AbstractThe Cox proportional hazards regression model is commonly used to measure relative associations between independent variables and time until a well‐defined event in the presence of censored event times. A requisite assumption of this model is that the relative hazards are constant over time. That is, it is assumes that for two different specifications of covariates, the relative ratio of the hazard rates is constant over the time of interest. Although this assumption is generally robust to deviations, interpretation of model results may be misleading and invalid when this assumption is substantially violated. A single point in time where the relative hazard ratios change significantly is known as a change‐point. Estimating the change‐point may be clinically or scientifically important and can be used for stratified analysis. This paper describes common methods for testing the proportional hazards assumption, estimating the change‐point, and conducting a stratified analysis. Methods are illustrated through an analysis on time‐to‐decannulation among paediatric tracheostomy patients.