Laboratory of Biostatistics, Epidemiology, Clinical Research and Health Economics, UPRES EA2415, University of Montpellier, Montpellier, France.
Laboratory of Clinical Pharmacy, Faculty of Pharmacy, University of Montpellier, Montpellier, France.
Sci Rep. 2017 Mar 14;7:44428. doi: 10.1038/srep44428.
Given that drug abuse and dependence are common reasons for hospitalization, we aimed to derive and validate a model allowing early identification of life-threatening hospital admissions for drug dependence or abuse. Using the French National Hospital Discharge Data Base, we extracted 66,101 acute inpatient stays for substance abuse, dependence, mental disorders or poisoning associated with medicines or illicit drugs intake, recorded between January 1, 2009 and December 31, 2014. We split our study cohort at the center level to create a derivation cohort and a validation cohort. We developed a multivariate logistic model including patient's age, sex, entrance mode and diagnosis as predictors of a composite primary outcome of in-hospital death or ICU admission. A total of 2,747 (4.2%) patients died or were admitted to ICU. The risk of death or ICU admission was mainly associated with the consumption of opioids, followed by cocaine and other narcotics. Particularly, methadone poisoning was associated with a substantial risk (OR: 35.70, 95% CI [26.94-47.32], P < 0.001). In the validation cohort, our model achieved good predictive properties in terms of calibration and discrimination (c-statistic: 0.847). This allows an accurate identification of life-threatening admissions in drug users to support an early and appropriate management.
鉴于药物滥用和依赖是住院的常见原因，我们旨在开发和验证一种模型，以便早期识别因药物依赖或滥用而危及生命的住院患者。我们使用法国国家住院数据基础，提取了 2009 年 1 月 1 日至 2014 年 12 月 31 日期间与药物或非法药物摄入相关的物质滥用、依赖、精神障碍或中毒的 66101 例急性住院患者。我们以中心水平将研究队列分为推导队列和验证队列。我们开发了一个多变量逻辑模型，包括患者的年龄、性别、入院方式和诊断，作为住院期间死亡或 ICU 入院的复合主要结局的预测因子。共有 2747 名（4.2%）患者死亡或被收入 ICU。死亡或 ICU 入院的风险主要与阿片类药物的使用有关，其次是可卡因和其他麻醉品。特别是，美沙酮中毒与较大的风险相关（OR：35.70，95%CI[26.94-47.32]，P<0.001）。在验证队列中，我们的模型在校准和区分方面具有良好的预测性能（C 统计量：0.847）。这允许对吸毒者的危及生命的入院进行准确识别，以支持早期和适当的管理。
