Development and validation of a nomogram for predicting heterotopic ossification following spinal cord injury
Yulei Xie,Junwei Zhang,Xiaoqin Jin,Shujia Liu,Wei Song
DOI: https://doi.org/10.1016/j.clineuro.2024.108348
Abstract:Objective: Heterotopic ossification (HO) following spinal cord injury (SCI) can severely compromise patient mobility and quality of life. Precise identification of SCI patients at an elevated risk for HO is crucial for implementing early clinical interventions. While the literature presents diverse correlations between HO onset and purported risk factors, the development of a predictive model to quantify these risks is likely to bolster preventive approaches. This study is designed to develop and validate a nomogram-based predictive model that estimates the likelihood of HO in SCI patients, utilizing recognized risk factors to expedite clinical decision-making processes. Methods: We recruited a total of 145 patients with SCI and presenting with HO who were hospitalized at the China Rehabilitation Research Center, Beijing Boai Hospital, from June 2016 to December 2022. Additionally, 337 patients with SCI without HO were included as controls. Comprehensive data were collected for all study participants, and subsequently, the dataset was randomly partitioned into training and validation groups. Using Least Absolute Shrinkage and Selection Operator regression, variables were meticulously screened during the pretreatment phase to formulate the predictive model. The efficacy of the model was then assessed using metrics including receiver-operating characteristic (ROC) analysis, calibration assessment, and decision curve analysis. Results: The final prediction model incorporated age, sex, complete spinal cord injury status, spasm occurrence, and presence of deep vein thrombosis (DVT). Notably, the model exhibited commendable performance in both the training and validation groups, as evidenced by areas under the ROC curve (AUCs) of 0.756 and 0.738, respectively. These values surpassed the AUCs obtained for single variables, namely age (0.636), sex (0.589), complete spinal cord injury (0.681), spasm occurrence (0.563), and DVT presence (0.590). Furthermore, the calibration curve illustrated a congruence between the predicted and actual outcomes, indicating the high accuracy of the model. The decision curve analysis indicated substantial net benefits associated with the application of the model, thereby underscoring its practical utility. Conclusions: HO following SCI correlates with several identifiable risk factors, including male gender, youthful age, complete SCI, spasm occurrence and DVT. Our predictive model effectively estimates the likelihood of HO development by leveraging these factors, assisting physicians in identifying patients at high risk. Subsequently, correct positioning to prevent spasm-related deformities and educating healthcare providers on safe lower limb mobilization techniques are crucial to minimize muscle injury risks from rapid iliopsoas muscle extension. Additionally, the importance of early DVT prevention through routine screening and anticoagulation is emphasized to further reduce the incidence of HO.