| Summary: | 碩士 === 國立陽明大學 === 衛生資訊與決策研究所 === 93 === Study Purpose: More and more evidences have shown already, the information technology to the modern public health will be more important. Using information science and technology to help public health personnel to oppose the emerging infectious diseases is a major subject nowadays. This study utilized the administrative data in the past three years to set up an automatic syndromic surveillance system and verified the system utilizing the first half year of 2003.
Materials and Methods: This study utilized the administrative data of Taipei city collected from 2000 to 2002 and classified the patient's ICD-9-CM codes into fever, respiratory, and gastrointestinal syndromes established by the U.S.A.'s CDC and ESSENCE. We used linear regression with autoregressive error (abbreviated as Autoreg) and ARIMA statistic model to establish our prediction models for each syndrome. This research used eight kinds of statistical models to predict the counts of each three syndromes of the first half year of 2003. Influenza-like illness and diarrhea data provided by sentinel physicians and "SARS-related events" defined by the magazine of Taiwan Public Health Association were used as the gold standard for evaluation the performance of the system.
Results: This study established a web-based automatic syndromic surveillance system which could utilize the Internet to upload the data files, implement statistical analysis regularly, insert new data into database automatically , offer various types of statistical graphs and GIS (geographical information system) maps, and present the latest epidemic situation materials on the webpage. The evaluation of the system is as follows: First, in the whole city prediction from January 1, 2003 to June 30, 2003, the sensitivity was 0.69 on average, all of the positive predictive values for four models of using ARIMA were 1, and the timeliness of the system was delayed two days on average. Second, in the assessment of the "fever syndrome" in SARS epidemic period in 12 districts from Mar 10, 2003 to Jun 30, 2003, all of the sensitivity rates of the four models using ARIMA were 1, the timeliness of the system was 5 days earlier on average.
Conclusions and Suggestions: The whole city data may be useful for monitoring the naturally occurring outbreaks (e.g., influenza-like illness); the specific district data with fever syndrome may be useful for detecting SARS-related events. Because of the differences in demographics and medical resources in the individual district, how to choose an adequate statistical model and a significant-level value for the individual district is a great challenge of this kind of research.
|
|---|
