Bazmi Elham, Behnoush Behnam, Hashemi Nazari Saeed, Khodakarim Soheila, Behnoush Amir Hossein, Soori Hamid
Department of Epidemiology, Scho ol of Public Health and Safety, Shahid Beheshti University of Medical Sciences, Tehran, Iran.
Department of Forensic Medicine, Tehran University of Medical Sciences, Tehran, Iran.
Arch Acad Emerg Med. 2020 May 17;8(1):e59. eCollection 2020.
Seizure is a common complication of tramadol poisoning and predicting it will help clinicians in preventing seizure and better management of patients. This study aimed to develop and validate a prediction model to assess the risk of seizure in acute tramadol poisoning.
This retrospective observational study was conducted on 909 patients with acute tramadol poisoning in Baharloo Hospital, Tehran, Iran, (2015-2019). Several available demographic, clinical, and para-clinical characteristics were considered as potential predictors of seizure and extracted from clinical records. The data were split into derivation and validation sets (70/30 split) via random sampling. Derivation set was used to develop a multivariable logistic regression model. The model was tested on the validation set and its performance was assessed with receiver operating characteristic (ROC) curve.
The mean (standard deviation (SD)) of patients' age was 23.75 (7.47) years and 683 (75.1%) of them were male. Seizures occurred in 541 (60%) patients. Univariate analysis indicated that sex, pulse rate (PR), arterial blood Carbone dioxide pressure (PCO), Glasgow Coma Scale (GCS), blood bicarbonate level, pH, and serum sodium level could predict the chance of seizure in acute tramadol poisoning. The final model in derivation set consisted of sex, PR, GCS, pH, and blood bicarbonate level. The model showed good accuracy on the validation set with an area under the ROC curve of 0.77 (95% CI: 0.67-0.87).
Representation of this model as a decision tree could help clinicians to identify high-risk patients with tramadol poisoning-induced seizure and in decision-making at triage of emergency departments in hospitals.
癫痫发作是曲马多中毒的常见并发症，对其进行预测将有助于临床医生预防癫痫发作并更好地管理患者。本研究旨在开发并验证一种预测模型，以评估急性曲马多中毒患者癫痫发作的风险。
本回顾性观察性研究对伊朗德黑兰巴哈洛医院909例急性曲马多中毒患者（2015 - 2019年）进行。将几个可用的人口统计学、临床和辅助临床特征视为癫痫发作的潜在预测因素，并从临床记录中提取。通过随机抽样将数据分为推导集和验证集（70/30分割）。推导集用于建立多变量逻辑回归模型。该模型在验证集上进行测试，并通过受试者工作特征（ROC）曲线评估其性能。
患者的平均（标准差）年龄为23.75（7.47）岁，其中683例（75.1%）为男性。541例（60%）患者发生癫痫发作。单因素分析表明，性别、脉搏率（PR）、动脉血二氧化碳分压（PCO）、格拉斯哥昏迷量表（GCS）、血碳酸氢盐水平、pH值和血清钠水平可预测急性曲马多中毒患者癫痫发作的可能性。推导集中的最终模型包括性别、PR、GCS、pH值和血碳酸氢盐水平。该模型在验证集上显示出良好的准确性，ROC曲线下面积为0.77（95%CI：0.67 - 0.87）。
将该模型表示为决策树可帮助临床医生识别曲马多中毒所致癫痫发作的高危患者，并在医院急诊科分诊时进行决策。
