MEASUREMENT UNCERTAINTY


Original patient result


Classification based on decison limit/reference range


Range of MU values Result +/– MU


Classification based on decison limit/reference range


Save result


possibly due to the availability of more frequent testing, anticoagulation control was improved using the “worse” method. For now, the clinical outcome models are useful, but require refining.


Compare result


Biological variation Biological variation has been the mainstay of setting APS since their early application in EQA schemes. Underlying the APS derived from BV are within individual and within group variability. From there, specifications for allowable analytical imprecision and bias have been derived (the history is reviewed nicely in Sandberg6


). Change


Determine Max MU to retain misclassification below allowable limit


Fig 2. Monte Carlo simulation of increasing uncertainty to patient results.


targets is based on reducing the risk of excessive uncertainty in final patient results affecting clinical interpretation. Performance specifications are set according to the models presented at the Milan conference in 2014.6


Model 1,


directly (Model 1a) or indirectly (Model 1b), sets limits for MU based on clinical outcome. The absence of sufficient data that directly links assay performance to outcome is well recognised. Model 1 is appropriate when measurands are directly associated with diagnosis and treatment decisions in specific diseases. For measurands under tight metabolic control, biological variation targets can be used according to those published in the EFLM Biological Variation database (https://biologicalvariation.eu/), defining Model 2. In the absence of being able to apply Models 1 or 2, performance specifications based on current state of the art – the best analytical performance that can achieved with current methodology – can be used.


Clinical outcome models (Model 1)


Clinical outcome studies are those that are designed to identify performance requirements for analytical methods based on the outcome to the patient. Although this at first glance would appear to be a simple concept, in reality published data is scarce. To aid with generating reliable data, Model 1 was split into two different models representing direct clinical outcome studies (Model 1a) and indirect clinical outcome studies (Model 1b – using simulation or decision theory).


30


While clinical laboratory results certainly guide clinical decision-making, the outcome of the patient is determined by intermediate steps after the generation of the result, including, of course, intervention decisions. Results guide that process but are not solely responsible for the outcome of the patient if unexpected decisions are made. Many measurands are used in combination with many other diagnostic procedures, some from our labratories and some not. A major development in supporting Model 1b (simulation modelling for clinical outcome) has seen the development of data- driven methods to simulate the impact of degrading analytical performance on patient classification against clinical decision limits or reference intervals. Conceptually, the process itself is quite intuitive using Monte Carlo simulation (Fig 2) although it has for some time been difficult to implement. Recent publications, and work by the EFLM, have provided an online resource that allows us in our local laboratory to investigate clinical outcome effects due to MU. From there (indirectly) we can derive performance specifications based on our local patient population.7


The purpose


is to determine how “bad” uncertainty can be before we see an unacceptable proportion of patients misclassified. It has been shown that worse analytical


performance, depending on the clinical situation, can still improve patient outcomes. Studies comparing point-of- care INR measurement with laboratory measurements showed deceased analytical performance of the POCT device. However, the way it was used,


No change


Review diagnostic accuracy measures


n Calculations from biological variation


In the laboratory it is commonplace to quantify bias and imprecision to manage our quality processes. However, this is less important to our clinical users for interpretation of results. Clinical focus is on the overall variability of the result. The concern is with how big that variability is and whether that changes their understanding of the interpretation of the result.


This concept aligns with more of a top-down, total error (TE) concept and was first described in the 1960s. This is the theory developed by Dr James Westgard over the following decades and became the mainstay of quality control and V&V activities that we are familiar with today. MU differs somewhat by including metrological traceability – and the meaning of results when compared to a reference system. Bias is explicitly handled in the TE formula whereas the MU approach assumes correction of bias if it is significant. This has led to a lot of discussion regarding “which is best”. Irrespective of those discussions, the calculation for the maximum allowable uncertainty (MAU) is simply 0.5 x the within subject biological variation (CVI). To be applied to patient results, and have an associated confidence (customarily ~95%) a coverage factor, k, is applied, usually equal to 2 so that the MAU = 2 x 0.5 x CVI.


n The issue with troponin In an excellent paper on troponin clinical utility and APS, it is discussed in detail about the challenges of using BV data blindly in APS derivation. Most BV data published had not been collected in a protocol that reflected the clinical utility of the assay. For example, troponin results are required quickly, and repeats performed in short periods of time. It is of little clinical utility to use BV data that has been collected at weekly intervals over a period of months and then try to apply


APRIL 2024 WWW.PATHOLOGYINPRACTICE.COM


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56  |  Page 57