Let’s start a post to discuss and list our current prediction problems of interest. I’ll update this as new problems get suggested and discussed.
One of our goals as a community of patient level prediction researchers is to establish best practices for prediction, and specifically best practice (e.g., how to evaluate performance, how to frame problem and standardizing output)) for each type of problem we discussed in the plp call.
The prediction types and current prediction problems are:
- Drug safety:
- Sprint trials (https://www.sprinttrial.org/public/dspHome.cfm) - e.g., warfarin and GI bleed
- new complications or confirm potential complications in post surveillance
- hypotension and fall in elderly due to hypotensive effects of medications
- Disease progression:
- newly diagnosed diabetes who will have complications
- predicting complications and hospitalization within 30 days of major surgery - also for certain subgroups (ethics - need to be careful this wont be used to discriminate)
- Others in literature?
- Treatment choice:
3a) evaluating propensity score model best practice - funding potential?
- problems from literature?
3b) Helping clinicians pick medicine to prescribe when unsure - problems from literature?
- Treatment adherence:
- disparities in treatment leading to discrepancies in adherence for secondary peripheral cardiovascular disease
- Treatment response:
- drugs used for secondary peripheral cardiovascular disease
- Health status:
- trauma exposure
- smoking
- alcohol
- BMI
Please let me know if I missed any problems mentioned in the call and I look forward to hearing more suggestions from the team.