Dataset Information

Machine learning approaches to predict the need for intensive care unit admission among Iranian COVID-19 patients based on ICD-10: A cross-sectional study.

ABSTRACT:

Background & aim

Timely identification of the patients requiring intensive care unit admission (ICU) could be life-saving. We aimed to compare different machine learning algorithms to predict the requirements for ICU admission in COVID-19 patients.

Methods

We screened all patients with COVID-19 at six academic hospitals in Tehran comprising our study population. A total of 44,112 COVID-19 patients (≥18 years old) were included, among which 7722 patients were hospitalized. We used a Random Forest algorithm to select significant variables. Then, prediction models were developed using the Support Vector Machine, Naıve Bayes, logistic regression, lightGBM, decision tree, and K-Nearest Neighbor algorithms. Sensitivity, specificity, accuracy, F1 score, and receiver operating characteristic-Area Under the Curve (AUC) were used to compare the prediction performance of different models.

Results

Based on random Forest, the following predictors were selected: age, cardiac disease, cough, hypertension, diabetes, influenza & pneumonia, malignancy, and nervous system disease. Age was found to have the strongest association with ICU admission among COVID-19 patients. All six models achieved an AUC greater than 0.60. Naıve Bayes achieved the best predictive performance (AUC = 0.71).

Conclusion

Naïve Bayes and lightGBM demonstrated promising results in predicting ICU admission needs in COVID-19 patients. Machine learning models could help quickly identify high-risk patients upon entry and reduce mortality and morbidity among COVID-19 patients.

SUBMITTER: Karimi Z

PROVIDER: S-EPMC11369020 | biostudies-literature | 2024 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning approaches to predict the need for intensive care unit admission among Iranian COVID-19 patients based on ICD-10: A cross-sectional study.

Karimi Zahra Z Malak Jaleh S JS Aghakhani Amirhossein A Najafi Mohammad S MS Ariannejad Hamid H Zeraati Hojjat H Yekaninejad Mir S MS

Health science reports 20240902 9

<h4>Background & aim</h4>Timely identification of the patients requiring intensive care unit admission (ICU) could be life-saving. We aimed to compare different machine learning algorithms to predict the requirements for ICU admission in COVID-19 patients.<h4>Methods</h4>We screened all patients with COVID-19 at six academic hospitals in Tehran comprising our study population. A total of 44,112 COVID-19 patients (≥18 years old) were included, among which 7722 patients were hospitalized. We used ...[more]

PMID: 39229475

Similar Datasets

Project description:Rationale: Determining when an intensive care unit (ICU) patient is ready for discharge to the ward is a complex daily challenge for any ICU care team. Patients who experience unplanned readmissions to the ICU have increased mortality, length of stay, and cost compared with those not readmitted during their hospital stay. The accuracy of clinician prediction for ICU readmission is unknown.Objectives: To determine the accuracy of ICU physicians and nurses for predicting ICU readmissionsMethods: We conducted a prospective study in the medical ICU of an academic hospital from October 2015 to September 2017. After daily rounding for patients being transferred to the ward, ICU clinicians (nurses, residents, fellows, and attendings) were asked to report the likelihood of readmission within 48 hours (using a 1-10 scale, with 10 being "extremely likely"). The accuracy of the clinician prediction score (1-10) was assessed for all clinicians and by clinician type using sensitivity, specificity, and area under the curve (AUC) for the receiver operating characteristic curve for predicting the primary outcome, which was ICU readmission within 48 hours of ICU discharge.Results: A total of 2,833 surveys was collected for 938 ICU-to-ward transfers, of which 40 (4%) were readmitted to the ICU within 48 hours of transfer. The median clinician likelihood of readmission score was 3 (interquartile range, 2-4). When physician and nurse likelihood scores were combined, the median clinician likelihood score had an AUC of 0.70 (95% confidence interval [CI], 0.62-0.78) for predicting ICU readmission within 48 hours. Nurses were significantly more accurate than interns at predicting 48-hour ICU readmission (AUC, 0.73 [95% CI, 0.64-0.82] vs. AUC, 0.60 [95% CI, 0.49-0.71]; P = 0.03). All other pairwise comparisons were not significantly different for predicting ICU readmission within 48 hours (P > 0.05 for all comparisons).Conclusions: We found that all clinicians surveyed in our ICU, regardless of the level of experience or clinician type, had only fair accuracy for predicting ICU readmission. Further research is needed to determine if clinical decision support tools would provide prognostic value above and beyond clinical judgment for determining who is ready for ICU discharge.

Project description:BackgroundKnife crime has increased considerably in recent years in Northern Europe. Affected patients often require immediate surgical care due to traumatic organ injury. Yet, little is known about clinically relevant routine laboratory parameters in stab injury patients and how these are associated with intensive care unit (ICU) admission, hospitalization and number of surgeries.MethodsWe retrospectively analyzed 258 stab injury cases between July 2015 and December 2021 at an urban Level I Trauma Center. Annual and seasonal incidences, injury site, injury mechanism, Injury Severity Score (ISS), and surgical management were evaluated. First, correlations between routine laboratory parameters for hematology, coagulation, and serum biochemistry (peak, and Δ (change from admission to peak within 3 days following admission)) and length of hospital stay, ICU stay, and number of surgeries were assessed using Spearman's rank correlation coefficients. Second, multivariable Least Absolute Shrinkage and Selection Operator (LASSO) regression analyses were conducted to identify parameters predictive of clinical outcomes. Third, longitudinal developments of routine laboratory parameters were assessed during hospital admission.ResultsIn 2021, significantly more stab injuries were recorded compared with previous years and occurred less during winter compared with other seasons. Mean ISS was 8.3 ± 7.3, and ISS was positively correlated with length of hospital and ICU stay (r = 0.5-0.8, p < 0.001). Aspartate transaminase (AST) (Δ) (r = 0.690), peak C-reactive protein (CrP) (r = 0.573), and erythrocyte count (Δ) (r = 0.526) showed the strongest positive correlations for length of ICU stay for penetrating, thoracoabdominal, and organ injuries, respectively. No correlations were observed between routine laboratory parameters and number of surgeries. For patients with penetrating injuries, LASSO-selected predictors of ICU admission included ISS, pH and lactate at admission, and Δ values for activated partial thromboplastin time (aPTT), K+, and erythrocyte count. CrP levels on day 3 were significantly higher in patients with penetrating (p = 0.005), thoracoabdominal (p = 0.041), and organ injuries (p < 0.001) compared with those without.ConclusionOur data demonstrate an increase in stab injury cases in 2021 and an important link between changes in routine laboratory parameters and ICU admission and hospitalization. Monitoring ISS and changes in AST, CrP, erythrocyte count, pH, lactate, aPTT, and K+ may be useful to identify patients at risk and adjust surgical and ICU algorithms early on.

Project description:ObjectiveTo compare the performance of machine learning models against the traditionally derived Combined Assessment of Risk Encountered in Surgery (CARES) model and the American Society of Anaesthesiologists-Physical Status (ASA-PS) in the prediction of 30-day postsurgical mortality and need for intensive care unit (ICU) stay >24 hours.BackgroundPrediction of surgical risk preoperatively is important for clinical shared decision-making and planning of health resources such as ICU beds. The current growth of electronic medical records coupled with machine learning presents an opportunity to improve the performance of established risk models.MethodsAll patients aged 18 years and above who underwent noncardiac and nonneurological surgery at Singapore General Hospital (SGH) between 1 January 2012 and 31 October 2016 were included. Patient demographics, comorbidities, preoperative laboratory results, and surgery details were obtained from their electronic medical records. Seventy percent of the observations were randomly selected for training, leaving 30% for testing. Baseline models were CARES and ASA-PS. Candidate models were trained using random forest, adaptive boosting, gradient boosting, and support vector machine. Models were evaluated on area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC).ResultsA total of 90,785 patients were included, of whom 539 (0.6%) died within 30 days and 1264 (1.4%) required ICU admission >24 hours postoperatively. Baseline models achieved high AUROCs despite poor sensitivities by predicting all negative in a predominantly negative dataset. Gradient boosting was the best performing model with AUPRCs of 0.23 and 0.38 for mortality and ICU admission outcomes respectively.ConclusionsMachine learning can be used to improve surgical risk prediction compared to traditional risk calculators. AUPRC should be used to evaluate model predictive performance instead of AUROC when the dataset is imbalanced.

Project description:BackgroundEffectively identifying deteriorated patients is vital to the development and validation of automated systems designed to predict clinical deterioration. Existing outcome measures used for this purpose have significant limitations. Published criteria for admission to high acuity inpatient areas may represent markers of patient deterioration and could inform the development of alternate outcome measures.ObjectivesIn this scoping review, we aimed to characterise published criteria for admission of adult inpatients to high acuity inpatient areas including intensive care units. A secondary aim was to identify variables that are extractable from electronic health records (EHRs).Data sourcesElectronic databases PubMed and ProQuest EBook Central were searched to identify papers published from 1999 to date of search. We included publications which described prescriptive criteria for admission of adult inpatients to a clinical area with a higher level of care than a general hospital ward.Charting methodsData was extracted from each publication using a standardised data-charting form. Admission criteria characteristics were summarised and cross-tabulated for each criterion by population group.ResultsFive domains were identified: diagnosis-based criteria, clinical parameter criteria, organ-support criteria, organ-monitoring criteria and patient baseline criteria. Six clinical parameter-based criteria and five needs-based criteria were frequently proposed and represent variables extractable from EHRs. Thresholds for objective clinical parameter criteria varied across publications, and by disease subgroup, and universal cut-offs for criteria could not be elucidated.ConclusionsThis study identified multiple criteria which may represent markers of deterioration. Many of the criteria are extractable from the EHR, making them potential candidates for future automated systems. Variability in admission criteria and associated thresholds across the literature suggests clinical deterioration is a heterogeneous phenomenon which may resist being defined as a single entity via a consensus-driven process.

Dataset Information

Machine learning approaches to predict the need for intensive care unit admission among Iranian COVID-19 patients based on ICD-10: A cross-sectional study.

Background & aim

Methods

Results

Conclusion

Publications

Machine learning approaches to predict the need for intensive care unit admission among Iranian COVID-19 patients based on ICD-10: A cross-sectional study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets