Impact of referral bias on prognostic studies outcomes: insights from a population-based cohort study on infective...

Prognostic studies derived from samples of patients managed in tertiary hospitals are subject to referral bias. We aimed to characterize this bias using the example of infective endocarditis.

We analyzed data from a French population-based cohort, which included 497 patients with infective endocarditis. Patients were admitted directly to a tertiary hospital (Group T), admitted to a non-tertiary hospital and referred to a tertiary hospital (Group NTT) or not (Group NT). We compared patients' characteristics, survival rates and prognostic factors between groups.

Compared with Group T (n = 291), NTT patients (n = 144) were more often males (81.3% vs. 72.5%; P = .046), injection drug users (9.7% vs. 4.5%; P = .033), and had more frequent surgical indications (78.5% vs. 64.3%; P = .003). Compared with Group NT (n = 62), NTT patients were more often males (81.3% vs. 67.7%; P = .034) and had surgical indications more often (78.5% vs. 19.4%; P < .001). One-year survival was higher in NTT + T patients than in NT patients (73.0% vs. 56.1%; P = .01). Prognostic factors and hazard ratios estimates varied across groups.

When derived from samples mixing patients admitted directly and those referred to tertiary hospitals, validity of characteristics description, survival estimates, and hazard ratios is threatened by referral bias.