Development and internal validation of a prediction model for the presence of tinnitus in a Dutch population-based cohort

Abstract

Objectives: In this study we aim to develop and internally validate a prediction model on tinnitus experience in a representative sample of the Dutch general population. Methods: We developed a multivariable prediction model using elastic net logistic regression with data from the Dutch Lifelines Cohort Study. This is a multigenerational cohort study on adults who are located in the northern parts of the Netherlands. The model was internally validated using 10-fold cross-validation. The outcome of the model was tinnitus presence, for which we used 24 candidate predictors on different domains (among others demographic, hearing specific, and mental health variables). We assessed the overall predictive performance, discrimination, and calibration of the model. Results: Data on 122.884 different participants were included, of which 7,965 (6.5%, 0 missing) experienced tinnitus. Nine variables were included in the final model: sex, hearing aids, hearing limitations, arterial blood pressure, quality of sleep, general health, symptom checklist of somatic complaints, cardiovascular risk factors, and age. In the final model, the Brier score was 0.056 and 0.787 in internal validation. Conclusion: We developed and internally validated a prediction model on tinnitus presence in a multigenerational cohort of the Dutch general population. From the 24 candidate predictors, the final model included nine predictors.