Dataset Information

Temporal Pattern Detection to Predict Adverse Events in Critical Care: Case Study With Acute Kidney Injury.

ABSTRACT: BACKGROUND:More than 20% of patients admitted to the intensive care unit (ICU) develop an adverse event (AE). No previous study has leveraged patients' data to extract the temporal features using their structural temporal patterns, that is, trends. OBJECTIVE:This study aimed to improve AE prediction methods by using structural temporal pattern detection that captures global and local temporal trends and to demonstrate these improvements in the detection of acute kidney injury (AKI). METHODS:Using the Medical Information Mart for Intensive Care dataset, containing 22,542 patients, we extracted both global and local trends using structural pattern detection methods to predict AKI (ie, binary prediction). Classifiers were built on 17 input features consisting of vital signs and laboratory test results using state-of-the-art models; the optimal classifier was selected for comparisons with previous approaches. The classifier with structural pattern detection features was compared with two baseline classifiers that used different temporal feature extraction approaches commonly used in the literature: (1) symbolic temporal pattern detection, which is the most common approach for multivariate time series classification; and (2) the last recorded value before the prediction point, which is the most common approach to extract temporal data in the AKI prediction literature. Moreover, we assessed the individual contribution of global and local trends. Classifier performance was measured in terms of accuracy (primary outcome), area under the curve, and F-measure. For all experiments, we employed 20-fold cross-validation. RESULTS:Random forest was the best classifier using structural temporal pattern detection. The accuracy of the classifier with local and global trend features was significantly higher than that while using symbolic temporal pattern detection and the last recorded value (81.3% vs 70.6% vs 58.1%; P<.001). Excluding local or global features reduced the accuracy to 74.4% or 78.1%, respectively (P<.001). CONCLUSIONS:Classifiers using features obtained from structural temporal pattern detection significantly improved the prediction of AKI onset in ICU patients over two baselines based on common previous approaches. The proposed method is a generalizable approach to predict AEs in critical care that may be used to help clinicians intervene in a timely manner to prevent or mitigate AEs.

SUBMITTER: Morid MA

PROVIDER: S-EPMC7109618 | biostudies-literature | 2020 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Temporal Pattern Detection to Predict Adverse Events in Critical Care: Case Study With Acute Kidney Injury.

Morid Mohammad Amin MA Sheng Olivia R Liu ORL Del Fiol Guilherme G Facelli Julio C JC Bray Bruce E BE Abdelrahman Samir S

JMIR medical informatics 20200317 3

<h4>Background</h4>More than 20% of patients admitted to the intensive care unit (ICU) develop an adverse event (AE). No previous study has leveraged patients' data to extract the temporal features using their structural temporal patterns, that is, trends.<h4>Objective</h4>This study aimed to improve AE prediction methods by using structural temporal pattern detection that captures global and local temporal trends and to demonstrate these improvements in the detection of acute kidney injury (AKI ...[more]

PMID: 32181753

Similar Datasets

Project description:BackgroundAcute kidney injury (AKI) survivors are at increased risk of major adverse kidney events (MAKEs), including chronic kidney disease (CKD), end-stage kidney disease (ESKD) and death. High-risk AKI patients may benefit from specialist follow-up, but factors associated with increased risk have not been reported.MethodsWe conducted a retrospective study of AKI patients admitted to a single centre between 2012 and 2016 who had a baseline estimated glomerular filtration rate (eGFR) >30 mL/min/1.73 m2 and were alive and independent of renal replacement therapy (RRT) at 30 days following discharge. AKI was identified using International Classification of Diseases, Tenth Revision codes and staged according to the Kidney Disease: Improving Global Outcomes criteria. Patients were excluded if they were kidney transplant recipients or if AKI was attributed to intrinsic kidney disease. We performed Cox regression models to examine MAKEs in the first year, defined as the composite of CKD (sustained 25% drop in eGFR), ESKD (requirement for chronic RRT or sustained eGFR <15 mL/min/1.73 m2) or death. We examined secondary outcomes (CKD, ESKD and death) using Cox and competing risk regression analyses.ResultsWe studied 2101 patients (mean ± SD age 69 ± 15 years, baseline eGFR 72 ± 23 mL/min/1.73 m2). Of these, 767 patients (37%) developed at least one MAKE (429 patients developed CKD, 21 patients developed ESKD, 375 patients died). MAKEs occurred more frequently with older age [hazard ratio (HR) 1.16 per decade, 95% confidence interval (CI) 1.10-1.24], greater severity of AKI (Stage 2 HR 1.38, 95% CI 1.16-1.64; Stage 3 HR 1.62, 95% CI 1.31-2.01), higher serum creatinine at discharge (HR 1.04 per 10 µmol/L, 95% CI 1.03-1.06), chronic heart failure (HR 1.41, 95% CI 1.19-1.67), liver disease (HR 1.68, 95% CI 1.39-2.03) and malignancy (non-metastatic HR 1.44, 95% CI 1.14-1.82; metastatic HR 2.26, 95% CI 1.80-2.83). Traditional risk factors (e.g. diabetes and cardiovascular disease) had limited predictive value.ConclusionsMore than a third of AKI patients develop MAKEs within the first year. Clinical variables available at the time of discharge can help identify patients at increased risk of such events.

Project description:Background and objectivesMajor adverse kidney events, a composite of death, new kidney replacement therapy, or persistent kidney dysfunction, is a potential patient-centered outcome for clinical trials in sepsis-associated kidney injury. We sought to determine the incidence of major adverse kidney events within 30 days and validate this end point in pediatric sepsis.Design, setting, participants, & measurementsWe conducted a retrospective observational study using the Pediatric Health Information Systems Plus database of patients >6 months to <18 years old with a diagnosis of severe sepsis/septic shock; orders for bacterial blood culture, antibiotics, and at least one fluid bolus on hospital day 0/1; and known hospital disposition between January 2007 and December 2011. The primary outcome was incidence of major adverse kidney events within 30 days. Major adverse kidney events within 30 days were validated against all-cause mortality at hospital discharge, hospital length of stay, total hospital costs, hospital readmission within 30 days and 1 year, and lowest eGFR between 3 months and 1 year after discharge. We reported incidence of major adverse kidney events within 30 days with 95% confidence intervals using robust SEM and used multivariable logistic regression to test the association of major adverse kidney events within 30 days with hospital costs and mortality.ResultsOf 1685 admissions, incidence of major adverse kidney events within 30 days was 9.6% (95% confidence interval, 8.1% to 11.0%), including 4.5% (95% confidence interval, 3.5% to 5.4%) death, 1.7% (95% confidence interval, 1.1% to 2.3%) kidney replacement therapy, and 5.8% (95% confidence interval, 4.7% to 6.9%) persistent kidney dysfunction. Patients with versus without major adverse kidney events within 30 days had higher all-cause mortality at hospital discharge (28% versus 1%; P<0.001), higher total hospital costs ($61,188; interquartile range, $21,272-140,356 versus $28,107; interquartile range, $13,056-72,697; P<0.001), and higher proportion with eGFR<60 ml/min per 1.73 m2 between 3 months and 1 year after discharge (19% versus 4%; P=0.001). Major adverse kidney events within 30 days was not associated with length of stay or readmissions.ConclusionsIn children with sepsis, major adverse kidney events within 30 days are common, feasible to measure, and a promising end point for future clinical trials.

Project description:Klotho and Fibroblast Growth Factor (FGF)-23 were recently postulated as candidate biomarkers and/or therapeutic targets in acute kidney injury (AKI). We examined whether urine Klotho and serum intact FGF23 levels were differentially and independently associated with major adverse kidney events (MAKE) in critically ill patients with and without AKI. Single-center, prospective, case-control study. ICU in a tertiary medical center. 54 AKI patients and 52 controls without AKI admitted to the ICU. None. AKI was defined by KDIGO criteria and included only AKI stage ≥2. Controls were matched by age, gender, and baseline eGFR. Paired serum and urine samples were obtained 24-48h after AKI diagnosis (cases) or ICU admission (controls). The primary outcome was 90-day MAKE, which was the composite of all-cause death, dependence on renal replacement therapy or a 50% or higher decrease in eGFR from baseline. Forty-four (41.5%) patients developed MAKE-90. Patients who developed MAKE-90 had more comorbidity, higher acuity of illness scores and more prevalent AKI. Levels of urine Klotho adjusted by creatinine (Cr) were lower and serum intact FGF23 levels were higher in AKI patients vs. ICU controls. In adjusted models, the highest vs. lowest tertile of urine Klotho/Cr was independently associated with an overall 95% lower risk of MAKE-90 (81% lower risk in patients with AKI). The highest vs. lowest tertile of serum intact FGF23 was associated with >300% higher risk of MAKE-90. Urine Klotho/Cr levels were significantly lower and serum intact FGF23 levels significantly higher in critically ill patients with AKI vs. matched-controls without AKI. When measured in the first 48h of ICU admission or AKI diagnosis, urine Klotho/Cr independently associated with major adverse kidney events, particularly in patients with AKI. These results show promise for testing these biomarkers -individually or in combination- as part of novel risk-prediction models of renal outcomes in the ICU.

Project description:BACKGROUND:Sodium-glucose cotransporter-2 inhibitors (SGLT2is) represent a new class of oral hypoglycemic agents used in the treatment of type 2 diabetes mellitus. They have a positive effect on the progression of chronic kidney disease, but there is a concern that they might cause acute kidney injury (AKI). METHODS AND FINDINGS:We conducted a systematic review and meta-analysis of the effect of SGLT2is on renal adverse events (AEs) in randomized controlled trials and controlled observational studies. PubMed, EMBASE, Cochrane library, and ClinicalTrials.gov were searched without date restriction until 27 September 2019. Data extraction was performed using a standardized data form, and any discrepancies were resolved by consensus. One hundred and twelve randomized trials (n = 96,722) and 4 observational studies with 5 cohorts (n = 83,934) with a minimum follow-up of 12 weeks that provided information on at least 1 adverse renal outcome (AKI, combined renal AE, or hypovolemia-related events) were included. In 30 trials, 410 serious AEs due to AKI were reported. SGLT2is reduced the odds of suffering AKI by 36% (odds ratio [OR] 0.64 [95% confidence interval (CI) 0.53-0.78], p < 0.001). A total of 1,089 AKI events of any severity (AEs and serious AEs [SAEs]) were published in 41 trials (OR 0.75 [95% CI 0.66-0.84], p < 0.001). Empagliflozin, dapagliflozin, and canagliflozin had a comparable benefit on the SAE and AE rate. AEs related to hypovolemia were more commonly reported in SGLT2i-treated patients (OR 1.20 [95% CI 1.10-1.31], p < 0.001). In the observational studies, 777 AKI events were reported. The odds of suffering AKI were reduced in patients receiving SGLT2is (OR 0.40 [95% CI 0.33-0.48], p < 0.001). Limitations of this study are the reliance on nonadjudicated safety endpoints, discrepant inclusion criteria and baseline hypoglycemic therapy between studies, inconsistent definitions of renal AEs and hypovolemia, varying follow-up times in different studies, and a lack of information on the severity of AKI (stages I-III). CONCLUSIONS:SGLT2is reduced the odds of suffering AKI with and without hospitalization in randomized trials and the real-world setting, despite the fact that more AEs related to hypovolemia are reported.

Dataset Information

Temporal Pattern Detection to Predict Adverse Events in Critical Care: Case Study With Acute Kidney Injury.

Publications

Temporal Pattern Detection to Predict Adverse Events in Critical Care: Case Study With Acute Kidney Injury.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets