Hi, I am creating a markov model to obtain the cost per life year gained. Not sure how to get the life years/survival from the model. please help and thanks.

Typically, a Markov model will have health states and events. All the non-dead states will have either a 1 for LYs or a utility (between 0-1) for QALYs. Note those values would be adjusted for a cycle length less than a year.

Then, when the cohort analysis is run, the cohort will move among the states, accumulating LYs and/or QALYs based on the percentage of the cohort in each state.

I suggest you look at the following demo page on our website - Markov Model Demos - for more details.

Thanks for your replying. My model has cycle length of 3 months. Just to clarify, I would enter 0.25 for initial effectiveness, incremental effectiveness, and final effectiveness for non-dead states? and zero for the dead state?

Thanks again Andrew. I entered these values for the initial and incremental effectiveness for each of my non-dead Markov states. I have a couple of questions now:

1. Some of my Markov states are followed by a chance node and these patients can have a chance to die or be well after this particular state. So do I enter these values (0.25) under rewards for the states that come after the chance node?

For example one of my states is stroke which is then followed by chance node and there's a probability that patients are dead or patients are alive following this state. The dead or alive branches are jump states and they would feed back into these states back into the markov node.

2. I want to run a probabilistic sensitivity analysis. How do I enter the distributions for each of my probability and cost variables?

1. The nodes following the Health States are your patient pathways. Patients will have different probabilities of going down each of those paths. The Health State is where you accumulate rewards for all the patients who start the cycle in that Health State.

If you have events (which would be at the chance nodes or even further to the right) you can accumulate rewards here but they would specifically the rewards associated with that event. So, if one of the branches is Die, is there an event reward associated with that?

Usually we see Health States as the underlying state of a patient (Alive, Unwell, Local Caner etc), rather than "stoke" which is usually an event. Please ensure having a Health State of "Stroke" makes sense.

In the TreeAge software, there is a nice example called "Markov 3 State - Traditional" which you can find in the example models. This shows capturing event based utilities. (Look in the Files view under Example models > Healthcare)

2. Running Probabilistic Sensitivity Analysis is well described in both the Help files and the demonstration videos in the website. Please take a look at: Sensitivity Analysis tutorials

## Comments

Andrew MunzerTypically, a Markov model will have health states and events. All the non-dead states will have either a 1 for LYs or a utility (between 0-1) for QALYs. Note those values would be adjusted for a cycle length less than a year.

Then, when the cohort analysis is run, the cohort will move among the states, accumulating LYs and/or QALYs based on the percentage of the cohort in each state.

I suggest you look at the following demo page on our website - Markov Model Demos - for more details.

saraHThanks for your replying. My model has cycle length of 3 months. Just to clarify, I would enter 0.25 for initial effectiveness, incremental effectiveness, and final effectiveness for non-dead states? and zero for the dead state?

Andrew MunzerYes. That would count LYs. You would instead use utilities multiplied by those same factors for QALYs.

However, the Final Effectiveness should be 0 since that is AFTER the last cycle.

Those values would be affected by half-cycle correction should you choose to apply that.

saraHThanks again Andrew. I entered these values for the initial and incremental effectiveness for each of my non-dead Markov states. I have a couple of questions now:

1. Some of my Markov states are followed by a chance node and these patients can have a chance to die or be well after this particular state. So do I enter these values (0.25) under rewards for the states that come after the chance node?

For example one of my states is stroke which is then followed by chance node and there's a probability that patients are dead or patients are alive following this state. The dead or alive branches are jump states and they would feed back into these states back into the markov node.

2. I want to run a probabilistic sensitivity analysis. How do I enter the distributions for each of my probability and cost variables?

Regards,

Belinda OrmeSarah

1. The nodes following the Health States are your patient pathways. Patients will have different probabilities of going down each of those paths. The Health State is where you accumulate rewards for all the patients who start the cycle in that Health State.

If you have events (which would be at the chance nodes or even further to the right) you can accumulate rewards here but they would specifically the rewards associated with that event. So, if one of the branches is Die, is there an event reward associated with that?

Usually we see Health States as the underlying state of a patient (Alive, Unwell, Local Caner etc), rather than "stoke" which is usually an event. Please ensure having a Health State of "Stroke" makes sense.

In the TreeAge software, there is a nice example called "Markov 3 State - Traditional" which you can find in the example models. This shows capturing event based utilities. (Look in the Files view under Example models > Healthcare)

2. Running Probabilistic Sensitivity Analysis is well described in both the Help files and the demonstration videos in the website. Please take a look at: Sensitivity Analysis tutorials

Please sign in to leave a comment.