INRODUCTION TO TREEAGE PRO - InaHEAinahea.org/files/precongress/L3 Introduction to TreeAge Pro Markov... · INRODUCTION TO TREEAGE PRO: Markov ... – 3) Add utility weights ... purpose

INRODUCTION TO TREEAGE PRO:Markov

Asrul Akmal Shafie BPharm, Pg Dip Health Econs, PhD

[email protected]

Adjunct Associate ProfessorMonash University

Board MemberHTAsiaLink

Associate Professor & Program ChairmanUniversiti Sains Malaysia

mailto:[email protected]

W e l e a d

• In previous exercise, the values at the end of the branches (such as HIV infected) were specified as life expectancies: the average length of life someone in that terminal state is expected to survive.

• However, for many modeling problems, it is the detail that happens in those life times that is important.

• We will now describe how to represent those lifetimes as a series of states in a Markov Processes.

PART II: Markov Processes

W e l e a d

• To represent people who are HIV negative, you will make a simple two‐state Markov process where people are either Alive or Dead (shown on left)

• For people who are HIV positive, you will construct a slightly more complicated process where the patients areinitially HIV Positive, may progress to full‐blown AIDS, and may die from either state (Dead).


W e l e a d


• Markov processes are constructed in TreeAge using a Markov Node.

• There is a direct correspondence between the states of a Markov process, transitions between states, and the branches of a Markov node.

W e l e a d

• States (such as Live and Dead)– are represented as the branches of the Markov node, and all states are defined

from that node.

• Transitions between states– are represented by a chance node from each state; the branches of the chance

node describe which state the person starts out the next time cycle. Forexample, in the illustration, for a person in the Live state, taking the Die branch sends them to the Dead state, taking the StayLive branch sends them to theLive state


W e l e a d

• Building Markov Processes for the purpose of constructing costeffectiveness analyses in TreeAge requires an understanding ofseveral concepts:– Markov State Value:

• People in a Markov process accrue value by living in a state.• For example, if a patient lives in a state for 1 year, then that patient

accumulates 1 more year of life.• For the whole cohort, the Markov process accumulates value by summing

the portion of the cohort alive in each state times the value of living in thatstate, summed over all states.

• Therefore, each state must have a value or utility that represents how muchadded benefit there is for being in each state for one period. Called theincremental utility, it would be 1 for states that represent perfect health, orif no quality adjustment is being used.

• For states representing illness or less than perfect health, the value of the incremental utility would be a number between 0 and 1 that wouldrepresent the quality adjustment for living in that state.


W e l e a d


– ½ Cycle correction:• A Markov process breaks up a continuous process (life) into discrete,

sequential time periods.• All transitions in a Markov process occur at the beginning or end of the

time period (in TreeAge, they occur at the end).• As shown below, counting transitions at the beginning or end under‐ or

over‐ estimates the like expectancy represented in the Markov Process.• To “correct” this, most software packages, including TreeAge, allow for a

special definition of the first time period, which is essentially ½ a cycle inlength.

• It does this by assigning only ½ the value of being in that particular state in the first time period.

W e l e a d

Decision Problem


W e l e a d

– Discounting:• When models are constructed for the purpose of cost effectiveness

analyses, future costs and outcomes should be discounted.• Future costs and benefits are discounted by a “discount rate”, using the

formula: 1/(1+r)t

• In the following exercise, you will:– 1) Convert the terminal nodes of the prior tree to Markov Processes– 2) Add a discount rate to appropriate discount future costs and benefits– 3) Add utility weights for each state to account for quality of life


W e l e a d

Decision Problem• You will modify the HIV screening decision tree that you

built previously into a Markov decision tree, doing so in two stages.

• First, you will change the Not Infected terminal nodes into a 2‐state Markov model (Live and Dead) using a constant risk of death by “natural” causes as the transition probability.

• Then you will modify the HIV Infected nodes into a 3‐state Markov model (HIVpos, AIDS, and Dead) using constant risks of HIV positivity progressing to AIDS and of death from AIDS.

• At this point you will also use a table to allow the use of age‐specific mortality rates for death from other causes.


W e l e a d

Decision Problem

• The length of each cycle in the Markov models will be 1 year.

• Therefore we will need to add probabilities to the tree for transitions from one state to another based on the likelihood of those transitions occurring during a year’s time.

• Also you will implement a half cycle correction, to account for transitions occurring (on average) at the midpoint of each cycle.


W e l e a d

Decision Problem

• The 2‐stage model that substitutes for the Not Infected nodes will be the same throughout the tree, allowing the use of the “cloning” feature of TreeAge Pro.

• Similarly, the 3‐stage model that substitutes for the HIV Infected nodes can also be “cloned”.


W e l e a d

Markov Structure: Not Infected• The structure of the 2‐stage Markov model substituted for one of the

original Not Infected nodes is shown in Figure 1.• Figure 1: Two‐stage Markov Model – first substitution for Not

Infected node


W e l e a d

Markov Structure: Not Infected

• TreeAge Pro has a special Markov node, represented by a purple “M” within a circle .

• The branches that lead off a Markov node designate all of the Markov states (and only Markov states).

• In this example, there are two health states: (1) Live and (2) Dead.• Dead is a terminal node because it is an absorbing state (i.e. there is

no chance of leaving this state).• The nonabsorbing state (Live) is a chance node, off which you can

create a subtree (also called a cycle tree) that reflects events occurring during a cycle.


W e l e a d

Markov Structure: Not Infected• The last branch at the end of each pathway will be a terminal node

and the name to the right of each terminal node must be set to one of the Markov state names.

• These “jump states” specify what state to go to for the next cycle. (Note that in a regular decision tree, these “payoff” variables represent the outcome of the analysis.

• In a Markov model, outcomes are defined elsewhere – not at theterminal node.)


W e l e a d

Markov Structure: Not Infected• The probabilities assigned to each branch off the Markov node

represent the probability of being in each of the Markov states at time 0.

• In this example, we assume that all patients are alive when HIV screening results become available.

• Thus all patients begin in the Live state, and the probability is 1.• Since the cycle length is one year, probabilities shown in cycle trees

(i.e. the tree structures to the right of Markov states) represent one‐ year probabilities, and will be evaluated each cycle.


W e l e a d

Building the Tree• You should be familiar with building a decision tree in

TreeAge Pro (see table).


Add a branch OPTIONS menu, ADD BRANCH option (or double click on node, or Ctrl+A)

Label a branch Click on the branch, text insertion box appears

Insert probability variable

Click directly below the branch (or Tab from branch label)

Change node type OPTIONS menu, CHANGE NODE TYPEoption(or click icon on Tool Bar, or Ctrl+T)

W e l e a d

Building the Tree• Using Figure 1 as a guide, enter the structure of the Markov

model for the Not Infected node off the Positive test node.• First change the node type of the Not Infected node to

Markov, then add Live and Dead states to two branches off the Markov node.

• Enter 1 as the probability of Live and 0 for Dead (since all of the cohort begins the model alive).

• Note that when you change the nodes at the end of the Markov cycle trees to terminal nodes, TreeAge Pro will prompt you as to which of the Markov states you want to use as a “jump state”.


W e l e a d

Building the Tree• Do not build beyond the first Not Infected node for now

(you will be cloning it later).• Note that when you enter probabilities using variables

rather than numbers, TreeAge Pro will prompt you to create the variable.

• Click OK to create each variable.• Remember that all branches off Markov nodes and chance

nodes need to have probabilities.


W e l e a d

Utilities• You will now assign utilities.• In a standard decision tree, the utility values (or payoff

values in TreeAge Pro) are assigned at the end of the tree at each of the terminal nodes.

• In a Markov model, the terminal node specifies what state to go to for the next cycle, rather than the payoff.

• Utilities in a Markov model are assigned at the first generation branches off the Markov node (i.e. the Markov state branches).


W e l e a d

Utilities• There are three types of utilities in a Markov model:

– The Initial utility of a state is the utility a person (or a portion of a cohort) accrues due to being in a state when the model starts (evaluated only once)

– The Incremental utility of a state is the utility a person accrueseach time she or he arrives in that state as the model runs (evaluated many times)

– The Final utility of a state is the utility a person accrues if in that state when the model finishes running (evaluated only once).


W e l e a d

Utilities• You will define these utilities using utility and cost variables (uNoHIV,

CostTest) to allow quality adjustment of life expectancy and a discount variable (r) that will allow discounting of life expectancy and of costs over time.

• The variable uNoHIV is the utility of the Live state. If this utility valueis equal to 1, then the expected value equals life expectancy.

• By setting utilities less than or equal to 1, one can calculate quality adjusted life expectancy.

• For discounting, you will use a TreeAge Pro variable (_stage) that represents the Markov cycle number.

• Note that TreeAge Pro has a built‐in discounting function; for thepurpose of this tutorial, you will manually enter the formulas fordiscounting.


W e l e a d

Utilities• To enter utility values, go to each of the first generation

branches (states) off the Not Infected Markov node (Liveand Dead)

• Pull down the VALUES menu, and choose MARKOV STATEINFORMATION.

• An editing box will appear, which will allow you to enter the three costs. For the Live node, enter CostTest for INITIAL then0 for INCREMENTAL and FINAL.


W e l e a d

Utilities• To enter effectiveness, use the drop down menu off SELECT REWARD SET

(see Figure 2 below) to select 1 (Effect).


W e l e a d

Utilities• For INITIAL use the expression: 0.5 * uNoHIV (the first cycle gets credit

for a half year ‐ the “half cycle correction factor”).• For INCREMENTAL enter uNoHIV/(1+r)^_stage• For FINAL type 0 (see Figure 3 below).• For the Dead node, cost and effectiveness are 0 for INITIAL,

INCREMENTAL, and FINAL.


W e l e a d

Assigning Probabilities and Defining Variables• Where you assign values to probabilities is important.

Some variables will be assigned globally (affecting thewhole tree); others will be assigned locally (affecting aportion of the tree).

• At the Positive test node you will define the utility locallyfor not being HIV infected (uNoHIV) as being equal to theutility of drug toxicity (uTox) to account for lost quality oflife due to false positive diagnosis and treatment.


W e l e a d

Assigning Probabilities and Defining Variables1. Assigning Probabilities:

• Select the Positive test node, pull down the VALUES menu,and choose VARIABLES AND TABLES (or click the V= button onthe Tool Bar).

• Select uNoHIV and click on DEFINE VARIABLE, then AT SELECTEDNODE(S), then enter uTox.

• Create the variable by clicking OK.


W e l e a d

Assigning Probabilities and Defining Variables2. Defining Variables

• To define variables numerically, pull down the VALUES menu and choose VARIABLES AND TABLES (or click the V= button on the Tool Bar).

• Select pDie, r, uNoHIV, and uTox while holding down the CTRL key.

• Click the DEFINE VARIABLE button, then select DEFAULT FORTREE.


W e l e a d

Assigning Probabilities and Defining Variables• Enter the following values when prompted and then click

OK or press Enter:


VariableuTox

uNoHIVr

pDie

Value0.9

0.920.03

0.005

W e l e a d

Creating Clones• The cloning feature of TreeAge Pro allows you to make an exact

duplicate of a tree structure to attach at other locations in the tree.• This is different from the copy and paste feature in that when you

copy and paste a tree, you can then make changes to the pasted treewithout affecting the original tree.

• When you create a clone, you are only allowed to make changes to the “clone master” (and each change will be made to all copies of that clone).

• Since the model for Not Infected persons are the same regardless of testing, you can make the model you have already built a clone master, and attach clones to the other Not Infected nodes.


W e l e a d

Creating Clones• Select the Not Infected Markov node, pull down the

OPTIONS menu and choose SELECT SUBTREE (the subtreenodes will become solid colors).

• Open the EDIT menu and choose CREATE CLONE.• You’ll be prompted to name your clone master.• Type No HIV and click OK.• The Not Infected branch is thickened to indicate that it

represents a clone master


W e l e a d

Creating Clones• Select the Not Infected nodes off both the Negative test

and No Screen nodes by clicking one of them then the other while holding down the SHIFT key.

• Convert them to Markov nodes, then pull down the EDITmenu and choose ATTACH CLONE.

• The attached clones will be shown in gray (not black) and you will not be able to directly edit them.

• Your tree should now look like this (Figure 4):


W e l e a d

PART II: Markov ProcessesCreating Clones• Figure 4:

W e l e a d

Changing Preferences and Setting the Markov TerminationCriteria• The default is to report the expected values in dollar units

with 0 significant digits.• To change this, pull down the EDIT menu, choose NUMERIC

FORMATTING then COST/EFF, and change the DECIMAL PLACES to 2, then click OK and CLOSE.

• To see variable definitions and Markov information in the tree, pull down the EDIT menu, select PREFERENCES then VARIABLES/MARKOV INFO, and then click SHOW DEFINITIONS, EXPAND NODE TO FIT VARIABLE, and SHOW MARKOV INFORMATION.


W e l e a d

• A Markov stopping criterion must be set at each Markov node before tree evaluation

• To do this, select a Not Infected branch, pull down the VALUES menu, and select MARKOV TERMINATION (or click the M= button on the toolbar).

• Clicking OK will accept the default, which will typically run the model until almost all of the cohort are in the Dead state.

• Now set the default Markov termination value for the other 2 Not Infected branches.

• Termination criteria must be set individually for all 3 Not Infected Markov nodes.


W e l e a d

Cost‐Effectiveness and Markov Analysis• The discounted cost per quality adjusted life year gained

can be obtained by selecting the root (or decision) node, pulling down the ANALYSIS menu and choosing COST‐ EFFECTIVENESS.

• Two graph display options are provided; just click OK. For a text report, click the ACTIONS button then TEXT REPORT.

• To return to the tree, click OK on the text report, then FILE CLOSE the graph without saving it.


W e l e a d

Cost‐Effectiveness and Markov Analysis• To look at a Markov trace, select one of the Markov nodes,

pull down the ANALYSIS menu, and select MARKOV COHORT(QUICK).

• A box will appear with two options: SUMMARY REPORT and GRAPH.

• Select SUMMARY REPORT.• TreeAge Pro creates a report which includes the stage

number, the effectiveness and cost values for that stage(weighted averages), cumulative sums for effectiveness and costs, and the probability of being in each health state at that stage.


W e l e a d

Cost‐Effectiveness and Markov Analysis• This report can be copied to the clipboard by selecting

EXPORT followed by COPY TO CLIPBOARD, or exported to Excel, or saved to a text file.

• Note that summing any particular column of probabilities will give the total expected time in that state.

• A Markov trace is a very useful tool for debugging purposes.


W e l e a d

Cost‐Effectiveness and Markov Analysis• This report can be copied to the clipboard by selecting

EXPORT followed by COPY TO CLIPBOARD, or exported to Excel, or saved to a text file.

• Note that summing any particular column of probabilities will give the total expected time in that state.

• A Markov trace is a very useful tool for debugging purposes.


W e l e a d

Sensitivity Analysis• To perform a one‐way sensitivity analysis, select the decision node,

pull down the ANALYSIS menu, select SENSITIVITY ANALYSIS, then ONE WAY.• You can select the variable on which to perform the analysis, low and

high values, as well as the number of intervals.• Keep in mind that TreeAge Pro runs the model once for each interval;

choosing a large number of intervals may make the analysis take a longer time.

• Selecting TEXT REPORT shows results for each value of the variable;selecting GRAPH will produce a graph of results versus the variablechosen for sensitivity analysis.


W e l e a d

Markov Structure: HIV Infected• The next Markov model that you will substitute for the HIV Infected

nodes is shown in Figure 5 (below).• It includes three health states: HIVpos, AIDS, and Dead.• Begin by creating three branches off the Markov node you create by

changing the HIV Infected node from a terminal node to a Markov node.

• Then label the three branches with the three health states and continue to construct the model as shown in Figure 5.

• We assume that all persons begin the model in the HIVpos state and that only those in the AIDS state will die of AIDS.

• Again remember that all branches off Markov nodes and chance nodes need probabilities (see Figure 5 below).


W e l e a d

Markov Structure: HIV Infected


W e l e a d

Markov Structure: HIV Infected• You will again assign utilities.• Go to each of the first generation branches off the HIV

Infected Markov node, pull down the VALUES menu, and choose MARKOV STATE INFORMATION (or click the M= icon on the toolbar).

• An editing box will appear, allowing you to enter the Initial, Incremental and Final utilities for Reward 1 (effectiveness) and Reward 2 (cost) for each state.

• Enter the utilities given in the table below.


W e l e a d

Markov Structure: HIV Infected


U t i l i t y Assignments

Branch Reward Initial Incremental Final

HIVpos 1(Effect) .5*uHIVpos uHIVpos/(1+r)^_stage 02(Cost) CostTest+CostInf*0.5 CostInf/(1+r)^_stage 0

AIDS 1(Effect) 0 uAIDS/(1+r)^_stage 0

2(Cost) 0 CostAIDS/(1+r)^_stage 0

Dead 1(Effect) 0 0 0

2(Cost) 0 0 0

W e l e a d

Creating Tables in TreeAge Pro (may skip this step if table already exists on network)• You will input a table of age‐specific mortality rates for the

population (from US life tables), allowing incorporation ofnon‐AIDS causes of death as a function of age by defining atable named “muASR” and entering mortality rates from anExcel spreadsheet


W e l e a d

Creating Tables in TreeAge Pro• Open Excel, then open the file muASR.xls (which you will

need to get from the instructors).• This file contains age‐specific mortality rates.• Copy the two columns of data onto the clipboard by

pressing Ctrl‐C or selecting COPY from the EDIT menu.• Next, you will create the table within TreeAge.


W e l e a d

Creating Tables in TreeAge Pro• A table consists of two columns.• The “Index” (first column) is the value that you will use to

refer to the corresponding number in the second column(“Value”).

• TreeAge Pro interprets the term “tablename[Index]” as the Value associated with the particular Index (note that the Index value has to be enclosed in [brackets]).

• You could enter Index/Value pairs one at a time by pulling down the TABLE menu and selecting ADD ENTRY.

• For now you will paste the copied Excel spreadsheet.


W e l e a d

Creating Tables in TreeAge Pro• Pull down the VALUES menu, and choose VARIABLES AND TABLES, then

click the TABLES button.• Now click NEW TABLE. You will need to give the table a name that will

be used to refer to it in your tree (Name) and a filename (File) for thetable file.

• Type muASR for both the table Name and File, then click OK.• To enter values into a table, click on OPEN TABLE – PASTE TABLE.• Then click CANCEL in the Add Table Entry dialogue box.• Note the entries in the Table: muASR window, click the SAVE button,

then close it.


W e l e a d

Creating Tables in TreeAge Pro• TreeAge Pro will automatically save the table file

(“muASR.tbl”) in the “Tables” subdirectory within the “TreeAge Pro” directory on the computer you are using.

• If you are going to use another computer with TreeAge Pro on the hard drive, you need to copy the table file onto a disk and then put it in the Tables subdirectory of the new computer, or save the tree as a package file (*.pkg), which saves both the tree file and its associated table files and allows the table files to be used with the tree when thepackage file is opened.


W e l e a d

Assigning Probabilities and Defining Variables• At this point you will redefine pDie as a function of age and

define pDieAIDS which combines the annual mortality of AIDS and the age related risk of dying from other causes.

• The equation for pDie (shown below) converts the annual rates from the life table (in the muASR table) to annual probabilities.

• The first cycle through the model, _stage is equal to 0.Hence, the rate selected from the table is that of thestarting age of the cohort (which will be 43).


W e l e a d

Assigning Probabilities and Defining Variables• The second cycle through the model, _stage is equal to 1

and the annual rate selected from the muASR table is that associated with 44‐year‐olds.

• The overall mortality rate for AIDS is equal to the sum of the rates of dying of other causes and the excess disease‐ specific mortality rate (defined as muAIDS).

• The equation for pDieAIDS (shown below) coverts the sum of these rates to annual probabilities.

• pDie = 1‐exp(‐muASR[Age+_stage])• pDieAIDS = 1‐exp(‐muASR[Age+_stage]‐muAIDS)


W e l e a d

Assigning Probabilities and Defining Variables• To define these variables, pull down the VALUES menu,

choose VARIABLES AND TABLES (or click the V= button on the Tool Bar).

• First select pDie from the variable list and the select pDieAIDS while holding down the Ctrl key.

• Click on DEFINE VARIABLE – DEFAULT FOR TREE and define these variables by the above equations.

• As before, when you enter the names of unrecognized variables, TreeAge Pro will want to create the variable.Click OK.


W e l e a d

Assigning Probabilities and Defining Variables• Since the probability of transition to HIV+ to AIDS will be

different depending on detection and treatment, you’ll define it locally at the Positive test node (Figure 6 below).

• Select the Positive test node, pull down the VALUES menu,and choose DEFINE VARIABLES (or click the V= button on theTool Bar).

• Select pGetAIDS and click on DEFINE VARIABLE – AT SELECTEDNODE and enter pAIDSwRx.

• Again TreeAge Pro will ask you if you want to create thevariable. Click OK.


W e l e a d

Assigning Probabilities and Defining Variables


W e l e a d

Assigning Probabilities and Defining Variables• Another way to change locally defined variables is to

double‐click the variable box under the node where the definition changes.

• Since patients with undetected HIV (either through a falsenegative test or no screening being done) will have differentutilities than those who know they are HIV (+) and highercosts once their disease is detected (but no costs prior todetection), we need to modify locally defined variables atthe HIV infected nodes off the Negative Test and No Screenbranches.


W e l e a d

Assigning Probabilities and Defining Variables• Double‐click the variable box below the HIV Infected node off the Negative Test

branch.• This brings up the Variable Definitions box (if the SHOW INHERITED box is checked,

uncheck it).• Double‐click the line that begins “CostInf =…”, bringing up the Define Variable box.• Cut the definition “CostNotDetect” by using EDIT – CUT or Ctrl+X, then click OK.• Confirm that you want to eliminate the definition by clicking OK. This takes you

back to the Variable Definitions box.• Click the ACTIONS button then UNDEFINED VARIABLES and select CostAIDS, bringing up

the Define Variable box.• Paste, using EDIT – PASTE or Ctrl+V, to input “CostNotDetect”, then click OK.


W e l e a d

Assigning Probabilities and Defining Variables• You’re back to Variable Definitions.• Click ACTIONS then UNDEFINED VARIABLES

• Select uHIVpos, then type uHIVundetected in the input box and click OK, then OK again to create the variable.

• Copy these variables by clicking ACTIONS then COPYVARIABLES.


W e l e a d

Assigning Probabilities and Defining Variables• Now double‐click the variable box below the HIV Infected

node off No Screen and click ACTIONS then PASTE VARIABLES.• Then, delete the definition of CostInf: double‐click the line

that begins “CostInf =…”, then delete the definition.• Confirm that you want to eliminate the definition by

clicking OK.• Close the Variable Definitions box by clicking the X in the

top right corner.


W e l e a d

Assigning Probabilities and Defining Variables• The remaining variables will be defined numerically.• Click the V= button on the Tool Bar.• Select the following variables while holding down the Ctrl

key: Age, CostAIDS, CostDetect, CostNotDetect, muAIDS, pAIDSwRx, pGetAIDS, uAIDS, uHIVpos, and uHIVundetected (CostDetect and CostNotDetect need to be modified to reflect yearly, rather than lifetime, costs).

• Click the DEFINE VARIABLE button, then DEFAULT FOR TREE.


W e l e a d

Assigning Probabilities and Defining Variables• Enter the following values when prompted and then click

OK or press Enter:


Variable

uHIVundetected

Value

0.91uHIVpos 0.88uAIDS 0.73

pGetAIDS 0.06pAIDSwRx 0.034

muAIDS 0.03CostNotDetect 27230 (AIDS in undetected HIV+)

Cost Detect 8250 (Costs of detected HIV+)CostAIDS 24750 (AIDS in previously detected HIV+)

Age 43

W e l e a d

Creating Clones• Select the HIV Infected Markov node, pull down the

OPTIONS menu and choose SELECT SUBTREE (the subtreenodes will become solid colors).

• Now pull down the EDIT menu and choose CREATE CLONE. TreeAge Pro will prompt you for a name for your clone master.

• Type HIV and then click OK.


W e l e a d

Creating Clones• The HIV Infected branch is thickened to indicate that it

represents a clone master.• Convert the two remaining HIV Infected terminal nodes to

Markov nodes.• Select the newly changed nodes, then pull down the EDIT

menu and choose ATTACH CLONE, select HIV then click OK.


W e l e a d

Creating Clones• Your tree should now look like Figure 7


W e l e a d

Setting Markov Termination Criteria and Running the Model• Once again, a Markov stopping criterion must be set at

each of the HIV Infected Markov nodes before the program will run.

• To do this, select an HIV Infected branch, pull down the VALUES menu, and select MARKOV TERMINATION (or click the M= button on the toolbar).

• Clicking OK will accept the default. Now set the default Markov termination value for the other HIV Infected branches.

• Termination criteria must be set individually for all 3 HIV Infected Markov nodes.


W e l e a d

Setting Markov Termination Criteria and Running the Model• Click the root node, then pull down the ANALYSIS menu,

choosing COST‐EFFECTIVENESS.• For the two graph display options, click OK. For a text

report, choose ACTIONS then TEXT REPORT.• To return to the tree, click OK on the text report, then FILE

CLOSE the graph without saving the graph.• Sensitivity analyses and Markov traces can be performed as

before (select the decision node, pull down the ANALYSIS menu, select SENSITIVITY ANALYSIS then ONE WAY, and then choose the variable and range to be examined).


W e l e a d

Setting Markov Termination Criteria and Running the Model• The graph should look something like:


Thank you

INRODUCTION TO TREEAGE PRO - InaHEAinahea.org/files/precongress/L3 Introduction to TreeAge Pro Markov... · INRODUCTION TO TREEAGE PRO: Markov ... – 3) Add utility weights ... purpose

Documents