Previously submitted to: JMIR Dermatology (no longer under consideration since Jan 21, 2026)
Date Submitted: Aug 22, 2025
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
A Novel Algorithm for Identifying Generalized Pustular Psoriasis Flares Using Natural Language Processing and Outpatient Electronic Health Records
ABSTRACT
Background:
Generalized pustular psoriasis (GPP) is a systemic, neutrophilic, inflammatory disease characterized by chronic symptoms and periods of flaring, which can be potentially life-threatening. The rarity of GPP and inconsistent documentation complicate real-world identification of GPP flares.
Objective:
We developed an algorithm to identify outpatient encounters indicative of GPP flares using structured data and unstructured clinical notes from electronic health records (EHRs).
Methods:
De-identified patient-level EHR data (2017–2024) from six specialty dermatology networks in the USA were accessed. Eligible outpatient encounters had (1) an International Classification of Diseases, 10th Revision, Clinical Modification diagnosis code for GPP, (2) a dermatologist-defined diagnosis status of “inadequately controlled” or “worsening” disease (gold standard indicator of flare), and (3) available clinical notes. Natural language processing (NLP) of clinical notes identified the most common terms (eg, clinical manifestations, treatment) by flare status.
Results:
Multivariable logistic regression models and leave pair-out cross-validation selected models yielding the highest area under the receiver-operating characteristic plot (AUROC) (discriminative ability for flares) from 1,354 encounters (679 patients: 81% female, 84% Caucasian, mean age 58 years). The final model combined NLP terms and EHR variables and yielded a higher discriminative ability (AUROC=0.72; sensitivity=0.68; specificity=0.66) than either NLP terms or EHR data alone.
Conclusions:
Although not fully discriminative, the model could be used in future to identify GPP flares from EHR data. Clinical Trial: N/A
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.