choice-metrics.com

Dear Professor, I'm here again to seek help, thank you in advance. My experimental background is oncologists making decisions on the treatment methods for cancer patients.
Because oncologists do not make decisions for themselves, but for patients, they will face different patient profiles.

Q1: My question is about how to design scenarios. Can it be implemented in Ngene software?
For example, I got 5 attributes for patient profiles (such as age>65 or not, with or without comorbid conditions, …), and 7 attributes for the discrete choice experiment.
I saw from the literature that the author's handling method is: “by obtaining descriptive information from physicians on attributes of patients from their practices, then developing attribute distributions to identify average splits. However, the author did not explain the details of how to operate.
I realized that if we assume that all 5 patient profile attributes are only 2 categories, there will be 32 scenario combinations, and then adding 7 DCE attributes (assuming I will generate 36 selection sets and divide them into 3 blocks) will make the experimental design and subsequent investigation very complex.

Q2: How to conduct subsequent analysis?
Do these 5 patient profile attributes still need to be included in the subsequent analysis?
If necessary, how to include it? Is it necessary to perform an interaction analysis between these attributes and DCE's attributes?

Thank you for patiently guiding me.

The patient profiles will enter the utility functions as scenario variables and have the same value across alternatives. The attributes will also enter the utility functions and can have different values across alternatives. So in generating the experimental design in Ngene as well as in model estimation you combine both in the utility functions.

You do not need to include all possible combinations of patient profiles and attribute levels, you only need to use a subset of them.

Below is an example script for Ngene with age and comorbid as scenario variables, and with side_effects and cost as attributes.

Code: Select all

design
;alts = chemo, radiotherapy, neither
;rows = 12
;eff = (mnl,d)
;model:
U(chemo)        = asc_chemo
                + b1            * senior[1,0]         ? 0 = below 65 (base), 1 = above 65
                + b2            * comorbid[1,0]       ? 0 = without (base), 1 = with comorbid conditions
                + b3.dummy[0|0] * side_effects[1,2,0] ? 0 = mild (base), 1 = moderate, 2 = severe
                + b4            * cost[100,200,300]
                /
U(radiotherapy) = asc_rt
                + b5            * senior[senior]      ? scenario variable
                + b6            * comorbid[comorbid]  ? scenario variable
                + b7            * side_effects
                + b4            * cost
$

Michiel

Hi,

Thanks for this really helpful topic. I have a similar question. I am also using a scenario attribute but I am using a single profile, binary choice design. People will imagine they are experiencing a symptom (scenario attribute) and then decide whether to have a test (Yes/No) that differs by four attributes. I'm wondering if this still needs to enter the model as a scenario attribute or whether it can just be included as an attribute like the other four within the utility model for having a test?

Thanks in advance!

Patient profiles are scenario variables but are added only to the utility function of the Yes OR the No alternative. But they are not attributes of Yes or No. Of course, the resulting mathematical model with attributes or with scenario variables may look the same in the case of 2 alternatives. So in that case, the variables enter the utility function exactly the same same as an attribute, yes.

Michiel

Thank you, Professor. I replied too late. I will refer to your example for design.