A sequential text search algorithm is superior to adminstratively coded data in estimating wound dehiscence as a surgical patient safety indicator

Abstract

Abstract Introduction: The federal government has defined 20 Patient Safety Indicators by administrative discharge coding, including postoperative wound dehiscence. This project demonstrates the shortcomings of this method and utilizes an alternative technique to identify events. Methods: From 1998–2004, all coded discharge summaries at our institution were electronically screened for 362 codes to indicate an abdominopelvic procedure, comprising the denominator-at-risk. Screening of this subset for the code reclosure of postoperative disruption of abdominal wall (54.61) was compared with Boolean sequential text searching for or evisceration. Results were de-identified by an honest broker using proprietary HIPAA compliant software and reviewed. Results: 14,862 unique patients had 18,319 visits containing 28,532 matching discharge codes. 91 events were identified using the coded definition of dehiscence, but only 47 were true positives (52%). False positives resulted from diagnostic miscoding of patients whose wounds were intentionally left open at the initial operation or who had delayed primary wound closures. Sequential text searching identified 535 possible events of which 98 (p Conclusions: Nearly half of the cases identified by the administrative definition of dehiscence were miscoded, thus estimation of the cost and quality burden of this complication by screening discharge codes is flawed. If the goal is to identify actual adverse events, sequential text search detects more true cases and is superior.