Dataset Information

Machine learning versus physicians' prediction of acute kidney injury in critically ill adults: a prospective evaluation of the AKIpredictor.

ABSTRACT: BACKGROUND:Early diagnosis of acute kidney injury (AKI) is a major challenge in the intensive care unit (ICU). The AKIpredictor is a set of machine-learning-based prediction models for AKI using routinely collected patient information, and accessible online. In order to evaluate its clinical value, the AKIpredictor was compared to physicians' predictions. METHODS:Prospective observational study in five ICUs of a tertiary academic center. Critically ill adults without end-stage renal disease or AKI upon admission were considered for enrollment. Using structured questionnaires, physicians were asked upon admission, on the first morning, and after 24?h to predict the development of AKI stages 2 or 3 (AKI-23) during the first week of ICU stay. Discrimination, calibration, and net benefit of physicians' predictions were compared against the ones by the AKIpredictor. RESULTS:Two hundred fifty-two patients were included, 30 (12%) developed AKI-23. In the cohort of patients with predictions by physicians and AKIpredictor, the performance of physicians and AKIpredictor were respectively upon ICU admission, area under the receiver operating characteristic curve (AUROC) 0.80 [0.69-0.92] versus 0.75 [0.62-0.88] (n?=?120, P?=?0.25) with net benefit in ranges 0-26% versus 0-74%; on the first morning, AUROC 0.94 [0.89-0.98] versus 0.89 [0.82-0.97] (n?=?187, P?=?0.27) with main net benefit in ranges 0-10% versus 0-48%; after 24?h, AUROC 0.95 [0.89-1.00] versus 0.89 [0.79-0.99] (n?=?89, P?=?0.09) with main net benefit in ranges 0-67% versus 0-50%. CONCLUSIONS:The machine-learning-based AKIpredictor achieved similar discriminative performance as physicians for prediction of AKI-23, and higher net benefit overall, because physicians overestimated the risk of AKI. This suggests an added value of the systematic risk stratification by the AKIpredictor to physicians' predictions, in particular to select high-risk patients or reduce false positives in studies evaluating new and potentially harmful therapies. Due to the low event rate, future studies are needed to validate these findings. TRIAL REGISTRATION:ClinicalTrials.gov, NCT03574896 registration date: July 2nd, 2018.

SUBMITTER: Flechet M

PROVIDER: S-EPMC6697946 | biostudies-literature | 2019 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning versus physicians' prediction of acute kidney injury in critically ill adults: a prospective evaluation of the AKIpredictor.

Flechet Marine M Falini Stefano S Bonetti Claudia C Güiza Fabian F Schetz Miet M Van den Berghe Greet G Meyfroidt Geert G

Critical care (London, England) 20190816 1

<h4>Background</h4>Early diagnosis of acute kidney injury (AKI) is a major challenge in the intensive care unit (ICU). The AKIpredictor is a set of machine-learning-based prediction models for AKI using routinely collected patient information, and accessible online. In order to evaluate its clinical value, the AKIpredictor was compared to physicians' predictions.<h4>Methods</h4>Prospective observational study in five ICUs of a tertiary academic center. Critically ill adults without end-stage ren ...[more]

PMID: 31420056

Similar Datasets

Project description:BackgroundSepsis-associated acute kidney injury (SA-AKI) is common in critically ill patients, which is associated with significantly increased mortality. Existing mortality prediction tools showed insufficient predictive power or failed to reflect patients' dynamic clinical evolution. Therefore, the study aimed to develop and validate machine learning-based models for real-time mortality prediction in critically ill patients with SA-AKI.MethodsThe multi-center retrospective study included patients from two distinct databases. A total of 12,132 SA-AKI patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV) were randomly allocated to the training, validation, and internal test sets. An additional 3,741 patients from the eICU Collaborative Research Database (eICU-CRD) served as an external test set. For every 12 h during the ICU stays, the state-of-the-art eXtreme Gradient Boosting (XGBoost) algorithm was used to predict the risk of in-hospital death in the following 48, 72, and 120 h and in the first 28 days after ICU admission. Area under the receiver operating characteristic curves (AUCs) were calculated to evaluate the models' performance.ResultsThe XGBoost models, based on routine clinical variables updated every 12 h, showed better performance in mortality prediction than the SOFA score and SAPS-II. The AUCs of the XGBoost models for mortality over different time periods ranged from 0.848 to 0.804 in the internal test set and from 0.818 to 0.748 in the external test set. The shapley additive explanation method provided interpretability for the XGBoost models, which improved the understanding of the association between the predictor variables and future mortality.ConclusionsThe interpretable machine learning XGBoost models showed promising performance in real-time mortality prediction in critically ill patients with SA-AKI, which are useful tools for early identification of high-risk patients and timely clinical interventions.

Project description:BACKGROUND:There has been a global increase in the incidence of acute kidney injury (AKI), including among critically-ill surgical patients. AKI prediction score provides an opportunity for early detection of patients who are at risk of AKI; however, most of the AKI prediction scores were derived from cardiothoracic surgery. Therefore, we aimed to develop an AKI prediction score for major non-cardiothoracic surgery patients who were admitted to the intensive care unit (ICU). METHODS:The data of critically-ill patients from non-cardiothoracic operations in the Thai Surgical Intensive Care Unit (THAI-SICU) study were used to develop an AKI prediction score. Independent prognostic factors from regression analysis were included as predictors in the model. The outcome of interest was AKI within 7?days after the ICU admission. The AKI diagnosis was made according to the Kidney Disease Improving Global Outcomes (KDIGO)-2012 serum creatinine criteria. Diagnostic function of the model was determined by area under the Receiver Operating Curve (AuROC). Risk scores were categorized into four risk probability levels: low (0-2.5), moderate (3.0-8.5), high (9.0-11.5), and very high (12.0-16.5) risk. Risk of AKI was presented as likelihood ratios of positive (LH+). RESULTS:A total of 3474 critically-ill surgical patients were included in the model; 333 (9.6%) developed AKI. Using multivariable logistic regression analysis, older age, high Sequential Organ Failure Assessment (SOFA) non-renal score, emergency surgery, large volume of perioperative blood loss, less urine output, and sepsis were identified as independent predictors for AKI. Then AKI prediction score was created from these predictors. The summation of the score was 16.5 and had a discriminative ability for predicting AKI at AuROC?=?0.839 (95% CI 0.825-0.852). LH+ for AKI were: low risk?=?0.117 (0.063-0.200); moderate risk?=?0.927 (0.745-1.148); high risk?=?5.190 (3.881-6.910); and very high risk?=?9.892 (6.230-15.695), respectively. CONCLUSIONS:The function of AKI prediction score to predict AKI among critically ill patients who underwent non-cardiothoracic surgery was good. It can aid in early recognition of critically-ill surgical patients who are at risk from ICU admission. The scores could guide decision making for aggressive strategies to prevent AKI during the perioperative period or at ICU admission. TRIAL REGISTRATION:TCTR20190408004, registered on April 4, 2019.

Project description:ObjectivesCritically ill patients with coronavirus disease 2019 have variable mortality. Risk scores could improve care and be used for prognostic enrichment in trials. We aimed to compare machine learning algorithms and develop a simple tool for predicting 28-day mortality in ICU patients with coronavirus disease 2019.DesignThis was an observational study of adult patients with coronavirus disease 2019. The primary outcome was 28-day inhospital mortality. Machine learning models and a simple tool were derived using variables from the first 48 hours of ICU admission and validated externally in independent sites and temporally with more recent admissions. Models were compared with a modified Sequential Organ Failure Assessment score, National Early Warning Score, and CURB-65 using the area under the receiver operating characteristic curve and calibration.SettingSixty-eight U.S. ICUs.PatientsAdults with coronavirus disease 2019 admitted to 68 ICUs in the United States between March 4, 2020, and June 29, 2020.InterventionsNone.Measurements and main resultsThe study included 5,075 patients, 1,846 (36.4%) of whom died by day 28. eXtreme Gradient Boosting had the highest area under the receiver operating characteristic curve in external validation (0.81) and was well-calibrated, while k-nearest neighbors were the lowest performing machine learning algorithm (area under the receiver operating characteristic curve 0.69). Findings were similar with temporal validation. The simple tool, which was created using the most important features from the eXtreme Gradient Boosting model, had a significantly higher area under the receiver operating characteristic curve in external validation (0.78) than the Sequential Organ Failure Assessment score (0.69), National Early Warning Score (0.60), and CURB-65 (0.65; p < 0.05 for all comparisons). Age, number of ICU beds, creatinine, lactate, arterial pH, and Pao2/Fio2 ratio were the most important predictors in the eXtreme Gradient Boosting model.ConclusionseXtreme Gradient Boosting had the highest discrimination overall, and our simple tool had higher discrimination than a modified Sequential Organ Failure Assessment score, National Early Warning Score, and CURB-65 on external validation. These models could be used to improve triage decisions and clinical trial enrichment.

Project description:OBJECTIVES:Acute kidney injury is a common complication of major surgery. However, acute kidney injury occurring within the first 48 hours after surgery (early acute kidney injury) and therefore likely related to the surgery itself is possibly different from acute kidney injury occurring after 48 hours (late acute kidney injury). The aim of this study was to describe the epidemiology and identify differences in risk factors and outcomes between early and late acute kidney injury following major surgery. DESIGN:Retrospective cohort study. SETTING:Academic Medical Center. PATIENTS:Patients admitted to ICU following noncardiac major surgery. INTERVENTIONS:None. MEASUREMENTS AND MAIN RESULTS:We analyzed data from 3,499 patients and defined acute kidney injury according to full Kidney Disease: Improving Global Outcomes criteria and classified as early (48?hr or less) or late (> 48?hr to 7 d) based on time from surgery. Separate multivariable logistic regression models were fit to identify risk factors of early acute kidney injury compared with no acute kidney injury and risk factors of late acute kidney injury compared with no acute kidney injury. Overall 41.7% (1,459/3,499) developed early acute kidney injury versus 14.4% (504/3,499) late acute kidney injury. Most acute kidney injury occurred within 48 hours following surgery and 12 hours was the peak interval. Risk factors for early acute kidney injury included increased age, body mass index, decreased estimated glomerular filtration rate, and anemia, whereas late acute kidney injury cases were closely associated with postoperative factors, like sepsis, mechanical ventilation, positive fluid balance, blood transfusions and exposure to diuretics, vasopressors, and nonsteroidal anti-inflammatory drugs. After adjusting for age, body mass index, estimated glomerular filtration rate, comorbidities, surgery type, both early acute kidney injury (odds ratio [95% CI], 1.84 [1.50-2.27]) and late acute kidney injury (odds ratio [95% CI], 1.42 [1.09-1.85]) were associated with higher 1-year mortality compared with patients without acute kidney injury. We found similar results in a validation cohort of 10,723 patients admitted between 2008 and 2014. CONCLUSIONS:Most surgery-related acute kidney injury occurred within 48 hours of surgery. Acute kidney injury occurring within the first 48 hours was associated with underlying health, whereas acute kidney injury occurring after 48 hours was related to postoperative complications or drugs. Design of clinical and experimental interventions for acute kidney injury in this population should consider these differences.

Dataset Information

Machine learning versus physicians' prediction of acute kidney injury in critically ill adults: a prospective evaluation of the AKIpredictor.

Publications

Machine learning versus physicians' prediction of acute kidney injury in critically ill adults: a prospective evaluation of the AKIpredictor.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets