«6-8 out of 10 ECG247 arrhythmia alerts are wrong»

Jarle Jortveit, cardiologist and medical director at ECG247

We sometimes get asked why ECG247 alerts about arrhythmias that are not arrhythmias (false positive test results). The explanation is that when arrhythmias are rare in the examined population (low prevalence), even a test with high sensitivity and specificity will generate many false positives in absolute numbers – because there are so many healthy people being tested. Then the positive predictive value (PPV) becomes low, and it may seem that many alerts are wrong, even if the test itself is good. This is about the occurrence of arrhythmias, not about poor test quality.

A concrete example is illustrated in the figure above

• Scenario A (low incidence):
  Prevalence 0.5% (50 out of 10,000 have arrhythmia), sensitivity 99%, specificity 95%
  True positives: 49
  False positives: 497
  PPV = 49 / (49 + 497) ≈ 9% → ≈ 9 out of 100 alerts are true (~9/10 are false)

• Scenario B (high incidence):
  Prevalence 10% (1000 out of 10,000 have arrhythmia), same test
  True positives: 990
  False positives: 450
  PPV = 990 / (990 + 450) ≈ 69% → majority of alerts are true (~3/10 are false)

It is very important that ECG247 detects as many arrhythmias as possible (high sensitivity). To achieve this, we must accept false positive test results. All warnings of possible arrhythmia must therefore be evaluated by a physician before final diagnosis and initiation of treatment.