A random survival forest (machine learning) model was developed and validated to predict a composite outcome of all-cause mortality and heart failure hospitalization in 850 patients with confirmed ATTR-CM, using internal-external cross-validation across three European referral cohorts (Switzerland and Austria, 2014–2025).
The model achieved Harrell C-indices of 0.72–0.77 and 3-year AUCs of 0.73–0.80 across held-out cohorts, outperforming both the NAC and Mayo Clinic staging systems by 2–10% in C-index and 3–19% in 3-year AUC; performance held up in patients on disease-modifying therapy.
- All centers are specialized European referral centers, limiting generalizability to community settings. - The cohort is predominantly male (88.2%), so findings may not apply equally to women. - External validation outside Europe (e.g., North America, Asia) has not been performed.
This ML-based model offers better individualized risk stratification in ATTR-CM than current staging tools (NAC, Mayo), and may help guide intensity of follow-up and therapy decisions. Clinicians should note it has not yet been validated outside specialized European centers.
Explore related topics
