Background: Understanding real-world outcomes of patients receiving treatment for inflammatory rheumatic diseases is of paramount importance. Real-world observational studies offer a unique opportunity to investigate the long-term effectiveness and safety associated with different therapeutic interventions. By capturing data from routine clinical practice, these studies bridge the gap between controlled clinical trials and the complexities of managing these diseases in everyday healthcare settings.

Objectives: In this study we explored the feasibility of analyzing structured and unstructed data from anonymized electronic health records (EHRs) with the help of artificial intelligence (AI), particularly natural language processing (NLP) and machine learning. Healthcare professionals would visualize these data through dashboards that include information about the most important treatments and outcomes of patients with rheumatoid arthritis (RA).

Methods: Anonymized EHRs from VieCuri Medical Center (Venlo, the Netherlands) were processed using NLP and machine learning to generate an Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) database that was validated in-house, as previously described [1-3]. Data of patients from the Rheumatology department with a diagnosis of RA from 2019-2023 were analyzed regarding variables such as demographics, clinical characteristics, comorbidities, procedures, disease-specific scores, complications, laboratory parameters, or medications. Dashboards for result visualization were generated using PowerBI software (Microsoft).

Results: A total of 2190 patients with RA were analyzed, with a mean follow up of 3.3 years. The mean age was 65.5 years and 64% (n = 1403) were females. The most frequently reported comorbidities [4] were hypertension (23%, n = 696), cancer (21%, n = 628) and diabetes mellitus (12%, n = 354). Deep venous thrombosis was detected in 2.5% (n = 76) of patients and pulmonary embolism in 2.6% (n = 79). Regarding therapeutic patterns, use of methotrexate as single therapy or in combination with other DMARDs was found in 64% (n = 1400) of patients, of which the majority (66%, n = 930) were on a weekly dose of ≤ 15 mg. Hydroxycholoroquine use was found for 49% (n = 1079) of patients, whereas biological/biosimilar use appeared for 16% (n = 346) of patients. By following disease activity scores (DAS) per patient, sustained remission was detected for 27% of patients, defined as two consecutive DAS scores < 2.6 and no DAS score > 3.2 during follow up, with a mean of 470 days from RA diagnosis to remission.

Conclusion: With the help of AI methodologies, it is feasible to collect structured and unstructured data about pharmacotherapeutic interventions and treatment outcomes of patients with a diagnosis such as RA. The main advange of our approach lies in its seamless integration with existing workflows. The healthcare professionals involved in this study were not burdened with the need to record additional data beyond their routine EHR entries.

REFERENCES: [1] Van de Meulebroucke C et al. What Can We Expect Following Anterior Total Hip Arthroplasty on a Regular Operating Table? A Validation Study of an Artificial Intelligence Algorithm to Monitor Adverse Events in a High-Volume, Nonacademic Setting. J Arthroplasty. 2019 Oct;34(10):2260-2266.

[2] Hens D et al. Validation of an Artificial Intelligence driven framework to automatically detect red flag symptoms in screening for rare diseases in electronic health records: hereditary transthyretin amyloidosis polyneuropathy as a key example. J Peripher Nerv Syst. 2023 Mar;28(1):79-85.

[3] Moya A, et al. Detection of transthyretin amyloid cardiomyopathy by automated data extraction from electronic health records. ESC Heart Fail. 2023 Dec;10(6):3483-3492.

[4] Filipowicz-Sosnowska A. Comorbidities and multimorbidity in rheumatic diseases. Reumatologia. 2019;57(1):1-2.

Acknowledgements: NIL.

Disclosure of Interests: Dries Hens: None declared, Piet van Riel Pfizer Netherlands, Manon Merkelbach: None declared, Ruud Simons: None declared, Nils Cornelis: None declared, Tim Jansen: None declared.