Most disasters are readily identifiable but health care disasters may not be as readily recognized, particularly early in their evolution. Syndromic surveillance, one kind of health care disaster, is the term coined for systems that use data that precede diagnosis to recognize disease outbreaks with sufficient specificity to warrant a public health response. Algorithms designed for early detection of health care disasters that rely solely on temporal trends may not be specific enough and several investigators have begun to incorporate geospatial clustering into the pattern recognition algorithms in order to improve specificity. Knowing the specific geographic location of the patient's home address would allow easy identification of a patient in many cases, however, we propose to develop and characterize pattern recognition algorithms that operate on deidentified data in order to protect patients' privacy. We will carry out this work using actual data from two different cities in order to increase the chance that the approach developed can be generalized. The overall approach would be to utilize established pattern recognition algorithms developed at Children's Hospital in Boston, and modify them as needed to accommodate use of deidentified data. Then we will compare the detection characteristics of the algorithm using identifiable data and deidentified data in order to determine the effect of using deidentified data.
Bourgeois, Florence T; Shannon, Michael W; Valim, Clarissa et al. (2010) Adverse drug events in the outpatient setting: an 11-year national analysis. Pharmacoepidemiol Drug Saf 19:901-10 |
Bourgeois, Florence T; Mandl, Kenneth D; Valim, Clarissa et al. (2009) Pediatric adverse drug events in the outpatient setting: an 11-year national analysis. Pediatrics 124:e744-50 |