Table 5 Analysis of longitudinal ECG data

From: Explainable AI associates ECG aging effects with increased cardiovascular risk in a longitudinal population study

Model

Group

n

HR (95% CI)

p-value

Baseline

Correct prediction

1702

Reference

–

 

Overestimation

874

1.44 (1.16–1.80)

0.001

 

Underestimation

614

0.83 (0.59–1.18)

0.305

Follow-up

Correct prediction

1729

Reference

–

 

Overestimation

849

1.43 (1.14–1.80)

0.002

 

Underestimation

612

0.74 (0.53–1.02)

0.064

Serial ECG

Correct prediction, correct prediction

1165

Reference

–

 

Overestimation, overestimation

551

1.65 (1.25–2.17)

< 0.001

 

Correct prediction, overestimation

288

1.52 (1.04–2.21)

0.029

 

Overestimation, correct prediction

311

1.34 (0.96–1.86)

0.081

 

Underestimation, underestimation

351

0.68 (0.42–1.10)

0.117

 

Correct prediction, underestimation

249

0.91 (0.58–1.43)

0.686

 

Underestimation, correct prediction

253

1.41 (0.88–2.27)

0.154

  1. This table presents HRs for “Overestimation” and “Underestimation” across “Baseline,” “Follow-up,” and “Serial ECG” scenarios. While “Baseline” and “Follow-up” scenarios utilize a single ECG, “Serial ECG” incorporate two ECGs recorded 5–6 years apart, simulating a longitudinal screening scenario. To investigate the influence of serial ECGs on the association between aging effects and increased cardiovascular risk, “Follow-up” is compared to “serial ECG,” both having a follow-up w.r.t. mortality data of up to 16 years.
Back to article page