Page 25 - AIH-1-1

Artificial Intelligence in Health NLP in EHR

Figure 1. Flow of the systematic literature review.

The exclusion criteria involved articles that applied NLP in 2.2.2. Classification of articles
areas other than EHR. In this review article, we used the classification listed in
In the second step, the articles selected in the first phase Table 1.
underwent further classification through a thorough
examination of abstracts, methods, limitations, and 3. Reporting of findings
future scope. The inclusion criteria for articles in this
phase were as follows: The results of the SLR were presented at this stage. We
(i) Use of NLP to identify or predict certain diseases. propose the use of pragmatic-level EHR to introduce more
(ii) Use of NLP to identify certain aspects of EHR. automation. In addition, we discuss the implications for
researchers, practitioners, and patients.
The exclusion criteria for articles were as follows:
(i) Lack of clear methods. 3.1. NLP levels
(ii) Related to biomedical terminologies. The included studies employed different NLP levels within
(iii) Involving classification. EHR, and their respective findings are systematically
(iv) Focused on the prediction of morbidity. presented based on the NLP levels. A summary of the
(v) Related to smoking status in EHR. discussed studies is presented in Table 2.
(vi) Related to image extraction.
3.1.1. Phonetic level
In the third phase, the selected articles from the second
phase were highlighted to indicate information about the Tissot and Dobson conducted a study that combined
[19]
NLP level that they focused on, their contribution, future phonetic and morphological levels, executed in Portuguese.
scope, and limitations. Since article text or notes, progress There is potential for expanding the study to larger corpora,
or visit notes, pathology, and radiology are integral parts utilizing machine learning algorithms, and incorporating
of EHR, titles containing these terms were included in the information beyond drug names.
review. These articles were then shared with Dr. Hegler
Tissot who specializes in EHR to identify the NLP levels. 3.1.2. Morphological level
After individual verification, the articles were finalized for Feller et al. assessed HIV risk from clinical notes in
[20]
the review. English. The study encountered several limitations, such

Volume 1 Issue 1 (2024) 19 https://doi.org/10.36922/aih.2147