Development of interpretable machine learning models to predict in-hospital prognosis of acute heart failure patients.

AIMS: In recent years, there has been remarkable development in machine learning (ML) models, showing a trend towards high prediction performance. ML models with high prediction performance often become structurally complex and are frequently perceived as black boxes, hindering intuitive interpretation of the prediction results. We aimed to develop ML models with high prediction performance, interpretability, and superior risk stratification to predict in-hospital mortality and worsening heart failure (WHF) in patients with acute heart failure (AHF). METHODS AND RESULTS: Based on the Kyoto Congestive Heart Failure registry, which enrolled 4056 patients with AHF, we developed prediction models for in-hospital mortality and WHF using information obtained on the first day of admission (demographics, physical examination, blood test results, etc.). After excluding 16 patients who died on the first or second day of admission, the original dataset (n = 4040) was split 4:1 into training (n = 3232) and test datasets (n = 808). Based on the training dataset, we developed three types of prediction models: (i) the classification and regression trees (CART) model; (ii) the random forest (RF) model; and (iii) the extreme gradient boosting (XGBoost) model. The performance of each model was evaluated using the test dataset, based on metrics including sensitivity, specificity, area under the receiver operating characteristic curve (AUC), Brier score, and calibration slope. For the complex structure of the XGBoost model, we performed SHapley Additive exPlanations (SHAP) analysis, classifying patients into interpretable clusters. In the original dataset, the proportion of females was 44.8% (1809/4040), and the average age was 77.9 ± 12.0. The in-hospital mortality rate was 6.3% (255/4040) and the WHF rate was 22.3% (900/4040) in the total study population. In the in-hospital mortality prediction, the AUC for the XGBoost model was 0.816 [95% confidence interval (CI): 0.815-0.818], surpassing the AUC values for the CART model (0.683, 95% CI: 0.680-0.685) and the RF model (0.755, 95% CI: 0.753-0.757). Similarly, in the WHF prediction, the AUC for the XGBoost model was 0.766 (95% CI: 0.765-0.768), outperforming the AUC values for the CART model (0.688, 95% CI: 0.686-0.689) and the RF model (0.713, 95% CI: 0.711-0.714). In the XGBoost model, interpretable clusters were formed, and the rates of in-hospital mortality and WHF were similar among each cluster in both the training and test datasets. CONCLUSIONS: The XGBoost models with SHAP analysis provide high prediction performance, interpretability, and reproducible risk stratification for in-hospital mortality and WHF for patients with AHF.

Impact of catheter ablation on functional tricuspid regurgitation in patients with atrial fibrillation.

BACKGROUND: Atrial fibrillation (AF) plays the main role in atrial functional tricuspid regurgitation (TR). However, the effectiveness of catheter ablation (CA) for atrial functional TR together with the mechanisms of improvement of atrial functional TR have not been fully evaluated. METHODS: We retrospectively investigated consecutive 2685 patients with AF who received CA from February 2004 to December 2019 in Kyoto University Hospital, Kyoto, Japan. The current study population consisted of 2331 patients with available transthoracic echocardiographic (TTE) data before CA (2110 patients without significant TR and 221 patients with significant TR). Among the 221 patients with significant TR, there were 64 patients with functional TR and follow-up TTE at 6-18 months after CA for AF, in whom we compared echocardiographic parameters from baseline to follow-up. RESULTS: Patients with significant TR were older, and more often women, and had more persistent AF than those without significant TR. Among the 64 patients with functional TR, TR severity and TR jet area significantly improved at follow-up (TR jet area: 5.8 [4.0-7.6] cm2 to 2.1 [1.1-3.1] cm2, P < 0.001). Moreover, mitral regurgitation jet area, left atrial area, mitral valve diameter, right ventricular end-diastolic area, right atrial area, and tricuspid valve diameter decreased at follow-up. CONCLUSIONS: TR severity and jet area improved after CA in patients with AF and significant TR. The improvement of TR might be associated with reverse remodeling of the right heart.

Prognostic impact of catheter ablation in patients with asymptomatic atrial fibrillation.

BACKGROUND: Catheter ablation for asymptomatic atrial fibrillation (AF) remains controversial. The aim of the present study was to explore the prognostic impact of catheter ablation in asymptomatic AF patients. METHODS: We performed a post-hoc analysis of 537 risk-matched pairs of AF patients receiving first-time catheter ablation or conservative management. The primary outcome measure was a composite of cardiovascular death, heart failure (HF) hospitalization, ischemic stroke, or major bleeding. The study patients were divided into asymptomatic and symptomatic patients, and were further divided according to the presence or absence of previous AF-related complications (ischemic stroke or HF hospitalization). RESULTS: Most baseline characteristics were well balanced between the catheter ablation versus conservative management groups. The median follow-up period was 5.3 years. Catheter ablation as compared to conservative management was associated with significantly lower incidence of the primary outcome measure in the asymptomatic AF patients (14.7% versus 25.4% at 8-year, log-rank P = 0.008). However, the advantage of catheter ablation was significant only in the high-risk subset of patients with the previous AF-related complications (19.2% versus 55.6% at 8-year, log-rank P = 0.006), but not in those without (13.9% and 17.3%, P = 0.08). On the other hand, among the symptomatic AF patients, catheter ablation was associated with significantly lower incidence of the primary outcome measure regardless of the previous AF-related complications. CONCLUSIONS: In the post-hoc analysis of the matched AF cohort, catheter ablation as compared with conservative management was associated with better long-term clinical outcomes among asymptomatic AF patients only when the previous AF-related complications were present.

Matched comparison of catheter ablation versus conservative management for atrial fibrillation.

It is still controversial whether catheter ablation for atrial fibrillation (AF) could improve clinical outcomes in general AF population. Among 4398 patients with diagnosis of AF in the outpatient department of Kyoto University Hospital between January 2005 and March 2015, we identified 537 pairs of patients who received first-time catheter ablation (ablation group) or conservative management (conservative group), matched for age, gender, AF duration, AF type, AF symptoms, and previous heart failure (HF). The primary outcome measure was a composite of cardiovascular death, HF hospitalization, ischemic stroke, or major bleeding. Most baseline characteristics were well balanced between the 2 groups, except for the higher prevalence of low body weight, history of malignancy, and severe chronic kidney disease in the conservative group. Median follow-up duration was 5.3 years. The cumulative 5-year incidence of the primary outcome measure was significantly lower in the ablation group than in the conservative group (5.2% versus 15.6%, log-rank P < 0.001). Even after adjusting for the imbalances in the baseline characteristics, the lower risk of the ablation group relative to the conservative group for the primary outcome measure remained highly significant (HR 0.32, 95% CI 0.21-0.47, P < 0.001). Ablation compared with conservative management was also associated with significantly lower risks for the individual components of the primary outcome. In this matched analysis in AF patients, ablation as compared with conservative management was associated with better long-term clinical outcomes, although we could not deny the possibility of selection bias and unmeasured confounding.

Inferior vena cava thrombus due to hyperhomocysteinemia.

We report a case of a 39-year-old man with inferior vena cava (IVC) thrombus after withdrawal of IVC filter. He complained of 1-week history of swelling in his right leg, and computed tomography (CT) demonstrated occlusive thrombus in right femoral and popliteal veins. Venous thrombectomy was attempted after IVC filter placement. The filter was withdrawn without IVC filter-associated thrombus and the patient was discharged on warfarin. On day 21 after discharge, he complained of progressive swelling and CT revealed a significant amount of thrombus in IVC where the filter had been placed. Despite adequate anticoagulant therapy, CT showed increased thrombus extending from infra-renal IVC to common iliac vein. Systemic work-up for the risk factors of venous thromboembolism revealed the presence of hyperhomocysteinemia (total homocysteine 83.1 µmol/L; normal range 5-15 µmol/L) due to unbalanced diet with deficiency of folic acid and vitamin B12. Four weeks after administration of folic acid, vitamin B6/B12 as the treatment of hyperhomocysteinemia in addition to warfarin, swelling of his leg improved with significant resolution of thrombus by CT. CT also demonstrated a severe stenosis of the IVC just below the renal veins. At 6 months, he complained of mild post-thrombotic sequelae with no evidence of recurrent IVC thrombus. .