LA3CK Score for Predicting in-Hospital Mortality after Minor Bleeding in Patients with Atrial Fibrillation

Backgrounds and Purposes Minor bleeding was found to be closely related with death in in-hospital patients with atrial brillation (AF (cid:0) and the factors affecting mortality after minor bleedings were then explored. Methods This study was divided into two parts. First, 15219 AF patients without major bleeding were included at their rst hospitalizations from our centers since 2008 to 2018 to explore the association of minor bleeding with in-hospital mortality. Then factors related with in-hospital death were further explored in 789 AF patients with minor bleeding but without major bleeding and joined into one score predicting mortality after the minor bleedings during hospital stay. Results Patients with minor bleedings were at high risk of mortality (Adjusted OR:4.61, 95% CI: 3.77-5.65, p<0.001), especially if they were older (Adjusted OR:1.05, 95% CI: 1.03-1.06, p<0.001), or complicated with chronic lung diseases, chronic kidney disease stage III-V, anemia, malignant tumor, acute myocardial infarction. The c-index was 0.681 using ROC curve when these risk factors were joined into the score LA3CK. The c-index was 0.81 when this score was extended to the whole in-hospital AF population. Conclusions In the in-hospital AF patients, risk of mortality increased in patients with minor bleeding. The score LA3CK score had a good eciency in predicting in-hospital mortality after minor bleedings.


Introduction
Atrial brillation (AF) is a common clinical arrhythmia with ageing. 1 Bleedings often accompany with AF, which are caused not only by adverse effects of antithrombotic drugs used for preventing stroke, 2 but also the complications easily leading to risk of bleeding, such as lung diseases, active malignancy, etc.
Non-major bleeding and minor bleeding were shown to be associated with subsequent clinical events, such as risk of death, ischemic stroke and myocardial infarction. [3][4][5] However, factors affecting mortality after minor bleedings were unclear. The physicians were still confused which kind of patients had higher risk of mortality after minor bleeding. We sought to explore the factors that have impact on the in-hospital mortality after minor bleedings in AF patients using data from clinical practice after con rming the relationship of minor bleeding and mortality during hospital stay. Then these factors were joined into one score predicting mortality after the minor bleedings in the in-hospital AF patients.

Study Patients
Selection process of the data was as previously described. 6  This study was divided into two parts. First, after excluding 469 patients with major bleeding, 15219 AF patients were divided into two groups, minor bleeding group (N = 789) and no minor bleeding group (N = 14430). The association of minor bleeding with in-hospital mortality was explored in the whole population. Then further exploration of factors related with in-hospital mortality was performed in 789 AF patients with minor bleeding but without major bleeding.

De nitions of Bleeding
The corresponding copy of the inpatient records were obtained to con rm the diagnoses of any bleeding events, which were composite of major bleeding and minor bleeding. Major bleeding was de ned as described in previous studies. 7,8 (1). Reduction in haemoglobin level of ≥ 2 g/L within 24 hours; (2).
Transfusion of ≥ 2 units (U) of blood during the whole hospital stay; (3). Symptomatic bleeding in a critical area or organ; (4). Life-threatening bleeding (fatal bleeding, bleeding with a total decrease in haemoglobin level of ≥ 5 g/L, bleeding requiring inotropic agents or necessitating surgery). Bleeding that did not satisfy the de nition of major bleeding was considered as minor bleeding.

Statistical Analysis
Data were expressed using mean ± SD for continuous variables and number (percentage %) for categorical variables. Continuous and categorical variables were compared using Chi-square test and unpaired Student's T test, respectively. Number of deaths in patients with minor bleeding was calculated.
Logistic regression analyses using stepwise method were used to explore the association of minor bleeding with risk of mortality. Thereafter, factors affecting mortality after minor bleeding were explored using univariate and multivariate logistic regression in the minor bleeding group. Factors used in the multivariate analysis included sex, age, hypertension (HTN), diabetes mellitus (DM), dyslipidemia, heart failure (HF) history, acute myocardial infarction (AMI), anemia, chronic lung diseases (CLD), chronic kidney disease (CKD) stage III-V, which was de ned according to The National Kidney Foundation/Kidney Disease Outcome Quality Initiative (NKF/K-DOQI), stroke/transient ischemic attack (TIA), malignant tumor (including active malignancy and malignant history). At last, factors with signi cant impact on the mortality after minor bleeding were joined into one score and the c-index was explored using ROC curve. P < 0.05 was considered as statistical signi cance. All statistical analyses were performed using SPSS 23.0 software (SPSS Inc., Chicago, IL, USA).

Page 4/11
Baseline characteristics of the AF patients were shown in Table 1. Of these patients, 789 patients experienced minor bleeding and their distribution location sites were listed in Fig. 1. Patients who had minor bleeding were more likely to be older and had stroke/TIA history, dyslipidemia, DM history, HF history, AMI, CKD stage III-V, major bleeding history, CLD history, malignancy and anemia compared to those without minor bleeding. Of the patients with minor bleeding, the comparison of those died and survived was listed in Table 1, as well.   LA3CK score These risk factors were joined into one simple LA3CK score (Table 4). L: Chronic lung disease (1 point); A3: Age ≥ 80 years old /Acute myocardial infarction/Anemia (1 point each); C: cancer (including both malignant tumor history and active malignant tumor) (1 point); K: chronic kidney dysfunction (1 point). Total score ranged from 0 to 6 scores. The c-index of LA3CK score was 0.681 (95% CI:0.641-0.721) using ROC curve. (Fig. 2A). Patients with score ≥ 2 increased the risk of mortality by 4-fold after minor bleeding compared to those with score < 2 according to logistic regression. When we extend this score in the whole in-hospital AF population, we found the c-index of 0.811. (Fig. 2B).

Discussion
Risk of mortality increased after minor bleeding in the in-hospital AF patients. Factors increased risk of mortality after minor bleeding included CLD, old age ≥ 80 years old), AMI, anemia, malignant tumor and CKD stage III-V. The c-index was 0.681 (95% CI:0.641-0.721) using ROC curve when these risk factors were joined into the LA3CK score.
Most studies focused on the increased risk of clinical outcomes caused by major bleeding. 3,9 Actually, the minor bleeding was 1.6 times common compared with major bleeding in the current population, though less common compared with other trials testing oral anticoagulation. 4 Similar to the results from the randomized control trial data, 4 the minor bleeding was independently associated with subsequent risk of death using the data from clinical practice in AF patients. This study was performed after excluding the patients with major bleeding no matter if the major bleeding occurred following minor bleeding, thus we believed the possibility of impact caused by major bleeding was minimized to the utmost.
An increased risk of mortality after minor bleeding was found with ageing. Therefore, we believed that more attention should be taken for those with higher age, even if they had minor bleeding. Besides, the bleeding risk related with complications should be taken more seriously, as most of the bleedings were

Conclusion
In the in-hospital AF patients, risk of mortality increased in patients with minor bleeding. The LA3CK score had a good e ciency in predicting in-hospital mortality after minor bleedings.
Declarations caused by the complicated diseases (ie. hemoptysis caused by bronchiectasis, ocular hemorrhage related with hypertension or DM, tumor related bleeding, etc).
The risk factors derived from this exploratory analysis were involved into the LA3CK score, showing a good e ciency in predicting death after minor bleeding. This score was mainly constituted of underlying complicated diseases. Therefore, the score could be promptly deduced by a careful clinical evaluation, as only anemia caused by hemoglobin decrease was derived from laboratory examinations, which was also very important and necessary for diagnosing minor bleeding. Though unstandardized anticoagulation therapy was con rmed increasing risk of mortality, this factor changed a lot with the treatment process, which made it di cult to be involved into a score. The little need for instrumental and laboratory evaluation to calculate the individual risk of death facilitates the physicians and patients for clinical decisions. Furthermore, it would be helpful for risk strati cation for predicting risk of death after minor bleeding using this speci cally derived calibration methods.

Limitations
There are several limitations to be mentioned for the current analysis. First, this is a retrospective analysis collecting information from three central hospitals in Beijing. It is prudential to extend the results from the present study to a more extensive community because the dataset represented the selective in-hospital population. Second, de nitions of variables were retrospectively reviewed, but were considered strictly built as uniform diagnosis and treatment guidelines were used in clinical practice in the Beijing District. Third, minor bleeding preadmitting into the hospitals was usually neglected by the patients, therefore, only major bleeding history was available for these patients. However, minor bleeding was particularly described in the records, and the minor bleeding information in hospital was accurate and believable. Fourth, mild to moderate anemia was usually not recorded in the case reports, therefore, the anemia information for those with mild to moderate anemia was not fully collected due to the data collection method which was mainly dependent on the diagnosis from the case reports. Besides, it is hard to differentiate the reason of anemia for all the in-hospital patients, as some may be caused by in-hospital bleeding and some may have long anemic history. Fifth, it is necessary to point out that tuberculosis was not included in the CLD due to diversity of the disease and lung cancer was not included in the CLD to avoid duplication recording with the malignant tumor. Sixth, malignant tumor included both malignant tumor history and active malignant. Above all, information from the case reports was mostly comprehensive and accurate which increased the robustness of the study. Seventh, this study mainly captured events that occurred during hospitalization. Therefore, whether exploration or generalization of in-hospital data could be made needed further exploration.