ABSTRACT
Over the past 5 years, there has been an increase in the development of EHR-based models for predicting suicidal behaviour. Using the McGinn (2000) framework for creating clinical prediction rules, this study discusses the broad validation of one such predictive model in a context external to its derivation. Along with reporting performance metrics, our paper high-lights five practical challenges that arise when trying to undertake such a project including (i) validation sample sizes, (ii) availability and timeliness of data, (iii) limited or incomplete documentation for predictor variables, (iv) reliance on structured data and (v) differences in the source context of algorithms. We also discuss our study in the context of the current literature.
