Coronavirus Isolation Hospitalized Patients Developing Machine Learning Models to Predict Delirium Occurrence
Jul 14, 2024
Professor Park Hye-yeon of the Department of Psychiatry at Seoul National University Bundang Hospital has established a machine learning (machine learning) model that predicts delirium occurrence in patients hospitalized in isolation with COVID-19.
Delirium is a disease that shows psychotic symptoms such as sleep disorders, hallucinations and hearing, hyperactivity or anxiety at the same time as cognitive impairments such as attention and language skills. Unlike degenerative dementia, which develops symptoms for at least several months, it is characterized by sudden occurrence in a short period of time. Since there is no effective treatment, it is important to predict it in advance and control risk factors early to prevent it.
Such delirium is relatively common, with 10-15% of all hospitalized patients experiencing it. Delirium worsens the medical course of inpatients and causes falls, extending the period of isolation hospitalization, which has been a problem due to the high rate of COVID-19 patients who have been re-emerging until recently.
Recent studies have shown that the percentage of delirium in severely ill COVID-19 patients ranges from 55% to 70%, of which about 30% are reported to experience delirium for months or longer.
Accordingly, Professor Park Hye-yeon's team conducted a study to develop a model for predicting the occurrence of delirium in patients hospitalized with COVID-19 by utilizing 93 delirium factors such as drugs taken, underlying diseases, and imaging/blood tests of 878 people who were hospitalized with COVID-19 isolated at four hospitals.
As a result, we succeeded in developing a machine learning model that can quickly and accurately identify delirium incidence and patient-specific risk factors by entering clinical information of COVID-19 patients such as ▲ biosignal ▲ medication ▲ blood test results at the beginning of hospitalization. The accuracy of predicting delirium occurrence of this model is 87.3%, which is expected to be effectively used to select high-risk groups for delirium at the beginning of hospitalization in patients hospitalized in isolation with COVID-19.
It also has a function that goes beyond simple predictions to identify individual risk factors for patients and suggest what factors can be controlled and mediated, including drugs. According to the study, there are nine major factors that increase the risk of delirium in COVID-19 patients, among which drugs (antipsychotics, antibiotics, sedatives, antipyretics), mechanical ventilation (artificial respiration), and blood sodium reduction are particularly dangerous, and machine learning warns you if there is an abnormality in these indicators.
Professor Park Hye-yeon said, "Delirium is also evident in patients in isolation due to acute infectious diseases such as COVID-19, and this prolongs the isolation hospitalization period due to deterioration of the medical course and falls." If this predictive model is used, it will be possible to identify risk factors for each patient in advance and control drugs, thereby minimizing the occurrence of delirium." We will then conduct a verification study to ensure that machine learning models can be utilized in real-world clinical settings.
Meanwhile, this study was conducted with the support of the Ministry of Health and Welfare's 'Patient-centered Medical Technology Optimization Research Project', and was recently published in the SCIE international journal 'Digital Health'.
Delirium is a disease that shows psychotic symptoms such as sleep disorders, hallucinations and hearing, hyperactivity or anxiety at the same time as cognitive impairments such as attention and language skills. Unlike degenerative dementia, which develops symptoms for at least several months, it is characterized by sudden occurrence in a short period of time. Since there is no effective treatment, it is important to predict it in advance and control risk factors early to prevent it.
Such delirium is relatively common, with 10-15% of all hospitalized patients experiencing it. Delirium worsens the medical course of inpatients and causes falls, extending the period of isolation hospitalization, which has been a problem due to the high rate of COVID-19 patients who have been re-emerging until recently.
Recent studies have shown that the percentage of delirium in severely ill COVID-19 patients ranges from 55% to 70%, of which about 30% are reported to experience delirium for months or longer.
Accordingly, Professor Park Hye-yeon's team conducted a study to develop a model for predicting the occurrence of delirium in patients hospitalized with COVID-19 by utilizing 93 delirium factors such as drugs taken, underlying diseases, and imaging/blood tests of 878 people who were hospitalized with COVID-19 isolated at four hospitals.
As a result, we succeeded in developing a machine learning model that can quickly and accurately identify delirium incidence and patient-specific risk factors by entering clinical information of COVID-19 patients such as ▲ biosignal ▲ medication ▲ blood test results at the beginning of hospitalization. The accuracy of predicting delirium occurrence of this model is 87.3%, which is expected to be effectively used to select high-risk groups for delirium at the beginning of hospitalization in patients hospitalized in isolation with COVID-19.
It also has a function that goes beyond simple predictions to identify individual risk factors for patients and suggest what factors can be controlled and mediated, including drugs. According to the study, there are nine major factors that increase the risk of delirium in COVID-19 patients, among which drugs (antipsychotics, antibiotics, sedatives, antipyretics), mechanical ventilation (artificial respiration), and blood sodium reduction are particularly dangerous, and machine learning warns you if there is an abnormality in these indicators.
Professor Park Hye-yeon said, "Delirium is also evident in patients in isolation due to acute infectious diseases such as COVID-19, and this prolongs the isolation hospitalization period due to deterioration of the medical course and falls." If this predictive model is used, it will be possible to identify risk factors for each patient in advance and control drugs, thereby minimizing the occurrence of delirium." We will then conduct a verification study to ensure that machine learning models can be utilized in real-world clinical settings.
Meanwhile, this study was conducted with the support of the Ministry of Health and Welfare's 'Patient-centered Medical Technology Optimization Research Project', and was recently published in the SCIE international journal 'Digital Health'.
|
bellho@sportschosun.com
