- Oral presentation
- Open Access
Continuous Monitoring of Urine Output And Hemodynamic Disturbances Improves Early Detection of Acute Kidney Injury During First Week of Icu Stay
© Flechet et al.; 2015
- Published: 1 October 2015
- Acute Kidney Injury
- Urine Output
- Mean Arterial Blood Pressure
- Continuous Monitoring
- Good Calibration
Acute Kidney Injury (AKI) is associated with increased morbidity and mortality in critically ill patients . Early detection and treatment may improve outcome. Previously, we developed a logistic regression (LR) model for early detection of AKI based on routinely collected data available at baseline, ICU admission and at the end of the first day (LR_BAD1) . Continuous monitoring parameters may provide additional predictive power, in particular, urine output and hemodynamic parameters, whose management influences kidney perfusion.
To assess if adding continuous monitoring variables recorded during the first 24h of ICU stay, to a model to predict AKI in the first week of ICU admission, can improve the predictive performance.
The model was built and validated in a subset of 1778 ICU patients from the EPaNIC trial . Patients with end-stage renal disease, those with AKI on the first day of ICU stay and those without available hemodynamic monitoring data during the first day were excluded. AKI was defined by the creatinine criteria from the KDIGO guidelines.
The LR_BAD1 model used only covariates selected at baseline, ICU admission and at the end of the first day. in the LR_BAD1+ model, we have added features extracted from hourly measures of urine and minute-by-minute measures of heart frequency (HF) and mean arterial blood pressure (MABP). Moreover, the cumulative dose of inotropes administered to each patient during the first 24h of ICU stay was used as additional covariate. The predictive power was assessed by ROC, decision and calibration curves analysis using 300 bootstraps replicates.
Demographics - 1778 ICU patients.
Age (years): median [IQR]
Apache II: median [IQR]
ICU length of stay (hours): median [IQR]
Baseline serum creatinine (mg/dl): median [IQR]
Type of admission: elective/emergency (%)
1155 (64.96) / 623 (35.04)
Diagnostic group: cardiac/transplant/non cardiac surgery and trauma-burns/medical-others (%)
1186 (66.70) / 131 (7.37) / 385 (21.65) / 76 (4.27)
Male gender: n (%)
Sepsis on ICU admission: n (%)
Diabetes: n (%)
Incidence of AKI within first week: n (%)
Statistics of LR_BAD1 and LR_BAD1+.
0.80 +- 0.01
0.84 +- 0.01
0.07 +- 0.01
0.08 +- 0.01
0.87 +- 0.09
0.88 +- 0.08
0.00 +- 0.01
0.00 +- 0.01
0.22 +- 0.01
0.27 +- 0.02
Early detection of AKI can be improved by routinely monitored information of urine and hemodynamics. Hence, AKI can be already predicted with high discrimination and good calibration, only by using routinely collected ICU data.
GM receives funding from FWO (1846113N). GVdB receives long-term research financing via the Flemish government Methusalem-program.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.