PhD Scientific Days 2023

Health Sciences - Posters Q

The Retrospective Evaluation of the Pharmaceutical Therapies at Karolina Hospital via Text Mining Technology

Text of the abstract

Introduction: In the Karolina Hospital, our Health Information System (HIS) generates a large amount of unstructured data for each care event that we can access as text files. Text mining is the process of examining this large collection of documents to discover new information.
Aims: Our objective was to reconstruct information from all 2022 medical records for an evaluation of antibiotic stewardship and to calculate the Medication Risk Score (MERIS).
Method: We used Python for the extraction of the data from the 6049 unique text files. To create keywords and convert drug names to ATC codes, we used the drug database of the National Institute of Pharmacy and Nutrition. We located the text file sections pertaining to drug therapy and organized the actual treatments in a dictionary. The documents that contained the chosen therapy data were also searched for the eGFR value. On the collected data, we applied the MERIS algorithm and classified the therapies into low-, medium-, and high-risk groups based on the results. We searched for the antibiotics used in the same treatments to determine the proportion of their usage.
Results: In all the health records, we found 2999 that matched the selection criteria. In 91% of the selected cases, we found an eGFR value as well. We calculated the MERIS score of all treatment with the following distribution: 52.3% low-, 34.7% medium- and 13% high-risk. Antibiotics were used in 46.3% of cases. The most used antibiotic drug was amoxicillin/clavulanic acid, and the proportion of fluoroquinolones was only 5.2%.
Conclusion: The distribution of MERIS scores was matched with the literary data, which proves that the calculator program works well, and the proportion of antibiotic drug usage shows that our antibiotic stewardship activity has been successful in the last few years.