Risk factors and predictive models for frozen shoulder

Abstract

AbstractThis study aims to explore the risk factors associated with frozen shoulder (FS) and develop a predictive model for diagnosing FS, in order to facilitate early detection of the condition. A total of 103 patients diagnosed with FS and admitted to the Department of Joint Surgery at Suining Central Hospital between October 2021 and October 2023 were consecutively included in the study. Additionally, 309 individuals without shoulder joint diseases, matched for age and gender, who visited the department during the same time, were included as the control group.The complete recording of clinical data for all patients was followed by the utilization of statistical tests such as the Mann–Whitney U test, sample t test, and chi-square test to compare different groups. Additionally, multivariate binary logistic regression analysis was employed to identify risk factors associated with the occurrence of FS in patients, leading to the establishment of a prediction model and derivation of a simplified equation. The diagnostic effectiveness of individual indicators and prediction models was assessed through the use of receiver operating characteristic (ROC) curve analysis. In the sample of 103 individuals, 35 were identified as male and 68 as female, with an average age range of 40–70 years (mean age: 54.20 ± 6.82 years). The analysis conducted between different groups revealed that individuals with a low body mass index (BMI), in conjunction with other factors such as diabetes, cervical spondylosis, atherosclerosis, and hyperlipidemia, were more susceptible to developing FS. Logistic regression analysis further indicated that low BMI, diabetes, cervical spondylosis, and hyperlipidemia were significant risk factors for the occurrence of FS. These variables were subsequently incorporated into a predictive model, resulting in the creation of a simplified equation.The ROC curve demonstrated that the combined indicators in the predictive model exhibited superior diagnostic efficacy compared to single indicators, as evidenced by an area under the curve of 0.787, sensitivity of 62.1%, and specificity of 82.2%. Low BMI, diabetes, cervical spondylosis, and hyperlipidemia are significant risk factors associated with the occurrence of FS. Moreover, the utilization of a prediction model has demonstrated superior capability in forecasting the likelihood of FS compared to relying solely on individual indicators. This finding holds potential in offering valuable insights for the early diagnosis of FS.