Mortality prediction of patients in intensive care units using machine learning algorithms based on electronic health records

Choi, Min Hyuk; Kim, Dokyun; Choi, Eui Jun; Jung, Yeo Jin; Choi, Yong Jun; Cho, Jae Hwa; Jeong, Seok Hoon

doi:10.1038/s41598-022-11226-4

Cited by 21 publications

(17 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…al. [10], our model outperformed traditional clinical scores not only in the prediction of early mortality but also in the prediction of late mortality. A multi-dimensional time series data-driven research approach was used, as well as stratification over the different groups of features, to identify feature groups with the highest predictive performance.…”

Section: Discussionmentioning

confidence: 76%

“…There is increasing interest in the literature in using modern ML algorithms, such as random forests (RF) and gradient boosting machines, since it has been shown that they can predict more accurately clinical outcomes (e.g., mortality) in comparison with classical logistic regression [30]. Some of the recent works on mortality prediction in ICU patients show the superiority of the ML-based predictive models against the traditional clinical scores [10,31,32]. These studies are not focused on a specific diagnostic group of patients and only use a limited feature set [8,11,12,33].…”

Section: Related Workmentioning

confidence: 99%

“…Mortality is a major end point in epidemiological and interventional studies in the ICUs [7]. Published studies so far mainly focus on the prediction of in-hospital or early mortality, irrespective of the primary diagnosis and related comorbidities [8][9][10][11][12], whereas studies focusing on late mortality prediction are quite rare [13,14]. It is generally accepted that the initial diagnosis of a patient and the evaluation of the reasons for ICU admission could significantly affect the overall survival; thus, it would be interesting to study individually the different disease outcomes and identify the specific disease-related clinical features with prognostic significance.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Outcome Prediction in Critically-Ill Patients with Venous Thromboembolism and/or Cancer Using Machine Learning Algorithms: External Validation and Comparison with Scoring Systems

Danilatou

Nikolakakis

Antonakaki

et al. 2022

IJMS

View full text Add to dashboard Cite

Intensive care unit (ICU) patients with venous thromboembolism (VTE) and/or cancer suffer from high mortality rates. Mortality prediction in the ICU has been a major medical challenge for which several scoring systems exist but lack in specificity. This study focuses on two target groups, namely patients with thrombosis or cancer. The main goal is to develop and validate interpretable machine learning (ML) models to predict early and late mortality, while exploiting all available data stored in the medical record. To this end, retrospective data from two freely accessible databases, MIMIC-III and eICU, were used. Well-established ML algorithms were implemented utilizing automated and purposely built ML frameworks for addressing class imbalance. Prediction of early mortality showed excellent performance in both disease categories, in terms of the area under the receiver operating characteristic curve (AUC−ROC): VTE-MIMIC-III 0.93, eICU 0.87, cancer-MIMIC-III 0.94. On the other hand, late mortality prediction showed lower performance, i.e., AUC−ROC: VTE 0.82, cancer 0.74–0.88. The predictive model of early mortality developed from 1651 VTE patients (MIMIC-III) ended up with a signature of 35 features and was externally validated in 2659 patients from the eICU dataset. Our model outperformed traditional scoring systems in predicting early as well as late mortality. Novel biomarkers, such as red cell distribution width, were identified.

show abstract

Section: Discussionmentioning

confidence: 76%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Outcome Prediction in Critically-Ill Patients with Venous Thromboembolism and/or Cancer Using Machine Learning Algorithms: External Validation and Comparison with Scoring Systems

Danilatou

Nikolakakis

Antonakaki

et al. 2022

IJMS

View full text Add to dashboard Cite

show abstract

“…Furthermore, machine learning techniques can be used to improve the performance of the severity scoring system. Previous studies found that machine learning algorithms improved ability to predict hospital mortality of ICU patients when compared with conventional severity model 27 and also a good performance to predict short and long length of ICU stay 28 .…”

Section: Discussionmentioning

confidence: 99%

Validating the APACHE IV score in predicting length of stay in the intensive care unit among patients with sepsis

Zangmo

Khwannimit

2023

Sci Rep

View full text Add to dashboard Cite

The Acute Physiology and Chronic Health Evaluation (APACHE) IV model can predict the intensive care unit (ICU) length of stay (LOS) in critically ill patients. Thus, this study aimed to validate the performance of the APACHE IV score in predicting ICU LOS among patients with sepsis. This retrospective study was conducted in the medical ICU of a tertiary university between 2017 and 2020. A total of 1,039 sepsis patients were enrolled. Patients with an ICU stay of 1 and > 3 days accounted for 20.1% and 43.9%. The overall observed and APACHE IV predicted ICU LOS were 6.3 ± 6.5 and 6.8 ± 6.5, respectively. The APACHE IV slightly over-predicted ICU LOS with standardized length of stay ratio 0.95 (95% CI 0.89–1.02). The predicted ICU LOS based on the APACHE IV score was statistically longer than the observed ICU LOS (p < 0.001) and were poorly correlated (R2 = 0.02, p < 0.001), especially in patients with a lower severity of illness. In conclusions the APACHE IV model poorly predicted ICU LOS in patients with sepsis. The APACHE IV score needs to be modified or we need to make a new specific model to predict ICU stays in patients with sepsis.

show abstract

“…AI systems have leveraged clinical data to predict mortality [1][2][3][4][5] , length of stay (LOS) 6 , sepsis [7][8][9] and the occurrence of specific disease diagnoses 10 . As a growing number of AI systems are sought to be deployed in clinical settings, a defining challenge for AI in healthcare is how to responsibly deploy models that have been developed 11,12 .…”

Section: Introductionmentioning

confidence: 99%

Diagnosing and remediating harmful data shifts for the responsible deployment of clinical AI models

Subasri

Krishnan

Dhalla

et al. 2023

Preprint

View full text Add to dashboard Cite

Harmful data shifts occur when the distribution of data used to train a clinical AI system differs significantly from the distribution of data encountered during deployment, leading to erroneous predictions and potential harm to patients. We evaluated the impact of data shifts on an early warning system (EWS) for in-hospital mortality that uses electronic health record (EHR) data from patients admitted to a general internal medicine service. We found model performance to differ across subgroups of clinical diagnoses, sex and age. To explore the robustness of the model, we evaluated potentially harmful data shifts across demographics, hospital types, seasons, times of hospital admission, and whether the patient was admitted from an acute care institution or nursing home, without relying on model performance. Interestingly, we found that models trained on community hospitals experience harmful data shifts when evaluated on academic hospitals, whereas the models trained on academic hospitals transfer well to the community hospitals. To improve model performance across hospital sites we employed transfer learning, a strategy that stores knowledge gained from learning one domain and applies it to a different but related domain. We found hospital type-specific models that leverage transfer learning, perform better than models that use all available hospitals. Furthermore, we monitored data shifts over time and identified model deterioration during the COVID-19 pandemic. Typically machine learning models remain locked after deployment, however, this can lead to model deterioration due to data shifts that occur over time. We used continual learning, the process of learning from a continual stream of data in a sequential manner, to mitigate data shifts over time and improve model performance. Overall, our study is a crucial step towards the deployment of clinical AI models, by providing strategies and workflows to ensure the safety and efficacy of these models in real-world settings.

show abstract

Mortality prediction of patients in intensive care units using machine learning algorithms based on electronic health records

Cited by 21 publications

References 39 publications

Outcome Prediction in Critically-Ill Patients with Venous Thromboembolism and/or Cancer Using Machine Learning Algorithms: External Validation and Comparison with Scoring Systems

Outcome Prediction in Critically-Ill Patients with Venous Thromboembolism and/or Cancer Using Machine Learning Algorithms: External Validation and Comparison with Scoring Systems

Validating the APACHE IV score in predicting length of stay in the intensive care unit among patients with sepsis

Diagnosing and remediating harmful data shifts for the responsible deployment of clinical AI models

Contact Info

Product

Resources

About