| Objectives:1. To develop the potential and suitable indicators for early detection of epidemics or outbreaks of infectious diseases in rural Hubei Province, so as to provide primary information for the construction of syndromic surveillance system in health facilities.2. To establish and operate the syndromic surveillance system, for exploring the feasibility of the system used in rural areas.3. To assess the timeliness of the syndromic surveillance system, so as to explore the capability of the system in early detection of infectious disease epidemics or outbreaks.4. To determine the optimum threshold, and comprehensively evaluate the warning efficacy of each time-series model, so as to provide theoretical support for the automatic warning of the system.5. To preliminary evaluate the performance of the system, for providing evidences for implementing the system better in next stage.Methods:1. The surveillance indicators were confirmed by analyzing historical data, reviewing of literatures and expert consultation meeting.2. Qianjiang City and Shayang County were selected as study areas. The surveillance units included county hospitals, township hospitals and village health stations. The surveillance information were collected and reported daily into a web-based system by health staffs. Multiple quality control measures were taken during implementation. Data reported from April1st to March31st2013were used to evaluate the data quality and generate descriptive statistics. 3. From April1st2012to March31st2013, data on initial visit outpatients collected from the study towns were extracted from the database of the syndormic surveillance system. Information on infectious disease cases were obtained from the China Information System for Disease Control and Prevention (CISDCP) for the same period and same region. Cross-correlation function was used to examine the correlation, and the lag periods were set as from-21days to+21days.4. By using semi-synthetic simulated data, the warning efficacy of time-series models under different threshold levels were examined. These models included the cumulative sum control model (CUSUM), the C1, C2and C3models in Early Aberration Reporting System (EARS), the moving average model (MA), and the exponentially weighted moving average model (EWMA). The optimum threshold of each model was selected according to Youden’s index and the lag time of warning. The warning efficacy of each model was comprehensively evaluated by using receiver operator characteristic curve (ROC), area under curve (AUC), and dynamic monitoring operator characteristic curve (AMOC). After then, the initial visit fever outpatients in Shayang County from October1st2012to March31st2013were tested for outbreak detection by using the optimum threshold of each model. The warning signals were also verified retrospectively.5. The performance of the system was evaluated by investigating the subjective feelings of the primary data collectors, and by describing the experience that has acquired in using the system. The performance attributes included usefulness, flexibility, acceptability, portability, stability and cost.Results:1. Fourteen kinds of infectious disease were chose as specific diseases for monitoring. The first initial surveillance scheme was to monitor19initial symptoms, and the second one was to monitor8initial syndromes. Finally, scheme one was used and10targeted symptoms were determined.2. The formal study was conducted in154hospitals. The total missing report rate was2.68%, and the total late report late was16.53%. The total inconsistent rate was17.68%and30.20%in the first and second consistency check respectively. Totally,206,727targeted outpatient records were reported,98.94%came from study areas,50.06%were males,42.12%aged from40to65years,71.87%were the initial visits, and85.51%visited clinics within three days after symptoms onset. The most common symptom was cough (67.79%), followed by sore throat (37.06%), headache (15.79%) and fever (14.65%). There was a temporal trend of increases during summer and winter in fever and respiratory related symptoms/syndromes, and an increase during June to September in gastrointestinal related symptoms/syndromes.3. The maximum cross-correlation coefficient between fever outpatients and total infectious disease cases was found when the lag was4days (r=0.453). The maximum cross-correlation coefficient (r=0.414～0.572) between respiratory infectious disease related symptoms or syndromes and respiratory infectious disease cases was identified by moving symptomatic data backward by2to10days. The maximum cross-correlation coefficient (r=0.189～0.240) between gastrointestinal infectious disease related symptoms or syndromes and gastrointestinal infectious disease cases was identified by lag was7to8days. The maximum cross-correlation coefficient (r=0.292～0.521) between hand-foot-mouth disease cases related symptoms/syndromes (i.e. rash, fever with rash, fever and aged≤5years old, rash and aged≤5years old, fever with rash and aged≤5years old) and hand-foot-mouth disease cases was identified by lag was10to16days.4. The optimum threshold of CUSUM, EARS’Cl, EARS’C2, EARS’C3, MA and EWMA was3.5,1.5,2.0,3.0,2.0and3.5respectively. CUSUM model was most biased in favor of the upper left corner in ROC chart and the AUC value was highest, and it was also most biased in favor of the lower left corner in AMOC chart. Totally, warning signals were triggered in16days, but no real outbreak was found.5. A total of146data collectors were investigated.97.26%of the data collectors thought that the system was useful. The surveillance system was rather flexible. Among the respondents,88.36%reflected that the daily time spent on data reporting was less than30minutes,60.96%declared that they reported all the required data. All of the respondents thought that they were qualified for the project works and felt the system’s operation was simple.98.63%were satisfied with the surveillance activity, and94.52%were willing to continue to participate in the project. The development and implementation methods of the system including the methods for developing monitoring indicators, data collection approaches, monitoring processes, quality control measures, and early warning models, could be used in other setting.98.63%of the data collectors thought that the system was stable. During the first year of implementation, shutdowns occurred twice because of system maintenance. The cost for establishment of the surveillance system mainly lies in the establishment and maintenance of system platforms, purchase of hardware facilities, payments for data collectors, trainings and supervisions.Conclusions:1. Compare with syndromes, symptoms are more suitable for monitoring in resource-poor settings.2. The study successfully established and implemented a syndromic surveillance system in health facilities in rural areas of Hubei Province. The data quality was rather good, and the system could be used as a potential proxy for changes of disease patterns.3. Compare with CISDCP, the syndromic surveillance system could detect the infectious disease epidemic and outbreak earlier by several days.4. With the optimum threshold, the warning efficacy of each model is rather good, and the CUSUM model has the highest efficacy. We can jointly use several models to decide which alarm needs to be investigated.5. The performance of the syndromic surveillance system is rather good, which has some applicability in rural areas.Innovations:1. We first confirmed the indicators for syndromic surveillance system in health facilities in rural areas. The study could provide methods and evidances for the development of surveillance indicators, which has portability and replicability.2. We first constructed and implemented a syndromic surveillance system in rural China, which provides relevant experience and evidence for other developing countries or areas to conduct similar research or practice.3. We evaluated the warning efficacy of each time-series model in different thresholds, which provides theoretical support for the automatic warning of the system.4. We preliminary evaluated the performance of the system, and explored the applicability in rural China. |
PDF Full Text Request