Detection of Atrial Fibrillation on Stroke Units: Comparison of Manual versus Automatic Analysis of Continuous Telemetry

Cerebrovasc Dis. 2020;49(6):647-655. doi: 10.1159/000511563. Epub 2020 Nov 18.

Abstract

Background: Detection of atrial fibrillation (AF) is one of the primary diagnostic goals for patients on a stroke unit. Physician-based manual analysis of continuous ECG monitoring is regarded as the gold standard for AF detection but requires considerable resources. Recently, automated computer-based analysis of RR intervals was established to simplify AF detection. The present prospective study analyzes both methods head to head regarding AF detection specificity, sensitivity, and overall effectiveness.

Methods: Consecutive stroke patients without history of AF or proof of AF in the admission ECG were enrolled over the period of 7 months. All patients received continuous ECG telemetry during the complete stay on the stroke unit. All ECGs underwent automated analysis by a commercially available program. Blinded to these results, all ECG tracings were also assessed manually. Sensitivity, specificity, time consumption, costs per day, and cost-effectiveness were compared.

Results: 216 consecutive patients were enrolled (70.7 ± 14.1 years, 56% male) and 555 analysis days compared. AF was detected by manual ECG analysis on 37 days (6.7%) and automatically on 57 days (10.3%). Specificity of the automated algorithm was 94.6% and sensitivity 78.4% (28 [5.0%] false positive and 8 [1.4%] false negative). Patients with AF were older and had more often arterial hypertension, higher NIHSS at admission, more often left atrial dilatation, and a higher CHA2DS2-VASc score. Automation significantly reduced human resources but was more expensive compared to manual analysis alone.

Conclusion: Automatic AF detection is highly specific, but sensitivity is relatively low. Results of this study suggest that automated computer-based AF detection should be rather complementary to manual ECG analysis than replacing it.

Keywords: Atrial fibrillation; Automated detection; Manual detection; SRAclinic; Stroke.

Publication types

  • Comparative Study

MeSH terms

  • Action Potentials
  • Adult
  • Aged
  • Aged, 80 and over
  • Algorithms*
  • Atrial Fibrillation / complications
  • Atrial Fibrillation / diagnosis*
  • Atrial Fibrillation / physiopathology
  • Automation
  • Electrocardiography*
  • Female
  • Germany
  • Heart Conduction System / physiopathology*
  • Heart Rate*
  • Hospitalization*
  • Humans
  • Inpatients
  • Male
  • Middle Aged
  • Predictive Value of Tests
  • Prospective Studies
  • Reproducibility of Results
  • Risk Factors
  • Signal Processing, Computer-Assisted*
  • Stroke / diagnosis
  • Stroke / etiology*
  • Stroke / physiopathology
  • Telemetry*
  • Young Adult