Identifying Important Risk Factors for Survival in Patient With Systolic Heart Failure Using Random Survival Forests
Key result
A random survival forests model predicted all-cause mortality similarly to a conventional Cox proportional hazards model (C-index 0.705 vs 0.698) in patients with systolic heart failure.
Why the study?
Does random survival forests modeling predict survival similarly to a conventional Cox proportional hazards model in adult patients with systolic heart failure?
Population
2231 adult patients with systolic heart failure who underwent cardiopulmonary stress testing
Comparison
Random survival forests (RSF) modeling vs Conventional Cox proportional hazards model
Design
Cohort
Follow-up
mean 5 years
Authors
Discussion
Loading...
Member takes
Implication
Comparable performance does not support replacing Cox models; leaves open advantages of random survival forests in systolic heart failure.
Key Points
- To evaluate random survival forests as an intuitive and robust machine learning method for variable selection and survival prediction in systolic heart failure.
- Analyzed 2,231 adult patients with systolic heart failure who underwent cardiopulmonary stress testing, followed for a mean duration of 5 years.
- Built a random survival forest using 2,000 bootstrap trees to evaluate 39 demographic, comorbidity, and exercise testing variables for predicting all-cause mortality.
- A total of 742 patients died during follow-up, with peak oxygen consumption, serum urea nitrogen, and treadmill exercise time identified as the top three survival predictors.
- The random survival forest achieved an out-of-bag C-index of 0.705, compared to 0.698 for the conventional Cox proportional hazards model.
Study Design
Type
Cohort (n=2,231)
Structured PICO
Does random survival forests modeling predict survival similarly to a conventional Cox proportional hazards model in adult patients with systolic heart failure?
P
Population
2231 adult patients with systolic heart failure who underwent cardiopulmonary stress testing
I
Intervention
Random survival forests (RSF) modeling
C
Comparator
Conventional Cox proportional hazards model
O
Outcome
All-cause mortalityhard clinical
Main Result
Absolute Event Rate: 0.705% vs 0.698%
Random survival forests perform as well as traditional Cox proportional hazard models for predicting survival in heart failure patients and may offer a more intuitive approach for identifying risk factors.
Cite This Study
Hsich et al. (2010) conducted a cohort in systolic heart failure (n=2,231). Random survival forests (RSF) vs. Cox proportional hazards model was evaluated on Predictive accuracy for all-cause mortality (C-index). A random survival forests model predicted all-cause mortality similarly to a conventional Cox proportional hazards model (C-index 0.705 vs 0.698) in patients with systolic heart failure.