The Decomposition of Time-Varying Hazard Into Phases, Each Incorporating a Separate Stream of Concomitant Information

Risk Factors for Death or Heart Transplant for Patients with Right Ventricular Dysfunction After the Norwood Procedure: A Secondary Analysis of the Single Ventricle Reconstruction Trial

“…Blackstone et al 1986). In other words, we decomposed the cumulative hazard function into phases of time-varying risk.…”

“…Non-parametric Kaplan-Meier plots were constructed to guide the goodness of fit of the parametric models. From the cumulative hazard function, up to three individual hazard phases were identified and decomposed, as previously described by Blackstone, et al (Blackstone et al 1986). The three possible phases are the early phase, which represents decelerating risk period; the constant phase, which represents a hazard phase with constant hazard risk; and, finally, a late phase, which represent an accelerating risk period ( Figure 10).…”

“…The three possible phases are the early phase, which represents decelerating risk period; the constant phase, which represents a hazard phase with constant hazard risk; and, finally, a late phase, which represent an accelerating risk period. Reproduce with permission from (Blackstone & al, 1986), Copyright…”

“…We included the period of time from the Norwood procedure to the stage II procedure, which combined hospital morality, thirty-day mortality, and the interstage mortality into one time period. By employing multiphase parametric hazard modeling, we wanted to identify if the hazard of death or heart transplant varied over time, as well as identifying different risk factors for death or transplant associated with each risk phases (Sergeant, Blackstone & Meyns 1997;Blackstone et al 1986). Despite combining different risk periods into the time to event analysis, we identified only a single hazard phase, a constant phase.…”

“…Since the post-Norwood period may likely be a high-risk period, we needed to adjust for left censoring in our analysis. We adjusted for left censoring by using a methodology employing parametric survival analysis developed by Blackstone et al (Blackstone et al 1986) allowing us to make distribution assumptions of the added risk period. We then included the risk period prior to the post-Norwood echocardiogram, which is the period of time from the Norwood procedure to the post-Norwood echocardiogram ( Figure 9) and then by making distribution assumptions on this added risk period, we adjusted for left censoring reducing the bias associated with this (Cain et al 2011).…”