Clinical prediction modelling is the backbone of precision medicine. However, a recent meta-analysis revealed that most prediction models are not validated in new, real-life settings (external validation) and consequently not considered for implementation in clinical practice. External validation, however, is costly and time-consuming, so a tool which can assess the robustness of a model in different settings would improve the likelihood of models being implemented in clinical practice and at lower costs.
This project aims to develop a user-friendly simulation model based on electronic health records of different boroughs to assess the robustness of clinical prediction models under different external validation scenarios.
The simulation model will be developed and validated using an existing prognostic model to predict the risk of developing psychoses and using data from the NIHR CRIS system allowing to compare simulation results with existing external validation results. New machine learning models will be developed and assessed using the new tool. The final simulation model will then be implemented as a user-friendly programme on an online platform allowing researchers to carry out their assessment of prognostic models to support translational research
Literature review (External validation, psychosis)
Attending relevant modules of our Data Science MSc
Establishing Patient and public focus group
Developing simulation framework
Extracting CRIS-EHR using NLP
Programming simulation tool
Validation using EHR
Refinement of simulation tool
Assessment of different machine learning models
Implementation on cloud database
Writing of thesis
Writing publications and reports to patients and public