L13 Flashcards
Why are DCEs used?
Is often difficult to attach a monetary value to a HC benefit - eg. how much is a kidney worth???
2 approaches to benefit valuation?
revealed preferences (value is revealed by observing the spending they do (ie. inference rather than asking)) and stated preferences (see last lecture)
What is the theoretical basis for DCEs?
People derive utility from underlying attributes of a good or service therefore to understand the value of a G/S we must look at the value of these different attributes
Explain why DCEs are not perfect?
Some randomness to choices such as tastes/measurement error
DCE equation for total utility?
TU = systematic (explained) amount + random (unexplained) component
What is the systematic component?
The explained utility derived from different attributes of a product
What, in words, was Lancaster’s theory?
That could model utility for a good, k, with a regression eqn. with each attribute X, betas telling its relative explanatory power and error for random component of choices
2 ways DCEs might be used in health econ?
1) HC preferences (eg. on childbirth services)
2) HC outcomes and preferences on allocation of resources towards services and research
7 steps for carrying out a DCE?
1) Specify Q of interest
2) Define attributes and levels
3) Create experimental design
4) Develop and test the questionnaire
5) Administer the questionnaire and collect answers
6) Data analysis
7) Results interpretation
Socks Don’t Create DADs Running
3 things that should be specified when specifying Q of interest in a DCE?
Alternative options of interest (eg. phone vs. physical consultation)
Choice format (eg. multiple choice, yes/no)
Who will be making the choice (eg. gen public/doctors etc.)
How should the definition of attributes be carried out? How may attributes be established?
Should identify IMPORTANT attributes of the good/service (eg. length of consultation, convenience level etc.)
The attributes should be meaningful and reflect differences between the good or service
(may use focus groups or literature to find these attributes)
How would the creation of experimental design be theoretically vs in practice?
Theoretically, would have all possible combos of levels and attributes but this can result in hundreds of Qs
Practically, use a FRACTIONAL FACTORAL design and infer an estimation of values of different characteristics of a G/S (is much more convenient)
Additional considerations that may be made in the developing and testing phase of a DCE?
Opt-out scenario
Number of questions asked to each person
‘warm-up- Qs
Main step in develop and testing phase of DCE?
Pilot the questionnaire to check burden and understanding
How might questionnaires vary in their administration?
eg. online/face-to-face
Problem with online surveys?
Often observe laziness tf inaccurate answers
What is the data analysis stage?
Regression analysis, model depends on Qs being asked (eg. binary variables)
What can the results in the final stage of a DCE be used for? (3)
Design new services
Predict Demand of existing services
Estimate the value placed on a service’s attribute
What 5 attributes did Bessen et al. (2014) include in their study which looked at ‘what are the preferences of breast cancer survivors for modes of delivery of follow up services?’?
1) Provider (GP/nurse etc.)
2) Frequency (6mnths, 9 or yrly)
3) Location (hospital, GP, breast cancer clinic etc.)
4) Method of delivery (F2F, telephone etc.)
5) Attributes of drop-in clinic (eg. secondary prevention, psychological support, side-effects treatment)
Explain how Bessen et al.’s survey worked?
254 possible choices broken down into 18 binary (A/B) choice sets split into 3 blocks of 6 questions
tf each person asked to choose between 6 pairs of programmes that varied in the 5 attributes!
What did Bessen et al.’s results show? (3)
Strong preference for:
1) Breast cancer physician
2) Breast cancer follow up clinic, either F2F or both
3) Slight preference for 6mnth regularity