Using real world patient data in your model - bootstrapping

If you have real world patient data which is available to describe the characteristics of the patients (trials), you may want to use that information in a model.

This needs to occur in two stages. First, you need to load the patient profile data into the model and second you need to assign that data to the individual trials that will run through your model.

Please see the attached Example14.2 - Bootstrapping- dist.trex.

How to load real world data into a model

Open the model and look at the Table View. The table called Profile has been created from profile data for 10 patients. Each row is a different patient profile. Each column is a different patient characteristic, column 1 is for Start Age and column 2 is for Gender.

The data for the 10 patients could have been loaded into the table Profile by either of the following methods:

1. If a table of data exists in Excel, with a row for each patient and a column for each characteristic, then copy and paste the data including the index row and the column headings into a TreeAge Pro table. Use the toolbar at the top of the table data to paste the data.

2. Create the table and manually enter the data for each patient.

How to incorporate real world data into your model

The Example model includes a uniform distribution, Trial0Profile, which will randomly sample a profile number for each trial sent through the model. The sampled profile number is then used as the index value to pull characteristics data for the appropriate row in the Profile table. Table lookups pull data from the Profile table based on the profile row and the appropriate characteristic column.

Characteristics can be assigned to either variables or trackers. Variables are a little easier in that a variable definition at the root node can assign the appropriate characteristic based on the proper row and table. However, trackers have the advantage of including the patient characteristics into the output data from the Microsimulation. If trackers are used, then you need to assign the trackers at all the places where a trial can enter the model (usually every health state with a non-zero initial probability and for every strategy).

In this model we use two trackers called Trial1StartAge and Trial2Gender. These trackers are assigned values using the following syntax to extract information from the table:

Trial1StartAge = Profile[Trial0Profile; 1];

where Trial0Profile is the index and 1 is the column from the table Profile where the Start Age is stored. Similarly:

Trial2Gender = Profile[Trial0Profile; 2];

where Trial0Profile is the index and 2 is the column from the table Profile where the Gender is stored.

Now, anywhere in the model where the StartAge and Gender are required, the model can refer to Trial1StartAge and Trial2Gender and the real world data will be incorporated.

Note that this model only demonstrates the bootstrapping technique. There is no disease model structure. However, additional structure could be added to a model like this to use bootstrapping data for a real decision model.

If you need additional information about this article, please send us a support ticket.

Was this article helpful?
0 out of 0 found this helpful
Have more questions? Submit a request



Article is closed for comments.