Theory 92/Selection & Bias/Clinical epidemiology
Hospital Cohorts Are Selected
The claim
Associations observed in hospital cohorts are biased by referral, so a factor that increases admission can appear protective or harmful without affecting disease.
01Mechanism
Admission requires both the disease and the referral pathway; conditioning on admission induces associations absent in the population.
02Prediction
Associations should differ between hospital and population samples.
03Refutation
Identical associations in both would remove referral bias.