**Introduction/Hypothesis**
We have previously applied artificial intelligence (AI) to an electrocardiogram (ECG) to detect cardiac amyloidosis (CA). In this validation study, we observe the post-development performance of the AI enhanced ECG to detect CA with respect to multiple potential confounders.

**Methods**
Amyloid patients diagnosed after algorithm development (January 1, 2019-December 31, 2021) with a 12 lead ECG were identified (n=394) and were required to have CA. A near 18:1 age- and sex- matched control group was identified (n=6,860). Area under the receiver operating characteristic (AUC) was determined for the cohort and subgroups.

**Results**
The average age was 69.7 years, 26.6% were female, and most patients were white (91.7%). The AI-ECG for CA had an AUC of 0.85 (95% Confidence Interval [CI]: 0.83—0.88) for the cohort with minor performance difference between sexes, age, and CA subtype; a slight decrease from the original study (AUC 0.91). White, black, and patients of “other” races demonstrated similar algorithm performance (AUC 0.82-0.86). Hispanic patients exhibited an AUC of 0.71 (CI: 0.40—1.00) with smaller sample size. Algorithm performance shift over time was not observed. Low ECG voltage and infarct pattern exhibited high AUC (>0.92), while left ventricular hypertrophy (LVH) and left bundle branch block (LBBB) demonstrated lesser performance (AUC 0.62 and 0.74 respectively).

**Conclusion**
The AI-ECG for the detection of CA maintained overall strong performance with respect to patient age, sex, race, and amyloid subtype. Lower performance was noted in LBBB and LVH and further validation in more ethnically diverse cohorts is necessary.



Post-development performance and validation of the Artificial Intelligence enhanced electrocardiogram for the detection of cardiac amyloidosis