Study on the random forest algorithm for predicting the risk of acute pain in elderly patients with knee osteoarthritis after arthroscopy

Objective: To explore the predictive value of random forest algorithm in the early acute pain risk of elderly patients with knee osteoarthritis after knee arthroscopy. Methods: The elderly patients with knee osteoarthritis scheduled by knee arthroscopy were prospectively analyzed,and 200 patients who met the inclusion and exclusion criteria were selected as the research subjects.The general information of patients was collected before 1 day of surgery,and the condition of patients was statistically analyzed after 3 days of surgery.The protective and risk factors of early postoperative acute pain in elderly patients with knee osteoarthritis were analyzed through univariate and multivariate analysis.The importance of each variable was ranked using the random forest algorithm to establish a predictive model for early postoperative acute pain. Results: The incidence rate of acute pain in elderly patients with knee osteoarthritis after surgery was 61.42%(121/197).The differences of the degree of preoperative articular cartilage injury,postoperative quadriceps function exercises,postoperative PCIA,addition of analgesic drugs,age,preoperative pain level during activity,preoperative knee function(KSS) score,fearful visual analog scale(FAVS),pain catastrophizing scale(PCS),and total score of the hospital anxiety and depression scale(HADS) were statistically significant between two groups(P<0.05 to P<0.01).The postoperative quadriceps function exercise(OR=0.357,P<0.01),postoperative PCIA(OR=0.261,P<0.01) and addition of analgesic drugs(OR=0.430,P<0.01) were the protective factors for early postoperative acute pain in elderly patients with knee osteoarthritis.The preoperative joint cartilage injury severity(OR=5.607,P<0.01),age(OR=4.555,P<0.01),FAVS(OR=5.214,P<0.01),PCS(OR=4.138,P<0.01) and total HADS score(OR=6.798,P<0.01) were the risk factors for acute pain.Based on logistic regression analysis,a random forest model was constructed to predict the risk of early postoperative acute pain in elderly patients with knee osteoarthritis.The area under the ROC curve was 0.892,the sensitivity was 77.27%,and the specificity was 88.89%. Conclusions: The postoperative quadriceps function exercises,postoperative PCIA,addition of analgesic drugs,degree of preoperative articular cartilage injury,age,FAVS,PCS,and HADS total score are the influencing factors of early postoperative acute pain in elderly patients with knee osteoarthritis.Constructing a predictive model for prediction and importance analysis is beneficial for guiding the precise implementation of prevention and control measures in clinical practice.