What matters most to older adults in treatment decision making: a discrete choice experiment.

Methods <br>
We performed a cross-sectional survey using a Discrete Choice Experiment (DCE) on older adults treatment decision making.
A fractional-factorial design was used, selecting a subset of combinations to maximize information on main effects and key interactions (D-efficiency) for a conditional logit model in Ngene 1.3 (ChoiceMetrix). <br>
Participants were volunteers from the Netherlands, aged 50+. <br><br>

Statistical Analysis<br>
A conditional logit model for the main effect of the attributes was derived. All attributed were treated as categorical variables with dummy-coding. The utility of all attributes was individually tested using a Wald chi square test.
Interaction testing was not included in the main model, but - based on literature and consensus of the study team - specific pre-specified interactions between attributes and personal characteristics (gender and physical pain, age and maintaining independence, educational level and societal costs, gender and independence) were tested using Wald tests.
Further, data was analyzed using latent class analysis. Class memberships were analyzed to define characteristics of the class members.<br><br>

Datafiles: <br>
1. Original raw data file with choice data (stata corp 2018) <br>
2. Long data file (stata corp 2018) <br>
3. DO file, syntax (stata corp 2018) <br>