Clustering Analysis of Patients Kdigo Scoring Time Series Data Based on DTW Algorithm

Acute kidney injury (AKI) is a clinical syndrome caused by rapid decline of renal function in a short time. Although renal dysfunction is a decisive factor of AKI, the subtypes of different trajectories have not been fully studied, so it is important to determine the clinical subtypes. AKI patients in the open source database of critical medicine mimic3 were taken as the research object. A 72-hour kdigo scoring track was created. Then, dynamic time warping (DTW) is used to calculate the similarity of heterogeneous kdigo trajectories, and hierarchical clustering (HAC) is used to identify trajectory-based subtypes. Two trajectory-based subtypes of AKI were identified, namely, continuous deterioration and delayed improvement. The results of subtypes were used in multiple logistic regression analysis, and a risk nomograph was established. It was concluded that sofa, vaso, vent, chf, hypertension, resprate, spo2, creatinine, chloride, bun, ph were independent risk factors for the development of AKI patients. This study plays a reference role in the development of AKI and can provide help for clinical staff.