18 Trust but Verify: Repeat CBC Sample Measurements Help Identify Automated Hematology Analyzer Errors. (11th January 2018)

• Record Type: Journal Article
• Authors: Lyashchenko, Alex K; Wontakal, Sandeep N; Reynafarje, Giselle M; Hod, Eldad; Kratz, Alexander
• Is Part Of: American journal of clinical pathology. Volume 149(2018)Supplement 1
• Page Start: S172
• Page End: S173
• Publication Date: 2018-01-11
• Subjects: Diagnosis, Laboratory -- Periodicals; Pathology -- Periodicals; 616.07
• DOI: 10.1093/ajcp/aqx149.387
• Languages: English
• ISSNs: 0002-9173

Abstract

Abstract: Introduction: The precision and accuracy of modern automated hematology analyzers suggest that with proper instrument calibration, clinically significant errors in the measurement of hemoglobin (Hgb) concentration and white blood cell (WBC) count should virtually never occur. Following the identification of a CBC sample for which the Hgb result differed by more than 4 g/dL on successive analyses, we evaluated the reproducibility of Hgb and WBC results obtained using automated hematology analyzers at our institution. Methods: We compared duplicate measurements on approximately 27, 000 CBC samples that had been automatically analyzed twice by a Sysmex XN-9000 system equipped with four XN-10 analyzer modules (Sysmex, Kobe, Japan). The initial measurement on these samples produced a flagged impedance platelet count result. For such samples, which comprised approximately 10% of all CBCs analyzed during the study period, a second set of CBC results is automatically obtained within two minutes by the same XN-10 analyzer, this time using a fluorescent platelet counting method (PLT-F). Clinically significant Hgb errors were defined as a >1 g/dL difference between the paired Hgb measurements. To account for the broader range of WBC count values, WBC count errors were defined by the simultaneous presence of an absolute difference of >2, 000/uL and a relative difference of >10% between the two results. Results: Despite coefficients of variation of <1.0% for Hgb and <3.0% for the WBC count, along with continued optimal instrument performance on quality control evaluations, we found that Hgb errors and WBC count errors occurred in approximately 1/800 and 1/1, 700 of the reanalyzed samples, respectively (33 Hgb and 16 WBC errors in 26, 607 samples). To test whether these errors were limited to samples flagged for platelet counts, or related to sample quantity or instrument measurement mode, we reanalyzed 3, 938 nonflagged samples with sufficient quantity and normal appearance. Such retesting identified three Hgb and two WBC count errors, corresponding to rates of 1/1, 300 and 1/2, 000, respectively. Importantly, comparable rates were observed on all four of our analyzers and on a similar instrument at an affiliated institution. Discussion: We find that although the reproducibility of Hgb and WBC results is generally excellent, rare clinically significant discrepancies occur in up to 1 in 800 CBC tests. Such an error rate may be low enough to escape widespread recognition, but is high enough to have a negative effect on patient care. Our study demonstrates the utility of repeat patient sample testing in the evaluation of analyzer performance and suggests a potential role for delta checks in minimizing the reporting of incorrect results by identifying a subset of samples for retesting. The causes of the erroneous Hgb and WBC count results in some patient samples remain to be determined.