Photo Credit: Anyaberkut

The following is a summary of “Early identification of suspected serious infection among patients afebrile at initial presentation using neural network models and natural language processing: A development and external validation study in the emergency department,” published in the June 2024 issue of Emergency Medicine by Choi, et al.

For a study, researchers sought to develop and externally validate predictive models utilizing neural networks and natural language processing (NLP) to identify suspected serious infections in patients initially afebrile upon presentation to the emergency department (ED).

The retrospective study included adult patients who presented to the ED without fever. Four artificial neural network models were developed using patient demographics, vital signs, laboratory test results, and textual data extracted from initial ED physician notes using the term frequency-inverse document frequency (TF-IDF). Models were trained, internally validated using data from one hospital, and externally validated using data from a different hospital. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC) with 95% CIs.

The study datasets comprised 150,699 patients for training, 37,675 for internal validation, and 85,098 for external validation. In the internal validation dataset, the AUCs (95% CIs) for Models 1 (demographics + vital signs), 2 (demographics + vital signs + initial ED physician note), 3 (demographics + vital signs + laboratory tests), and 4 (demographics + vital signs + laboratory tests + initial ED physician note) were 0.789 (0.782–0.796), 0.867 (0.862–0.872), 0.881 (0.876–0.887), and 0.911 (0.906–0.915), respectively. In the external validation dataset, the AUCs (95% CIs) for Models 1, 2, 3, and 4 were 0.824 (0.817–0.830), 0.895 (0.890–0.899), 0.879 (0.873–0.884), and 0.913 (0.909–0.917), respectively. Model 1 was applicable immediately after ED triage, Model 2 after initial physician notes (median time: 28 min), and Models 3 and 4 after initial laboratory results (median time: 68 min).

The study successfully developed and validated neural network models enhanced by NLP to detect suspected serious infections in patients initially afebrile upon presentation to the ED. Integration of information extracted from initial ED physician notes significantly improved model performance, enabling early identification of suspected serious infections during ED visits.

Reference: sciencedirect.com/science/article/abs/pii/S0735675724001153