Bayesian Analysis of the ECG

Effective ECG interpretation depends not only on the technical analysis of the tracing but also on the ability to place the test within the patient's clinical reality. This is where Bayesian ECG analysis stands out: by combining the pretest probability that the patient has the suspected condition with the electrocardiographic findings, it allows the post-test probability of the suspected diagnosis to be calculated. Unlike traditional approaches based only on sensitivity and specificity, this tool offers a broader and more individualized view, increasing diagnostic accuracy and strengthening clinical decision-making.

Bayesian Analysis

Bayesian analysis offers three tools to aid ECG interpretation:

Bayes tree
Fagan nomogram
Direct calculation

Clinical Case

A 75-year-old man with hypertension and diabetes presents to the emergency department with pressure-like retrosternal chest pain radiating to the jaw and left arm. Based on the clinical context, the pretest probability of occlusion myocardial infarction (OMI) is estimated at 80%. However, the ECG shows no ST-segment elevation. Using the following data on ST-segment elevation, Bayesian ECG analysis can be applied.

Sensitivity: 43.6%
Specificity: 96.5%
Positive likelihood ratio: 12.5
Negative likelihood ratio: 0.58

Bayes tree

Bayes tree with a pretest probability of 80%, sensitivity of 44% and specificity of 96% — Bayes tree.

Formula for total negative results: 45.1% false negatives + 19.3% true negatives = 64.4% — Calculating the post-test probability from the Bayes tree. After building the Bayes tree, the first step to obtain the post-test probability is to calculate the total of negative results.

Post-test probability from the Bayes tree: 45.1% false negatives ÷ 64.4% negatives ≈ 70% — Calculating the post-test probability from the Bayes tree. Once the total of negative results is known, simply solve the ratio “proportion of false negatives/total negative results” to find the post-test probability. Note that the formula for the post-test probability in this example is valid only for a negative test.

Fagan nomogram

Fagan nomogram linking a pretest probability of 80% and a likelihood ratio of 0.58 to a post-test probability of about 70% — Fagan nomogram. Note the same result.

Direct calculation

Converting pretest probability into pretest odds:

Converting the pretest probability of 80% into pretest odds: 0.80 ÷ (1 − 0.80) = 4 — Calculating the post-test probability by direct calculation.

Converting pretest odds into post-test odds:

Post-test odds equal pretest odds 4 × negative likelihood ratio 0.58 = 2.32 — Calculating the post-test probability by direct calculation. Note that for negative tests the negative likelihood ratio is used in the second formula, whereas for positive tests the positive likelihood ratio is used.

Converting post-test odds into post-test probability:

Converting post-test odds into probability: 2.32 ÷ (1 + 2.32) ≈ 70% — Calculating the post-test probability by direct calculation. Note the same result.

This clinical case shows that, even without ST-segment elevation on the ECG, the patient still has a high probability of OMI. This reinforces the importance of Bayesian ECG analysis as a tool to increase diagnostic accuracy and, consequently, provide more appropriate and effective treatment.