A few of the well-established algorithms used within TreeAge Pro are presented here.

• Markov State Reward Algorithm
• Chance Node EV Algorithm
• Probabilistic Senstivitiy Analysis Simulation
• Individual Patient Simulation

 

In a Markov model, rewards (cost and/or effectiveness) are calculated by cycle and by health state based on the following calculations.

1. (Reward calc for state/cycle) = (% of cohort starting cycle in state) * (reward entered for state)
2. (Reward calc for cycle) = ∑_states(Reward calc for state/cycle)
3. (Reward calc for Markov model) = ∑_cycles(Reward calc for cycle)

Transition rewards are calculated in a similar way. However, only the portion of the cohort starting the cycle in the state multiplied times the combined probability of reaching that node in the transition subtree.

 

When a chance node has multiple branches to represent possible outcomes, the expected value (EV) of the chance node calculated as follows.

1. (EV for chance node) = ∑_branches(EV for branch) * (Probability for branch)

 

PSA simulation runs as follows.

1. Run the simulation…
1. Draw a sample from each parameter distribution.
2. Substitute the samples into the model.
3. Calculate EVs for each strategy in the model.
2. Aggregate the individual iterations into statistics.
3. Present numeric reports and graphs based both on aggregate statistics and on individual iterations.

 

Microsimulation runs as follows.

1. Run each trial through each strategy in the model.
1. Sample individual characteristics for that trial.
2. At each chance node, draw a random number between 0 and 1. Based on the draw, move to one of the branches of the chance node.
3. Continue running through the model until the individual patient history is complete.
2. Aggregate all the patient histories into statistics.
3. Present numeric reports and graphs based both on aggregate statistics and on individual iterations.