The aim of this study is to investigate the characteristics of physical function that influence the achievement of the MCID in postoperative ODI scores for patients with lumbar degenerative diseases using Random Forest. Understanding physical function affecting MCID achievement allows for a more accurate predictive model.
The study design was a retrospective cohort study. The subjects were patients who underwent lumbar interbody fusion or decompression surgery for lumbar degenerative disease at our hospital and were followed up for one year postoperatively. Patients who underwent reoperation, had spinal surgery for bone tumors, infections, or trauma, underwent spinal corrective surgery for adult spinal deformities, had severe comorbidities, or had dementia were excluded. The main outcome was the achievement of the MCID in the ODI one year after surgery. The following variables were assessed preoperatively: gender, surgical procedure, history of comorbidities (hypertension, heart disease, diabetes, psychiatric disorders), social background (living situation, history of alcohol consumption, smoking), preoperative ODI, age, BMI, the physical (PCS), mental (MCS), and social (RCS) components of the MOS 36-Item Short-Form Health Survey, presence of motor and sensory deficits, Functional Reach Test (FRT), 30-second chair stand test (CS30), Visual Analogue Scale (VAS) for low back pain and leg pain, range of motion of the trunk (flexion and extension), trunk muscle strength [trunk flexion and extension torque, and the ratio of extension to flexion strength (E/F ratio)]. Random forest analysis was performed using the achievement of MCID as the outcome and the other variables as predictors, and the features importance were ranked according to their contribution to achieving MCID.
A total of 150 patients (69 females, mean age ± standard deviation（SD）: 69.4 ± 11 years, mean BMI ± SD: 25.4 ± 3.8) were included in the study. Of these, 91 patients achieved MCID, while 59 did not. The random forest model showed an OBB estimate of error rate of 31% for predicting ODI outcomes. The factors most strongly associated with achieving MCID, in order of importance, were preoperative ODI, PCS, leg pain, trunk extension torque, extension flexibility, and low back pain.
This study was the first to clarify the priority of physical function characteristics. In the future, it will be necessary to consider constructing prediction models based on the physical functions identified in this study.
This could provide valuable information for predicting postoperative satisfaction in patients with lumbar degenerative disease.
Postoperative lumbar degenerative disease
Oswestry Disability Index
