Hi, Thank you for your videos, they are very useful! Is there a way to know the time in an state when the transition matrix allows the transition from Diseased to Healthy and vice versa? I suppose it is possible, but it would be necessary to build as many markov models as cycles we investigate, I have tried but I have not succeeded, do you know any way? Thanks!!
Hi Susana - it would certainly be possible. If you need to know time since someone entered the Healthy state but it is possible to re-enter the Healthy state then you will need to create tunnel states for the Healthy state as well. It is possible, and you do not have to have as many states as cycles provided you are content at some point to approximate all those who have been in a state for at least a certain number of cycles to be identical in terms of transition probabilities and payoffs. The other thing to consider is whether you really want people to go back to Healthy or if you want to include a Recovered state - people who have had a disease and recovered might not be identical to people who have never had the disease.
This is a very informative video. Thanks for making tunnel state understanding so simple. But I have two questions, Firstly that can anybody directly reach disease2 or disease3 from health state healthy and Secondly why didn't you consider staying in disease2 and disease1 as you have considered in case of disease3
Hi Dwaipayan, the idea is that you want to know exactly how long someone has been in the diseased state. If you only let people enter into disease1 and you insist they must move onto the next state at the end of the cycle, then when they are in diseaseX they must have spent X cycles in the diseased state. At some point we no longer get any benefit from adding more tunnel states (e.g. because hardly anybody is reaching the "end of the tunnel", or because after a while it is reasonable to re-adopt the Markov memoryless property). At that point we need to make sure we don't have any of our cohort disappearing, so they have to remain in the final state. Hope that helps 🙂
2 ปีที่แล้ว
Great example! Thanks for sharing! I will apply this for my students. Just a quick note: maintaing the general "disease" (the sum of all others) would only be able in case of both costs and utility values being th the same (independent of the stage of the disease). Otherwise, we would also need to split the sum of costs and qalys by each disease stage, right?
Yes! Tunnel states are best used when the transition probabilities or per-cycle payoffs change depending on how long it is since you entered a health state in a very predictable way, e.g., you have fitted a parametric model for the amount of time people remain in a state, or you know that after 6 months their resource use goes down. If you have different disease stages they should probably be different health states (and maybe each have their own tunnel states if needed)!
Is it possible to program that all patients that spent four consecutive cycles in state A move to state B? Thank you
Hi, Thank you for your videos, they are very useful!
Is there a way to know the time in an state when the transition matrix allows the transition from Diseased to Healthy and vice versa? I suppose it is possible, but it would be necessary to build as many markov models as cycles we investigate, I have tried but I have not succeeded, do you know any way? Thanks!!
Hi Susana - it would certainly be possible. If you need to know time since someone entered the Healthy state but it is possible to re-enter the Healthy state then you will need to create tunnel states for the Healthy state as well. It is possible, and you do not have to have as many states as cycles provided you are content at some point to approximate all those who have been in a state for at least a certain number of cycles to be identical in terms of transition probabilities and payoffs.
The other thing to consider is whether you really want people to go back to Healthy or if you want to include a Recovered state - people who have had a disease and recovered might not be identical to people who have never had the disease.
@@TMSnowsill Thank you for your answer!!
This is a very informative video.
Thanks for making tunnel state understanding so simple.
But I have two questions, Firstly that can anybody directly reach disease2 or disease3 from health state healthy and
Secondly why didn't you consider staying in disease2 and disease1 as you have considered in case of disease3
Hi Dwaipayan, the idea is that you want to know exactly how long someone has been in the diseased state. If you only let people enter into disease1 and you insist they must move onto the next state at the end of the cycle, then when they are in diseaseX they must have spent X cycles in the diseased state. At some point we no longer get any benefit from adding more tunnel states (e.g. because hardly anybody is reaching the "end of the tunnel", or because after a while it is reasonable to re-adopt the Markov memoryless property). At that point we need to make sure we don't have any of our cohort disappearing, so they have to remain in the final state. Hope that helps 🙂
Great example! Thanks for sharing! I will apply this for my students. Just a quick note: maintaing the general "disease" (the sum of all others) would only be able in case of both costs and utility values being th the same (independent of the stage of the disease). Otherwise, we would also need to split the sum of costs and qalys by each disease stage, right?
Yes! Tunnel states are best used when the transition probabilities or per-cycle payoffs change depending on how long it is since you entered a health state in a very predictable way, e.g., you have fitted a parametric model for the amount of time people remain in a state, or you know that after 6 months their resource use goes down. If you have different disease stages they should probably be different health states (and maybe each have their own tunnel states if needed)!
@@TMSnowsill Many thanks!