数据上报质量对基于医疗机构的症状监测系统预警灵敏度影响的研究

Influence of data quality on early warning sensitivity of syndromic surveillance system based on medical institutions

目的 评价湖北省潜江地区基于医疗机构症状监测系统的数据上报质量对该系统预警灵敏度的影响。方法 计算医疗机构症状监测系统运行期间数据质量评估指标（迟报率、缺报率）；使用半合成模拟暴发数据及预测模型ROC曲线下面积（AUC）评估多种预警模型的预测能力，筛选最佳预警模型；采用时间序列的广义相加模型（GAM）对数据质量评估指标与系统灵敏度进行曲线拟合及阈值效应分析。结果 2012年4月1日至2014年1月31日潜江地区累计上报总症状179 905例，迟报8 744次，平均每月迟报416次，总迟报率为16.45%；缺报2 566次，缺报率为4.83%。与其他预警模型（累积和模型、休哈特模型、指数加权移动平均模型、早期异常报告系统模型）相比，移动平均法模型的预测效果最佳（AUC=0.93）；与其他模型相比，AUC差异有统计学意义（P〈0.001）。症状监测系统运行期间，系统预警暴发的灵敏度介于84.89%-97.25%之间。缺报率对监测系统灵敏度有影响，即当缺报率〉2.78%时，系统灵敏度迅速下降，未观察到迟报率对灵敏度有影响。结论 湖北省潜江地区医疗机构症状监测系统的数据报告质量影响该系统的预警灵敏度，在此次设定参数的前提下，主要是数据的缺报率对系统灵敏度有影响。

Objective To evaluate the influence of data quality on the sensitivity of early warning syndromic surveillance system based on medical institutions in Qianjiang, Hubei province and explore the relationship between data quality and sensitivity of early warning of the system. Methods The delay reporting rate and underreporting rate were calculated for the evaluation of the data quality. Data obtained from semi-synthetic simulated outbreak and area under the curve （AUC） were used in combination to test the sensitivity of early warning of various models and select the optimal model. Time-series generalized additive model （GAM） was used to analyze the curve fitting and threshold effect between data quality and early warning sensitivity of the system. Results A total of 179 905 cases were reported from April 1, 2012 to January 31, 2014, in which 8 744 were not reported timely （16.45%）. Averagely 416 reporting were delayed in each month. There were 2 566 cases which were underreported （4.83%）. Compared with other early warning models, i.e. Cumulative Sum （CUSUM）, Shewhart, Exponentially Weighted Moving Average （EWMA）, Early Aberration Reporting System （EARS-3C）, the MA model had the maximum area under the curve （AUC=0.93）, and the difference was significant （P〈0.001）. The early warning sensitivity ranged from 84.89% to 97.25% during the operation period of the syndromic surveillance system. Underreporting had influence on early warning sensitivity, when underreporting rate was over 2.78%, the sensitivity would decrease obviously. No obvious associations were observed between the delay reporting rate and early warning sensitivity of the system. Conclusion The data quality had influence on the early warning sensitivity of the syndromic surveillance system based on medical institution in Qianjiang. In the context of this study, underreporting had the main influence on the sensitivity of early warning.