Turk Kardiyol Dern Ars. Ahead of Print: TKDA-05898 | DOI: 10.5543/tkda.2026.05898
Automated Two-component Pre-CTPA ECG Risk Phenotype and 365-day Mortality among ECG-evaluated Patients with Computed Tomography-confirmed Pulmonary Embolism
Hasan Burak İşleyen1, Sevil Tuğrul Yavuz2, Sercan Bulut3, Oğuzhan Abanoz4, Esra Danışman4, Necla Zeynep Eren4, Nuray Kahraman Ay4, Nusret Açıkgöz4, Ramazan Özdemir4, Ali Arda Sözen4, Mahmut Uluganyan41Department of Cardiology, İstanbul Nişantaşı Üniversitesi, İstanbul, Türkiye
2Department of Cardiology, İstanbul Başakşehir Çam ve Sakura Şehir Hastanesi, İstanbul, Türkiye
3Department of Cardiology, Bağcılar Training and Research Hospital, İstanbul, Türkiye
4Department of Cardiology, Bezmialem Foundation University Hospital, İstanbul, Türkiye
Objective: Electrocardiographic abnormalities are common in acute pulmonary embolism, but their prognostic value in public waveform-linked cohorts remains incompletely characterized. We evaluated whether an interpretable automated two-component pre-CTPA ECG risk phenotype was associated with 365-day mortality among ECG-evaluated patients with CTPA-confirmed pulmonary embolism.
Method: This secondary analysis used a retrospective cohort derived from MIMIC-IV-Ext-PE linked to MIMIC-IV-ECG. The source population included 1,534 unique PE-positive patients after patient-level indexing. A total of 650 patients had an ECG within 6 hours before CTPA, 571 were admission-linked, and 567 passed waveform signal-quality screening. The exposure assigned 1 point for tachycardia >100 beats/min and 1 point for V3 T-wave inversion. The primary Cox model was adjusted for age, sex, Charlson comorbidity index, PE subtype, ICU admission initiated at or before CTPA, and early physiologic derangement after excluding heart rate from the derangement definition.
Results: The primary cohort included 567 patients and 126 deaths within 365 days. Crude 365-day mortality increased across score levels: 15.0% for score 0, 26.8% for score 1, and 41.7% for score 2. In the primary model, each 1-point increase in the score was associated with higher 365-day mortality (HR 1.70 [1.36–2.13], P<0.001). The exposure-level proportional hazards diagnostic did not show clear evidence of violation (P=0.060). A stabilized inverse-probability-of-selection weighted Cox model yielded a similar estimate (HR 1.59, 95% CI 1.28–1.99; P<0.001), although measured selection imbalance was reduced but not eliminated.
Conclusion: Among ECG-evaluated patients with CTPA-confirmed pulmonary embolism, an automated two-component pre-CTPA ECG risk phenotype was associated with higher 365-day mortality. The phenotype should be interpreted as an early prognostic marker requiring further validation rather than as a validated clinical decision tool.
Keywords: Automated waveform analysis, ECG, electrocardiography, MIMIC-IV, mortality, pulmonary embolism, risk stratification.
Corresponding Author: Hasan Burak İşleyen
Manuscript Language: English