Citation

Tignanelli CJ, Silverman GM, Lindemann EA, Trembley AL, Gipson JL, Beilman G, Lyng JW, Finzel R, McEwan R, Knoll BC, Pakhomov S, Melton GB. J. Trauma Acute Care Surg. 2020; ePub(ePub): ePub.

Affiliation

Christopher J. Tignanelli, MD, Department of Surgery, University of Minnesota, Minneapolis, MN; Institute for Health Informatics, University of Minnesota, Minneapolis, MN; Department of Surgery, North Memorial Health Hospital, Robbinsdale, MN Greg M. Silverman, BS, Institute for Health Informatics, University of Minnesota, Minneapolis, MN Elizabeth A. Lindemann, BS, Institute for Health Informatics, University of Minnesota, Minneapolis, MN Alexander L. Trembley, NRP, BSM, North EMS, Robbinsdale, MN Jon L. Gipson, MD, Department of Surgery, North Memorial Health Hospital, Robbinsdale, MN Gregory Beilman, MD, Department of Surgery, University of Minnesota, Minneapolis, MN John W. Lyng, MD, NRP, Department of Emergency Medicine, North EMS, North Memorial Health Hospital, Robbinsdale, MN Raymond Finzel, BS, Institute for Health Informatics, University of Minnesota, Minneapolis, MN Reed McEwan, MS, Institute for Health Informatics, University of Minnesota, Minneapolis, MN Benjamin C. Knoll, BS, Institute for Health Informatics, University of Minnesota, Minneapolis, MN Serguei Pakhamov, PhD, Institute for Health Informatics, University of Minnesota, Minneapolis, MN Genevieve B. Melton, MD, PhD, Department of Surgery, University of Minnesota, Minneapolis, MN; Institute for Health Informatics, University of Minnesota, Minneapolis, MN.

Copyright

(Copyright © 2020, Lippincott Williams and Wilkins)

DOI

10.1097/TA.0000000000002598

PMID

31977990

Abstract

BACKGROUND: Incomplete prehospital trauma care is a significant contributor to preventable deaths. Current databases lack timelines easily constructible of clinical events. Temporal associations and procedural indications are critical to characterize treatment appropriateness. Natural language processing (NLP) methods present a novel approach to bridge this gap. We sought to evaluate the efficacy of a novel and automated NLP pipeline to determine treatment appropriateness from a sample of prehospital EMS motor vehicle crash (MVC) records.

METHODS: 142 records were utilized to extract airway procedures, intraosseous (IO)/intravenous (IV) access, packed red blood cell (PRBC) transfusion, crystalloid bolus, chest compression system, tranexamic acid (TXA) bolus, and needle decompression. Reports were processed using four clinical NLP systems and augmented via a word2phrase method leveraging a large integrated health system clinical note repository to identify terms semantically similar with treatment indications. Indications were matched with treatments and categorized as indicated, missed (indicated but not performed), or non-indicated. Automated results were then compared with manual review and precision and recall were calculated for each treatment determination.

RESULTS: NLP identified 184 treatments. Automated timeline summarization was completed for all patients. Treatments were characterized as indicated in a subset of cases including: 69% (18 of 26) for airway, 54.5% (6 of 11) for IO access, 11.1% (1 of 9) for needle decompression, 55.6% (10 of 18) for TXA, 60% (9 of 15) for PRBC, 12.9% (4 of 31) for crystalloid bolus, and 60% (3 of 5) for chest compression system. The most commonly non-indicated treatment was crystalloid bolus (22 of 142 patients). Overall, the automated NLP system performed with high precision and recall with over 70% of comparisons achieving precision and recall of greater than 80%.

CONCLUSION: NLP methodologies show promise for enabling automated extraction of procedural indication data and timeline summarization. Future directions should focus on optimizing and expanding these techniques to scale and facilitate broader trauma care performance monitoring.

LEVEL OF EVIDENCE: Diagnostic Tests or Criteria, Level III.


Keywords: Social Transition

Language: en