From January 2000 to December 2009, a 10-year period, we retrospectively recruited all individuals with a principal discharge diagnosis of DHF (International Classification of Disease, Ninth Revision, codes: 428.0, 428.1, 428.20, 428.21, 428.22, 428.23, 428.30, 428.31, 428.32, 428.33, 428.41, 428.43, 428.9), included in the INCA study,23 AMI (410.01, 410.10, 410.11, 410.12, 410.21, 410.31, 410.41, 410.51, 410.70, 410.71, 410.72, 410.81, 410.82, 410.91), included in the CASTUO study,24 and stroke (433.0, 433.01, 433.10, 433.11, 433.21, 433.30, 434.00, 434.01, 434.10, 434.11, 434.90, 434.91, 436), included in the ICTUS study25 and who were consecutively admitted to the district hospital. For the present analysis, we selected the 3 diagnoses (DHF, stroke, and AMI) constituting the most frequent causes of admission for cardiovascular disease.26,27 The information sources were the hospital coding service and the Spanish National Death Index. Mortality data were obtained from this index in 2011 for 100% of patients. This study was approved by the appropriate ethics committee. The only patients excluded were those with DHF who also had severe mitral or aortic valve disease.

Atrial fibrillation diagnosed via an electrocardiogram (ECG) at admission or during hospitalization meeting the following characteristics: irregular R-R interval, absence of P waves, and variable atrial cycle length. Atrial fibrillation was considered present if it was noted as a secondary diagnosis in the discharge sheet–independently of the type or time of onset–and was additionally confirmed in the discharge ECG report. Chronic kidney disease (CKD) was defined as the presence of a glomerular filtration rate less than 60 mL/min/1.73 m2 for at least 3 months or its previous diagnosis in the medical records. Chronic obstructive pulmonary disease (COPD) was considered to be present in patients with a previous diagnosis of emphysema or chronic bronchitis. Peripheral arterial disease was considered to be present if it was previously diagnosed in patients’ medical records. The main study variables were in-hospital and postdischarge all-cause mortality. The median postdischarge follow-up was 6.2 years (interquartile range, 3.9-8.8).

Variables associated with in-hospital all-cause mortality were studied using relative risk (RR) and its corresponding 95% confidence interval (95%CI) in a binary logistic regression model adjusted for age, sex, hypertension, diabetes mellitus, hypercholesterolemia, active smoking, history of stroke, COPD, CKD, and peripheral arterial disease, and discharge diagnosis. In addition, model calibration was evaluated using the Hosmer-Lemeshow test and discrimination using the C statistic. Variables associated with postdischarge all-cause mortality were studied using hazard ratios (HRs) in a Cox regression model adjusted for various risk factors (age, sex, hypertension, diabetes mellitus, hypercholesterolemia, active smoking, history of stroke, COPD, CKD, and peripheral arterial disease, and discharge diagnosis). HRs are presented with their corresponding 95%CIs. The “Enter” method was chosen in both models. In addition, in the hierarchical model, the first-degree interaction between AF and discharge diagnosis (DHF, AMI, or stroke) was studied using likelihood ratio statistics and the backward elimination method (chunk test). The proportional hazards assumption was tested for the analyses. The incidence rates were calculated with Kaplan-Meier survival analysis and were estimated using the !COI V 2008.02.29 JM Domenech macro (Autonomous University of Barcelona). The distributions were compared between the groups using the log-rank test. SPSS version 20 statistical software (IBM, United States) was used for the analyses.

The study population included 6613 patients (2177 with AMI, 2208 with DHF, and 2228 with stroke) with a mean age of 73.3 years. In total, 1686 patients (25.5%) had AF. Atrial fibrillation was significantly more common in patients with DHF than in those with AMI or stroke. The baseline characteristics of the sample according to discharge diagnosis are shown in Table 1. The baseline characteristics of the patients diagnosed with AF vs those without AF are shown in Table 2. Patients with AF were older and had a higher prevalence of comorbidities (hypertension, COPD, and CKD).

There were 819 deaths (12.4%). Patients who died during hospitalization were significantly (all P < .05) older (79 ± 9 years vs 73 ± 11 years) and more likely to be female and have AF (29.6% vs 24.9%), COPD (20.4% vs 17.3%), and CKD (14.3% vs 9.1%). However, they had a significantly lower prevalence of hypercholesterolemia (18.7% vs 46.4%) and active smoking (6.9% vs 14.9%).

In-hospital mortality was higher in patients with AF than in those without this condition (241 [14.3%] vs 572 [11.6%]; P < .001). After stratification according to discharge diagnosis (Table 3), in-hospital mortality was higher in patients admitted for AMI who had AF vs those who did not have AF (16.7% vs 11.4%; P = .017) and in patients admitted for stroke who had AF vs those who did not (20.9% vs 13.6%; P < .001); in contrast, no significant difference was seen in those admitted for DHF (10.0% vs 9.4%; P = .589).

In a well-calibrated adjusted model with high discriminatory power (Table 4), AF was not an independent predictor of mortality (adjusted RR = 1.072; P = .732) in the total cohort or according to discharge diagnosis (AMI: adjusted RR = 1.115; P = .607; DHF: adjusted RR = 1.003; P = .983; stroke: adjusted RR = 1.280; P = .073). There was no significant interaction.

Patients who died during postdischarge follow-up were significantly (all P < .05) older (77 ± 9 years vs 71 ± 12 years) and more likely to be male and have diabetes mellitus (44.5% vs 34.6%), AF (31.0% vs 21.1%), stroke (14.1% vs 12.2%), peripheral arterial disease (5.2% vs 2.4%), COPD (24.5% vs 12.7%), and CKD (14.3% vs 5.7%). However, they had a significantly lower proportion of hypercholesterolemia (35.5% vs 53.4%) and active smoking (9.6% vs 18.2%).

The postdischarge mortality incidence rate was 11.9 (95%CI, 11.0-12.8) per 100 patient-years for patients with AF vs 7.1 (95%CI, 6.8-7.5) for those without AF, giving an incidence rate ratio of 1.66 (95%CI, 1.52-1.82; P < .001). After stratification according to discharge diagnosis (Table 3), patient mortality was higher in those admitted for AMI who had AF than in those who did not have AF (incidence rate of 11.4 vs 4.6 per 100 person-years; P < .001) and in those admitted for stroke who had AF than in those who did not (incidence rate of 12.8 vs 7.0 per 100 person-years; P < .001); there was no significant difference for those admitted for DHF (11.4 vs 12.6; P = .116).

In the adjusted model (Table 5), AF was an independent predictor of mortality (adjusted HR, 1.668; P < .001) in the total cohort, as well as in patients admitted for AMI (adjusted HR, 1.494; P = .001) or stroke (adjusted HR, 1.426; P < .001). However, there was no significant difference in patients admitted for DHF (adjusted HR, 0.964; P = .603), indicating a highly significant interaction. A model that also included the complete medical treatment until discharge showed no differences from that shown in Table 5 (data not shown). In the subgroup of patients with DHF whose left ventricular ejection fraction was known (n = 805), there was no significant interaction between AF and ejection fraction (P for interaction = .331) in terms of mortality.

Figure presents the Kaplan-Meier survival curve and shows that patients with a more unfavorable clinical course (with significant overlap in clinical course among the groups) were those with DHF (independently of AF presence), as well as those with AMI and AF or stroke and AF (log-rank test, P < .001). In contrast, the patients with a significantly more favorable clinical course were (in order) those with AMI without AF and those admitted for stroke without AF.

Figure.

Kaplan-Meier survival analysis according to discharge diagnosis and presence of atrial fibrillation. AF, atrial fibrillation; AMI, acute myocardial infarction; DHF, decompensated heart failure.

(0.11MB).

This study has several limitations. First, during the prolonged study recruitment period, there were important changes in the treatment recommendations that could have influenced the findings. Second, given the retrospective study design, it was not possible to characterize the type of AF or its time of onset. Third, there was no information on left ventricular ejection fraction in more than half of patients and a significant proportion of patients lacked information on DHF etiology. However, due to its large sample size, this study afforded an excellent opportunity to explore the hypothesis of a differential effect of AF on mortality. Fourth, 35% of patients with AF included in this study received anticoagulants. This figure, although suboptimal, is in line with that of other studies highlighting the global problem of undertreatment with anticoagulants.46 In addition, given that our study covers a 10-year period, it must be remembered that this percentage represents a mean. There was a growing trend for anticoagulant prescription (first 2 years vs last 2 years: 26.6% vs 41.4%; P for trend < .001). Finally, because this study is based on a single-center registry conducted in a secondary hospital, the results might not be generalizable to the entire health care system.