Sequential Survival Models

Greetings all. I have been invited to contribute to a research project using Cox regression to evaluate association between a target variable and hard clinical outcomes. The research team is proposing sequential Cox models, with each adding additional variables. I haven’t seen anything about model fit or choosing the most appropriate model for the data. As a non-statistician, I would love this groups thoughts on this approach to observational (“real-world”) data analysis including resources where appropriate.

Thank you in advance.

If the order of variable entry is completely pre-specified, using subject matter knowledge or for example entering variables by increasing cost of collecting those variables, this can be quite reasonable. Otherwise a fully pre-specified model should be used. Details are in my RMS book and course notes. But model assumptions (e.g., proportional hazards) should be checked. Also be sure to not assume linearity for continuous predictors. You can use restricted cubic spline functions for them, for example.