Prognostic Significance of Newly Diagnosed Atrial Fibrillation After Acute Myocardial Infarction: A Study of 184,980 Medicare Patients

Abstract

We aimed to determine whether newly diagnosed atrial fibrillation (AF) predicted cardiovascular events and death after myocardial infarction (AMI) in a large nationwide cohort of patients. All Medicare beneficiaries aged >65 years who were discharged alive after a diagnosis of AMI between January 1, 2007 and December 31, 2008 were identified. Main exposure was a diagnosis of AF during admission or within 90 days after discharge. Primary outcome was a composite of recurrent AMI, stroke and all-cause mortality. Secondary outcomes were each of recurrent AMI, stroke and all-cause mortality. We used Cox proportional hazards regression to assess the relationship between AF and time-to-event outcomes with follow up ending at 3 years. Of 184,980 patients, 9.1 % had AF; 40.6 % were male; 82.8 % were non-Hispanic whites. Mean age was 79.1 ± 8.1 years. Overall, 15.7 % had subsequent AMI, 5.7 % had stroke and 43.9 % died during a mean follow up of 26.4 months. AF was associated with a significantly increased risk of the primary outcome (Hazard ratio (HR) = 1.10; 95 % confidence interval (CI): 1.07–1.12). AF was also separately associated with significantly increased risk of recurrent AMI (HR = 1.09; 95 % CI: 1.04–1.14), stroke (HR = 1.29; 95 % CI: 1.21–1.37), and death (HR = 1.09; 95 % CI: 1.06–1.12). Neither age, race nor sex modified the effects of AF on primary or secondary outcomes. In conclusion, AF is a significant predictor of adverse cardiovascular outcomes and mortality after AMI. Further studies are needed to understand mechanisms by which AF alters outcomes in survivors of AMI.

1. Introduction

One out of every 4 deaths in the United States is due to heart disease [1]. Coronary artery disease is the underlying cause of mortality in 1 out of every 7 deaths [1]. The American Heart Association estimates that that an American dies every 40 s from myocardial infarction [1].

Atrial fibrillation (AF) is the most common supraventricular arrhythmia complicating acute myocardial infarction (AMI), with an incidence of 6–21 % [2]. There appears to be a bidirectional relationship between AF and AMI, mediated at least in part by the common risk factors they share [3,4]. In AMI, the presence of AF is an independent marker of poor prognosis. New-onset AF after AMI has been associated with an increase in short and long-term mortality, stroke and recurrent AMI [5–10]. Previous paroxysmal or chronic AF may be equally as predictive of poor outcomes after AMI [11,12].

Most of the studies evaluating the impact of AF after AMI have been conducted outside of the United States (Europe and Asia). The findings of these studies may not necessarily be generalizable to the United States because of different healthcare systems and practices. Previous studies have shown that differences in treatments and healthcare practices between regions and countries may lead to international differences in disease outcomes [13,14]. Furthermore, the prevalence of AF in the United States has been reported to be the highest in the world, as high as 1.5 to 2 times that of Europe and Asia [15]. Thus, there is a knowledge gap regarding the prognostic significance of AF after AMI in the United States, particularly among the elderly population who are at greatest risk for both conditions.

In the present study, we aimed to determine whether newly diagnosed AF predicted all-cause mortality and cardiovascular events (recurrent AMI and stroke) in a large nationwide cohort of patients enrolled in Medicare who were discharged alive after an AMI.

2. Methods

2.1. Study cohort

In this study, we identified all Medicare beneficiaries aged >65 years discharged alive after a diagnosis of AMI (ICD-9 410.x1) between January 1, 2007 and December 31, 2008. The date of inpatient admission for AMI served as the index admission date, while the index discharge date was defined as the date that the patient was discharged from the hospital. Beyond being diagnosed with an AMI, further inclusion criteria were: (1) having complete claims information in the 12 months prior to the index admission date; (2) not having had an AMI within 12 months prior to the index admission date; (3) being enrolled in Medicare Parts A and B during the 12 months prior to the index stay; (4) being enrolled in Medicare Part D during the 6 months prior to index; (5) after index discharge, having complete claims information and being enrolled in Medicare Parts A, B and D until either the patient’s death or 3 years after the index discharge date, whichever occurred first. The final cohort consisted of 184,980 patients. This study was authorized by the institutional review board of the University of Iowa.

Primary Predictor of Interest and Outcomes: Our overarching goal was to determine the impact of newly diagnosed AF on clinical outcomes within this cohort of patients discharged following an AMI hospitalization. Newly diagnosed AF was defined as having (1) at least one inpatient stay with a primary or first secondary ICD-9 diagnosis code of 427.31, or (2) at least 2 outpatient stays with a primary or first secondary ICD-9 diagnosis code of 427.31, in the time window from the index admission date to 90 days after discharge [16]. The primary outcome was a composite of recurrent AMI, stroke and all-cause mortality. Secondary outcomes were each of recurrent AMI, stroke and all-cause mortality separately. Recurrent AMI was identified using ICD-9 codes 410.x1, and stroke was identified using codes 430, 431, 434.00, 434.01, 434.10, 434.11, 434.90, 434.91, 435.0, 435.1, 435.3, 435.8, 435.9, 436 and 997.02. To determine the date of death, the Centers for Medicare and Medicaid Services (CMS) used a variety of sources, including the Medicare Common Working File, date of death submitted by family members, and benefit information collected from the Railroad Retirement Board and the Social Security Administration.

2.2. Statistical analyses

First, we compared patient demographics, socioeconomic status and health outcomes according to the presence of a clinical diagnosis of AF, using a chi-square test of independence between the given variable and AF status. We expected a number of variables to be statistically significant due to large sample size, thus we calculated the standardized difference as a measure of magnitude of differences between groups.

$$d=\frac{100\mathrm{*}\left({\bar{x}}_1-{\bar{x}}_2\right)}{\sqrt{\left(s_1^2+s_2^2\right)/2}},$$

where ${\bar{x}}_1$ and ${\bar{x}}_2$ are the sample means for the AF and non-AF groups, respectively, and $s_1^2$ and $s_2^2$ are the respective sample variances.

We assessed the effects of AF on recurrent AMI, stroke and survival using time-to-event analyses. To do so, for each patient in the cohort, we measured time (in days) from the index discharge date to recurrent AMI, stroke or death, whichever occurred first. We also measured time to recurrent AMI, stroke and death, separately. We followed patients until either their death or 3 years after index discharge. If a patient had not experienced the given outcome within 1096 days (3 years) after the index discharge date, the observed value of the outcome was recorded as 1096 and was considered censored.

We created Kaplan-Meier curves for each of the outcomes to assess the unadjusted effect of AF status. Next, we fit Cox proportional hazards (PH) regression models to assess the relationship between AF status and the time-to-event outcomes, while adjusting for a wide range of confounding variables. Each model controlled for >100 covariates, including patient demographics, pre-index and at-index comorbidities, procedures during hospitalization, insurance variables, socioeconomic variables, and medications (including warfarin) taken before and after the index stay. Medications were identified using pharmacy claims and represent whether a particular medication was filled by the patient. A full list of covariates and how they were measured can be found in Appendix A. In the event that a patient died before suffering the given outcome, then that variable was considered censored at the time of death.

We were also interested in whether the effect of AF varied across sex, age or race groups. Race was divided into 4 categories: white, black, Hispanic and other or unknown. In all models, white was used as the baseline. We fit Cox PH regression models which included two-way interactions between AF and each of age, sex and race.

3. Results

Table 1 presents measures of patient demographics, including sex, race and age, as well as baseline comorbidities, complications and procedures during the index stay, measures of socioeconomic factors, medications and 3-year post-index cardiovascular outcomes for the full, AF and non-AF cohorts. Based on the χ² test of independence and a 0.05 level of significance, all variables with the exception of overweight/obese, stroke at index and three-year post-index recurrent AMI were significantly related to AF status. The AF cohort had a significantly higher proportion of males compared to the non-AF cohort (41 % vs. 39 %). Patients in the AF cohort were also more likely to be non-Hispanic white (89 % vs. 82 %), less likely to be black (5 % vs. 9 %), and tended to fall within older age categories. They were also more likely to have heart failure at baseline (54 % vs. 52 %) and suffer ventricular arrhythmias during index admission (10 % vs. 8 %). The AF cohort was less likely to receive stents (22 % vs. 28 %), but more likely to receive pacemakers (3 % vs. 1 %) during index admission. Further, the AF cohort was more likely to be prescribed beta blockers (59 % vs. 50 %), ACE inhibitors (39 % vs. 38 %) and ARBs (19.3 % vs.18.5 %), but less likely to receive prescriptions for clopidogrel (16 % vs. 19 %) in the 6 months preceding index admission. Warfarin and anti-arrhythmic drugs were prescribed much more often among the AF cohort in the 90 days post-index (48 % vs. 11 %, and 27 % vs. 8 %, respectively). Conversely, the non-AF patients were more likely to have co-morbid conditions including hypertension (4 % vs. 3 %), chronic kidney disease (33 % vs. 29 %), and depression (8 % vs. 7 %). The non-AF patients were also more likely to suffer complications during index admission (cardiac arrest, cardiogenic shock, renal failure and pneumonia), although they suffered fewer ventricular arrhythmias.

Table 1.

Baseline characteristics and 3-year outcomes for the full, AF and non-AF cohorts. The p-value is based on a χ² test of independence between AF status and the given variable.

Characteristic	Full cohort count (%)	AF cohort (%)	Non-AF cohort (%)	p-Value	Standardized difference
Total patients	184,980	16,740 (9.1)	168,240 (90.9)	-	-
Male sex	75,062 (40.9)	6448 (38.5)	68,614 (40.8)	<0.0001	−4.63
White race	153,229 (82.8)	14,839 (88.6)	138,390 (82.3)	<0.0001	18.19
Black race	15,409 (8.3)	884 (5.3)	14,525 (8.6)	<0.0001	−13.21
Hispanic race	10,776 (5.8)	666 (4.0)	10,110 (6.0)	<0.0001	−9.33
Age: 66–70	33,563 (18.1)	2205 (13.2)	31,358 (18.6)	<0.0001	−14.99
Age: 71–75	34,305 (18.6)	2762 (16.5)	31,543 (18.8)	<0.0001	−5.91
Age: 76–80	37,232 (20.1)	3517 (21.0)	33,715 (20.0)	0.0028	2.40
Age: 81–85	35,998 (19.5)	3700 (22.1)	32,298 (19.2)	<0.0001	7.18
Age: 85 and older	43,882 (23.7)	4556 (27.2)	39,326 (23.4)	<0.0001	8.85
Baseline comorbidities
Hypertension	6801 (3.7)	521 (3.1)	6280 (3.7)	<0.0001	−3.41
Diabetes mellitus	71,238 (38.5)	5642 (33.7)	65,596 (40.0)	<0.0001	−11.01
Ischemic heart disease	172,248 (93.1)	15,345 (91.7)	156,903 (93.3)	<0.0001	−6.04
Chronic kidney disease	59,463 (32.2)	4853 (29.0)	54,610 (32.5)	<0.0001	−7.52
Stroke	14,522 (7.9)	1314(7.9)	13,208 (7.9)	0.9955	0.00
Heart failure	96,207 (52.0)	9055 (54.1)	87,152 (51.8)	<0.0001	4.59
Depression	14,112 (7.6)	1184(7.1)	12,928 (7.7)	0.0045	−2.34
Overweight/obese	12,636 (6.8)	1122 (6.7)	11,514(6.8)	0.4895	−0.56
Sleep apnea	7201 (3.9)	700 (4.2)	6501 (3.9)	0.0428	1.62
Complications during index admission
Cardiac arrest	4253 (2.3)	328 (2.0)	3925 (2.3)	0.0021	−2.58
Ventricular arrhythmia	14,688 (7.9)	1600 (9.6)	13,088 (7.8)	<0.0001	6.32
Cardiogenic shock	5924 (3.2)	401 (2.4)	5523 (3.3)	<0.0001	−5.34
Renal failure	30,761 (16.6)	2273 (13.9)	28,488 (16.9)	<0.0001	−9.34
Pneumonia	7612 (4.1)	521 (3.1)	7091 (4.2)	<0.0001	−5.87
Procedures during index admission
Cardiac catheterization	97,806 (52.9)	8596 (51.4)	89,210 (53.0)	<0.0001	−3.35
Stent placement	51,157 (27.7)	3614(21.6)	47,543 (28.3)	<0.0001	−15.47
CABG	65,828 (35.6)	5029 (30.0)	60,799 (36.1)	<0.0001	−12.98
Pacemaker implantation	2881 (1.6)	478 (2.9)	2403 (1.4)	<0.0001	9.87
Socioeconomic factors
Receives low income subsidy	11,070 (6.0)	1065 (6.4)	10,005 (6.0)	0.0308	1.73
Eligible for Medicaid	68,476 (37.0)	5491 (32.8)	62,985 (37.4)	<0.0001	−9.72
Medications
Beta blockers (180 days pre-AMI)	94,240 (51.0)	9866 (59.0)	84,374 (50.2)	<0.0001	17.71
ACE inhibitors (180 days pre-AMI)	69,480 (37.5)	6464 (38.6)	63,016 (37.5)	0.0032	2.39
ARB (180 dayspre-AMI)	34,371 (18.6)	3225 (19.3)	31,146 (18.5)	0.0170	1.92
Clopidogrel (180 days pre-AMI)	34,961 (18.9)	2662 (15.9)	32,299 (19.2)	<0.0001	−8.67
Warfarin (90 days post-AMI)	26,083 (14.1)	8051 (48.1)	18,032 (10.7)	<0.0001	89.95
Anti-arrhythmias (90 days post-AMI)	18,438 (10.0)	4499 (26.9)	13,939 (8.3)	<0.0001	50.36
Measures of negative health outcomes 3 years post-index
AMI	29,032 (15.7)	2560 (15.3)	26,472 (15.7)	0.1338	−1.22
Stroke	10,579 (5.7)	1385 (8.3)	9194 (5.5)	<0.0001	11.12
Death	81,280 (43.9)	7922 (47.3)	73,358 (43.6)	<0.0001	7.48

Fig. 1a–d present the Kaplan-Meier curves for the unadjusted time to the composite primary outcome as well as the individual components of the primary outcome - recurrent AMI, stroke, and death, respectively, stratified by AF status. As can be seen in Fig. 1a, patients with AF experienced lower probability of survival without recurrent AMI or stroke compared to the non-AF cohort. Patients with AF also experienced lower probability of stroke-free survival and overall survival compared to the non-AF cohort (Fig. 1c–d). However, there was no difference in the probability of survival without recurrent AMI between the AF and non-AF cohort. The significant differences in baseline characteristics between the AF and non-AF cohort make it difficult to draw conclusions from the unadjusted Kaplan-Meier estimates.

Fig. 1.

Kaplan-Meier curves displaying the relationship between the number of days post-index discharge and the probability of (a) recurrent AMI- and stroke-free survival (b) recurrent AMI-free survival (c) stroke-free survival, and (d) survival, respectively, stratified by AF status.

After multivariable adjustment using Cox proportional hazards regression models, we found that AF was associated with a significant increase in the probability of primary outcome - recurrent AMI, stroke or death (HR = 1.10; 95 % CI: (1.08, 1.13); p < 0.0001, Table 2). The above risk was consistent in analyses evaluating each of the individual outcomes separately - recurrent AMI (HR = 1.10; 95 % CI: (1.05, 1.15); p = 0.0003), stroke (HR = 1.31; 95 % CI: (1.22, 1.39); p < 0.0001) and all-cause death (HR = 1.09; 95 % CI: (1.06, 1.12); p < 0.0001). The full model results are presented in Table 2. A full list of covariates can be found in Appendix A.

Table 2.

The hazard ratio and its 95 % CI and associated p-value for selected predictors derived from a Cox proportional hazards regression models for each of the three outcomes of interest. Each of the regression models control for all the covariates listed in the Methods section.

Predictor	HR (95 % CI)	p-Value	HR (95 % CI)	p-Value	HR (95 % CI)	p-Value	HR (95 % CI)	p-Value
	Time to AMI, stroke or death		Time to AMI		Time to stroke		Time to death
AF	1.10 (1.08, 1.13)	<0.0001	1.10 (1.05, 1.15)	<0.0001	1.31 (1.22, 1.39)	<0.0001	1.09 (1.06, 1.12)	<0.0001
Male	1.13 (1.11, 1.14)	<0.0001	1.12 (1.09, 1.15)	<0.0001	0.85 (0.82, 0.89)	<0.0001	1.15 (1.13, 1.17)	<0.0001
Black race (Ref = white)	1.01 (0.99, 1.04)	0.3747	1.02 (0.97, 1.06)	0.4718	1.43 (1.34, 1.52)	<0.0001	0.99 (0.97, 1.02)	0.4371
Hispanic race (Ref = white)	0.91 (0.88, 0.94)	<0.0001	0.96 (0.91, 1.01)	0.1434	1.19 (1.09, 1.29)	<0.0001	0.87 (0.84, 0.90)	<0.0001
Other/unknown race (Ref = white)	0.90 (0.87, 0.94)	<0.0001	0.98 (0.92, 1.05)	0.5677	1.18 (1.06, 1.32)	0.0031	0.87 (0.83, 0.91)	<0.0001
Age 71 to 75 (baseline = 66 to 70)	1.11 (1.08, 1.13)	<0.0001	1.06 (1.02, 1.10)	0.0035	1.11 (1.04, 1.19)	0.0030	1.15 (1.12, 1.18)	<0.0001
Age 76 to 80 (baseline = 66 to 70)	1.26 (1.23, 1.29)	<0.0001	1.09 (1.05, 1.13)	<0.0001	1.22 (1.14, 1.30)	<0.0001	1.38 (1.34, 1.41)	<0.0001
Age 81 to 85 (baseline = 66 to 70)	1.48 (1.44, 1.51)	<0.0001	1.19 (1.14, 1.24)	<0.0001	1.36 (1.28, 1.46)	<0.0001	1.67 (1.62, 1.71)	<0.0001
Age over 85 (baseline = 66 to 70)	1.90 (1.85, 1.94)	<0.0001	1.41 (1.35, 1.47)	<0.0001	1.47 (1.37, 1.58)	<0.0001	2.25 (2.19, 2.31)	<0.0001
Baseline comorbidities
Hypertension	0.95 (0.92, 0.98)	0.0036	0.98 (0.93, 1.04)	0.5708	1.13 (1.04, 1.24)	0.0060	0.91 (0.88, 0.95)	<0.0001
Diabetes mellitus	0.98 (0.96, 1.01)	0.1321	1.06 (1.01, 1.10)	0.0109	0.98 (0.92, 1.05)	0.6109	0.96 (0.93, 0.98)	<0.0001
Chronic kidney disease	1.10 (1.08, 1.12)	<0.0001	1.15 (1.10, 1.19)	<0.0001	1.03 (0.97, 1.10)	0.3436	1.10 (1.08, 1.13)	<0.0001
Stroke	1.09 (1.06, 1.11)	<0.0001	0.86 (0.82, 0.91)	<0.0001	1.58 (1.48, 1.68)	<0.0001	1.11 (1.08, 1.14)	<0.0001
Heart failure	1.27 (1.25, 1.29)	<0.0001	1.27 (1.24, 1.31)	<0.0001	1.10 (1.05, 1.15)	0.0001	1.33 (1.31, 1.36)	<0.0001
Depression	1.03 (1.00, 1.05)	0.0357	0.96 (0.91, 1.00)	0.0550	0.97 (0.90, 1.05)	0.4446	1.05 (1.02, 1.08)	0.0004
Overweight/obese	0.89 (0.86, 0.91)	<0.0001	0.92 (0.88, 0.96)	0.0004	0.90 (0.83, 0.97)	0.0064	0.86 (0.84, 0.89)	<0.0001
Sleep apnea	0.93 (0.90, 0.96)	<0.0001	0.93 (0.88, 0.99)	0.0219	1.00 (0.90, 1.10)	0.9505	0.92 (0.89, 0.96)	<0.0001
Stent placementa	1.34 (1.30, 1.38)	<0.0001	1.60 (1.52, 1.38)	<0.0001	1.11(1.03, 1.20)	0.0095	1.30 (1.26, 1.35)	<0.0001
CABGa	0.56 (0.55, 0.58)	<0.0001	0.49 (0.47, 0.52)	<0.0001	0.77 (0.71, 0.82)	<0.0001	0.55 (0.53, 0.57)	<0.0001
			Medications
Beta blockers	1.00 (0.98, 1.01)	0.6085	1.09 (1.06, 1.12)	<0.0001	1.08 (1.03, 1.12)	0.0008	0.96 (0.95, 0.98)	<0.0001
ACE inhibitors	0.98 (0.97, 1.00)	0.0132	1.03 (1.00, 1.05)	0.0453	0.99 (0.95, 1.04)	0.7697	0.97 (0.95, 0.98)	<0.0001
ARB	0.91 (0.89, 0.92)	<0.0001	0.97 (0.94, 1.00)	0.0830	1.02 (0.97, 1.07)	0.5560	0.88 (0.86, 0.89)	<0.0001
Clopidogrel	1.06 (1.05, 1.08)	<0.0001	1.19 (1.16, 1.23)	<0.0001	1.17 (1.11, 1.23)	<0.0001	1.03 (1.01, 1.05)	0.0029
Warfarinb	0.84 (0.82, 0.85)	<0.0001	0.90 (0.86, 0.93)	<0.0001	1.07 (1.00, 1.14)	0.0409	0.79 (0.77, 0.81)	<0.0001

^aProcedure during index admission.

^bPresent 90 days after index; HR, hazard ratio; CI, confidence interval; ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker.

Based on results not shown, for each of the studied outcomes, interactions between AF and sex, race and age were not significant. Therefore, only the main effects of these variables are listed in Table 2. In addition to AF, a number of patient characteristics were associated with an increased risk of adverse outcomes. Males had a significantly higher hazard of suffering the primary outcome (HR = 1.13; 95 % CI: (1.11, 1.14); p < 0.0001), recurrent AMI (HR = 1.12; 95 % CI: (1.09, 1.15); p < 0.0001) and death (HR = 1.15; 95 % CI: (1.13, 1.17); p < 0.0001) but lower hazard of stroke (HR = 0.85; 95 % CI: (0.82, 0.89); p < 0.0001). Compared to their white counterparts, Hispanics were less likely to suffer recurrent AMI, stroke or death (HR =0.91; 95 % CI: (0.88, 0.94); p < 0.0001) and death (HR = 0.87; 95 % CI: (0.84, 0.90); p < 0.0001), but were more likely to suffer a stroke (HR = 1.19; 95 % CI: (1.09, 1.29); p < 0.0001). Similarly, blacks were more likely to suffer a stroke (HR = 1.43; 95 % CI: (1.34, 1.52); p < 0.0001) compared to whites. Finally, age was associated with an increased risk of each of the studied outcomes. For instance, controlling for all other covariates in the model, we estimate that the risk of stroke for patients over 85 years of age is 1.47 (95 % CI: (1.37, 1.58); p < 0.0001) times that of patients 66 to 70 years of age.

4. Discussion

In this analysis of a nationwide cohort of Medicare patients in the United States who had been discharged alive after AMI, we report the prognostic effect of AF on cardiovascular outcomes and mortality. After a mean follow up of 26.4 months post-AMI, subjects with AF had significantly higher incidence of a composite of cardiovascular events (stroke and recurrent AMI) and all-cause mortality. A summary of the findings of this study can be seen in the central illustration (Fig. 2). This is one of the largest studies analyzing AF in post-AMI patients, and the first in the Medicare population. AF prevalence was 9.1 % in our study population, which is within the range reported by a prior systematic review (6–21 %) [2].

Fig. 2.

Central Illustration summarizing the methods of the study, Kaplan-Meier curves of primary and secondary end-points and Cox regression analysis focusing on effects atrial fibrillation on each of the outcomes of interest.

Similar to our findings, previous analyses of large registries in Sweden and Denmark found that that AF was associated with an increase in a composite of cardiovascular events and mortality [5,11]. Batra et al. found that AF was associated with a 28 % increase in a composite of cardiovascular disease (stroke, AMI) and mortality among survivors of AMI after 90 days follow-up [11]. Bang et al. studied new-onset AF among survivors of AMI, following them for a median of 5 years [5]. New-onset AF was associated with a 1.9 times increased risk of all-cause mortality, 2.5 times increased risk of non-fatal stroke, and 1.8 times increased risk of non-fatal re-infarction [5]. Furthermore, a number of other studies have shown that the occurrence of new-onset AF after AMI increased the risk of cardiovascular events or death [7,8,10,17].

Our study differs from some previously published data. For example, in the TRILOGY ACS trial, baseline AF did not predict a composite of cardiovascular events and death [18]. Consuegra et al. found that neither chronic AF nor new-onset AF predicted long-term mortality after ST elevation AMI, although new-onset AF did predict in-hospital mortality [19]. Interestingly, Lau et al. found that long-term prognosis of new-onset AF patients appeared to be inversely proportional to severity of acute coronary syndrome [12]. An Effect modification by race on the risk for mortality particularly in African American patients with AF was previously described [20]. In this study, the effect of AF on mortality and cardiovascular outcomes was not significantly affected by sex, race or age.

Anticoagulation has a key role in minimizing the risk for mortality and cardiovascular events in patients with AF [21]. In the current study, while most of the patients with AF were above 75, only around half of those with newly diagnosed AF received anticoagulation. Such a low rate of anticoagulation use might have contributed to the increased risk for adverse outcomes in patients with AF.

Studies that have analyzed different subtypes of AF (new-onset, paroxysmal or chronic AF) have reported inconclusive findings. Saito et al. found that new-onset, but not chronic AF, was associated with a relative risk of recurrent AMI of 3.16 in the long term [8]. Contrarily, Lau et al. reported that previous AF increased the risk of long-term mortality while new-onset AF increased the risk of in-hospital mortality [12]. This finding suggests that new-onset AF may be a marker of severity of acute illness in AMI patients, thus the higher in-hospital mortality. We could not evaluate this finding in our study since we only included patients who were discharged alive. Among survivors of AMI, Batra et al. found no difference in outcomes by AF subtype after a 3 month follow-up [11]. We did not distinguish AF by subtype in our study.

Major causes of AF-related deaths following AMI are linked to the development of heart failure as well as increased risk of sudden cardiac death (SCD) [22]. AF can directly facilitate and lower the threshold for ventricular arrhythmias, leading to SCD [23,24]. Patients with AF have a higher prevalence of underlying comorbidities including older age, hypertension, hypotension, diabetes mellitus, previous AMI, ventricular arrhythmias and heart failure [6,7]. Whether the increased mortality associated with AF in post-AMI patients is related to the impact of AF, or to the effect of other comorbidities and the underlying myocardial injury caused by AMI, remains an active research question [25–27]. In results not shown, we found that there was no difference in the top 5 causes of death between the AF and non-AF cohort. Thus, while AF may have some effect on the probability of death and other outcomes, these results suggest that AF may not have a large impact on the cause of death.

Mechanisms that promote AF in the setting of MI are complex and multifactorial, and our understanding of the pathophysiology remains incomplete. Potential mechanisms that have been implicated include atrial infarction or ischemia, increased atrial pressures, inflammation, pericarditis, elevated catecholamine, autonomic nervous system changes and metabolic abnormalities [28–30]. Multiple studies have demonstrated the role of acute left atrial ischemia in the etiology of AF in AMI patients [31,32]. MI can lead to atrial fibrosis which affects intercellular conduction and promotes AF vulnerability [33]. Elevated atrial pressure causes increased atrial stretch which shortens the atrial refractory period and significantly increases the vulnerability to AF [34]. AF may also develop from ventricular dysfunction in the setting of AMI as atrial pressure rises secondary to elevated left ventricular end-diastolic pressure [35].

5. Limitations

Our study has certain limitations. Firstly, the study was conducted between January 1, 2007 and December 31, 2008. Less than half of the study population were not on anticoagulation. Additionally, due to the timing of the study, it was not possible to measure the impact of use of direct oral anticoagulants (DOACs) on AF related cardiovascular complications. Given the retrospective nature of the study, there is the possibility that our results were affected by confounding. However, the large sample size allowed us to control for a large number of possible confounders related to patient demographics, comorbidities, drugs and treatments, and socioeconomic status. Secondly, our study included only patients >65 years old, thus the findings cannot be generalized to the entire population. Since AF is much more common in the older population, we believe that our study is clinically relevant. Thirdly, inclusion of Medicare part D (prescription benefit plan) enrollees only could impact generalizability if such patients are systematically different than the Medicare beneficiaries who do not enroll in a prescription benefit plan. Moreover, beneficiaries with prescription drug coverage may also have greater ability to adhere to prescription medications compared to patients without prescription coverage. Furthermore, considering that our study sample was an administrative database, we lacked granular details such as left ventricular ejection fraction, creatinine, hemoglobin, or international normalized ratio/time in therapeutic range in warfarin users. The use of ICD codes may have led to significant inaccuracies in estimation of diagnoses and events. To improve accuracy, we used validated algorithms whenever possible, for example, in determining AF, AMI or stroke diagnoses (See Appendix A for definitions). Finally, since we only assessed survivors of AMI, there may have been a selection bias in this group compared to all AMI patients with AF. A previous study found that new-onset AF was associated with increased in-hospital mortality among AMI patients [12]. If this holds true in our population, then we may have underestimated the effect of AF on outcomes.

6. Conclusion

In patients who survive an AMI, AF is associated with increased risk of cardiovascular events and all-cause mortality. Further studies are needed to understand mechanisms by which AF alters outcomes in post-AMI patients. Our study results may help clinicians better risk-stratify patients who survive an AMI.