Presentation (CTI)

OBJECTIVES: The Hurley staging system classifies hidradenitis suppurativa (HS) severity and is critical for guiding treatment decisions and evaluating patient outcomes. However, documentation of the Hurley Score is inconsistent in real-world data (RWD) sources, limiting the size of available HS patient cohorts for research. The purpose of our study was to develop and validate a prediction model to estimate the Hurley Score (eHurley) from clinical notes in the electronic medical records (EMR).
METHODS: We used the OM1 PremiOM™ HS, a linked claims and EMR dataset on HS patients in the US. Physician-recorded Hurley scores and predictive features were extracted from clinical notes. Candidate tokens were prioritized by clinical relevance and absolute difference in Hurley Scores between notes with and without the token, yielding 85 baseline predictors for the estimation model. A logistic regression binary classifier (HS Stage 1 vs. HS Stage ≥2) was trained on an 80/20 patient-level train/test split. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUROC).
RESULTS: Hurley Scores were successfully extracted from approximately 3,500 patients across 5,500 encounters, which served as the labeled training and test set. The logistic regression model achieved an AUROC of 0.84 on the training set and 0.81 on the test set. The predicted probability distribution of the estimation set closely mirrored that of the test set, supporting the validity of scores generated. The eHurley model was subsequently applied to generate scores for approximately 15,000 additional patients across 57,000 encounters, representing a 4.3-fold increase in patients and a 10.5-fold increase in available encounters.
CONCLUSIONS: This study developed and validated a prediction model to estimate Hurley scores from clinical notes with good performance. Application of the model to real-world data sets may allow estimated Hurley scores to be used for research purposes.