Jean-Philippe Lafrance1, Donald R Miller. 1. Center for Health Quality, Outcomes, and Economic Research, Edith Nourse Rogers Memorial Veterans Hospital, Bedford, MA 01730, USA. jplaf@bu.edu
Abstract
BACKGROUND: Existing acute kidney injury (AKI) definitions are not well adapted for database studies, leading to a great variety of methods used in research. Variations in time before hospitalization used to assess baseline kidney function when identifying episodes of AKI may lead to different case samples and mortality risks in database studies, but the magnitude of these effects is not known. STUDY DESIGN: A retrospective cohort study. SETTINGS & PARTICIPANTS: 1,126,636 veterans hospitalized at least once within the US Department of Veterans Affairs health care system between 2000 and 2005. PREDICTOR: AKI was defined by comparing (using ratio [≥1.5] or difference [increase of 0.3-0.5 mg/dL]) the highest serum creatinine level during hospitalization with the lowest level during 4 different baseline assessment periods (in-hospital only and 3, 6, or 12 months preadmission). OUTCOMES & MEASUREMENTS: In-hospital mortality risk was estimated using multivariable logistic regression models. RESULTS: Using the ratio definition, the cumulative incidence of AKI ranged from 12.5% (in-hospital only) to 18.3% (12 months preadmission). Newly added cases had milder AKI and lower mortality risk. The discriminative power increased slightly (C statistic increased from 0.846 to 0.855; P = 0.001) by extending the baseline period to at least 3 months. Both the ratio and difference definitions did not perform well in patients with chronic kidney disease stages 4 and 5. LIMITATIONS: Possibility of residual confounding and under-representation of women (4.5%). CONCLUSIONS: Many additional AKI cases may be identified by extending the baseline assessment period; however, added cases may be less severe with lower mortality risk. The relative strengths of these biases and combined effects of reducing misclassification (identification of more AKI cases) and increasing risk dilution (identifying milder cases) may vary across populations. Consensus regarding how baseline kidney function should be determined in database studies should be reached.
BACKGROUND: Existing acute kidney injury (AKI) definitions are not well adapted for database studies, leading to a great variety of methods used in research. Variations in time before hospitalization used to assess baseline kidney function when identifying episodes of AKI may lead to different case samples and mortality risks in database studies, but the magnitude of these effects is not known. STUDY DESIGN: A retrospective cohort study. SETTINGS & PARTICIPANTS: 1,126,636 veterans hospitalized at least once within the US Department of Veterans Affairs health care system between 2000 and 2005. PREDICTOR: AKI was defined by comparing (using ratio [≥1.5] or difference [increase of 0.3-0.5 mg/dL]) the highest serum creatinine level during hospitalization with the lowest level during 4 different baseline assessment periods (in-hospital only and 3, 6, or 12 months preadmission). OUTCOMES & MEASUREMENTS: In-hospital mortality risk was estimated using multivariable logistic regression models. RESULTS: Using the ratio definition, the cumulative incidence of AKI ranged from 12.5% (in-hospital only) to 18.3% (12 months preadmission). Newly added cases had milder AKI and lower mortality risk. The discriminative power increased slightly (C statistic increased from 0.846 to 0.855; P = 0.001) by extending the baseline period to at least 3 months. Both the ratio and difference definitions did not perform well in patients with chronic kidney disease stages 4 and 5. LIMITATIONS: Possibility of residual confounding and under-representation of women (4.5%). CONCLUSIONS: Many additional AKI cases may be identified by extending the baseline assessment period; however, added cases may be less severe with lower mortality risk. The relative strengths of these biases and combined effects of reducing misclassification (identification of more AKI cases) and increasing risk dilution (identifying milder cases) may vary across populations. Consensus regarding how baseline kidney function should be determined in database studies should be reached.
Authors: Marianne Pons; Leonidas G Koniaris; Sharon M Moe; Juan C Gutierrez; Aurora Esquela-Kerscher; Teresa A Zimmers Journal: Surgery Date: 2018-05-03 Impact factor: 3.982
Authors: Morgan E Grams; Sushrut S Waikar; Blaithin MacMahon; Seamus Whelton; Shoshana H Ballew; Josef Coresh Journal: Clin J Am Soc Nephrol Date: 2014-01-23 Impact factor: 8.237
Authors: Steven M Handler; Pui Wen Cheung; Colleen M Culley; Subashan Perera; Sandra L Kane-Gill; John A Kellum; Zachary A Marcum Journal: J Am Med Dir Assoc Date: 2014-05-10 Impact factor: 4.669