September 1, 1986Journal of the American Statistical Association
The Decomposition of Time-Varying Hazard into Phases, Each Incorporating a Separate Stream of Concomitant Information
Authors
Discussion
Loading...
Member takes
Implication
Statistical modeling study demonstrates multiphase decomposition of time-varying risk after heart valve replacement, highlighting enhanced post-surgical risk stratification.
Key Points
- To establish a fully parametric statistical framework that decomposes time-varying hazard functions into additive, overlapping risk phases adjusted for patient-specific covariates.
- Decomposed time-varying hazard patterns into three additive, overlapping phases: early, constant-hazard, and late.
- Modeled each temporal phase using nested parametric equations scaled by separate logit-linear or log-linear concomitant covariate functions.
- Applied maximum likelihood estimation to calculate parametric estimates and confidence limits for hazard functions, survivorship functions, and event probabilities.
- Generated structured mathematical representations of non-constant post-intervention risk trajectories, addressing both early mortality and long-term reoperation events.
- Enabled patient-specific risk adjustments across distinct temporal phases, providing refined survivorship curves and probabilistic inferences for managing valvar heart disease.
Cite This Study
Blackstone et al. (1986) studied this question.