Study On A School-Based Smart Epidemic Syndromic Surveillance System:Development,Usability And Evaluation

Background COVID-19 has exposed the shortcomings of China’s infectious disease surveillance system.There are two prominent problems in the current school infectious disease syndromic surveillance system（SSS）:Relying on school physicians to collect data manually,ignoring the health information of students in attendance.About 30%of primary and secondary schools are equipped with school physicians,and the prevalence of attending school with illness seriously limits the popularization and function of school-based SSS.Objectives Aiming at the existing problems,based on the existing network platform,a school intelligent infectious disease SSS（AISSS）was designed and implemented.JSSS automatically collects absenteeism and body temperature data synchronously through face recognition and infrared temperature measurement.The research focuses on three key issues:Could the completeness and accuracy of the data collected by AISSS meet the surveillance requirements?How to further improve the accuracy of temperature automatically collected by AISSS?What is the surveillance effectiveness of indicators constructed based on data collected by AISSS?There were six phases to the study:1.The prototype was optimized to construct AISSS,and AISSS’advantages and disadvantages were preliminarily evaluated.2.To verify the influence of distance,light,motion,air temperature and high temperature objects on the accuracy of temperature measurement in AISSS,and to provide reference for constructing the standard environment for instrument operation.3.To analyze the completeness,accuracy and influencing factors of absenteeism and temperature collected by AISSS,and to understand the quality status and improvement direction of AISSS data collected in real scenes.4.To implement the AISSS data quality improvement plan,investigate the implementation effect of the plan,and provide reference for revising the system operation standard.5.To compare the differences in the surveillance effects of indicators based on the data reported by AISSS and school physicians on infectious diseases,and to construct a surveillance index system for AISSS.6.Other attributes of AISSS were evaluated to provide reference for a new round of system optimization.Methords To achieve the above objectives,the research was divided into six parts,with the following methods:1.The AISSS and prototype were qualitatively compared in terms of system architecture,functional modules,data flow and physical support,and their cost-effectiveness,security,simplicity,representativeness and standard application were evaluated.2.Five groups of experiments were arranged to verify the influence of distance,light,motion,air temperature and high temperature objects on the measurement accuracy of thermal imager.3.Primary schools A and B in Hangzhou and Middle school C in Shanghai were selected,with a total of 3538 students.The collected data included:Based on the daily absence rate（DAR1）,fever rate（DFR）,abnormal high temperature rate（DAHTR）and abnormal low temperature rate（DALTR）reported by AISSS,daily absenteeism rates reported by school physicians（DAR2）and weekly influenza virus positive rates（WPRIV）published by the China National Influenza Center.The time of data collection was from March 1,2021 to January 14,2022.The time series graph of these variables was analyzed,the correlation between variables was investigated,and the causes of data anomalies were investigated through observation,interview,thermal image analysis and other methods.4.The data quality improvement intervention plan was formulated,and school A and B（3111 and 3118 students in two semesters）were selected as the intervention population,and the implementation of the plan was investigated by observation,interview and thermography analysis.The temperature measurement data of the two schools collected by AISSS from September 1,2021 to June 24,2022 were tracked,and the changes of DFR,DAHTR and DALTR before and after the implementation of the intervention program were compared Cross-sectionally and longitudinally.5.The absenteeism and temperature of schools A and B（3111 and 3118students in two semesters）were collected systematically from September 1,2021 to June 24,2022,and the grade characteristics of absenteeism were analyzed.School A constructed three indexes of all-cause absenteeism rate for grades L1-2（DARA-L）,grades 3-6（DARA-H）,and whole school（DARA-T）,while school B constructed four indexes of all-cause absenteeism rate for grades 1-2（DARB-L）,grades 3-6（DARB-H）,whole school（DARB-T）and fever rate（DFRb）.The all-cause absenteeism rate（DARA-P and DARB-P）and sickness absenteeism rate（DARA-S and DARB-S）collected by school physicians during the same period were used as references,exponential smoothing model and infectious disease event investigation were used to analyze the sensitivity,specificity and Youden index of infectious disease surveillance.6.Based on the results of the above five studies,the method of combining qualitative and quantitative analysis was used to comprehensively evaluate the acceptability,data quality,flexibility,effectiveness,stability,standard application and timeliness of AISSS.Results1.AISSS adopted distributed architecture,including infrastructure layer,edge data processing layer,central data processing layer,data display layer,user access layer,and uses encryption,password,authentication,firewall,redundancy and other technologies to ensure system security.It used face recognition and thermal imager to synchronously collect absenteeism and body temperature,including three functional modules of data acquisition,statistics and management.AISSS redefined the operational definition of absence and fever:Students who do not perform face recognition within 1 hour of the school deadline were considered to be absent from school.Students who are present were classified as normal（36.0-37.2℃）,fever（37.3-39.4℃）and abnormal（≥39.5℃or<36.0℃）Third class.Data were analyzed at individual,class,grade and school levels,and real-time feedback is provided to parents（mobile phone）,head teachers（mobile phone）and school administrators（computer）.The main advantages of AISSS are:automatic data collection,simple operation;Covering attendance cases with good representation;Research and development costs are low.The main problems are:The system is open,so the data security challenge is big;It is difficult to randomly sample within the community;System operation standards need to be improved.2.Distance,sunlight,individual movement,ambient temperature and high temperature objects have significant influence on the accuracy of thermal imaging.The results show that:（1）the best measuring distance of thermal imager is 90-110cm;（2）Sunlight exposure to the human body and the instrument will lead to high temperature measured by the thermal imager;（3）More than 10 minutes of brisk walking can lead to high temperature measurement;（4）The accuracy of the thermal imager is reliable at 5-25℃;（5）High temperature objects will make the temperature reading high,non-metallic material baffle can avoid such effects.3.Instruments of school A were installed at the gate,with a sunshade and personnel on duty,equipment density of about 320 people/set;the instruments of school B were installed in the teaching building,with staff on duty and the equipment density was about 280 people/set;the instruments of school C were installed at the school gate and dormitory,unsupervised,equipment density of about 80 people/set.The number of absences reported by school physicians（3167,completeness 84.6%）only accounted for 36.5%of the number of absences reported by AISSS（8680,completeness 100%）,but in school A（r=0.809,p<0.001）or in school B（r=0.766,p<0.001）the two were positively correlated.The absenteeism was significantly correlated with the low level of the influenza activity(r _AB=0.507,P=0.004;r _BC=0.419,P=0.017;r _AC=0.452,P=0.012),and the difference was significant in influenza outbreak period.The completeness of temperature was good but the accuracy was different.Individual,space,environmental factors and data post-processing were the main causes of abnormal high temperature readings,while individual,operation ability and instrument settings were the main causes of abnormal low temperature readings.4.The content of data quality improvement plan included:（1）three technological upgrades,namely,increasing the human body anchoring function of thermal imager,adjusting the optimal temperature measurement distance to 50-60cm,and adding temperature measurement behavior standard prompt words on the screen;（2）System operation specifications were formulated,including instrument installation specifications（5 items）,student behavior specifications（6 items）,school supervision specifications（4 items）and information feedback specifications（3 items）.Among the specifications,13 items were well implemented,2 items were suspended for reasons,and 3 items were not fully implemented.The intervention effect was significant:the mean abnormal high temperature rate in A and B schools decreased from 1.38%and0.47%to 0.07%and 0.04%;the average abnormal low temperature rate decreased from 20.09%and 12.56%to 4.14%and 2.32%.5.The characteristics of absenteeism rates vary between grades 1-2 and 3-6.There was no significant correlation between DARA-L and DARB-L（r=-0.056,P=0.488）,while there was significant correlation between DARA-H and DARB-H（r=0.464,p<0.001）.The warning sensitivity of DARA-L,DARA-H,DARA-T,DARA-P,DARA-S were all 100%,specificity was 72.3%,85.5%,64.0%,63.6%,and64.3%,respectively;the Youden index was 72.3%,85.5%,64.0%,63.6%,and 64.3%,respectively.The sensitivity,specificity and Youden index of the combination of DARA-L and DARA-H were 100%,65.5%and 65.5%,respectively.The sensitivities of DARB-L,DARB-H,DARB-T,DARB-P,DARB-S and DFRb were all 100%,the specificity was 86.46%,91.6%,82.4%,80.3%,81.3%,and 82.4%,respectively;?the Youden index was 86.46%,91.6%,82.4%,80.3%,81.3%,82.4%.The sensitivity,specificity and Youden index of DARB-L and DARB-H were 100%,84.0%and84.0%,respectively.If DFRb was included,the sensitivity,specificity and Youden index were 81.5%,88.9%and 70.4%,respectively.The sensitivity,specificity and Youden index of DARB-T and DFRb were 100%,85.1%and 85.1%,respectively.6.Combined with the research results,our comprehensive evaluation conclusions on AISSS were as follows:Cost effectiveness:AISSS was based on the existing network platform,saving the cost of system research and development.Automatic data collection,greatly reducing human resource dependence.Representativeness:AISSS takes into account both attendance and absenteeism cases,and its in-school sample coverage is better than that of other similar systems,while its community sample representation will gradually improve with the increase of users.Simplicity:AISSS system structure,data type and organization are simple.Automatic data collection,operation without too much training.Acceptability:Schools,parents and students have a high degree of cooperation in the operation of AISSS.Data quality:AISSS absence collection has less human interference,unified time,real-time report,and its data was more complete and accurate than the data reported by school physicians.The completeness of temperature was good,but the accuracy varied greatly from school to school.The technical upgrade of thermal imager and the system operation standard could effectively reduce the difference.Effectiveness:The outbreak detection of all-cause absence reported by AISSS was better than that of all-cause absence reported by school physicians.Temperature and absence information are complementary,and the combination of the two can improve the accuracy of epidemic surveillance.Standard use:AISSS software and hardware standards meet the national requirements,and its absenteeism and body temperature are reasonably defined.The implementation effect of the preliminary system operation specification is encouraging.Timeliness:AISSS absenteeism and body temperature were collected simultaneously,and the data were analyzed and fed back in real time.The ability of early epidemic warning reported by AISSS was better than that reported by school physicians,and the inclusion of fever rate further strengthened the timeliness of surveillance.Flexibility:Similar to existing SSS,AISSS has excellent flexibility.Security:The system is relatively open,and the data security work is challenging,but generally controllable.Stability:At present,the system was running stably.In the future,we could cooperate with CDC to reduce possible risks.Conclusions1.AISSS had a reasonable definition of absenteeism and fever.The completeness,accuracy and timeliness of absenteeism were better than those reported by school physicians.The completeness and accuracy of temperature data were good,but the accuracy was easily affected.2.The ideal operating environment of AISSS is:indoor,the temperature is5-25℃,there is no strenuous movement of individuals,and there is a distance marker and baffle in front of the instrument.3.Measures such as promoting the standardization of instrument installation environment,improving the intelligence level of thermal imagers,and strengthening the awareness of individual temperature measurement behavior norms can significantly improve the quality of AISSS data.4.The absenteeism rate in each grade of primary school has significant phase characteristics,and lower grades（grades 1-2）and higher grades（grades 3-6）have different epidemic surveillance effectiveness.5.The information between fever and absenteeism is highly complementary,and the inclusion of fever rate can significantly improve the overall monitoring accuracy and timeliness of the system.Therefore,the quality of AISSS body temperature data needs to be improved,and its value in infectious disease surveillance needs to be explored urgently.6.AISSS is better than manual absenteeism surveillance system in cost effectiveness,on-campus representation,simplicity,acceptability,data quality（absenteeism）,effectiveness and timeliness,while there is room for improvement in community representation,standard application,security,and stability.