OBJECTIVES: Partitioned survival analysis (Part SA) is a useful modelling technique for assessing cost-effectiveness of interventions that impact survival. Part SA models typically comprise of three mutually exclusive health states: progression-free survival (PFS), progressed disease (PD) and death. Incident progression in Part SA models is defined as “proportion of patients who progressed from PFS to PD state during a model cycle”. The current study explores three different methods of estimating incident progression and their implications on cost effectiveness models. METHODS: A hypothetical Part SA model (Microsoft Excel) compared incident progression using three approaches: Approach 1 (assuming all patients progress before dying), Approach 2 (assuming same mortality for PFS and PD), and a ‘novel’ Approach 3 (assuming general mortality rates for PFS). A weekly cycle, 35-year time horizon model with a hypothetical Weibull distribution was developed to simulate PFS and overall survival (OS) curves. Cycle-specific hazard rates were determined from the Weibull function to estimate the proportion of patients that die and remain in the PFS health states over time. Mortality rates from Australian Life Tables and from the Weibull distribution were used. The proportion of patients in the PD health state was estimated as the difference between OS and PFS. Scenario analyses were conducted to compare the effect of rapidly progressing disease on incident progression. RESULTS: In the base case, proportion of cumulative progressed patients over time calculated by adding the incident progressed patients in each cycle estimated using the three different approaches was estimated as 97.1% (Approach 1), 54.4% (Approach 2) and 77.5% (Approach 3). Scenario analysis estimated cumulative progressed patients were similar using Approach 1 (100%) and Approach 3 (98.1%), while 59.8% patients progressed using Approach 2. CONCLUSIONS: The novel Approach 3 may lead to the closest approximation of the ‘true’ incident progressed patients in Part SA modelling.