Abstract Background Our laboratory information system (LIS) provides only cursory capability for monthly review of quality control measurements. We streamlined the major portion of this process using customized programming in R, as described below. Methods Month-long QC data were obtained using the LIS archive’s data retrieval program (QLIK). Data were read and processed using our R program (QC-R.R), for each analyte and QC level, to produce both Levey-Jennings (LJ) plots (for evaluation of possible temporal trends in QC data) as well as cumulative distribution (CD) plots (to evaluate whether QC data are normally distributed). Plots were sent to graphics files whereby they could be reviewed sequentially by a single forward-arrow click. Results Our laboratory utilizes two QC measurements for each of 157 analytes. Thus, monthly review involves evaluation of 314 datasets for QC results. With the R program outputs, results can be reviewed visually in rapid fashion, to determine whether there exist discrepancies between fixed limits and current data. An example CD plot is shown in Figure. In this example, QC data (points) were, appropriately, normally distributed (theoretical curved line based on data mean and SD). The currently operative fixed mean±2SD limits (vertical dashed lines) were, however, disparate from those of data (vertical solid lines). This set of results was thus marked for updating of mean and SD, with reference to a master list of results of calculations generated by the QC-R.R program. Time for review of monthly QC results was dramatically reduced, from more than 12 man-hours to less than 4 man-hours per month. Conclusions Limitations of our LIS for monthly QC review were bypassed using R-based programming to automate a significant fraction of the process. This led to a significant savings in labor attending to this important task. The use of CD plots is helpful for rapid data interpretation.
Read full abstract