653. Can a Statistical Model Predict the Risk of Vancomycin-Induced Nephrotoxicity (VIN)?
Session: Poster Abstract Session: Pharmacokinetics and Adverse Drug Reactions
Friday, October 21, 2011
Room: Poster Hall B1
Background: 

Previous research conducted at Baystate Medical Center identified initial trough concentrations ≥ 20mg/L, vancomycin (VAN) therapy ≥ 5 days, decubitus ulcers, malignancy, and concurrent use of ≥ 2 nephrotoxic drugs as independent risk factors for VIN in adult non-intensive care unit (ICU) patients. This led to the development of a statistical model used to predict the risk of VIN.

Methods: 

A retrospective study was designed to validate the model in adult patients receiving IV VAN > 48 hours. Exclusion criteria included ICU residence within 48 hours of first VAN administration, cystic fibrosis, active chemotherapy, IV contrast dye within 7 days of starting VAN through treatment duration, hemodialysis, vasopressor therapy within 48 hours of or during VAN therapy, and baseline serum creatinine ≥ 2mg/dL. Chart review was conducted to determine baseline characteristics. The model was applied to assign a risk score for development of VIN. The primary objective was to determine the sensitivity, specificity, and accuracy of the model.  Secondary objectives were an analysis of initial trough as a predictor of the risk for VIN, hospital length of stay, and patient management after development of VIN.

Results: 

Both the VIN and non-VIN groups consisted of 91 patients. A score of > 1 in the calculator corresponded with a sensitivity and specificity of 65% and 68% respectively, and an overall accuracy of 68%. An initial trough > 17 mg/L had 72% accuracy for predicting VIN, indicating the model was no better at predicting the risk of VIN than initial trough alone. Average length of stay was 13.6 days in VIN patients and 9.2 days in non-VIN patients (p=0.0007). After the onset of acute kidney injury, 26% of cases had no adjustment of VAN management. A nephrology consult was ordered for 41% of VIN patients, with VAN considered a cause or contributing factor in 26% of cases. Other management strategies included discontinuing VAN (25%), monitoring levels (25%), reducing the dose (24%), and extending the interval (3%).

Conclusion: 

An initial trough > 17mg/L predicted VIN as well as the statistical model. VIN patients experienced a statistically significant increased length of stay (13.6 v 9.2 days). Finally, there may be a need for more education regarding VIN and its management.


Subject Category: J. Clinical practice issues

Safia Kuriakose, PharmD1, Erica Tenholder, PharmD, BCPS1, Evan Horton, PharmD1,2, Seth T. Housman, PharmD, MPA1,3, Jane Garb, MS1 and Daniel Skiest, MD1,4, (1)Baystate Medical Center, Springfield, MA, (2)Massachusetts College of Pharmacy and Health Sciences, Worcester, MA, (3)Ctr. for Anti-Infect R&D, Hartford Hospital, Hartford, CT, (4)Baystate Medical Center-Tufts University School of Medicine, Springfield, MA

Disclosures:

S. Kuriakose, None

E. Tenholder, None

E. Horton, None

S. T. Housman, None

J. Garb, None

D. Skiest, None

Findings in the abstracts are embargoed until 12:01 a.m. EST Thursday, Oct. 20 with the exception of research findings presented at IDSA press conferences.