Predicting the probabilities of missed general practice appointments in England and Wales

Abstract

Missing general practice (GP) appointments represent an important challenge for healthcare systems. In England and Wales, reducing the number of missed appointments would benefit both the National Health Service (NHS) and the patients, avoiding delay in diagnosis and treatment. Since the COVID-19 pandemic, appointment mode has shifted substantially, and many GP practices have started scheduling online appointments in place of face-to-face meetings. In this context, our aim was to build and compare prediction models for the probability of missing a GP appointment, as a function of appointment’s characteristics and the level of deprivation of the area where the GP practice is located. We examined all English GP appointments in 2021 and used two different statistical approaches for prediction: a generalized linear model (logistic regression) and a machine learning approach (Extreme gradient boosting). Predictions were further validated with 2022 data. Both approaches provided comparable predictions in term of calibration, with the advantage that results from the logistic regression can be interpreted as odds ratios. Longer time between booking and appointment plays an important role, as well as deprivation. Deprived areas, which already tend to have lower healthcare standards, may also be losing more resources from cancelled and unattended appointments compared to their less deprived counterparts. Investigating the role of contextual factors behind these inequalities (both within and outside the healthcare system) would be an important step forward.