Experienced physicians noted unexpectedly elevated concentrations of lactate dehydrogenase in some patient samples, but quality control specimens showed no bias. To evaluate this problem, we used a “latent reference individual extraction method”, designed to obtain reference intervals from a laboratory database by excluding individuals who have abnormal results for basic analytes other than the analyte in question, in this case lactate dehydrogenase. The reference interval derived for the suspected year was 264–530 U/L, while that of the previous year was 248–495 U/L. The only change we found was the introduction of an order entry system, which requests precise sampling volumes rather than complete filling of vacuum tubes. The effect of vacuum persistence was tested using ten freshly drawn blood samples. Compared with complete filling, 1/5 filling resulted in average elevations of lactate dehydrogenase, aspartic aminotransferase, and potassium levels of 8.0%, 3.8%, and 3.4%, respectively (all p < 0.01). Microhemolysis was confirmed using a urine stick method. The length of time before centrifugation determined the degree of hemolysis, while vacuum during centrifugation did not affect it. Microhemolysis is the probable cause of the suspected pseudo-elevation noted by the physicians. Data-mining methodology represents a valuable tool for monitoring long-term bias in laboratory results.