JMIR Publications

ABSTRACT

Background:

Hip and knee osteoarthritis is incredibly prevalent worldwide, with large numbers of older adults undergoing joint replacement (arthroplasty) every year. A backlog of elective activity due to the COVID-19 pandemic, and also an ageing population, has led to significant issues with access to timely arthroplasty surgery. One potential method of increasing the efficiency of arthroplasty services is by improving the percentage of patients referred from primary care who are listed for surgery. Use of Artificial Intelligence techniques, specifically Machine Learning, provide a potential unexplored solution to select suitable patients correctly and rapidly for arthroplasty surgery.

Objective:

1. Develop a cohort of patients referred by GPs regarding assessment of suitability for hip or knee replacement from NHS Grampian data via the Grampian DaSH. 2. Determine demographic, clinical and/or imaging characteristics influential in the selection of patients to undergo hip or knee arthroplasty, with development of a tested and validated patient specific predictive model to guide arthroplasty referral pathways.

Methods:

The project will be delivered through two linked work packages conducted within the Grampian Data Safe Haven and Safe Haven Artificial Intelligence Platform. The first will include a cohort of individuals ≥16 years referred for consideration of elective primary hip replacement or primary knee replacement between January 2015 to January 2022. Linked pseudonymised NHS Grampian healthcare data will be acquired including patient demographics, medication, laboratory data, theatre records, text from clinical letters and radiological images/reports. Following creation of the dataset, machine learning techniques will be utilised to develop pattern classification & probabilistic prediction models based on radiological images. Supplemental demographic and clinical data will be used to improve the predictive capabilities of the models. The sample size is predicted to be approximately 2000 patients - sufficient for satisfactory assessment of the primary outcome. Cross-validation will be utilised for development, testing and internal validation. Evaluation will be performed through standard techniques such as the Area Under Curve (AUC) / C-statistic metric, Calibration characteristics (Brier Score), and a confusion/classification matrix.

Results:

The study was funded by the Chief Scientist Office (CSO) Scotland as part of a Clinical Research Fellowship which will run from 08/2021 to 08/2024 (Grant ref: CAF/21/06 – Appendix 3). Approval from the North Node Privacy Advisory Committee (NNPAC) was confirmed on 13/10/2021 (Appendix 4). The proposed start of data collection is March 2022, with the results expected to be published in Q1 2024. ISRCTN registration is currently underway.

Conclusions:

This project provides a first step towards delivering an automated solution for arthroplasty selection utilising routinely collected healthcare data. Following appropriate external validation and clinical testing this could significantly improve the proportion of referred patients that are selected to undergo surgery, with a subsequent reduction in waiting time for arthroplasty appointments.