Open problems · P3
Informative sampling in clinical lab data
Clinical labs are dense but ordered when something is wrong. Joint models of when a test is taken and what it shows are needed before routine health-system data can be used.
Why it matters
Health-system labs are the only source dense enough to see individual trajectories at scale. Naive analysis of them confuses being tested with being ill.
What exists
- A planned retrospective analysis with Alberta laboratory data (application in preparation).
- Population curves to serve as priors.
A first step
Adapt a joint visit-process and outcome model to a simulated health-system schedule and quantify the bias of ignoring the visit process.