Variable performance of 5 detection methods for identifying intravenous fluid contamination in basic metabolic panels at an academic medical center.

Abstract

Introduction: This study evaluated 5 methods for identifying intravenous fluid contamination in basic metabolic panel specimens.

Methods: Results for blood urea nitrogen, calcium, chloride, carbon dioxide, creatinine, glucose, potassium, and sodium were retrieved from metabolic panels tested in August 2025. All 5 methods were applied to 11 862 result sets with both preceding and subsequent results available within 48 hours. Chart review was conducted for all flagged specimens.

Results: The 5 methods collectively flagged 333 specimens, including 94 confirmed contaminations and 239 reflecting pathophysiologic changes or treatment effects. Alert rates ranged from 0.16% to 1.62%, and positive predictive values ranged from 23% to 100%. Among the 94 contaminated specimens, contaminants included 63 saline, 11 dextrose, 4 potassium, 5 PlasmaLyte A (Baxter) or lactated Ringer's solution, and 11 other fluids. The methods showed differing sensitivities across contaminant types. False alerts were associated with diverse clinical conditions or therapeutic interventions, including kidney impairment, diabetes, hypokalemia, substantial infusion volumes, and recent calcium injections. The methods differed in the likelihood of false alerts.

Discussion: All 5 methods showed low alert rates with variable positive predictive values. Their performance is influenced by patient population characteristics. Local verification is essential before adopting externally developed methods for detecting specimen contaminations.