Background: With increasing evidence on the impact of hyperglycemia on patient outcomes, hospitals are beginning to move towards using inpatient glucose control as metrics of performance (1-3). However, efforts to improve inpatient glycemic control should assure validity and reliability of laboratory values (4) as these measurements can potentially affect these metrics and how the metrics guide patient care.
Methods: This project was carried out in a large urban medical center as a part of a larger quality improvement project aimed at improving glycemic control for hospitalized adults. Three months of laboratory data for all glucose values > 1000mg/dl on a Basic Metabolic Panel (BMP) was extracted from the electronic medical record. This was done to identify targets for improvement in glycemic control for these patients at risk of developing diabetes-associated complications. Chart review was performed to extract relevant clinical data including laboratory and medication data. Three criteria for a likely falsely elevated glucose were created by team consensus: 1. glucose level > 1000 mg/dl on BMP; 2. a subsequent value of glucose by BMP < 200 mg/dl within 12 hours of the index value; 3. no intervening insulin administration and no more than three liters of intravenous fluid (IVF) administered in the hours between the two values. To evaluate if these events were likely false, we examined the proportion of events that occurred among patients with an elevated Hemoglobin A1C value, defined as A1C > 6.4 recorded in the EMR. Fisher’s exact testing was used for this comparison.
Results: Eighty-three instances of glucose level >1000 mg/dl over 3 months were included representing 67 unique patients. Out of these, 26 events (36%) met the criteria as a likely falsely elevated glucose. Eighteen of the events (69%) were based on phlebotomy testing that occurred during concomitant administration of dextrose-containing IVF. Of the 44 elevated glucose events with a documented elevated A1C, only 2 meet criteria for falsely elevated glucose event (5%). Results from Fischer’s exact test showed that likely falsely elevated glucose events were less likely to occur (i.e. did not meet criteria) in patients with an elevated A1C (p<0.0001)
Conclusions: As hospitals embark on more and more efforts to improve glycemic control in their admitted population, this study demonstrated that some of the highest glucose values in the hospital may, in fact, be falsely elevated glucose values. Additional studies are needed to look into why this might occur and how to prevent it.