Abstract: BACKGROUND: The affected limb after total knee arthroplasty is often accompanied by swelling, especially around the knee joint, which can seriously affect the early postoperative rehabilitation and functional exercise. Therefore, it is important to analyze the factors influencing the swelling of the affected limb, and the prediction model is based on the regression analysis of the independent influencing factors, and the total score derived from the combination of the influencing factors can visually calculate the probability of swelling in patients with total knee arthroplasty.
OBJECTIVE: A column line graph prediction model was developed based on LASSO regression analysis of the independent influences on the occurrence of postoperative swelling in patients undergoing total knee arthroplasty.
METHODS: The clinical data of patients who underwent total knee arthroplasty in the orthopedic joint ward of The Third Hospital of Shanxi Medical University from January 1, 2020 to June 30, 2021 and from January 1, 2018 to May 31, 2019 were retrospectively analyzed and collected. The data were divided into a training population (n=168) and a validation population (n=122) according to the pre and post time period of data collection. LASSO regression analysis was first performed based on the training population to screen for independent influences on the occurrence of postoperative swelling in patients undergoing total knee arthroplasty. Logistics univariate and multivariate regression was performed by screening the initial independent influencing factors. Finally, a column line graph prediction model of postoperative occurrence of swelling in total knee arthroplasty patients was produced by these influencing factors. Receiver operating characteristic curve and its area under the curve, C-index validation, and calibration curve were used to initially evaluate the model discrimination and calibration degree. Model validation was performed by the validation set. The C-index and calibration curve were used to further evaluate the performance of the column line graph model. Finally, decision curve analysis curves were used to see if the model could be used better in clinical practice.
RESULTS AND CONCLUSION: (1) LASSO regression analysis showed that the modeling population had significant significance in the duration of knee osteoarthritis, body mass index, and intraoperative blood loss. (2) Based on the influencing factors and existing theories, a column line graph prediction model for the occurrence of swelling in the affected limb after surgery in total knee arthroplasty patients was constructed with the receiver operating characteristic curve and its area under the curve (area under the curve=0.68) in the training set, and also the receiver operating characteristic curve in external validation population and its area under the curve (area under the curve=0.67), which finally showed that the model was effective in predicting the column line graph. (3) The C-index was [0.663 (95%CI: 0.487-0.839)] in the modeling cohort and[0.655, 95%CI: 0.537-0.772)] in the validation group, indicating that the prediction accuracy of the model was quite good. The calibration curve fit was good. (4) The decision curve analysis curve showed that the model would have good results in clinical use. (5) The final results indicate that this prediction model for the occurrence of swelling in the affected limb after total knee arthroplasty has good testing efficacy, which can help to clinically screen the possibility of swelling in the affected limb after total knee arthroplasty patients and give personalized interventions to different patients in a timely manner. 

Key words: total knee arthroplasty, recovery, swelling, nomogram, predictive model