Multi-state models: help?

Hi, I’m a statistics BS student familiar with survival analysis.
Can anyone explain to me what are multistate models, how they are used in survival analysis and what’s their relation with competing risks models?
Thank you in advice.


Less is more. Stay pure. Stay poor.
You can only be in one state. During the initial SARS-CoV-2 pandemic, people were using SIR, SEIR, etc. models. So a person can be susceptible, infected, and recovered. You can only be in one state. However interventions varied and people could get reinfected, so the world switch to 7-day average models.

For survival models, it is the same as the former. You are diseased then dead or recovered. You can't be sick and dead or sick and recovered. You can only be in one of these states at a time. However, in most all state models you can have competing risks. So you are sick, then the physician gives you a drug or not then you see if you get recovered or die. However, in this interim time period you can die for other not related causes - house fire or car accident. Thus, if you die from another cause you can't die from the disease. This is a competing event. However, we now don't know if you would have died from the disease given you did or did not receive treatment (counterfactual). So you need to control for these other causes of death that prevent us from knowing if you would have died from the disease or not. I haven't run many of these models. I believe the common approach is to look at sub-distributions.