Semi-automatic information extraction on the presence of paresis in neurosurgical patients from health records: a research using artificial intelligence

Adverse events identification in clinical documents is necessary for retrospective clinical research and evaluating medical care safety and cost-effectiveness. Since adverse events are usually reported with free text in medical records, special technologies are required to extract information about them from thousands of medical records.

MATERIALS AND METHODS: We introduced a technology to extract information based on natural language processing (patent RU 2751993C 1). The method is grounded on preselecting a lexicon specific to a particular adverse event and deciphering its use in microcontexts. The textual medical records, initially typed in by doctors using the keyboard into the electronic health records system «E-med» of the FSAI «NMRC named after ac. N. N. Burdenko» of the ministry of health of Russia from 2000 to 2017 were analysed.

RESULTS. The technology we proposed enabled us to solve the task of paresis identification with high quality (sensitivity = 0,947, specificity = 0,965, accuracy = 0,961, F1-score = 0,926, ROC AUC = 0,956 [0,941; 0,969]). Optimization of the proposed method (using only nouns when screening a vocabulary) enabled to reduce the time for its implementation from 13 to 6 hours with no major loss in quality.

CONCLUSION. Natural language processing can improve the quality of information extraction from medical texts, which, in particular, can be successfully applied in neurosurgical safety research.

1. Wang Y, Wang L, Rastegar-Mojarad M, et al. Clinical information extraction applications: a literature review. J Biomed Inform. 2018;77:34–49. doi:10.1016/j.jbi.2017.11.011

2. Danilov G, Shifrin M, Strunina U, Pronkina T, Potapov A. An Information extraction algorithm for detecting adverse events in Neurosurgery Using documents written in a Natural rich-in-Morphology language. Stud Health Technol Inform. 2019;262:194– 197. doi:10.3233/SHTI190051

3. Danilov G.V., Shifrin M.a., Potapov A.A., et al. Method for extracting information from unstructured texts written in natural language (patent rU2751993C1). published online july 21, 2021. accessed december 27, 2021. https://www.fips.ru/registers-doc-view/fips_servlet?dB=rUpaT&docNumber=2751993&TypeFile=html (In russ.).

4. Danilov G, Shifrin M, Strunina Y, et al. Detection of muscle weakness in medical texts using natural language processing. In: Studies in Health Technology and Informatics. Vol 270. IoS press; 2020:163– 167. doi:10.3233/SHTI200143

5. Danilov G, Shifrin M, Strunina Y, et al. Semiautomated approach for muscle weakness detection in clinical texts. In: Studies in Health Technology and Informatics. Vol 272. IoS press; 2020:55–58. doi:10.3233/SHTI200492

6. Young IJB, luz S, lone N. A systematic review of natural language processing for classification tasks in the field of incident reporting and adverse event analysis. Int J Med Inform. 2019;132:103971. doi:10.1016/j.ijmedinf.2019.103971

7. Gaebel J, Kolter T, Arlt F, Denecke K. Extraction of adverse events From Clinical documents To Support decision Making Using Semantic preprocessing. Stud Health Technol Inform. 2015;216:1030.

8. Campillo-gimenez B, Garcelon N, Jarno P, Chapplain JM, Cuggia M. Full-text automated detection of surgical site infections secondary to neurosurgery in rennes, France. Stud Health Technol Inform. 2013;192:572–575.

9. Tvardik N, Kergourlay I, Bittar A, Segond F, Darmoni S, Metzger M-H. Accuracy of using natural language processing methods for identifying healthcare-associated infections. Int J Med Inform. 2018;117:96–102. doi:10.1016/j.ijmedinf.2018.06.002