Context: The utilization of poison center data for research and surveillance depends on the complete and accurate coding of the data.
Objective: The intent of this study was to describe the coding that Texas poison centers used to identify calls relating to the recent H1N1 outbreak and the extent to which the coding accurately identified the calls.
Methods: Cases were all H1N1 calls added to the six Texas poison centers' common database during April 28-September 30, 2009. Cases were selected by identifying all records that had the required H1N1 code or text in the FreeArea1, Substance Verbatim, or PoisIndex code fields or had the terms "swine" or "H1N1" anywhere within the Notes field. The proportion of fields that were coded correctly was determined for each field alone and all three combined and for each of the six poison centers and by 2-month period.
Results: Of the 222 H1N1 calls identified, the FreeArea1 field was coded correctly in 67.1% cases, Substance Verbatim field in 73.9%, PoisIndex code field in 73.9%, all three fields together in 45.9%, and none of the three fields in 9.0%. Correct coding of all three fields ranged from 29.7 to 65.5% between the poison centers. All three fields were coded correctly in 49.7% of the April-May calls, 38.5% of June-July calls, and 29.6% of the August-September calls.
Discussion: All three of the fields were coded correctly in less than half of the H1N1 calls; however, at least one of the fields was useful in identifying 91% of the calls. Correct coding rates varied widely between the poison centers and declined over time.
Conclusions: Use of three different fields to code H1N1 calls identified the majority of such calls received. However, dependence on a single field would have missed a number of calls.