Related Experiment Videos
External validation of severity scoring systems for acute renal failure using a multinational database.
Shigehiko Uchino1, Rinaldo Bellomo, Hiroshi Morimatsu
1Department of Intensive Care, Austin Hospital, Melbourne, Australia.
Critical Care Medicine
|September 9, 2005
Summary
Current acute renal failure scoring systems lack broad applicability. Validation in a large, multi-center cohort revealed poor discrimination and calibration for predicting mortality in intensive care unit patients.
Area of Science:
- Nephrology
- Critical Care Medicine
- Clinical Epidemiology
Background:
- Several scoring systems exist to assess acute renal failure (ARF) severity.
- Previous validation studies were often limited to single or few centers, raising concerns about generalizability.
Purpose of the Study:
- To evaluate the widespread applicability and performance of established ARF severity scoring systems.
- To assess the discrimination and calibration of ARF-specific and general scoring systems in a large, multi-center cohort.
Main Methods:
- Prospective clinical investigation involving 1,742 intensive care unit patients with ARF.
- Calculation and testing of six ARF scoring systems (four ARF-specific, two general) using receiver operating characteristic curves and Hosmer-Lemeshow tests.
- Outcomes and demographic/clinical data were collected.
Main Results:
- All tested scoring systems demonstrated poor discrimination, with Area Under the Curve (AUC) values below 0.7.
- Calibration for predicting mortality was also found to be poor across multiple scoring systems.
- Blood lactate levels, used as a reference, also showed limited predictive value (AUC < 0.7).
Conclusions:
- Existing ARF scoring systems lack sufficient discrimination and calibration for predicting mortality in a broad, multi-center patient population.
- Development of more robust scoring systems may require large, multi-center databases for improved validation.