Laboratory Administration & Management of Pathology Practices

Contributed by Matthew Shashack, Ph.D.


Images hosted on other servers:

• Hoffmann and Waid initially introduced a method named average of normals in which the average of results that fall within or around the reference range are displayed on a control chart
• The average of normals concept led to the development of alternative methods and new algorithms for calculating average values, a deeper understanding of moving average (MA) error detection and its characteristics and guidelines for both determining optimal MA settings and for an MA approach that does not only rely on normals or mean calculations
• MA QC consists of 3 components (procedure settings)
  • Inclusion criteria, exclusion criteria and truncation limit
    • Aim to remove outliers, extremes and nonnormals for the calculation algorithm to reduce variation in the included assay results
    • Truncation limits are the defined upper and lower limits of patient values used in the MA calculation
    • Any relevant variables can be considered for inclusion criteria but this is often organized in MA management software
    • Exclusion criteria include tests with too much interpersonal variation as well as specific patient populations with increased test variance due to patient acuity
  • Calculation algorithm
    • First step in developing the calculation algorithm is intended to select the particular average values: the mean, the median, the exponentially weighted moving average (EWMA) or Bull’s algorithm (XbarB)
    • There is no guidance to specify which algorithm is the best
    • Calculating the mean is appropriate for MA when assay results are normally distributed and exhibit low variability
    • Median calculations are more effective for assays with nonnormally distributed results, particularly those with extreme values or outliers
    • Both the mean and median are recalculated continuously as new results are generated
    • EWMA calculates a new MA including previous measurements that are modified with a weighting factor
    • EWMA weighting factor is between 0 and 1 and a smaller factor assigns less importance to the new results
    • XbarB algorithm, also known as Bull's algorithm, was originally developed for batches of 20 measurements on a hematological analyzer to smooth the MA deviation, minimize the impact of outliers and make it suitable for non-Gaussian populations
    • XbarB algorithm is calculated per batch, thus delays in detecting bias can occur, especially for lower volume assays
    • Second step in developing the calculation algorithm is the selection of variables
    • MA algorithm variables include the batch size for mean, median and XbarB methods, whereas a weighting factor between 0 and 1 is required for the EWMA method
    • Larger batch sizes or lower weighting factor result in less variation in MA values and enable the detection of smaller systematic errors
    • Too large of a batch size can lead to delayed identification of larger systematic errors, as it requires more assay results
  • Control limits
    • Option 1: when assigned MA upper and lower control limits around the mean are exceeded, this indicates out of control status for the assay
    • Option 2: after applying the MA procedure to a training set from the laboratory, control limits are established
    • MA established from a training set (option 2) are nonsymmetrical, non-SD based and are designed to eliminate false MA alarms using the most stringent control criteria (Clin Chem Lab Med 2019;57:773)
• Optimal MA setting must be established before its application; optimization can be done in the following ways
  • Trial and error
  • Graphical
  • Power function analysis
  • TEa detection probability
  • ANPed (average number of patient samples affected until error detection)
  • Bias detection curves and MA validation charts
• See Diagrams / tables

• Optimal MA settings are implemented in software packages available in laboratories that support the MA application
• Features that are desirable in the laboratory software to perform MA QC are
  • MA inclusion and exclusion criteria selection
  • Support the selected calculation algorithm
  • Support of SD based on accuracy or MA reference range based control levels
  • Performs MA calculation for every new set of assay results
  • MA displayed as Levey-Jennings or accuracy plots
  • Real time notification of MA alarm (push set up)
  • Simple reset of MA after MA alarm by end user
• Most hemocytometer analyzers include basic MA packages that support at least the XbarB calculation algorithm
• Various middleware systems, such as Remisol Advance (Beckman Coulter), Instrument Manager (Data Innovations), Roche Middleware Solutions (Roche Diagnostics), Atellica Data Management System (Siemens Healthcare Diagnostics), AlinIQ (Abbott Diagnostics) and LIS like GLIMS, support the MA application
• Some software packages include only a single calculation algorithm, such as mean or XbarB calculations, while others support a broader range of algorithms
• These software packages are utilized for MA QC (i.e., real time MA calculations, graphical representation of MA and alarming in the case of error detection that supports statistical QC and long term assurance of the laboratory test being performed) (Clin Chem Lab Med 2019;57:773)

• In a study where 6 parameters (calcium, bicarbonate, FT4, MCV, MCHC and reticulocytes count) were assessed using the MA generator application (Huvaros, versions used between 2019 and 2021) in General Laboratory Information Management System, version 9.5.18 or 9.5.25, LIS and a graphical comparison of bias detection curve were performed
• Result was enhanced patient safety by a continuous system control service, complementing periodic IQC monitoring, prevention of the release of erroneous results to the clinic and faster detection when analytical errors occur
• Proved to be a valuable tool for interanalyzer comparisons and trend monitoring, contributing to improved long term test stability (Clin Chem Lab Med 2022;60:1719)
• MA QC can be integrated into risk based traditional QC plans in small volume medical laboratories, where MA can detect critical size bias
• MA can provide additional security between 2 QC plans without waiting for the interval provided by a traditional QC plan, thus confirming the existence of an error requiring corrective action (Biochem Med (Zagreb) 2022;32:020711)
• Average of deltas (AoD), formed by the combination of delta check (used to detect systematic error) and MA, enhances error detection with relatively few paired patient samples (Clin Chem 2021;67:1019)
• After optimization and validation, MA QC can be successfully implemented in the LIS and can be used for continuous QC (Biochem Med (Zagreb) 2022;32:010705)

• Though MA QC is emerging as a promising QC tool, its broad acceptance is hindered by
  • Complexity of determining optimal MA settings
  • Dearth of evidence based guidelines for its application, especially for low volume tests
• MA QC is gaining acceptance as a supplement to conventional QC, enhancing error detection rather than replacing the use of QC samples (Biochem Med (Zagreb) 2022;32:010705)

Which of the following is a benefit of patient moving averages (MA) over traditional internal quality control (QC)?

1. Only laboratories with high specimen volume can utilize patient moving averages
2. Patient moving averages can improve patient safety by serving as a continuous system monitor
3. Patient moving averages eliminate the need for internal quality controls
4. Unlike traditional internal quality controls, patient moving averages do not require inclusion and exclusion criteria

B. Patient moving averages can improve patient safety by serving as a continuous system monitor that potentially enables faster detection of analytical errors. Answer C is incorrect because internal quality controls are a regulatory requirement and MA is a supplement (not replacement) of internal QC. Answer D is incorrect because MA selection criteria include removing outliers, extremes and nonnormals as well as the use of truncation limits to define upper and lower limits of patient values used in the moving average calculation. Answer A is incorrect because both large and small volume laboratories can utilize MA.

Comment Here

Reference: Moving average quality control

Using the figure shown above, what is the impact of removing the truncation limit?

1. Moving average alert frequency decreases because all results are included in the moving average calculation
2. Moving average alert frequency increases because all patient data is included in the moving average calculation
3. Moving average alert frequency is unchanged
4. Moving average becomes more accurate because no patient results are removed

B. Moving average alert frequency increases when there is no truncation limit and all patient data, including extreme outliers, are included in the moving average calculation. Answer A is incorrect because removing the truncation limit will include extreme patient results in the MA calculation. This will skew the MA calculations and an increase in alerts. Answer D is incorrect because including extreme patient results can skew the calculated MA and lead to an increase in alerts. Answer C is incorrect because removing the truncation limit will allow all patient results to be included in the MA calculation, which results in a change of the alert frequency.

Comment Here

Reference: Moving average quality control