Probablistic Neural Network Estimation of Postoperative PROMS after Hip and Knee Replacement Surgeries

Description

This presentation discusses the utilization of modeling approaches to predict post-operative patient-reported outcome measures (PROMs) specifically for hip and knee surgeries. Fabio De-Mello introduces the concept of Oxford hip and knee scores, detailing their structure and significance in evaluating surgical success. He emphasizes the variation in these scores over the past decade within the UK population and acknowledges the multitude of factors contributing to this variation, which current estimation models cannot fully address.



De-Mello proposes implementing a classifier neural network model to enhance the estimation of post-operative outcomes by accounting for uncertainties more effectively than traditional regression models. He illustrates using a hypothetical patient example, demonstrating how the model allows for a probabilistic assessment of expected patient outcomes and the likelihood of achieving a minimal important difference (MID) in outcome scoring. The presentation also compares the performance metrics of this new model against existing models, showcasing improvements in predictive accuracy and uncertainty measures.



During the Q&A segment, interactions include inquiries about the practical implementation of this model in clinical settings, including patient communication regarding expected outcomes based on these predictions. De-Mello discusses the integration of various patient attributes in the modeling process, further emphasizing the model's ability to guide decision-making in surgical treatment by quantifying uncertainty and expected benefits.

DOI: 10.1302/3114-221128

Specialties

Conferences