The Evaluation Of Early Warning Performance Of Integrated Syndromic Surveillance System And Its Application Analysis

Objectives:(1) To understand the demographic characteristics of target populations in study areas of the Intergrated Surveillance System for infectious diseases (ISSC) in Shayang County, Hubei. To explore correlations in data streams between intergrated surveillance and case report surveillance.(2) To evaluate the warning capability of the ISSC system in real epidemics, and the effects of warning in single-stream surveillance and multi-stream (including series-stream and parallel-stream) surveillance strategies.(3) To test the theoretical warning capability of the ISSC system in simulated epidemics, where the number of real epidemic is scarce.(4) To understand the application value of the ISSC by quantitatively estimating the benefits of control measures based on early warnings of the ISSC system.Methods:(1) The target population in both ISSC system and case report system were depicted. Time correlation was analyzed using the Cross Correlation Coefficient Function; and spatial correlation was tested using the Spatial Autocorrelation Method and the Geographically Weighted Regression Model.(2) Three real epidemics in study areas acted as the’golden standard’for early warning evaluation. The start and end time points of epidemics were determined. Warnings in data streams were generated by EARS-3C model retrospectively. The warning capability of strategies of single-stream, series-stream, and parallel-stream were compared. Validity was assessed using sensitivity, specificity, and positive predictive value. Timeliness was assessed using detection time ratio.(3) A Susceptible-Exposed-Infectious-Recovered model generated simulated epidemics, which were then converted into corresponding simulated syndromic datasets through the healthcare-behaviors model. Converted syndromic datasets were superimposed onto the baselines to create the testing datasets. The EARS-3C model was used to generate warning signals. Validity was assessed using sensitivity and specificity. Timeliness was assessed using detection time ratio.(4) An impulse susceptible-exposed-infectious-recovered model was used to fit the real epidemics in study areas. Potential epidemics under different interventions were simulated and compared to evaluate intervention effectiveness, and to assess the warning capability of the ISSC system indirectly.Results:(1) The time trends of chief complaints data did not lag behind cases reported data, except data on sore throats, headaches, and FGS syndrome, where the lag times were more than0day. The time trends of antibiotics and antidiarrheal drug sales surveillance correlated with related diseases significantly. Moreover, antibiotics data was previous to cases reported data. The time trends of notifiable cases in children (6-12years old) could be reflected ahead by3days through the school absenteeism surveillance. The incidences of notifiable diseases and target symptoms correlated strongly in north study areas, except sore throats and headaches.(2) In the evaluation based on the real epidemics, positive predictive values were18.9-70.0%for all surveillance strategies. Specificities were more than90%except88.4%and87.2%for C3model and3Cs series model in series-stream surveillance strategies. Sensitivities were100%except66.7%for all models in chief complaints single-stream surveillance strategy. The detection time ratios of all surveillance strategies (chief complaints single-stream, drug sales single-stream, school absenteeism single stream, parallel-stream, and series-stream) were0.0%～1.3%、19.0%～70.6%、5.1%～54.0%、0.0%-35.5%、0.0%～70.6%, respectively.(3) In the evaluation based on the simulated epidemics, sensitivities and specificities of single-stream surveillance strategies were11.4～48.7%and91.5～98.8%respectively. Series-stream strategies were inferior in sensitivity (13.0～50.0%vs43.2～59.6%), but superior in specificity (89.7～98.4%vs79.2～94.4%) relative to parallel-stream strategies. The detection time ratios of single-stream surveillance of chief complaints, school absenteeism, and drug sales were30.9～42.3%,36.4-42.7%, and43.7-49.7%; and the timeliness of parallel-stream surveillance strategies were superior to series-stream strategies (detection time ratio:28.8～41.2%vs36.4～50.7%).(4) Compared to the traditional warning standard, the school absenteeism system generated simultaneous signals for the varicella epidemic, but3days in advance for the mumps epidemic and2-4days in advance for the influenza-like illness epidemic. The estimated attack rates of these three epidemics under the interventions based on ISSC warnings were7.1～28.3%、5.0～41.8%、4.6～14.6%respectively. With the same interventions and responses, the estimated attack rates of these three epidemics based on traditional warning standard were7.1～28.3%、6.2～62.9%、6.7～22.3%respectively. The estimated excess protection rates of control measures based on early signals were0.0%,15.3-44.1%, and29.0-37.0%for the three epidemic respectively.Conclusions:(1) The data in the ISSC system correlated geographically with the notifiable case report data in the study area within the surveillance period. Time trends of some data streams in the ISSC system were in advance of case report system. The ISSC system was able to warn epidemics at early stages.(2) According to the warning evaluation based on real epidemics in the study area within the surveillance period, school absenteeism stream presented the most validity, and chief complaints stream presented the most timeliness in single-stream surveillance strategies. Moreover, series-stream was effective for validity, and parallel-stream was effective for timeliness in multi-stream surveillance strategies.(3) The evaluation results based on simulated outbreaks showed that, in the study areas within the surveillance period, validity was superior in school absenteeism surveillance, timeliness was superior in chief complaints surveillance, series-stream strategies were beneficial in improving warning validity, and parallel-stream strategies were beneficial for improving warning timeliness. The results were similar with those based on real epidemics, so that the results of evaluation based on simulated epidemics were deemed to be stable and reliable.(4) The school absenteeism system could generate warning signals in advance of2-4days compared to traditional warning standards in two epidemics. The control measures based on early warning signals could reduce the attack rate of epidemics, and the school absenteeism system was considered to play a significant role in prevention in the study schools within the surveillance period.