (56.8)	4079 (32.7)	—
Atrial flutter	395 (2.8)	438 (3.5)	—

Data are presented as number (%) unless otherwise specified. AF, atrial fibrillation; ICD-10, International Classification of Diseases, 10th revision; SD, standard deviation.

a

Self-reported by patients or their surrogates. Numbers do not sum to group totals because of missing data in <1% of patients.

b

Numbers do not sum to group totals because of missing data in <0.1% of patients.

c

Defined based on ICD-10 codes.

The association between POAF and HF hospitalization remained significant but was attenuated in a sensitivity analysis excluding incident HF hospitalization within 1 year of the index surgical hospitalization (HR 1.15; 95% CI 1.01–1.31). Otherwise, our findings across other sensitivity analyses were similar to those of our primary analysis (see Supplementary material online, Table S3).

Non-cardiac surgery cohort

Among 2 929 854 patients who underwent non-cardiac surgery, 23 793 (0.8%) developed incident POAF and 188 620 (6.4%) had a prior diagnosis of AF (Table 3). The incident HF hospitalization rate per 1000 person-years was 66 (95% CI 64–69) for patients with POAF, 85 (95% CI 84–86) for patients with previously diagnosed AF, and 17 (95% CI 17–17) for patients without AF. The risk for HF hospitalization among patients with POAF was lower than patients with previously diagnosed AF but higher than those without AF (Figure 3). In a Cox model, we found associations with HF hospitalization for both POAF (HR 2.02; 95% CI 1.94–2.10) and previously diagnosed AF (HR 2.32; 95% CI 2.28–2.36). When compared with patients with previously diagnosed AF, those with POAF had a lower hazard for HF hospitalization (HR 0.84; 95% CI 0.81–0.88).

Figure 3

Cumulative rates of heart failure hospitalization after non-cardiac surgery, stratified by status of atrial fibrillation. Cumulative rates differed significantly among groups (P < 0.001 by log-rank test).

Table 3

Baseline characteristics of patients undergoing non-cardiac surgery, stratified by the occurrence of post-operative atrial fibrillation

Characteristic	Post-operative AF (N = 23 763)	Previously diagnosed AF (N = 188 620)	No AF (N = 2 717 471)
Age, mean (SD), years	72.0 (12.3)	74.0 (11.8)	55.5 (18.5)
Female sex	11 433 (48.1)	86 361 (45.8)	1 680 117 (61.8)
Racea
White	19 638 (83.3)	161 568 (86.2)	1 921 097 (71.3)
Black	1821 (7.7)	11 425 (6.1)	386 132 (14.3)
Hispanic	1284 (5.5)	9792 (5.2)	270 589 (10.0)
Asian	320 (1.4)	1455 (0.8)	43 395 (1.6)
Other	504 (2.1)	3153 (1.7)	73 845 (2.7)
Payment sourceb
Medicare	17 362 (73.1)	150 311 (79.7)	1 068 354 (39.3)
Medicaid	1120 (4.7)	5475 (2.9)	396 460 (14.6)
Private insurance	4127 (17.4)	27 088 (14.4)	1 035 359 (38.1)
Self-pay	345 (1.5)	1604 (0.9)	90 645 (3.3)
No charge	59 (0.3)	354 (0.2)	19 007 (0.7)
Other	744 (3.1)	3756 (2.0)	106 432 (3.9)
Vascular risk factors
Hypertension	17 412 (73.3)	148 856 (78.9)	1 265 472 (46.6)
Diabetes	6079 (25.6)	52 295 (27.7)	485 646 (17.9)
Body mass index
<30 kg/m2	19 115 (80.4)	152 081 (80.6)	2 167 611 (79.8)
30–40 kg/m2	3276 (13.8)	25 537 (13.5)	349 396 (12.9)
>40 kg/m2	1372 (5.8)	11 002 (5.8)	200 464 (7.4)
Ischaemic heart disease	2784 (11.7)	19 171 (10.2)	141 779 (5.2)
Valvular heart disease	2633 (11.1)	24 040 (12.8)	67 544 (2.5)
CHA2DS2-VASc score, mean (SD)	3.1 (1.5)	3.2 (1.4)	2.0 (1.4)
Type of AFc
Paroxysmal AF	9373 (39.4)	62 603 (33.2)	—
Persistent AF	417 (1.8)	4782 (2.5)	—
Chronic AF	402 (1.7)	28 200 (15.0)	—
Unspecified AF	12 086 (50.9)	74 470 (39.5)	—
Atrial flutter	1485 (6.3)	18 565 (9.8)	—

Data are presented as number (%) unless otherwise specified. AF, atrial fibrillation; ICD-10, International Classification of Diseases, 10th revision; SD, standard deviation.

a

Self-reported by patients or their surrogates. Numbers do not sum to group totals because of missing data in <1% of patients.

b

Numbers do not sum to group totals because of missing data in <0.1% of patients.

c

Defined based on ICD-10 codes.

The association between POAF and HF hospitalization remained significant but was attenuated in a sensitivity analysis excluding incident HF hospitalization within 1 year of the index surgical hospitalization (HR 1.49; 95% CI 1.38–1.61). Otherwise, our findings across other sensitivity analyses were similar to those of our primary analysis (see Supplementary material online, Table S3).

Discussion

In a large cohort of 3 million patients across 11 states in the USA, we found that incident POAF was associated with an increased risk for incident HF hospitalization among patients undergoing both cardiac and non-cardiac surgeries (Structured Graphical Abstract). These findings persisted across multiple sensitivity analyses. These data indicate that POAF, which is often considered to be a transient response to surgery, may in some cases be a manifestation of preclinical HF.

Prior work has shown that POAF is associated with an increased risk for stroke and mortality.³ Despite the known pathophysiological link between AF and HF, limited data exist on the association between POAF and future HF. In this context, our study now adds novel findings that patients undergoing cardiac surgery who experienced POAF had a 50% higher hazard of subsequent HF hospitalization and that patients undergoing non-cardiac surgery who experienced POAF had a two-fold higher hazard of subsequent HF hospitalization.

Our findings suggest that POAF may be a marker for underlying subclinical HF. This finding agrees with recent studies showing that, among patients presenting with initially unexplained dyspnoea, a history of AF is associated with the presence of underlying HF.^14,15 Our study findings have important implications for the management of patients who develop POAF. For example, among those who develop POAF, it may be reasonable to consider whether the patient might have latent HF and whether additional diagnostic testing, education about the potential symptoms of HF, and closer follow-up during the post-operative period are required. Those who develop POAF may warrant more aggressive treatment of other HF risk factors such as hypertension, diabetes, and atherosclerosis. The notion of screening for post-operative complications like POAF and linking its early identification to aggressive therapeutic interventions is consistent with the 2014 European Society of Cardiology guidelines on cardiovascular management in the post-operative setting.¹⁶ Future work should examine the disease trajectory and mortality of those who develop POAF and whether implementing intensified preventive strategies among patients who develop POAF can prevent future HF and/or reduce its associated costs and morbidity.¹⁷

Even when excluding HF hospitalization within 1 year of surgery, POAF remains associated with the risk of HF hospitalization. This suggests that POAF may be a marker for an underlying atrial myopathy that heralds future HF events. Our findings are also consistent with the possibility that POAF is an aetiologic contributor to HF. Prior work has elucidated a bidirectional relationship between AF and HF, whereby AF begets HF and AF begets HF.^5,18 Consistent with this notion, our results suggest that AF preceding surgery is associated with incident HF events, and prior work has shown that clinically apparent AF in non-surgical settings is associated with incident HF after adjustment for other risk factors.^14,15,19 Post-operative atrial fibrillation may be a sign of existing preclinical cardiac pathology, and POAF itself may also lead to adverse cardiac remodelling that further increases the risk of subsequent HF.²⁰ Our work supports continued investigation of the mechanistic link between AF and HF, including an investigation of whether selected strategies of treatment for POAF (for example, rhythm vs. rate control) can reduce the risk of HF hospitalization. Recent work has shown that early rhythm control in patients with AF is associated with a reduced risk of adverse cardiovascular outcomes compared with usual care.²¹ Although differences in the EAST trial in hospitalization for worsening HF do not reach statistical significance, the findings are consistent with a possible reduction in this outcome. Taken together, these data support a clinical trial examining rhythm control on HF-related endpoints.

We found that the association between POAF and HF hospitalization was stronger among non-cardiac surgery patients than among cardiac surgery patients. Patients undergoing cardiac surgery more frequently had risk factors for HF including hypertension, diabetes, and coronary artery disease compared with those who underwent non-cardiac surgery. Moreover, patients who underwent cardiac surgery experienced direct myocardial injury by virtue of the surgery itself, which may lead to direct physical irritation and increased risk for arrhythmia.²² The present data relating POAF to risk of incident HF suggest that surgery may be considered in some ways as an ‘atrial stress test’, wherein patients who cannot accommodate the changes in atrial loading, inflammation, and other perturbations develop POAF and are accordingly more likely to develop clinically overt HF events in the future as well. The ‘dose’ of stress from non-cardiac surgery to the atrium may be lower, so the incremental role of POAF as a marker and/or aetiologic contributor to HF hospitalization is likely more substantial among those undergoing non-cardiac surgery as compared with those undergoing cardiac surgery. Given that a predominant number of surgeries performed in the USA is non-cardiac in nature and the number of patients who experience POAF is subsequently the greatest after non-cardiac surgery, our findings are highly relevant from a population standpoint and justify additional investigation into POAF.

Our study had several limitations. First, because this was an observational study, there was a risk of residual confounding, and, therefore, we could not make any conclusions regarding causation. However, our findings may be of interest regardless of whether POAF is a risk marker of subclinical AF and/or contributes to the risk of future HF. Second, we relied on administrative claims data and ICD codes to identify medical conditions, which have inherent limitations. We attempted to mitigate these by using previously validated codes. Nevertheless, we lacked data on clinical details such as the means of POAF ascertainment and whether it was symptomatic. In the absence of continuous heart rhythm monitoring, we were unable to examine the degree to which the association between POAF and subsequent HF hospitalization was mediated by the progression of paroxysmal AF. Third, we did not have any data on the management strategies of POAF either during hospitalization or following hospital discharge and, therefore, could not comment on the impact of different strategies on POAF outcomes. Fourth, we did not have detailed clinical data such as left ventricular ejection fraction or left atrial size, which would have provided helpful additional context for our findings. For example, whether POAF is a stronger risk factor for HF with preserved ejection fraction, the pathophysiology of which involves left atrial abnormalities,²³,²⁴ is unknown and merits additional investigation.

Conclusions

We found that incident POAF was associated with an increased risk for incident HF hospitalization among patients undergoing both cardiac and non-cardiac surgeries. Our findings suggest that POAF may be a marker for subclinical HF and that POAF may potentially contribute to the development of HF. Future work is warranted to better understand this association and identify strategies to prevent HF hospitalization in these patients. In the meantime, clinicians should be aware that POAF may be a harbinger of HF.

Authors’ contributions

H.K. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis; P.G. and H.K. conceptualized and designed the study; P.G., M.K., U.K., and H.K. were involved in the acquisition, analysis, or interpretation of data and drafted the manuscript; all authors performed a critical revision of the manuscript for garnering important intellectual content; H.K. conducted a statistical analysis; M.S. and H.K. provided administrative, technical, or material support; H.K was involved in supervision of the study.

Supplementary material

Supplementary material is available at European Heart Journal online.

Funding

This work was supported by the American Heart Association 20CDA35310455 (P.G.); National Institute on Aging K76AG064428 (P.G.); Loan Repayment Program award L30AG060521 (P.G.); and the National Heart, Lung, and Blood Institute R01HL144541 (H.K.).

Author notes

Conflict of interest: P.G. receives personal fees for medicolegal consulting related to heart failure and has received honoraria from Akcea Inc. and Bionest Inc. H.K. serves as a PI for the NIH-funded ARCADIA trial (NINDS U01NS095869), which receives in-kind study drug from the BMS-Pfizer Alliance for Eliquis^® and ancillary study support from Roche Diagnostics; as Deputy Editor for JAMA Neurology; on clinical trial steering/executive committees for Medtronic, Janssen, and Javelin Medical; and on endpoint adjudication committees for NovoNordisk and Boehringer-Ingelheim. B.A.B. receives research support from the NIH/NHLBI, Axon, AstraZeneca, Corvia, Medtronic, GlaxoSmithKline, Mesoblast, Novartis, Tenax Therapeutics, and consults/serves on advisory boards for Actelion, Amgen, Aria, Boehringer-Ingelheim, Edwards, Eli Lilly, Imbria, Janssen, Merck, Novo Nordisk, VADovations.