Transdiagnostic mental health factors measuring anxious-depressive, compulsivity and intrusive thought and social withdrawal symptom dimensions have been identified as useful constructs that relate to a variety of cognitive constructs. However, the factor measurement relies on the administration of hundreds of questionnaire items which is both costly and a burden to participants. Additionally, it is unclear to what extent these factors generalise across datasets. Therefore, this study sought to optimise the measurement as well as assess the stability of these three factors. Using exploratory factor analysis on 209 questionnaire items, we replicated the same three-factor structure across a pooled dataset of n = 4782 participants, as well as within four independent subsets of the data. Using a machine learning approach, we reduced the number of items to 71 while still measuring the factors with high accuracy. We externally validated these factor scores by replicating previously observed associations with behaviour on a metacognition task. Our results support the generalisability of these symptom dimensions and provide a useful approach to optimising their measurement.