一种基于深度神经网络的临床记录ICD自动编码方法

随着国际疾病分类(international classification of diseases,ICD)编码数量的增加,基于临床记录的人工编码难度和成本大大提高,自动ICD编码技术引起了广泛的关注。提出一种基于多尺度残差图卷积网络的自动ICD编码技术,该技术采用多尺度残差网络来捕获临床文本的不同长度的文本模式,并基于图卷积神经网络抽取标签之间的层次关系,以加强自动编码能力。在真实医疗数据集MIMIC-III上的实验结果表明,该方法的P@k和Micro-F1分别为72.2%和53.9%,显著提高了预测性能。

Failure mode and effect analysis of the risk management of non-integral-dosage drug dispensing in PIVAS

To mitigate risks associated with the prescription examination,marking,dispensing,checking,and review of non-integral-dosage drugs in Pharmacy Intravenous Admixture Service(PIVAS),we formed a project team.Employing the failure mode and effect analysis(FMEA)management method,we identified potential risks in four critical steps of the non-integral-dosage drug dispensing process within PIVAS drug management:prescription verification,mixed allocation,and verification.For each step,we assigned scores for severity,incidence,and detectability,subsequently calculating the Risk Priority Number(RPN)to prioritize identified risks.Targeted measures for improvement were developed for steps with the highest RPN values.A total of 31 risk factors were documented in the management of non-integral-dosage drugs,with the dispensing process being particularly vulnerable.Specific measures were devised for eight high RPN risks.Following a 3-month optimization and improvement period,RPN values and incidences of internal differences were significantly reduced.The implemented measures demonstrated effective risk control.Notably,we established a comprehensive conversion system for partial-dose drug dispensing,directly translating into a volume of suction fluid for dispensing personnel based on doctor orders.This eliminated the need for manual secondary calculations,thereby standardizing and automating the dispensing of non-integral-dosage drugs in PIVAS.Simultaneously,our project team conducted a dissolution test on 23 types of drugs with non-integral dosage,revealing that the solvent volume increased for 11 types after dissolution.The dosage conversion for partial dosage was recalibrated based on the volume of the final solution to ensure dosage accuracy.Through the application of failure mode and effect analysis,we systematically managed the risks associated with non-integral-dosage drugs in PIVAS.This approach addressed safety concerns in the dispensing process,reduced errors,and ensured the safe and precise administration of medication to patients.