Dataset Information

Dynamic prediction of life-threatening events for patients in intensive care unit.

ABSTRACT:

Background

Early prediction of patients' deterioration is helpful in early intervention for patients at greater risk of deterioration in Intensive Care Unit (ICU). This study aims to apply machine learning approaches to heterogeneous clinical data for predicting life-threatening events of patients in ICU.

Methods

We collected clinical data from a total of 3151 patients admitted to the Medical Intensive Care Unit of Peking Union Medical College Hospital in China from January 1st, 2014, to October 1st, 2019. After excluding the patients who were under 18 years old or stayed less than 24 h at the ICU, a total of 2170 patients were enrolled in this study. Multiple machine learning approaches were utilized to predict life-threatening events (i.e., death) in seven 24-h windows (day 1 to day 7) and their performance was compared.

Results

Light Gradient Boosting Machine showed the best performance. We found that life-threatening events during the short-term windows can be better predicted than those in the medium-term windows. For example, death in 24 h can be predicted with an Area Under Curve of 0.905. Features like infusion pump related fluid input were highly related to life-threatening events. Furthermore, the prediction power of static features such as age and cardio-pulmonary function increased with the extended prediction window.

Conclusion

This study demonstrates that the integration of machine learning approaches and large-scale high-quality clinical data in ICU could accurately predict life-threatening events for ICU patients for early intervention.

SUBMITTER: Hu J

PROVIDER: S-EPMC9587604 | biostudies-literature | 2022 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Dynamic prediction of life-threatening events for patients in intensive care unit.

Hu Jiang J Kang Xiao-Hui XH Xu Fang-Fang FF Huang Ke-Zhi KZ Du Bin B Weng Li L

BMC medical informatics and decision making 20221022 1

<h4>Background</h4>Early prediction of patients' deterioration is helpful in early intervention for patients at greater risk of deterioration in Intensive Care Unit (ICU). This study aims to apply machine learning approaches to heterogeneous clinical data for predicting life-threatening events of patients in ICU.<h4>Methods</h4>We collected clinical data from a total of 3151 patients admitted to the Medical Intensive Care Unit of Peking Union Medical College Hospital in China from January 1st, 2 ...[more]

PMID: 36273130

Similar Datasets

Project description:ObjectiveThis study aims to evaluate the accuracy of the Acute Physiology and Chronic Health Evaluation (APACHE) II score on different days in predicting the mortality of critically ill patients to identify the best time point for the APACHE II score.MethodsThe demographic and clinical data are retrieved from the Medical Information Mart for Intensive Care (MIMIC)-IV dataset. APACHE II scores on days 1, 2, 3, 5, 7, 14, and 28 of hospitalization are calculated, and their performance is evaluated using the area under the receiver operating characteristic (AUROC) analysis. The cut-off for defining the high risk of mortality is determined using Youden's index. The APACHE II score on day 3 is the best time point to predict hospital mortality of ICU patients. The Hosmer-Lemeshow goodness-of-fit test is then applied to evaluate the calibration of the day 3 APACHE II score.ResultsWe recruited 6,374 eligible subjects from the MIMIC-IV database. Day 3 is the optimal time point for obtaining the APACHE II score to predict the hospital mortality of patients. The best cut-off for day 3 APACHE II score is 17. When APACHE II score ≥17, the sensitivity for the non-survivors and survivors is 92.8 and 82.2%, respectively, and the positive predictive value (PPV) is 23.1%. When APACHE II socre <17, the specificity for non-survivors and survivors is 90.1 and 80.2%, respectively, and the negative predictive value (NPV) is 87.8%. When day-3 APACHE II is used to predict the hospital mortality, the AUROC is 0.743 (P <0.001). In the ≥17 group, the sensitivity of non-survivors and survivors is 92.2 and 81.3%, respectively, and the PPV is 30.3%. In the <17 group, the specificity of non-survivors and survivors is 100.0 and 80.2%, respectively, and the NPV is 81.6%. The Hosmer-Lemeshow test indicated day-3 APACHE II has a high predicting the hospital mortality (X 2 = 6.198, P = 0.625, consistency = 79.4%). However, the day-1 APACHE II has a poor calibration in predicting the hospital mortality rate (X 2 = 294.898, P <0.001).ConclusionDay-3 APACHE II score is an optimal biomarker to predict the outcomes of ICU patients; 17 is the best cut-off for defining patients at high risk of mortality.

Dataset Information

Dynamic prediction of life-threatening events for patients in intensive care unit.

Background

Methods

Results

Conclusion

Publications

Dynamic prediction of life-threatening events for patients in intensive care unit.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets