Use of potassium concentrations as a quality-of-service metric for phlebotomists detects systematic preanalytical biases and facilitates their correction.

The detection and monitoring of preanalytical bias in the clinical laboratory can be challenging, as bias can be introduced at any point from specimen collection to sample processing, and those biases stemming from sample collection or handling before delivery to the laboratory can be difficult to measure and eliminate. In particular, variations in potassium concentrations due to preanalytical sources of error are a pervasive and clinically significant problem (1–3 ). Initial concern regarding spuriously increased potassium concentrations ( 5.2 mmol/L) occurring in the laboratory of the Dana Farber Cancer Institute was raised by clinicians reporting patients displaying increased potassium without any apparent clinical justification. Review of the medical records of the patients involved demonstrated that the majority had potassium values within the reference interval on the draw before the increased value, with a mean increase in potassium of 1.2 (0.7) mmol/L vs the prior value. Retesting on an alternative platform confirmed hyperkalemia. To assess for a preanalytical cause of these values, samples were redrawn on the same day from a series of patients with potassium values 5.2 mmol/L. This assessment demonstrated that in 27 of 28 patients tested, potassium was in the reference interval upon retesting with an independent sample. The mean degree of decrease in potassium upon redraw was 0.9 (0.7) mmol/L. The maximum decrease observed was 2.2 mmol/L. Altogether, these findings strongly suggested that a preanalytical source of error was responsible for the increases in potassium. After excluding hemolysis, potassium release from leukocytes, or cell lysis during centrifugation as contributing factors, we considered whether the phenomenon of spuriously increased potassium values might relate to the draw technique used by specific phlebotomists. Analysis of mean potassium values by specific phlebotomists demonstrated a significant effect (P 0.0001 by 1-way ANOVA) of the phlebotomist performing the draw (data not shown). As an example, a histogram of the potassium values from each draw from 2 phlebotomists displaying higher average potassium values was plotted relative to the potassium values across the institution (Figure 1). Demonstration that specific phlebotomists were strongly associated with increased rates of hyperkalemia prompted a systematic evaluation of phlebotomy practices at the institution. In particular, it was noted that several phlebotomists were either using squeeze balls or encouraging fist clenching during draws, despite these maneuvers being prohibited by institutional policy. Additionally, occasionally excessive vein tapping before draw and variations from the policy of using 20-gauge needles to minimize hemolysis were observed. In response to this, a period of direct observation and phlebotomist retraining was instituted in November through December 2013, emphasizing correction of the deviations from institutional protocol noted above. The observation that selected phlebotomists consistently displayed increased potassium values in their draws suggested that the degree of deviation in average potassium values could function as a quality-of-service metric reported back to each phlebotomist. After the period of retraining, phlebotomy staff had monthly meetings with supervisors where data regarding the mean potassium values and number of potassium values 5.2 mmol/L from their draws during the prior month were reviewed. Continued monitoring demonstrated that this program was successful in substantially lowering the incidence of pseudohyperkalemia across the institution. Without any alterations to the potassium assay itself, the percentage of potassium values 5.2 mmol/L reported each month fell by approximately 39% relative to the start of the study. For a subset of 5 phlebotFig. 1. The distribution of potassium values for 2 phlebotomists associated with significantly increased potassium values are displayed versus the institutional distribution. Clinical Chemistry 60:11 000 – 000 (2014) Letters to the Editor