Skip to Main content Skip to Navigation
Master Thesis

Détection automatique des infections du site opératoire

Abstract : Introduction Improving monitoring and prevention of surgical site infections (SSI) is part of the national nosocomial infection control program. Our goal was to implement an automated detection tool with a supervised machine learning to replace the current manual system. Method Two approaches were conducted to detect SSI following spine surgery and neurosurgery corresponding to 2133 and 2303 procedures respectively. The first approach uses the multiple sources of information available in the data warehouse of Bordeaux University Hospital. The second approach uses only free text. For each approach, we compared the precision of two algorithms namely logistic regression and random forests algorithms for a fixed recall value of 100%. Results The final model using all the data achieved the best performance for spine surgery with a precision of 94%. The model using free text data obtained correct results and was better for neurosurgery. Discussion The use of free text has the advantage of being replicable to other health institutions and applicable to various surgical specialties with stable performance. The performance of our algorithms needs to be evaluated on a test dataset.
Document type :
Master Thesis
Complete list of metadatas

Cited literature [41 references]  Display  Hide  Download

https://dumas.ccsd.cnrs.fr/dumas-02420229
Contributor : Isped Ub <>
Submitted on : Thursday, December 19, 2019 - 5:32:46 PM
Last modification on : Friday, October 23, 2020 - 4:41:01 PM
Long-term archiving on: : Friday, March 20, 2020 - 10:05:25 PM

File

M2SITIS_2019_QUÉROUÉ.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : dumas-02420229, version 1

Collections

Citation

Marine Quéroué. Détection automatique des infections du site opératoire. Santé publique et épidémiologie. 2019. ⟨dumas-02420229⟩

Share

Metrics

Record views

131

Files downloads

143