Efficient identification of postdischarge surgical site infections: Use ofautomated pharmacy dispensing information, administrative data, and medical record information

Although most surgical site infections (SSIs) occur after hospital discharg e, there is no efficient way to identify them. The utility of automated cla ims and electronic medical record data for this purpose was assessed in a c ohort of 4086 nonobstetric procedures following which 96 postdischarge SSIs occurred. Coded diagnoses, tests, and treatments were assessed by use of r ecursive partitioning, with 10-fold cross-validation, and logistic regressi on with bootstrap resampling, Specific codes and combinations of codes iden tified a subset of 2% of all procedures among which 74% of SSIs had occurre d. Accepting a specificity of 92% improved the sensitivity from 74% to 92%. Use of only hospital discharge diagnosis codes plus pharmacy dispensing da ta had sensitivity of 77% and specificity of 94%. All of these performance characteristics were better than questionnaire responses from patients or s urgeons. Thus, information routinely collected by health care systems can b e the basis of an efficient, largely passive, surveillance system for postd ischarge SSIs.