Background Self-reports of medical findings are indispensable in clinical practice and research but subject to recall bias. We analysed the recall accuracy of notifications about incidental findings (IFs) from a whole-body MRI examination and assessed determinants of recall error.

Methods Data from 3746 participants of a postal follow-up survey conducted on average 2.47 years after examination in the population-based Study of Health in Pomerania were analysed. Among those, 2185 (58.3%) underwent whole-body MRI at baseline, and findings of potential clinical relevance were disclosed in standardised postal letters. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated to determine the accuracy of self-reports. Poisson regression analysis was conducted to analyse predictors for false-positive and false-negative recall.

Results An IF was disclosed to 622 (28.5%) individuals; 81.5% had tumour relevance. The overall sensitivity and PPV of participants’ self-reports were 80% and 60%, respectively. PPvs were higher among women, better educated and married participants and among those with good verbal memory. Among MRI participants, lower educational level was associated with a higher risk of false-positive recall (risk ratio (RR) 1.44, 95% CI 1.01 to 2.03), while increasing age was associated with a higher risk of false-negative recall (RR 1.64, 95% CI 1.33 to 2.01).

Conclusions Most participants correctly recalled disclosed IFs. However, the probability of an event in case of a positive recall is barely above 50%. Therefore, relying on subjects’ recall of disclosed IFs will lead to a relevant proportion of errors. Clinicians and researchers should be aware of this problem and of participants’ characteristics which may moderate the probability of correct decisions based on recalled findings.