%0 Journal Article
%T Detecting false-positive disease references in veterinary clinical notes without manual annotations
%A Dan G. O¡¯Neill
%A Dave C. Brodbelt
%A David B. Church
%A Noel Kennedy
%J Archive of "NPJ Digital Medicine".
%D 2019
%R 10.1038/s41746-019-0108-y
%X A diagram showing how our method created labels for disease phrases in clinical text. Two patients are represented, one in each row. The patients are treated differently, as one patient received a relevant clinical code, and the other didn't. The bottom row represents a patient that was never coded with the disease that was mentioned in their notes. The area above each row¡¯s timeline represents events in the electronic medical record (EMR) system. The area below the timeline represents interpretations given by our method based on the events in the EMR. Each time the disease is mentioned in the EMR, our method labels the sentence with one of four labels. In the case of the bottom row, where the patient was never coded, the method is simple to apply: all sentences containing disease mentions are given the ¡®Never diagnosed¡¯ label. The top patient received a clinical code, indicating that the patient had been diagnosed with the disease which was mentioned in their notes. This row is slightly more complicated, as there are three potential labels that can be applied: ¡®Pre-window¡¯, ¡®During window¡¯ and ¡®After window¡¯. The label applied depends on temporal relationship in the EMR between the disease reference and the clinical cod
%K Epidemiology
%K Diagnosis
%K Signs and symptoms
%K Epidemiology
%K Diagnosis
%U https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6550178/