DOI 10.1016/j.semarthrit.2026.153033
Evaluating unsupervised and rule-based phenotyping methods versus administrative code counts for systemic sclerosis identification.
Аннотация
OBJECTIVE
Our study aims to evaluate the performance of an unsupervised machine learning approach, linear combinations of principal components (LPC); a rule-based approach, phenotype risk score (PheRS); and M34 International Classification of Disease (ICD) code counts for identifying systemic sclerosis (SSc) and predicting mortality from electronic health record (EHR).
METHODS
We conducted a single-center retrospective study. We developed SSc-specific PheRS and LPC score using 17 clinical features. We then evaluated these methods for identifying rheumatologist-diagnosed SSc and predict mortality.
RESULTS
Of 653 patients with at least one M34 code, 547 (84%) had rheumatologist-diagnosed SSc. When compared with randomly sampled matched controls, LPC outperformed PheRS in identifying rheumatologist-diagnosed SSc (AUROC 0.90 vs 0.87). However, in a three-method comparison among individuals with at least one M34 code, LPC was only marginally superior to PheRS and M34 code counts in distinguishing rheumatologist-diagnosed SSc from M34 non-SSc controls (AUROC 0.72 for LPC, 0.70 for PheRS and 0.70 for ICD code counts). Using at least two M34 codes had a sensitivity of 87.8% and positive predictive value of 87.0% in identifying SSc. Despite the limited performance gains for SSc identification, both LPC and PheRS had substantially higher performance in predicting mortality compared to M34 code counts in patients with SSc (AUROC 0.74 vs 0.72 vs 0.63).
CONCLUSION
M34 code counts achieved reasonable performance for SSc identification and comparable to those of LPC and PheRS. However, both LPC and PheRS demonstrated better performance for predicting mortality compared with M34 code counts in patients with SSc.
Предметные рубрики MeSH
Подготовка и перевод публикации
Обработка статьи · этапы, ошибки и управление
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Тип публикации указан по данным источника. Уровень доказательности автоматически не присваивается.