The invention discloses a postoperative
delirium prediction method and device based on a least square
support vector machine algorithm and
electronic equipment, and aims to solve the problems that existing scale evaluation is high in subjectivity and
instrument data is difficult to obtain. The method comprises the following steps: acquiring clinical and postoperative data of a patient from an EMR, HIS and LIS integrated
system of a hospital according to time, crowd, treatment and result rules; carrying out cleaning, variable
processing and 8: 2 division on a training
verification set, and constructing a sample set; clinical data are used as input, POD results are used as output, a radial basis kernel function and grid search are selected to optimize LSSVM parameters to construct a model, Brier scores, AUC and calibration curves are used for
verification, the POD probability of patients is output, and high-
risk groups are recognized. The device comprises
data acquisition,
processing and prediction modules, and the
electronic equipment comprises a storage and a processor. The method can solve the problem that in the prior art, the clinical practicability is limited when a
score table or a medical instrument is used for evaluating the postoperative POD prediction of the old
hip fracture patient.