Detection algorithms count the number of occurrences of a variable in a given spatial location over a given time period to look for anomalous patterns. Detection algorithms require structured data, that is, data in a format that can be interpreted by a computer. By far the most common structured data formats are relational database tables. An example of a structured data element is the number of units of cold and cough medicine sold over the last 24 hours in a particular county. Many other examples of structured data elements were described in earlier chapters and in Part IV.
Much data that could potentially be useful in biosurveillance are unstructured. These include symptoms reported by a patient when presenting at an emergency facility, physical and radiological findings recorded by a physician, and queries to healthcare related web sites. In order to use these data for biosurveillance, the information must be converted. We focus our discussion on the use of NLP to encode information from textual patient records for input to outbreak detection algorithms.
Was this article helpful?