Prediction models in clinical guidelines: a scoping review of clinical guideline guidance development documents

Objectives Probability or risk outcome estimates tailored to individual patient characteristics can facilitate more personalised care, for example, regarding treatment. Yet, the uptake of prediction models in clinical guidelines is lacking. It is unknown whether documents or handbooks for clinical guideline development provide any specific guidance, similar to clinical treatments, for how to assess and incorporate prediction models in clinical guidelines. We performed this systematic review of clinical guideline development documents and handbooks to explore the methodology regarding searching, selecting, critically appraising and implementing prediction models in clinical guidelines to enhance their usability in decision-making. Design and Setting We conducted a comprehensive systematic search for clinical guideline development documents of organisations included in the Guidelines International Network members directory and identified additional handbooks via PubMed and the TRIP database in October 2024. Data extracted from the retrieved documents included organisational details, methods and recommendations and guidance for prediction model inclusion. A narrative synthesis was performed. Results A total of 84 guidance documents were included in the final extraction, with eight (10.5%) guidance documents mentioning specific guidance for prediction models or related terms such as prognostic models or risk prediction. The included guidance related to searching, selecting and critically appraising prediction models for inclusion in clinical guidelines. No guidance on model performance or selection between competing models was provided. Conclusions The minority of identified handbooks included guidance regarding appraising and deciding whether and how to include prediction models in clinical guidelines. The absence of such guidance underscores the need for more extensive and standardised methodology for incorporating prediction models in clinical guidelines. This could help facilitate implementation of models into clinical guidelines, clinical practice and eventually impact decision making and subsequently patient outcomes.