Format

Send to

Choose Destination
J Med Syst. 2017 Aug 31;41(10):156. doi: 10.1007/s10916-017-0806-4.

Predicting Major Adverse Kidney Events among Critically Ill Adults Using the Electronic Health Record.

Author information

1
Division of Allergy, Pulmonary, and Critical Care Medicine, Department of Medicine, Vanderbilt University Medical Center, 1161 21st Ave S., T-1218 MCN, Nashville, TN, 37232-2650, USA. andrew.c.mckown@vanderbilt.edu.
2
Department of Biostatistics, Vanderbilt University Medical Center, Nashville, TN, USA.
3
Department of Anesthesiology, Vanderbilt University Medical Center, Nashville, TN, USA.
4
Department of Biomedical Informatics, Vanderbilt University School of Medicine, Nashville, TN, USA.
5
Division of Allergy, Pulmonary, and Critical Care Medicine, Department of Medicine, Vanderbilt University Medical Center, 1161 21st Ave S., T-1218 MCN, Nashville, TN, 37232-2650, USA.

Abstract

Prediction of major adverse kidney events in critically ill patients may help target therapy, allow risk adjustment, and facilitate the conduct of clinical trials. In a cohort comprised of all critically ill adults admitted to five intensive care units at a single tertiary care center over one year, we developed a logistic regression model for the outcome of Major Adverse Kidney Events within 30 days (MAKE30), the composite of persistent renal dysfunction, new renal replacement therapy (RRT), and in-hospital mortality. Proposed risk factors for the MAKE30 outcome were selected a priori and included age, race, gender, University Health System Consortium (UHC) expected mortality, baseline creatinine, volume of isotonic crystalloid fluid received in the prior 24 h, admission service, intensive care unit (ICU), source of admission, mechanical ventilation or receipt of vasopressors within 24 h of ICU admission, renal replacement therapy prior to ICU admission, acute kidney injury, chronic kidney disease as defined by baseline creatinine value, and renal failure as defined by the Elixhauser index. Among 10,983 patients in the study population, 1489 patients (13.6%) met the MAKE30 endpoint. The strongest independent predictors of MAKE30 were UHC expected mortality (OR 2.32 [95%CI 2.06-2.61]) and presence of acute kidney injury at ICU admission (OR 4.98 [95%CI 4.12-6.03]). The model had strong predictive properties including excellent discrimination with a bootstrap-corrected area-under-the-curve (AUC) of 0.903, and high precision of calibration with a mean absolute error prediction of 1.7%. The MAKE30 composite outcome can be reliably predicted from factors present within 24 h of ICU admission using data derived from the electronic health record.

KEYWORDS:

Acute kidney injury; Critical illness; Predictive modeling; Renal replacement therapy

PMID:
28861688
PMCID:
PMC5821255
DOI:
10.1007/s10916-017-0806-4
[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for Springer Icon for PubMed Central
Loading ...
Support Center