ECG Holter-based screening for sleep-disordered breathing in post-stroke patients: a large monocentric cohort study

Introduction Sleep-disordered breathing (SDB) is highly prevalent after stroke and is associated with poor functional outcomes and increased mortality, yet systematic screening remains challenging because of limited polysomnography accessibility. We evaluated three electrocardiogram (ECG)-based indices for SDB screening in a large consecutive post-stroke cohort. Methods This retrospective single-centre cohort study included 1,929 consecutive patients hospitalised for stroke or transient ischaemic attack at Saint-Étienne University Hospital (2010–2021) who were in sinus rhythm with ECG signal quality adequate for ECG-derived respiration (EDR) analysis; patients with atrial fibrillation, pacemaker rhythm, or severe ectopy (>10% of beats) were excluded a priori . EDR, the Thoracic Impedance Disturbance Index (TIDI) and the Very Low Frequency Increment (VLFI) were compared with the simultaneous oxygen desaturation index (ODI) as reference standard. Sensitivity, specificity, accuracy and predictive values were calculated at ODI thresholds of 15/h and 30/h; a combined TIDI+EDR concordant rule was assessed as a post-hoc exploratory analysis. Results The cohort comprised 61% males (mean age 65.9 years). Suspected SDB (ODI ≥ 5/h) was present in 74.6% of patients, moderate-to-severe SDB (ODI ≥ 15/h) in 38.9% and severe SDB (ODI ≥ 30/h) in 17.1%. For ODI ≥ 15/h ( n = 750), TIDI had the highest sensitivity (93.3%) and negative predictive value (90.8%), whereas EDR had higher specificity (80.2%) and accuracy (74.2%); the combined rule reached 85.2% specificity and 72.9% positive predictive value. For ODI ≥ 30/h ( n = 330), EDR showed specificity of 92.4% and accuracy of 85.2%, and the combined rule 95.4% and 86.9%, respectively. VLFI performed poorly across all metrics. Discussion ECG-based tools can identify post-stroke patients at high risk of SDB using routine cardiac monitoring. The complementary profiles of TIDI (rule-out) and EDR (rule-in) support a sequential screening approach, which requires prospective validation against polysomnography. VLFI is unsuitable for elderly stroke populations because of age-related autonomic dysfunction. Because patients with atrial fibrillation, pacemaker rhythm or severe arrhythmia were excluded, the findings may not generalise to this higher-arrhythmia-burden subgroup, who may also be at higher risk of SDB.