Related Experiment Videos
Refining predictive models in critically ill patients with acute renal failure.
Ravindra L Mehta1, Maria T Pascual, Carmencita G Gruta
1Division of Nephrology, University of California, San Diego Medical Center, San Diego, California 92103, USA. rmehta@ucsd.edu
Journal of the American Society of Nephrology : JASN
|April 19, 2002
Summary
High mortality in acute renal failure necessitates better risk prediction. This study developed a novel, disease-specific severity of illness equation for improved patient stratification and clinical trial analysis.
Area of Science:
- Nephrology
- Critical Care Medicine
- Biostatistics
Background:
- Acute renal failure (ARF) is associated with high in-hospital mortality rates.
- Effective risk-adjustment tools are crucial for quality improvement and clinical trial design.
Purpose of the Study:
- To develop and validate a disease-specific severity of illness equation for acute renal failure.
- To identify key clinical variables predictive of in-hospital mortality in ARF patients.
Main Methods:
- Multivariable logistic regression analysis of 605 ICU patients with ARF (1989-1995).
- Inclusion of demographic, historical, laboratory, and physiological variables.
- Evaluation of model discrimination using the area under the receiver operating characteristic curve (AUC).
Main Results:
- In-hospital mortality rate was 51.9%.
- Significant predictors of mortality included age, male gender, respiratory, liver, and hematologic failure, creatinine, BUN, urine output, and heart rate.
- The developed model demonstrated good discrimination (AUC=0.83) and outperformed existing models.
Conclusions:
- A validated, disease-specific severity of illness equation for ARF was developed.
- The model utilizes routinely available clinical variables for improved risk stratification.
- Further cross-validation is recommended for widespread clinical application, especially in clinical trials.