Course talk:CPSC522/Patient Centric Decision Support The Utility Assessment Method

I think your article was overall well written. I liked how you explained the problems with support systems that utility assessments solved. I found some of the information in the first two paragraphs repetitive. I was also a little confused about how utility assessments get rid of the bias in AI support systems. Aren't these support systems trained on historical data still? So even with the utility assessment won't there still be bias. Also, you could add some more information about what kind of utility assessments could be used.

Thank you, Kate. I've made efforts to eliminate redundancy wherever identified.

Regarding the utility assessment approach, I'm leaning towards employing a VAS-based method.

In response to your feedback on bias, I've made updates to the article. Despite our reliance on historical data, continuous updates serve as a means of mitigating this concern in our particular case.

The aim of the project is to advocate for a proposal that can be implemented. Suppose you are in a team "The other team members need to understand the proposal, what is involved for implementing it..." After reading your proposal, I would not know what to do. Where do we start? What would we actually implement?

For example, when do we do preference elicitation? Do we do it up-front (like when someone writes a will when they are healthy) or only when there is a decision to be made?

You advocate "assigning a utility value to each possible outcome" - surely there are always too many possible outcomes. How would you expect they do this? How would the team go about determining each possible outcome, and how would you suggest the patients are asked?

An example would help.

Thank for you useful comments Professor Poole. I updated my article with adding more detail in my approach.

In my opinion, it is not a good idea to ask patients direct questions. Not only they feel a pressure on them that can make them choose unwisely but also they might not have enough knowledge about some medical situations to decide between them. Some of the information in 'The Utility Assessment Approach' section is repeated in 'Incorporating Patient Preferences in Decision-Support Systems' section. I think you can better organize your page. Your page is full of abstract solutions, and I don't know how those methods can be implemented practically. Your approach takes a lot of time to implement. You need to ask every patient lots of questions. I think you can reduce this time by using a combination of patient-specific and historical data.

Thank you for your response, Fardad. I have revised my article to provide additional details regarding my proposed solution. Regarding your point about the limitations of directly asking patients, I acknowledge that every method has its drawbacks, and the choice of methodology depends on our specific goals. I fully concur with your suggestion of utilizing a combination of patient-specific and historical data. In fact, I explicitly mentioned in the bias section that we incorporate historical data into our approach and continuously update it.