CMR-derived left atrial reservoir strain and the LAVI/LVEF ratio are associated with 30-day readmission in atrial fibrillation: the CMRRADAR-AF retrospective cohort study

Abstract

Background Cardiovascular magnetic resonance (CMR) imaging allows simultaneous evaluation of cardiac structure, function, and tissue composition. Whether CMR-derived left atrial (LA) reservoir strain, the left atrial volume index–left ventricular ejection fraction (LAVI/LVEF) ratio, and atrial late gadolinium enhancement (LGE) can identify patients at risk for 30-day readmission after atrial fibrillation (AF)–related hospitalisation has not been investigated. Objectives This study aims to evaluate the association between CMR parameters and 30-day unplanned readmission following an AF-related index admission and to describe CMR differences across three clinical groups using a three-group cohort design. Methods This retrospective cohort study at Chulabhorn Hospital, Bangkok (IRB005/2569), enrolled 43 patients with AF who underwent CMR on 1.5 T or 3 T Philips scanners: 25 outpatients (OPD), 10 index-admitted patients without readmission (Adm-NoReadmit), and 8 index-admitted patients with 30-day unplanned readmission (Readmit). A CMR analysis was performed using CVI42 version 6.4. LA reservoir strain was derived by feature tracking from standard cine images. Results Among the admitted patients (n = 18), LA reservoir strain showed the highest apparent discrimination for readmission [area under the curve (AUC), 0.963; p < 0.001; exploratory internally derived cutpoint <11.1%: sensitivity, 88%; 95% confidence interval (CI), 47%–100%; specificity, 100%; 95% CI, 69%–100%]. Correcting for predictor selection, the optimism-corrected AUC was 0.919. The LAVI/LVEF ratio was also associated with readmission (AUC, 0.913; p = 0.002). Qualitative atrial hyperenhancement on non-dedicated LGE imaging completely separated readmitted from non-readmitted admitted patients (8/8 vs. 0/10); however, because it was assessed on an LV-optimised protocol not validated for atrial tissue, we report it only as an exploratory qualitative finding and do not interpret it as evidence of a fibrosis mechanism. The LACE score (length of stay, acuity of admission, Charlson comorbidity index, and emergency department visits) achieved an AUC of 0.719 (95% CI, 0.44–0.95; p = 0.125); an exploratory combined LACE+ LA reservoir strain model produced a numerically higher apparent AUC, but with only eight events, this estimate is unstable and likely overfitted and is not interpreted as evidence of incremental value. A three-group analysis revealed a non-monotonic pattern: The Adm-NoReadmit patient group had the highest LA reservoir strain (15.4%), exceeding the OPD group (8.5%), whereas the Readmit group had the lowest (6.6%). Conclusion In this exploratory feasibility analysis, CMR-derived LA reservoir strain and the LAVI/LVEF ratio were associated with 30-day unplanned readmission following an AF-related index admission. Because readmitted patients had a substantially higher prevalence of heart-failure history, these associations may reflect a shared structural–functional substrate rather than independent, AF-specific effects. Descriptive comparison across the three clinical groups showed a non-monotonic LA functional pattern, suggesting that CMR may aid risk stratification at discharge. However, prospective validation is needed.