Quality assurance and interpretation of laboratory data

image of Quality assurance and interpretation of laboratory data
Online Access: GBP25.00 + VAT
BSAVA Library Pass Buy a pass


Laboratory test results form part of the database from which a clinical diagnosis may be made. History, clinical examination and ancillary tests (laboratory tests, radiographs, etc.) are interpreted in conjunction with each other to obtain the best possible diagnosis. This chapter considers errors, units of measurement, interpretation of test results and rational laboratory data interpretation. Case examples are included.

Loading full text...

Full text loading...



Image of 2.1
2.1 Laboratory cycle: pre-analytical, analytical and post-analytical phases with the most common areas where errors can occur. QC = quality control.
Image of 2.3
2.3 Scatter plots obtained by analysing EDTA blood from a cat with (a–b) canine settings and (c–d) feline settings. An EDTA blood sample from a cat was submitted to a reference laboratory for haematology analysis. This sample was accompanied by a submission form that stated that the animal was a dog. (a–b) The analyser (Advia 120) scatter plots show the leucocyte and red cell scatter plots, respectively, that were obtained when the sample was analysed with the canine setting. (c–d) Leucocyte and red cell scatter plots obtained when the sample was analysed using the correct feline setting. Using the wrong setting caused an erroneous gating of the erythrocytes and leucocytes, leading to a falsely low mean cell volume (MCV), mean cell haemoglobin concentration (MCHC) and neutrophil count.
Image of 2.5
2.5 In the three circles, X is the true value and the dots represent results obtained from sequential analyses of the sample. (a) An accurate but imprecise method: the dots are widely but evenly distributed around the true value and the mean of all the values is equal to the true value. (b) A precise but inaccurate method: the dots are closely clustered together, showing good repeatability, but the results are consistently biased. (c) A method that is both accurate and precise: all the dots are close to the true values and are clustered closely together.
Image of 2.6
2.6 A Levey–Jennings plot system with an allowable error of ± 2SD from the mean (1). The first seven values appear ‘in control’ and are close to the mean, evenly distributed either side of it. The eighth value is outside the control limits (as indicated by the arrow), but subsequent values are within control limits.
Image of 2.7
2.7 A Levey–Jennings plot using the Westgard Multirules system. The first five readings are ‘in control’. Readings 6–9 all exceed 1SD on the same side of the mean, violating the 4 rule. This is an indicator of systematic error (bias). In addition, reading 7 violates the 1 rule. Following recalibration of the analyser, readings 10 and 11 are ‘in control’, but reading 12 exceeds 3SD, violating the 1 rule, reflecting random error. Troubleshooting revealed an error in the test procedure. Subsequent readings are ‘in control’.
Image of 2.9
2.9 Illustration of the problems with population-based reference intervals (RIs) when applied to an individual. If the test of interest has marked individuality, a result may fall within the population-based RI even though it is too high for that individual.
Image of 2.11
2.11 Examples of two analytes with a high and a low index of individuality, respectively. The four horizontal bars represent the range of values in four individuals. In the case of high individuality, a significant change in the analyte concentration caused by the disease may be missed and the result may fall within the RI. In this case, the use of subject-based RIs or RCVs may be beneficial. LRL = lower reference limit; URL = upper reference limit.
Image of 2.12
2.12 The distribution of values for an analyte. In (a) there is a Gaussian symmetrical distribution and the mean, median and mode are in the same central position. These data could be analysed by parametric methods, calculating the mean and 2SD to produce reference values. (b) The data points are not in a symmetrical distribution and the mode, median and mean are different. These data would be analysed by non-parametric methods (usually using percentiles) to produce reference values.
Image of 2.13
2.13 Establishment of reference intervals by using the central values with exclusion of the lowest and highest 2.5% of the reference values.
Image of 2.14
2.14 Demonstration of the problem of trying to establish cut-off points for any test between healthy individuals and diseased individuals. (a) The test results from healthy animals do not overlap the test results from diseased animals and so there is a clear cut-off indicated by line 1. This test has 100% sensitivity and specificity. (b) The test results from healthy animals overlap those from diseased animals. If the cut-off is set at line 1 the test is sensitive but not specific, because a high proportion of healthy animals will have results above the cut-off. Conversely, if the cut-off is set at line 3 the test becomes more specific (very few non-diseased animals have results above the cut-off), but is much less sensitive (a significant proportion of diseased animals have results below the cut-off. If line 2 is selected as the cut-off, the test has moderate sensitivity and specificity.
Image of 2.15
2.15 The receiver-operating characteristic (ROC) curve is a graphical representation of diagnostic sensitivity and specificity for a test at varying selected cut-off values, in this example numbered 1–6. Cut-off level 1 is the lowest cut-off plotted and gives a high sensitivity but low specificity. Cut-off level 6 is the highest cut-off plotted. The specificity is much higher (close to the -axis) but the sensitivity is lower. Cut-off 4 has the best compromise of sensitivity and specificity, lying closest to the top left corner of the graph. A good test has values close to the upper left corner of the plot. Test results around the diagonal dotted line would indicate a useless test.

More like this

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error