WSARE计算方法在传染病暴发早期 预警中的应用

探讨异常模式探测方法--WSARE(What's Strange About Recent Events)在传染病暴发早期预警中的应用价值,拓展传染病病例监测数据的多维聚集性探测统计方法.分别采用基于历史数据和贝叶斯网络为基线的WSARE算法,对2007年深圳市宝安区麻疹发病模拟实时监测预警.结果 表明WSARE算法能够早期探测到麻疹在特定人群的异常增高,在传染病暴发早期预警中具有重要应用价值.     

传染病暴发早期预警系统评价内容及其指标

为及时识别和应对传染病暴发与流行,近年来全球广泛开展了传染病暴发探测预警技术研究,探索利用不同类型和来源的监测数据进行传染病暴发早期预警,尝试并建立了各类传染病暴发早期预警系统(预警系统).预警系统作为一类有别于传统疾病监测系统的传染病暴发早期预警工具,有必要对其性能和运行效果开展专门的评价[1].本文系统回顾国内外有关文献,对预警系统评价的研究现况进行概述,并简要介绍预警系统主要评价内容及其指标,以期为我国开展相关研究提供参考和借鉴.     

国家传染病自动预警系统的设计与应用

提高早期发现传染病暴发的能力是公共卫生部门有效应对疫情的重要前提.近些年来,随着新发传染病及生物恐怖袭击事件引起的广泛关注,全球许多国家通过加强传统监测系统、增加新监测数据源和手段、研究有效的监测数据分析技术等方式,逐步探索并建立和完善了传染病监测与预警系统.中国疾病预防控制中心(CDC)早在2003年SARS暴发之前就着手开展了传染病流行预警技术研究,并成功建立了探测传染病暴发和流行的移动百分位数法.在此基础上,中国CDC又历时多年,先后开发完成了国家传染病自动预警信息系统,制定统一的预警信号响应流程,并在国内10个省份试点测试,于2008年4月在全国CDC系统中投入运行.该系统采用国内最为庞大的法定传染病监测数据,开发了简便实用的预警方法,实现预警信号的自动产生和及时发送,并对预警信号响应过程的信息进行及时收集.目前,该系统已成为各级CDC机构早期发现潜在传染病暴发的重要手段之一.     

移动百分位数法不同阈值在传染病暴发预警中的效果比较

目的 比较国家传染病自动预警系统(CIDARS)中移动百分位数法采用不同阈值对传染病预警效果的影响.方法 分别采用P50、P60、P70、P80和P90 5个阈值作为移动百分位数法的候选预警阈值,对全国范围2008年7月至2010年6月期间报告的19种法定传染病病例数进行暴发探测和结果的比较.以暴发探测起数最多、暴发探测时间最短和预警信号数最少作为移动百分位数法最优阈值的筛选标准.结果 细菌性和阿米巴性痢疾的最优阈值为P50,其他感染性腹泻和流行性腮腺炎的最优阈值为P60,甲型肝炎、流行性感冒和风疹的最优阈值为P70,流行性乙型脑炎的最优阈值为P80,猩红热、伤寒和副伤寒、戊型肝炎、急性出血性结膜炎、疟疾、流行性出血热、流行性脑脊髓膜炎、钩端螺旋体病、登革热、流行性和地方性斑疹伤寒、丙型肝炎和麻疹12种疾病的最优阈值为P90;对19种传染病分别采用最优剧值进行探测,与所有疾病均采用P50作为阈值相比,2年可减少64 840条(12.20%)预警信号,而暴发探测起数与暴发探测时间没有变化.结论 不同传染病采用移动百分位数法进行暴发探测的最优阈值不同,CIDARS可进一步优化各病种的预警阈值,从而在确保暴发探测准确性和及时性的前提下,减少预警信号数量.     

移动百分位数法不同阈值在传染病暴发预警中的效果比较

目的 比较国家传染病自动预警系统(CIDARS)中移动百分位数法采用不同阈值对传染病预警效果的影响.方法 分别采用P50、P60、P70、P80和P90 5个阈值作为移动百分位数法的候选预警阈值,对全国范围2008年7月至2010年6月期间报告的19种法定传染病病例数进行暴发探测和结果的比较.以暴发探测起数最多、暴发探测时间最短和预警信号数最少作为移动百分位数法最优阈值的筛选标准.结果 细菌性和阿米巴性痢疾的最优阈值为P50,其他感染性腹泻和流行性腮腺炎的最优阈值为P60,甲型肝炎、流行性感冒和风疹的最优阈值为P70,流行性乙型脑炎的最优阈值为P80,猩红热、伤寒和副伤寒、戊型肝炎、急性出血性结膜炎、疟疾、流行性出血热、流行性脑脊髓膜炎、钩端螺旋体病、登革热、流行性和地方性斑疹伤寒、丙型肝炎和麻疹12种疾病的最优阈值为P90;对19种传染病分别采用最优剧值进行探测,与所有疾病均采用P50作为阈值相比,2年可减少64 840条(12.20%)预警信号,而暴发探测起数与暴发探测时间没有变化.结论 不同传染病采用移动百分位数法进行暴发探测的最优阈值不同,CIDARS可进一步优化各病种的预警阈值,从而在确保暴发探测准确性和及时性的前提下,减少预警信号数量.     

2009-2012年浙江省传染病自动预警系统预警结果分析

目的分析中国传染病自动预警系统（预警系统）在浙江省的运行情况,探讨在传染病暴发早期探测中的效果。方法对2009-2012年浙江省预警系统移动百分位数法的传染病自动预警监测数据进行描述性分析,并比较2010年12月10日预警参数调整前后的预警效果。结果在全省范围内,预警系统19种传染病的39 304条预警信号,响应率为99.97%,平均响应时间为1.03h。其中309条信号（0.79%）经初步核实判断为疑似事件,经过进一步现场调查确认78起暴发,预警阳性率为0.20%。预警病种的报告病例数与预警信号数呈正相关变化（r=0.91,P〈0.05）。与预警参数调整前比较,预警参数调整后,每月预警信号数减少,预警阳性率上升（χ2=4.46,P〈0.05）。结论传染病自动预警系统运行良好,可初步实现传染病暴发的早期自动预警,通过调整预警参数可进一步改进预警应用效果,但需针对不同疾病特点选用适宜的预警方法。     

WSARE计算方法在猩红热暴发预警中的应用

目的用WSARE软件对猩红热监测数据进行分析,探测可能出现的暴发疫情。方法采用WSARE软件,应用基于历史数据基线的WSARE算法,对北京市2011年的猩红热监测数据进行分析,产生预警信号。结合原始监测数据,对预警信号进行描述和分级,评估出现暴发的可能性。结果共发出预警信号11次,双特征变量联合异常增高情况6次,单特征变量异常增高情况5次。出现高级别预警,暴发疫情存在可能性较大的有5月上、中旬西城区的学校和幼托机构;出现中级别预警,暴发疫情可能存在的有:5月中、下旬丰台区的幼托机构,5月下旬至6月初海淀区的幼托机构,6月中旬海淀区的幼托机构;出现低级别预警,暴发疫情存在可能性较小的有:6月上旬昌平区幼托机构,5月下旬朝阳区的学校。结论 WSARE计算方法是一种针对性强、灵活性强的时空预警方法。其可以实现对猩红热监测数据中的异常信号进行探测,早期提示可能出现的暴发疫情。     

浙江省传染病时空模型探测暴发疫情的效果评价

目的 分析时空模型探测在浙江省传染病暴发早期预警中的应用效果,为构建多点触发预警系统提供参考。方法 对2017—2021年浙江省传染病时空模型相关病种的预警信号进行描述性分析,并与同期突发公共卫生事件报告结果进行比较,以灵敏度和错误预警率评价时空模型的有效性。结果 时空模型预警系统共发出预警信号16 382条,涉及16种传染病,平均每周每县发出预警信号0.71条。响应率为100%,响应时间中位数为0.84 h。预警信号经初步核实和现场调查后,共确认5种疾病342起暴发疫情。时空预警模型灵敏度为66.02%。Ⅰ类疾病阳性预测值6.99%,错误预警率为0.0227%;Ⅱ类疾病阳性预测值3.44%,错误预警率为0.9450%。结论 浙江省时空预警系统运行良好,可实现传染病暴发的早期自动预警,但探测识别作用有限,仍需进一步调整和优化,以适应目前传染病监测预警的需要。     

前瞻性时空扫描统计量方法在手足口病暴发预警中的参数筛选

目的 根据手足口病的流行特点,比较和筛选时空扫描统计量方法较好的预警参数,从而更好的应用于手足口病暴发早期探测预警.方法 以真实的暴发事件作为参考数据,灵敏度、探测时间间隔和阳性预测值为评价指标,采用前瞻性时空扫描统计量方法,利用不同时空参数对山东省青岛市2009年6月手足口病病例数据逐日模拟预警探测,比较4种参数不同设定值的24种预警效果.结果 时间窗口设置为3天和基线设置为最近21天的病例数据进行时空扫描统计量预警运算,具有最优的预警效果.候选聚集区域范围的设定对预警效果无统计显著性.结论 采用前瞻性时空扫描统计量进行传染病暴发预警时,具有最优的参数,但相同的传染病在不同区域可能有不同的预警参数,下一步应该根据相同传染病在不同地区筛选最优参数方案.     

湖南省传染病自动预警系统预警效果分析

目的分析中国传染病自动预警系统(预警系统)在湖南省的应用情况,探讨在传染病暴发早期探测中的效果,为提高传染病预警能力及功效提供依据。方法利用传染病自动预警系统中,湖南省2008年1月1日-2011年12月31日发出的预警信号进行相关的统计分析。结果预警系统4年间共发出预警信号63 996次,涉及30种传染病,平均每县每周预警信号3.16条。全部的预警信号都得到响应,预警信号经过初步核实后,其中2 723条(1.25%)预警信号被判断为疑似事件,经过现场调查共确认暴发事件800起,预警系统的灵敏度为29.77%。结论预警系统可实现传染病暴发的早期自动预警,系统灵敏度逐步提升。但仍需进一步研究针对不同疾病的差异性、合理设置预警阈值,进一步减少假阳性和重复预警、提高早期探测暴发和预警的准确性。     

Epidemic features affecting the performance of outbreak detection algorithms

ABSTRACT: BACKGROUND: Outbreak detection algorithms play an important role in effective automated surveillance. Although many algorithms have been designed to improve the performance of outbreak detection, few published studies have examined how epidemic features of infectious disease impact on the detection performance of algorithms. This study compared the performance of three outbreak detection algorithms stratified by epidemic features of infectious disease and examined the relationship between epidemic features and performance of outbreak detection algorithms. METHODS: Exponentially weighted moving average (EWMA), cumulative sum (CUSUM) and moving percentile method (MPM) algorithms were applied. We inserted simulated outbreaks into notifiable infectious disease data in China Infectious Disease Automated-alert and Response System (CIDARS), and compared the performance of the three algorithms with optimized parameters at a fixed false alarm rate of 5% classified by epidemic features of infectious disease. Multiple linear regression was adopted to analyse the relationship of the algorithms' sensitivity and timeliness with the epidemic features of infectious diseases. RESULTS: The MPM had better detection performance than EWMA and CUSUM through all simulated outbreaks, with or without stratification by epidemic features (incubation period, baseline counts and outbreak magnitude). The epidemic features were associated with both sensitivity and timeliness. Compared with long incubation, short incubation had lower probability (beta*=-0.13, P<0.001) but needed shorter time to detect outbreaks (beta*=-0.57, P<0.001). Lower baseline counts were associated with higher probability (beta*=-0.20, P<0.001) and longer time (beta*=0.14, P<0.001). The larger outbreak magnitude was correlated with higher probability (beta*=0.55, P<0.001) and shorter time (beta*=-0.23, P<0.001). CONCLUSIONS: The results of this study suggest that the MPM is a prior algorithm for outbreak detection and differences of epidemic features in detection performance should be considered in automatic surveillance practice. KEYWORDS: Epidemic feature, Outbreak detection algorithms, Performance, Automated infectious disease surveillance.     

心血管病预测模型研究进展

20世纪六七十年代,大多数工业化国家冠心病的死亡率急剧上升,并达到高峰.为预测个人冠心病发病风险,有针对性地采取干预措施,西方国家开发出以弗明汉模型(Framingham model)为代表的多种心脑血管疾病预测模型,这些模型的应用方便了临床诊断和防治,提高了公众对疾病危险因素的认识,且有利于卫生资源合理分配.为此,国内一些学者参照相关研究开发出了适合国人的心脑血管疾病预测模型,也取得了较好的效果.本文就国内外主要的心血管病的预测模型研究综述如下.     

A modified CUSUM test to control postoutbreak false alarms

The cumulative sum (CUSUM) control chart is a method for detecting whether the mean of a time series process has shifted beyond some tolerance (ie, is out of control). Originally developed in an industrial process control setting, the CUSUM statistic is typically reset to zero once a process is discovered to be out of control since the industrial process is then recalibrated to be in control. The CUSUM method is also used to detect disease outbreaks in prospective disease surveillance, with a disease outbreak coinciding with an out-of-control process. In a disease surveillance setting, resetting the CUSUM statistic is unrealistic, and a nonrestarting CUSUM chart is used instead. In practice, the nonrestarting CUSUM provides more information but suffers from a high false alarm rate following the end of an outbreak. In this paper, we propose a modified hypothesis test for use with the nonrestarting CUSUM when testing whether a process is out of control. By simulating statistics conditional on the presence of an out-of-control process in recent time periods, we are able to retain the CUSUM's power to detect an out-of-control process while controlling the post–out-of-control false alarm rate at the desired level. We demonstrate this method using data on a Salmonella Newport outbreak that occurred in Germany in 2011. We find that in 7 out of 8 states where the outbreak was detected, the outbreak was detected at the same speed as an unmodified nonrestarting CUSUM while controlling the postoutbreak rate of false alarms at the desired level.     

甘肃省流感流行预警方法探研

目的 探索适合甘肃省流感流行的预警方法。方法 分别用简单控制图法、移动百分位数法、指数平滑法及累积和控制图法对2014－2015年度甘肃省流感样病例监测数据进行预警分析,结合灵敏度、特异度、阳性预测值、约登指数、Kappa值等指标,比较和评价4种方法的预警效果。结果 2014－2015年度甘肃省流感流行高峰起始时间为2014年第50周,流行高峰持续6周。4种预警方法中以累积和控制图法预警效果最优,能及时发出预警信号,灵敏度及特异度分别为66.67%和93.48%。结论 累积和控制图法适合甘肃省流感流行高峰预警。     

Optimizing the response to surveillance alerts in automated surveillance systems

Although much research effort has been directed toward refining algorithms for disease outbreak alerting, considerably less attention has been given to the response to alerts generated from statistical detection algorithms. Given the inherent inaccuracy in alerting, it is imperative to develop methods that help public health personnel identify optimal policies in response to alerts. This study evaluates the application of dynamic decision making models to the problem of responding to outbreak detection methods, using anthrax surveillance as an example. Adaptive optimization through approximate dynamic programming is used to generate a policy for decision making following outbreak detection. We investigate the degree to which the model can tolerate noise theoretically, in order to keep near optimal behavior. We also evaluate the policy from our model empirically and compare it with current approaches in routine public health practice for investigating alerts. Timeliness of outbreak confirmation and total costs associated with the decisions made are used as performance measures. Using our approach, on average, 80 per cent of outbreaks were confirmed prior to the fifth day of post-attack with considerably less cost compared to response strategies currently in use. Experimental results are also provided to illustrate the robustness of the adaptive optimization approach and to show the realization of the derived error bounds in practice.     

累积和模型在细菌性痢疾早期预警中的应用

目的探讨累积和(CUSUM)模型在细菌性痢疾早期预警监测中的应用。方法数据来源于2007年北京市法定传染病报告系统细菌性痢疾报告数据和北京市2007年细菌性痢疾暴发疫情处理记录。预警运算使用美国CDC早期异常报告系统软件。结果 330个街道或乡镇共发出预警信息数为3743条,平均每个街道或乡镇发出11.3条预警信息,其中最少的为0条,最多的为25条,中位数为12,四分位间距为7。及时预警了2007年疫情记录中的两起细菌性痢疾暴发。结论累积和模型运算过程简单,预警频次和因预警产生的审核、流调工作量均在可接受范围内,可以应用在日常细菌性痢疾监测和防控工作中。     

A Framework for Detecting and Classifying Outbreaks of Gastrointestinal Disease

Objective

To develop a methodological framework for detecting and classifying outbreaks of gastrointestinal disease on the island of Montreal, with the goal of improving early outbreak detection using simulated surveillance data.

Introduction

Outbreaks of waterborne gastrointestinal disease occur routinely in North America, resulting in considerable morbidity, mortality, and cost (Hrudey, Payment et al. 2003). Outbreak detection methods generally attempt to identify anomalies in time, but do not identify the type or source of an outbreak. We seek to develop a framework for both detection and classification of outbreaks using information in both space and time. Outbreak detection can be improved by using simulated outbreak data to build, validate, and evaluate models that aim to improve accuracy and timeliness of outbreak detection.

Methods

To generate outbreak data, we used a previously validated microsimulation model depicting waterborne outbreaks of gastrointestinal disease (Okhmatovskaia et al. 2010). The model is parameterized based on outbreak characteristics such as concentration and duration of contamination, and calibrated to produce realistic outbreak data (e.g., emergency department visits from GI-illness, laboratory reporting to public health) in space and time. We are interested in identifying unique space-time signatures in the data that would allow not only detection, but also classification based on outbreak type. For example, to be able to detect and classify an outbreak as due to a water plant failure versus an food-borne illness based on unique space-time patterns, even though symptoms and temporal outbreak patterns may be similar. For the detection step, we use a hidden Markov model (HMM) that accounts for spatial information through a spatially correlated random effect with an exponential decay. HMMs have been used previously in disease mapping (Green 2002) but not widely in space-time disease outbreak detection. For the classification step, we use a supervised clustering algorithm to classify the outbreak by source (e.g., water plant location) and type (e.g., disease).

Results

Preliminary results for the detection step show that the HMM can distinguish accurately between regions in an outbreak state versus those in a normal state at each time period. Ongoing work for the detection step includes further evaluation of the HMM accuracy as a function of outbreak characteristics. For the classification step, we are evaluating the suitability of different supervised clustering algorithms for identifying the type of outbreak from the HMM results.

Conclusions

If outbreaks are detected rapidly, interventions, such as boil-water advisories, are available to quickly and effectively limit the human and economic impacts. Traditional public health surveillance systems, however, frequently fail to detect waterborne disease outbreaks. Every disease outbreak has unique characteristics; simulation is the best method to estimate the capacity of syndromic surveillance to more efficiently detect different types of enteric disease outbreaks based on a variety of parameters. Outbreak detection can be improved with advances in data availability, such as syndromic surveillance data that will increase timeliness of detection, and space-time information to allow for simultaneous detection and classification of outbreaks by important characteristics (type of outbreak, source of outbreak).     

Spatial and Temporal Algorithm Evaluation for Detecting Over-The-Counter Thermometer Sale Increases during 2009 H1N1 Pandemic

Background

Spatial outbreak detection algorithms using routinely collected healthcare data have been developed since the late 90s to identify and locate disease outbreaks. However, current well-received spatial algorithms assume only one outbreak cluster present at the same point of time which may not be valid during a pandemic when several clusters of geographic areas concurrently occur. Based on a retrospective evaluation on time-series and spatial algorithms, this paper suggests that time series analysis in detection of pandemics is still a desirable process, which may achieve more sensitive performance with better timeliness.

Methods

In this paper, we first prove in theory that two existing spatial models, the likelihood ratio and the Bayesian spatial scan statistics, are not useful if multiple clusters occur at the same point of time in different geographic regions. Then we conduct a comparison between a spatial algorithm, the Bayesian Spatial Scan Statistic (BSS), and a time series algorithm, the wavelet anomaly detector (WAD), on the performance of detecting the increase of the over-the-counter (OTC) medicine sales during 2009 H1N1 pandemic.

Results

The experiments demonstrated that the Bayesian spatial algorithm responded to the increase of thermometer sales about 3 days later than the time series algorithm.

Conclusion

Time-series algorithms demonstrated an advantage for early outbreak detection, especially when multiple clusters occur at the same time in different geographic regions. Given spatial-temporal algorithms for outbreak detection are widely used, this paper suggests that epidemiologists or public health officials would benefit by applying time series algorithms as a complement to spatial algorithms for public health surveillance.     

Use of the local Knox statistic for the prospective monitoring of disease occurrences in space and time

The detection of clusters of events occurring close together both temporally and spatially is important in finding outbreaks of disease within a geographic region. The Knox statistic is often used in epidemiology to test for space-time clustering retrospectively. For quicker detection of epidemics, prospective methods should be used in which observed events in space and time are assessed as they are recorded. The cumulative sum (CUSUM) surveillance method for monitoring the local Knox statistic tests for space-time clustering each time there is an incoming observation. We consider the design of this control chart by determining the in-control average run length (ARL) performance of the CUSUM chart for different space and time closeness thresholds as well as for different control limit values. We also explain the effect of population density and region shape on the in-control ARL and discuss other distributional issues that should be considered when implementing this method.