Abstract
BACKGROUND: The standardization of uveitis nomenclature (SUN) working group is an international expert committee, which follows the aim to develop a standardized and internationally recognized terminology for the field of uveitis. This appears to be important in view of the demand for evidence-based medicine, especially for relatively rare diseases such as uveitis. METHODS: A databank of > 4000 uveitis patients was compiled using formal consensus techniques, for whom a majority consensus was previously achieved in the diagnosis. The patient data were analyzed within the subclasses of uveitis and divided into a training set and a validation set. Multinomial logistic regressions with LASSO regularization were carried out on the training set with machine learning (ML). The accuracy of the rules that were developed to express the criteria of ML, were assessed by a masked observer in a random sample of 10%. RESULTS: The estimations of total accuracy according to the uveitis classes in the validation set were high for all forms of uveitis: anterior uveitis 96.7% (95% confidence interval, CI 92.4-98.6%), intermediate uveitis 99.3% (95% CI 96.1-99.9%), posterior uveitis 98.0% (95% CI 94.3-99.3%), panuveitis 94.0% (95% CI 89.0-96.8%) and infectious posterior uveitis/panuveitis 93.3% (95% CI 89.1-96.3%). CONCLUSION: Classification criteria are presented, which show a high degree of accuracy (low misclassification rates) and are therefore suitable for future clinical and translational research.