高级检索

基于R语言的我国传染病网报信息审核程序包的开发与效能评价

Development and performance of a R-based evaluation package for online reporting of infectious disease information in China

    摘要
  • 摘要:
    目的 开发我国传染病网络报告信息审核程序包，为提高传染病报告信息审核效率与质量提供方法学支持。
    方法 基于传染病报告及时性、信息审核及时性、信息完整性、信息准确性、身份证件号有效性及重点疾病提醒6个维度构建审核规则，开发R语言程序包survalarm。采用survalarm对2018—2023年重庆市江津区医疗机构网络报告的传染病信息进行回顾性审核，从审核速度和准确性两个维度评估程序效能。
    结果 survalarm核心函数epiaudit（）经30次重复测试，在测试数据集中的平均执行时间为179.25±10.02s（ \overlinex ±s），处理通量达427.03张/s，较人工审核速度（0.10张/s）提升4270.30倍。survalarm在审核试验中，灵敏度为98.85%（95%CI: 93.28%～100.00%），特异度为99.58%（95%CI: 98.75%～99.86%），阳性预测值为96.63%（95%CI: 90.51%～98.96%），F1分数达97.73%（95%CI: 94.18%～99.42%）。2018—2023年江津区医疗机构共报告有效传染病网报卡76545张，网报及时率、完整率、准确率与身份证有效率分别为99.96%、98.37%、97.11%与95.12%；共发现4种信息报告不完整类型与12种信息报告不准确类型。2018—2023年江津区医疗机构传染病网报卡数与人工审核错误率Spearman相关系数为0.33（P= 0.004）。
    结论 高负荷的传染病审核工作与人工审核错误率上升存在关联，survalarm具备良好审核效能，其应用可有效节省人力、提升审核效率。

     

    Abstract:
    Objectives To develop an evaluation package for China's National Notifiable Disease Reporting System, and provide support to improve the efficiency and quality of infectious disease reporting.
    Methods The evaluation system was established based on six dimensions: timeliness of case reporting, timeliness of information verification, information completeness, information accuracy, validity of identity card code and key disease alerts. The R-based program package survalarm was developed accordingly. A retrospective analysis was conducted on the information of infectious disease reports from healthcare institutions in Jiangjin district, Chongqing, during 2018-2023 to evaluate survalarm's performance in terms of processing efficiency and verification accuracy.
    Results The core function epiaudit() completed 30 repeated tests on the dataset, with an average execution time of 179.25±10.02（ \overlinex ±s）seconds, achieving a processing throughput of 427.03 reports/s, indicating a 4 270.30-fold efficiency improvement compared with manual verification (0.10 reports/s). In the gold-standard dataset, survalarm showed a sensitivity of 98.85% (95% CI: 93.28% - 100.00%), specificity of 99.58% (95% CI: 98.75% - 99.86%), positive predictive value of 96.63% (95% CI: 90.51% - 98.96%), and a F1-score of 97.73% (95% CI: 94.18% - 99.42%). From 2018 to 2023, the healthcare institutions in Jiangjin reported 76 545 valid cards of infectious disease cases, with timelessness rate, completeness rate, accuracy rate, and identity card code validity rate of 99.96%, 98.37%, 97.11%, and 95.12%, respectively. Four types of incomplete reporting and 12 types of inaccurate reporting were identified. The Spearman correlation coefficient between the count of infectious disease online reporting cards and manual verifying error rate was 0.33 in Jiangjin health care institutions from 2018 to 2023 (P=0.004).
    Conclusion There was an association between high workload in manual infectious disease verifying and increased errors. Survalarm demonstrated good evaluation performance, which can effectively reduce human resources and improve the efficiency of evaluation of reported infectious disease information.

     

    • HTML全文
    • 参考文献(15)
    • 相关文章
    • 施引文献
  • 资源附件(0)

/

返回文章
返回

在线交流