Statistical tools
Exploring partitioned survival models
The dominant decision-modelling structure in oncology cost-effectiveness (NICE DSU TSD 19). Set progression-free and overall survival curves and watch the three health states — progression-free, progressed, dead — partition the cohort, with length-of-stay, QALYs and the ICER updating live. Built on the merlin family.
Two curves, fitted independently. Overall survival should sit above progression-free survival (more patients alive than progression-free).
Export to merlin · pendragon
The export fits the two endpoints with merlin and assembles the
partitioned-survival model with pendragon (pendragon_partsurv),
which emits the same state-occupancy object as a multi-state model, so the QALY / cost /
ICER value layer applies unchanged.
Red Door Analytics
Oncology cost-effectiveness, done properly
A partitioned-survival model is an area-under-the-curve construction: simple, but not internally coherent. Where the data support it, a multi-state (state-transition) model is the principled alternative — and fitting either to your data, with extrapolation and the health-economic outputs that follow, is what merlin and pendragon are for.
See also: Multi-state & competing-risks explorer· Survival DGM explorer· the merlin family· how we pre-specify an analysis →
Occupancy is the area under / between the progression-free and overall survival curves (partitioned survival; NICE DSU TSD 19). For teaching and study-design illustration; validate any analysis in merlin / pendragon.