Deep Learning Forecasts Waitlist Outcomes in MASH
Researchers developed a deep-learning competing risk model (DeepHit) to forecast death and transplant trajectories for 17,551 patients with MASH cirrhosis listed for liver transplant using SRTR data and external validation at University Health Network. DeepHit achieved competing event coherence (CEC) scores of 0.813, 0.811, 0.794 and 0.772 at 1, 3, 6 and 12 months respectively; random survival forests had higher concordance overall while DeepHit improved transplant Brier score at 12 months (0.206 vs 0.228).
Key Points
- 1Demonstrates DeepHit forecasts both death and transplant for 17,551 MASH waitlist patients
- 2Shows competing-risk modeling improves coherence across mutually exclusive outcomes versus single-risk methods
- 3Enables clinicians to stratify candidates using MELD, functional status, age, and blood type
Scoring Rationale
Strong registry-based validation and clinical applicability, tempered by modest methodological novelty over existing survival models.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
