Abstract: BACKGROUND: Cervical instability is a common orthopedic disease in young and middle-aged people, and is the early manifestation of cervical spondylosis, which has a great impact on the quality of life of patients. Therefore, early diagnosis of cervical instability to implement early intervention has positive clinical and social significance. 
OBJECTIVE: The clinical prediction model of cervical instability in young and middle-aged people was constructed based on machine learning to realize early screening of cervical instability in young and middle-aged people before X-ray examination.
METHODS: From September 2022 to October 2023, 155 young and middle-aged adults with cervical instability and 88 with non-cervical instability recruited through recruitment advertisements and spinal department outpatient of Wangjing Hospital, China Academy of Chinese Medical Sciences were selected as research subjects. The research subjects’ general information, living and working habits, discomfort symptoms, visual analog scale score, Neck Disability Index, and 36-Item Short Form Health Survey were collected on site based on questionnaires. The above information was used as predictive factors. After screening, six machine learning algorithms of Support Vector Machine, LightGBM, RandomForest, Logistic, AdaBoost, and XGBClassifier were used to train the model by ten-fold cross-validation method, and the clinical prediction model of cervical instability was constructed. Area under the curve was used as the main evaluation index. Univariate analysis was performed on the predictors, and SHAP method was used to rank the importance of the predictors. Correlation heat maps were used to show the degree of linear correlation between the predictors and the cervical instability.  
RESULTS AND CONCLUSION: (1) Among the six machine learning models, RandomForest model was chosen as the final prediction model, including nine predictors, such as age, body mass index, neck circumference/neck length, visual analog scale score, Neck Disability Index, bodily pain, general health, vitality, and mental health, area under the curve =0.725 4, and the calibration degree was good. It could be used as a reference tool for early screening of cervical instability in young and middle-aged people. (2) There were significant differences in age, visual analog scale score, Neck Disability Index, bodily pain, general health, and vitality between the two groups (P < 0.05). (3) The order of importance of predictors was age, Neck Disability Index, visual analog scale score, general health, body mass index, vitality, bodily pain, neck circumference/neck length, mental health, among which age, visual analog scale score, Neck Disability Index were positively correlated with cervical instability, while general health, body mass index, vitality, bodily pain, neck circumference/neck length, and mental health were negatively correlated with cervical instability.

Key words: cervical instability, clinical prediction model, machine learning, young and middle-aged people, predictor