建立臺灣地區急診為基礎的傳染病症候群偵測系統－ICD-9異常訊號偵測法、流行病學特徵、系統評比與建議

Abstract

According to the challenges of emerging infectious diseases (EID) and bioterrorist attack in the 21st Century, the traditional infectious disease surveillance systems have lacked active surveillance and failed to detect the 2001 anthrax attacks and the 2003 cross-country spread of severe acute respiratory syndrome (SARS) early. In order to establish scientific evidence-based active infectious disease surveillance system, we initiated a hospital emergency department (ED)-based timely and automatic syndromic surveillance system (ED-SSS) in Taiwan for the basis of further epidemiological investigation. Through the introduction of the real-time outbreak and disease reporting system (RODS) from the University of Pittsburg to Taiwan, we established both a nation-wide ED-SSS involving 189 hospitals and set up more integrative procedures of local ED-SSS for daily operation, signal investigation, checking data quality and specimens taken at one regional hospital with 5 branches in metropolitan Taipei. In this study, we built up the daily operation procedures to perform further implementation and signal investigation and conducted epidemiological syndrome data analysis, using two data set of ICD-9 codes of ED visits: (1) 2,692,325 visits of nation-wide ED-SSS in Taiwan, April 1, 2004 - March 31, 2005. (2) 130,313 visits in one regional Taipei hospital, October 1, 2005- April 30, 2006. Daily patterns and trends of nation-wide ED-SSS ICD-9 codes involving 11 syndrome groups were analyzed, according to the international research experiences and recommendations of infectious disease experts in Taiwan. We found increasing trends in gastro-intestinal (GI) syndromes during winter season, and appearing peaks of respiratory or influenza-like illness (ILI) in summer periods, plus two waves of enterovirus (EV)-like infections under 6 years of age. In addition, striking increase of ED-visits were shown in weekends (1.3-fold p<0.001), and holidays, especially 2.0-fold higher during Chinese Lunar New Year (p<0.001). Most ED visits were children aged <10 y/o (18.6%) and 20-39 year-old young adults (17.4%). Male vs. female ED visits were 1.12:1. Starting from July of 2004, case numbers of fever, respiratory and ILI syndromes gradually increase, indicating possible presence of respiratory infections in Taiwan that had been seldom reported and worthwhile future investigating. Therefore, we examined the sensitivity, specificity and positive-predictive values of different ICD-9 code syndrome groupings of ILI cases from the 1,281 cases through medical charts review at C-hospital in Taipei, June 24, 2004 – July 8, 2004. The modified version of ILI syndrome that deleted asthma, fever and dizziness had 75.5% sensitivity and 84.4% specificity. We then applied it to detect ILI cases in Taipei ED-SSS. Increased numbers and earlier detected EVI syndrome cases of nation-wide ED-SSS were found in central Taiwan, but there was no significant difference in patterns by different age and gender groups. Compared with sentinel physicians’ hand-foot-mouth disease surveillance, the ED-SSS was 2 weeks earlier in detecting the increased EVI syndrome cases in Taiwan. Additionally, the Taipei X-Hospital ED-SSS also showed similar pattern. Among the patterns of other syndrome groups, GI syndrome cases in both nation-wide and Taipei ED-SSS increased from Nov. of 2004. On the other hand, Taipei X-Hospital ED-SSS with different branches showed regional differences. Moreover, 30-39 year-old male adults showed one week earlier in the increasing of ILI syndrome cases implying possible important transmission role in this age group in a metropolitan city. With the limit historical ED-SSS data, both historical control limits (HCL) using past four-week background data and cumulative sum (CUSUM) process control methods were used for our aberration detection and their required conditions in ED-SSS. The weekday and weekend stratified HCL method showed 67% sensitivity to detect the Taiwan-CDC identified outbreaks and the modified CUSUM method was too sensitive to demonstrate signals on weekends and holidays and needed to be improved due to its higher sensitivity for the changes of continuous baseline. Undoubtedly, both methods are applicable to public health use for immediate prevention and control of infectious diseases but they need further researches to reduce false signals. Evaluation of surveillance systems found that ED-SSS, providing more information for further epidemiological investigation, had the best timeliness and better flexibility to integrate with other database for multi-disciplinary research than other infectious disease surveillance systems existed. It can serve two-way frequent communications between public health agencies and medical departments. With the large baseline data of community health, it can also apply to research on other domains in public health. Future efforts are needed for continuous investment of medical informatics, to recruit more high quality infection control public health personnel, and to improve different statistical prediction models for maximizing effectiveness for early EID detection.