Authors (including presenting author) :
Chan PK(1), Lee LS(1), Fung WC(1), Tang TCM(1), Wong LWA(1), Yeung SS(2), Ng YL(3), Chan VWK(1), Lau TW(1), Chiu KY(1)
Affiliation :
(1)Department of Orthopaedics and Traumatology, Queen Mary Hospital, The University of Hong Kong, (2)Department of Physiotherapy, MacLehose Medical Rehabilitation Centre, (3)Department of Occupational Therapy, MacLehose Medical Rehabilitation Centre
Introduction :
Knee joint replacement is a cost-effective procedure to reduce pain in patients suffering from knee osteoarthritis, but access to surgery is limited by long waiting times for initial assessment at orthopaedic specialist out-patient departments (SOPD). There is currently no standardized method to triage patients referred to O&T SOPD for knee pain, and the SOPD appointment is allocated based on limited information in referral letters without adequate prioritization based on severity.
Objectives :
(1) To identify factors associated with higher likelihood of requiring knee arthroplasty.
(2) To develop a prediction model for the need for knee arthroplasty prior to orthopaedic SOPD appointment allocation.
Methodology :
This retrospective cohort study included a total of 447 patients assessed at our clinic for knee pain from May 2019 to June 2021. Data was randomly divided into training (362 patients) and testing (85 patients) datasets using an 80:20 split. Based on clinical expertise and existing literature, 28 candidate predictors were identified, including demographics, comorbidities, radiological and clinical outcome measures, for the prediction of knee arthroplasty booking following orthopaedic surgeon assessment at SOPD. Statistical analysis was conducted using SPSS 27.0. Logistic regression analysis was performed to identify variables that improved prediction of knee arthroplasty. Model performance was quantified using area under receiver operating characteristics curve (AUC).
Result & Outcome :
Knee arthroplasty booking following orthopaedic surgeon assessment at SOPD was arranged for 44 (12.2%) and 14 (16.5%) patients in the training and testing datasets respectively. The final logistic regression model comprised four predictors, i.e., sex, age, pain on walking and range of motion, which were associated with higher likelihood of requiring knee arthroplasty. The model demonstrated promising predictive ability with an AUC of 0.820 and 0.762 for the training and testing datasets respectively.
This prediction model offers promising predictive ability to identify patients more likely to require knee arthroplasty, who can be prioritized for O&T SOPD appointments. This builds the foundation for future automated patient triage for SOPD booking to tackle the service bottleneck for knee osteoarthritis.
