Application of a rules-based natural language parser to critical value reporting in anatomic pathology

Am J Surg Pathol. 2012 Mar;36(3):376-80. doi: 10.1097/PAS.0b013e318245c9a4.

Abstract

Critical values in anatomic pathology are rare occurrences and difficult to define with precision. Nevertheless, accrediting institutions require effective and timely communication of all critical values generated by clinical and anatomic laboratories. Provisional gating criteria for potentially critical anatomic diagnoses have been proposed, with some success in their implementation reported in the literature. Ensuring effective communication is challenging, however, making the case for programmatic implementation of a turnkey-style integrated information technology solution. To address this need, we developed a generically deployable laboratory information system-based tool, using a tiered natural language processing predicate calculus inference engine to identify qualifying cases that meet criteria for critical diagnoses but lack an indication in the electronic medical record for an appropriate clinical discussion with the ordering physician of record. Using this tool, we identified an initial cohort of 13,790 cases over a 49-month period, which were further explored by reviewing the available electronic medical record for each patient. Of these cases, 35 (0.3%) were judged to require intervention in the form of direct communication between the attending pathologist and the clinical physician of record. In 8 of the 35 cases, this intervention resulted in the conveyance of new information to the requesting physician and/or a change in the patient's clinical plan. The very low percentage of such cases (0.058%) illustrates their rarity in daily practice, making it unlikely that manual identification/notification approaches alone can reliably manage them. The automated turnkey system was useful in avoiding missed handoffs of significant, clinically actionable diagnoses.

MeSH terms

  • Automation, Laboratory
  • Biopsy
  • Clinical Laboratory Information Systems*
  • Communication
  • Electronic Health Records
  • Humans
  • Interdisciplinary Communication*
  • Natural Language Processing*
  • Neoplasms / classification
  • Neoplasms / pathology*
  • Pathology, Clinical / methods*
  • Predictive Value of Tests
  • Terminology as Topic