Background

Alcohol withdrawal and its consequences are a common concern for the large numbers of patients who present to emergency departments (EDs) with alcohol use disorders. While the majority of patients who go on to develop alcohol withdrawal experience only mild symptoms, a small proportion will experience seizures or delirium tremens. The aim of this study was to develop a tool to predict the need for hospital admission in patients at risk for alcohol withdrawal using only objective criteria that are typically available during the course of an ED visit.

Methods

We conducted a retrospective study at an academic medical center. Our primary outcome was severe alcohol withdrawal syndrome (SAWS), which we defined as a composite of delirium tremens, seizure, or use of high benzodiazepine doses. All candidate predictors were abstracted from the electronic health record. A logistic regression model was constructed using the derivation dataset to create the alcohol withdrawal triage tool (AWTT).

Results

Of the 2,038 study patients, 408 (20.0%) developed SAWS. We identified eight independent predictors of SAWS. Each of the predictors in the regression model was assigned one point. Summing the points for each predictor generated the AWTT score. An AWTT score of 3 or greater was defined as high risk based on sensitivity of 90% and specificity of 47% for predicting SAWS.

Conclusions

We were able to identify a set of objective, timely, independent predictors of SAWS. The predictors were used to create a novel clinical prediction rule, the AWTT.

中文翻译：

---

预测戒酒综合症的多变量模型戒酒分类工具（AWTT）的推导和验证

背景

戒酒及其后果是向急诊科（ED）提出酗酒障碍的众多患者普遍关注的问题。虽然大多数继续戒酒的患者仅出现轻度症状，但一小部分患者会出现癫痫发作或del妄。这项研究的目的是使用仅在ED访视过程中通常可用的客观标准，开发一种工具来预测有戒酒风险的患者入院的需求。

方法

我们在一家学术医学中心进行了回顾性研究。我们的主要结局是严重的酒精戒断综合症（SAWS），我们将其定义为ir妄症，癫痫发作或高剂量苯二氮卓类药物的综合。从电子健康记录中提取所有候选预测变量。使用推导数据集构建逻辑回归模型，以创建酒精戒断分类法工具（AWTT）。

结果

在2038名研究患者中，有408名（20.0％）患上了SAWS。我们确定了SAWS的八个独立预测因子。回归模型中的每个预测变量都分配了一个点。对每个预测变量的点求和得出AWTT得分。基于预测SAWS的90％敏感性和47％的特异性，AWTT得分为3或更高被定义为高风险。

结论

我们能够确定出一套客观，及时，独立的SAWS预测因子。预测变量用于创建新的临床预测规则AWTT。