Partitioned Survival Analysis (NICE DSU TSD 19)
A method used to estimate the time patients spend in different health states (like progression-free and progressed disease) by using survival curves. It helps calculate health benefits and costs over time, especially in cancer treatments.
At a glance
Use when
Modeling diseases with clearly defined, irreversible health states (e.g., oncology); when clinical data are reported as survival curves; when a transparent, non-Markov approach is preferred.
Avoid when
Diseases with reversible health states or complex transition patterns; when transitions are not driven primarily by survival outcomes; when recurrent events or multiple competing risks need explicit modeling.
Inputs
Survival curves (e.g., progression-free survival and overall survival), utility values by health state, cost data by health state, time horizon, discount rates.
Outputs
Life years, quality-adjusted life years (QALYs), total costs, incremental cost-effectiveness ratios (ICERs), health state occupancy over time.
How it works
Partitioned Survival Analysis (PSA) is a survival-partitioning decision modeling technique that estimates the proportion of patients in distinct health states by combining multiple survival curves (e.g., progression-free survival and overall survival). It serves as an alternative to multi-state Markov models and is commonly applied in oncology cost-effectiveness analyses, particularly within NICE Technology Appraisals. The method is detailed in NICE Decision Support Unit Technical Support Document 19.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation
- Categories
- Cost-effectiveness Modelling
- Assumptions
- Health state occupancy is determined entirely by survival curves; patients transition sequentially through states; no back-transition between states; survival curves are accurately estimated and extrapolated.
- Strengths
- Simpler than multi-state models when transitions are sequential; avoids Markov cycle effects; directly uses standard clinical trial outputs (Kaplan-Meier curves); transparent and reproducible structure.
- Limitations
- Cannot model reversible transitions; relies heavily on accurate survival extrapolation; assumes independence between survival curves; may oversimplify complex disease pathways.
- Also known as
- Partitioned Survival Model, PSA, NICE DSU TSD 19, Survival Partitioning Method
Questions this answers
- › How much time do patients spend in each health state (e.g., progression-free, progressed, dead)?
- › What is the incremental cost-effectiveness of a treatment based on partitioned survival estimates?
- › How do survival curves translate into quality-adjusted life years (QALYs)?
- › How can non-Markovian survival data be used in economic models?
- › What are the cost and health outcomes over time when transitions are defined by survival functions?
- › How can uncertainty in survival extrapolation be incorporated into decision models?
References & sources
Similar by meaning
- Cranmer et al. comparison of partitioned survival analysis and state-transition multi-state modelling (oncology case study)
- NICE DSU Technical Support Document 14
- INES (INteractive model for Extrapolation of Survival and cost)
- State-Transition Modeling
- Change-point survival models for relative treatment effect extrapolation
Beta record. Based on the original catalogue summary; primary-source enrichment pending.