Turley Ben, Zamore Kenan, Holman Robert P
DC Department of Health, Washington, DC, USA.
CDC Foundation, Atlanta, GA, USA.
Addiction. 2025 Feb;120(2):296-305. doi: 10.1111/add.16686. Epub 2024 Oct 12.
Patient initiated transport refusal during Emergency Medical Service (EMS) opioid overdose encounters has become an endemic problem. This study aimed to quantify circumstantial and environmental factors which predict refusal of further care.
In this cross-sectional analysis, a case definition for opioid overdose was applied retrospectively to EMS encounters. Selected cases had sociodemographic and situational/incident variables extracted using patient information and free text searches of case narratives. 50 unique binary variables were used to build a logistic model.
Prehospital EMS overdose encounters in Washington, DC, USA, from July 2017 to July 2023.
Of EMS encounters in the study timeframe, 14 587 cases were selected as opioid overdoses.
Predicted probability for covariates was the outcome variable. Model performance was assessed using Stratified K-Fold Cross-Validation and scored with positive predictive value, sensitivity and F1. Prediction accuracy and McFadden's pseudo-R squared are also determined.
The model achieved a predictive accuracy of 78% with a high positive predictive value (0.83) and moderate sensitivity (0.68). Bystander type influenced the refusal outcome, with decreased refusal probability associated with family (nondescript) (-28%) and parents (-16%), while presence of a girlfriend increased it (+28%). Negative situational factors like noted physical trauma (-62%), poor weather (-14%) and lack of housing (-14%) decreased refusal probability. Characteristics of the emergency response team, like a prior crew member encounter (+20%) or crew experience <1 year (-36%), had a variable association with transport.
Refusal of emergency transport for opioid overdose cases in Washington, DC, USA, has expanded by 43.8% since 2017. Several social, environmental and systematic factors can predict this refusal. Logistic regression models can be used to quantify broad categories of behavior in surveillance medical research.
在紧急医疗服务（EMS）处理阿片类药物过量情况时，患者主动拒绝转运已成为一个普遍问题。本研究旨在量化预测拒绝进一步治疗的情况和环境因素。
在这项横断面分析中，阿片类药物过量的病例定义被回顾性应用于EMS接触情况。通过患者信息和病例叙述的自由文本搜索，提取选定病例的社会人口统计学和情境/事件变量。使用50个独特的二元变量构建逻辑模型。
2017年7月至2023年7月在美国华盛顿特区的院前EMS阿片类药物过量接触情况。
在研究时间段内的EMS接触情况中，14587例被选为阿片类药物过量病例。
协变量的预测概率为结果变量。使用分层K折交叉验证评估模型性能，并以阳性预测值、敏感性和F1评分。还确定了预测准确性和麦克法登伪R平方。
该模型的预测准确率为78%，具有较高的阳性预测值（0.83）和中等敏感性（0.68）。旁观者类型影响拒绝结果，与家人（无特定描述）（-28%）和父母（-16%）相关的拒绝概率降低，而有女朋友则会增加拒绝概率（+28%）。负面情境因素，如明显的身体创伤（-62%）、恶劣天气（-14%）和无住房（-14%）会降低拒绝概率。应急响应团队的特征，如之前与机组人员接触（+20%）或机组人员经验<1年（-36%），与转运有不同的关联。
自2017年以来，美国华盛顿特区阿片类药物过量病例拒绝紧急转运的情况增加了43.8%。几个社会、环境和系统因素可以预测这种拒绝。逻辑回归模型可用于量化监测医学研究中的广泛行为类别。
