DERMACLEAR Study, NLP System Increase Real-World Evidence of Clinical Practice

The DERMACLEAR study, which makes use of natural language processing (NLP), will increase the real-world evidence of clinical practice, according to a study¹ published in the Journal of the European Academy of Dermatology and Venereology. Furthermore, on average, the NLP system had a precision rate exceeding 95%, indicating its validity of use in the DERMACLEAR study.

AREE/AdobeStock

NLP is an artificial intelligence tool capable of analyzing large data sets and converting them into more structured, easily digestible data. The noninterventional DERMACLEAR study seeks to retrospectively collect patient data via electronic health records (EHR) from 7 dermatology departments in Spainin order to provide a comprehensive analysis of the proportion of patients with hidradenitis suppurativa (HS), psoriatic disease (PsO), chronic urticaria (CU), and atopic dermatitis (AD).

Investigators are also examining patient profiles, journeys, treatment patterns, and disease activity, as well as health care burdens, associated with these conditions. Patients 18 years of age and older with a clinical HS, PsO, CU, or AD diagnosis who had at least 1 outpatient visit to any of the 7 tertiary departments between June 2015 and June 2021 were eligible for inclusion. Patients lacking available EHR were excluded. To date, DERMACLEAR has gathered data from 54,458 patients.

In order to collect and manage data, the NLP system Medical Language API was used to identify and extract patient EHR information. Data was anonymized with an IOMED server capable of removing personal information and identifiers. Investigators carried out external and internal verification of the NLP system.

In this NLP system analysis, random samples for verification included records from 683 HS patients, 756 PsO patients, 821 CU patients, and 896 AD patients. Across all 7 study sites, the NLP system exceeded 95% accuracy in its verification process both internally and externally.

For most medical variables of interest, such as type of disease/disorder, type of drug, body area affected, outcome measure/diagnostic test, symptom, nonpharmacological treatment, and more, the NLP system was equal to or greater than 95% precise in identifying variables of interest based on internal verification.

“The high precision observed in this study is reassuring and enables physicians to trust the system to extract and process health information accurately,” according to Ortiz de Frutos et al. “This first-in-kind study will implement AI through NLP to process EHRs and determine the proportions of patients with HS, PsO, CU and/or AD in Spain, as well as describe patient profiles, patient journeys and the disease and healthcare burden....The DERMACLEAR study will increase the real-world evidence of clinical practice, obtaining a large amount of information on patients with HS, PsO, CU and/or AD attending hospitals in Spain. Results from the DERMACLEAR study will provide insight into the disease prevalence, clinical unmet needs and patient profile of patients with HS, PsO, CU and/or AD. This may further elucidate the role of a comprehensive follow-up and disease management of patients suffering from these dermatological diseases. The precision of the NLP system used in the DERMACLEAR study is high (≥95%) in identifying HS, PsO, CU and/or AD in EHRs, verifying that it is a valid instrument with clinical utility for the DERMACLEAR study.”

Reference

1. Ortiz de Frutos FJ, Giménez‐Arnau AM, Puig L, et al. The dermaclear study: verification results of a natural language processing system in dermatology. JEADV Clin Prac. Published online August 11, 2023. doi:10.1002/jvc2.217