Extracting Most Impacting Emergency Department Patient Flow By Embedding Laboratory-confirmed and Clinical Diagnosis on The Stiefel Manifold

Clément Pealat (1) , Guillaume Bouleux (1) , Vincent Cheutet (1)

1 DISP - Décision et Information pour les Systèmes de Production

Emergency departments (ED) in France are jeopardized each winter by the respiratory viruses. To limit the impact of those viruses, it is essential to have a better understanding of their impact on the patient flow. To tackle this, we propose in this work to use in conjunction ICD-10 code and laboratory-confirmed data with the aim of extracting a relevant patient flow. We first take benefice of the almost periodicity of both clinical diagnosis and laboratory-confirmed data and we embed next the underlying time series on the Stiefel manifold. The distance in the Stiefel manifold is finally used to extract clinical codes which are the nearest to the laboratory-confirmed time series. The results reveal that some of the respiratory and cardiac disorders codes have the same behaviours than that of the winter circulating viruses. At least, the Flag mean is employed to dispose of a picture of both the patient flow and the the length of stay for patients who might be infected by winter viruses.

Mots clés: Flag mean; Index Terms-Stiefel Manifold; Subspaces of different dimensions; Influenza; Respiratory Syncytial Virus; Patient Flow

Domaines: Intelligence artificielle [cs.AI]

Clément Pealat, Guillaume Bouleux, Vincent Cheutet. Extracting Most Impacting Emergency Department Patient Flow By Embedding Laboratory-confirmed and Clinical Diagnosis on The Stiefel Manifold. IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI 2019), May 2019, Chicago, United States. 4 p. ⟨hal-02141166⟩