Application of 3-D ∆ Check Graphs to HbA 1c Quality Control and HbA 1c Utilization

∆ checking is a laboratory information system (LIS)-based tool that detects patient and laboratory quality control errors. By using hemoglobin A 1c (HbA 1c ) data, we developed a novel approach to summarizing and presenting patient ∆ values to address limitations of current ∆ check algorithms. ∆ values were calculated from intrapatient pairs of HbA 1c (n = 55,327) measured during 2 years in a single referral or a university hospital laboratory. Threedimensional ∆-time (∆T) and percentile limit graphs were constructed. Cumulative distribution function analysis was used to explore clinical utilization. The ∆T graphs showed that HbA 1c ∆ values increase asymmetrically over time. Although the 2.5 to 97.5 and 5.0 to 95.0 percentile ∆ check limits were similar for both sites, the referral laboratory’s 0.5 to 99.5 percentile limits were wider. For acute patient care environments, we recommend limits of –3.5% and 1.8% for measurements between 0 and 60 days and –4.0% and 2.0% for measurements between 60 and 120 days. For the outpatient environment, we recommend limits of –4.2% and 2.1% and 5.0% and 2.5% for measurements between 0 and 60 days and 60 and 120 days, respectively. ∆ checking can be significantly improved with customization of limits set by population and interobservation period. Because LIS systems are incapable of these customizations, customers must become advocates for these modifications. ∆ checking involves the calculation and evaluation of sequential intrapatient differences (∆ values). ∆ checking is usually a laboratory information system (LIS)-based tool and can detect preanalytic and analytic errors and clinically unusual results. Numeric ∆ values are calculated with the formula: result (time2) – result (time1) . Percentage ∆ values are calculated by dividing the numeric ∆ value by result (time2) . 1 These ∆ values are then compared with thresholds stored in the LIS. Large ∆ values may arise from intrapatient variation and from instrument (analytic) variation and preanalytic error, including sample mix-up or patient/sample misidentification, and postanalytic error (eg, transcription error). If a ∆ value exceeds a predefined threshold, it can be investigated by confirming the specimen’s identity, reanalyzing the current specimen, reanalyzing quality control material, and even patient resampling and remeasure