Severe AKI early risk assessment model and device based on interpretable machine learning model and development method of model

Abstract

The invention provides a severe AKI early risk assessment model and device based on an interpretable machine learning model and a development method of the model. In data set establishment, based on a plurality of large electronic health record data sets, crowd selection, research variable inclusion and positive and negative sample set establishment are completed according to a research scheme; during data processing, data extraction and preprocessing are carried out, and construction of a statistical feature data set is completed based on development of a machine learning model; in model construction and evaluation, model construction and training are carried out based on the statistical feature data set, the performance of the model is further evaluated based on nine formulated indexes and three formulated modes, and a trained and calibrated prediction model is obtained; and in the model application, model migration, training and re-calibration are carried out for different application scenes based on the trained prediction model.

Description

technical field [0001] This application relates to machine learning, in particular to an early risk assessment model for severe AKI based on an interpretable machine learning model, a device and a development method thereof. Background technique [0002] Acute kidney injury (Acute kidney injury, AKI) is a clinical syndrome of great concern, with a high incidence of up to 50% in the intensive care unit (ICU), and is associated with high mortality. In 2013, 1.4 million to 2.9 million AKI patients were hospitalized in my country, with a total medical cost of about 13 billion US dollars and a hospital mortality rate of 12.4%. 16.1% of the critically ill patients gave up treatment and left the hospital, and about 65.3% of the patients died within 3 months after leaving the hospital. AKI has become a huge medical burden in our country, and there is a serious shortage of diagnosis and treatment. AKI was unrecognized at a high rate in both university-affiliated and local hospitals...

AI Technical Summary

Problems solved by technology

However, the results of some recent studies have shown that severe AKI (stage 2 / 3) is associated with a high risk of poor prognosis, while the occurrence of stage 1 has no clear correlation with the outcome and supports clinical judgment, and a large proportion of patients are transient Sex can improve with treatment

Excessive early attention may easily cause alarm fatigue and interfere with the treatment of other urgent diseases; most studies only use creatinine to judge the disease development stage of AKI. Significant time lags leading to model development failures; only a small fraction of models developed for prediction have been calibrated and subgroup analyzed, and external validation is limited to limited hospitals, regions, and the same country, which can lead to Model performance cannot be fully and effectively evaluated, leading to potential evaluation bias and unfairness; as far as we know, only a small number of models have been extended to actual application scenarios, and most models are only in the development stage of the model

At present, there is no clear research showing that the actual use of the model requires a large amount of local data for model migration or retraining; the purpose of developing the model is to serve the clinic better and more conveniently, but there are few studies Further develop the model into a platform / software that can be understood, trusted and easily operated by medical staff

Images