How effective is Muse for electrocardiogram (ECG) interpretation?

MUSE for ECG Interpretation: Effectiveness and Clinical Utility

MUSE (Marquette Universal System for Electrocardiography) is a moderately effective computer-based ECG interpretation system that is highly specific but only moderately sensitive, requiring mandatory physician overreading for safe clinical use.

Performance Characteristics

Based on validation studies, MUSE demonstrates:

• Specificity: 93.2-98.0% for rhythm interpretation 1
• Sensitivity: 60.6-79.1% depending on coding definitions used 1
• Revision rate: 7.8% of computer-based rhythm statements require correction (excluding pacemaker tracings) 2
• Primary rhythm diagnosis errors: 3.8% of cases 2

The highest sensitivity (79.1%) is achieved when AF/AFL appears in the first 3 ECG statements, with minimal loss of specificity (94.5%) 1.

Specific Diagnostic Performance

Atrial Fibrillation/Flutter Detection

• False-negative rate: 9.2% (misses nearly 1 in 10 cases of AF) 2
• False-positive rate: 1.1% when computer diagnoses AF in other rhythms 2
• Incorporating MUSE data increases AF/AFL diagnostic yield by 14.5% for prevalent cases and 7.5-16.1% for incident cases compared to administrative data alone 1

Sinus Rhythm Detection

• False-negative rate: Only 1.3% (excellent at detecting sinus rhythm) 2
• False-positive rate: 9.9% (incorrectly diagnoses sinus rhythm in nearly 1 in 10 non-sinus rhythms) 2

Problematic Areas

Paced rhythms remain highly problematic for MUSE interpretation and require particularly careful physician review 2.

Clinical Context and Guidelines

The American College of Physicians guidelines clearly state that computer analyses of ECGs are useful adjuncts but should not replace qualified physician interpretation 3. This is based on evidence showing computer ECG analysis correctly classifies only 0-94% of disorders, with arrhythmias being the most problematic 3.

Key Limitations

• Major interpretation errors requiring edits: 13.5% of MUSE interpretations in head-to-head comparisons 4
• Computer programs can decrease interpretation time and modestly reduce medical errors, but accuracy varies widely 3
• Physician overreading remains mandatory to correct computer-based diagnoses 2

Comparison to Modern AI Approaches

Recent AI-enabled ECG algorithms (2021) have shown superior performance to traditional computer programs like MUSE:

• AI algorithms required major edits in only 8.2% of cases versus 13.5% for standard computer programs 4
• AI interpretations were considered "ideal" in 70.5% of cases versus 63.9% for computer programs 4

Clinical Recommendations

Use MUSE as a screening and efficiency tool, not as definitive interpretation:

1. Always overread MUSE interpretations before making clinical decisions, particularly for:

   • Any paced rhythm
   • Suspected atrial fibrillation (given 9.2% false-negative rate)
   • Non-sinus rhythms where computer suggests sinus (9.9% error rate)

2. Trust MUSE more for:

   • Ruling in atrial fibrillation (98.9% specificity)
   • Detecting sinus rhythm (98.7% sensitivity)

3. Be particularly cautious with:

   • Pacemaker rhythms (high error rates)
   • Arrhythmia diagnoses (most problematic category)
   • Ectopic complex definitions (3.9% error rate)

Bottom Line

MUSE is a valuable adjunct that can enhance diagnostic yield and efficiency, but it cannot replace trained physician interpretation. Its moderate sensitivity means it will miss approximately 1 in 5 to 1 in 3 abnormalities depending on the condition, making physician confirmation essential for patient safety 3, 2.