As any good bench technologist knows, quality is in the details. But details bedevil quality in sources of error that affect the precision and accuracy of test results. In labs where automation is common and less sample handling is required for hands-on testing, light microscopy is still fraught with human error. Understanding sources of microscopic variation in hematology will help you develop a strategy that improves quality.
Generally, sources of error arise from preanalytical, analytical, or post-analytical phases that can create variation throughout the microscopic process. These errors may be random or systematic. A comprehensive quality program considers these sources of error and contains strategies to detect and correct process-related errors.
This can prove difficult with manual judgment based procedures such as preparing, staining, reading, and grading peripheral blood smears. Common sources of error include:
- Preanalytical: collection, storage, and labeling errors; preparation and
staining variation of wedge smears.
- Analytical: instrument measurement errors e.g. incorrectly flagging a sample for microscopic review or giving the bench technologist incorrect information; slide artifacts and errors in estimating numbers of cells.
- Post analytical: calculation, reporting, and grading variation can create significant errors.
Reporting conventions e.g. qualitative grading criteria vary between laboratories. A physician may receive a report from Lab A, for example, that describes “2+ anisocytosis” and compare it to a report from Lab B on the same patient that describes “Moderate macrocytes” and “Few microcytes.” Which is correct?
Each lab thinks it’s correct, but thinking itself is error-prone. Policies and procedures can describe cellular morphology, counting techniques, and how to grade the number of cells seen, but the reality of bench work is subject to cognitive bias that introduce error. This is different from instrument bias considered separately or as part of total analytical error. A cognitive bias is inherent in how we think and process information.
Cognitive biases are common to all human beings, including a wide range of effects based on heuristics and social experience. They are useful shortcuts in thinking, but biases related to scientific measurement can skew results.1 They are distinct from cultural, organizational, or self-interest bias. Even when we are aware of a cognitive bias that leads to an error in thinking or judgment it can be difficult to overcome, like a persistent optical illusion.2
Here are a few cognitive biases that affect microscopic bench work:
- Confirmation bias – a tendency to seek or interpret information in a way that supports one’s preconceptions or beliefs; for example, if a hemogram shows a low MCV, reporting greater numbers of microcytes.
- Bandwagon effect – a tendency to believe what a group thinks e.g. groupthink. This can affect interpretation of cellular morphology when a number of colleagues evaluate a smear.
- Anchoring effect – a tendency to rely too heavily on a past reference when making decisions; for example, a technologist may see blasts on a smear because of a history of leukemia or previous reports of blasts.
- Clustering illusion – a tendency to see patterns where none exist, such a rouleaux or platelet clumping in different sections of a wedge smear feathered edge.
All laboratories are also vulnerable to déformation professionnelle, translated from French as “conditioning by one’s job,”4 a common tendency to process information from one’s profession rather than a broader perspective.5 A laboratory may believe, for example, that it is important to distinguish degrees of grading criteria for red blood cell morphology, such as rare, few, 1+, etc. when a physician prefers a broader interpretation.
Simplify to Reduce
Shewhart and Deming described variation arising from two sources, common and special causes. Common causes contribute to the background noise variation in value distribution, while a special cause is unexpected, random in nature, and related to a defect.6 Common causes are not considered unusual and are usually eliminated by upgrading equipment or a process.
For example, our cognitive tendency to see patterns in red cell morphology is natural and leads to predictable errors in judgment between laboratories and technologists. This is a common cause. A special cause — also called assignable — is related to a specific event, such as a change in pH that affects stain quality. Other special causes include sampling and instrument measurement errors that can be discovered when a smear is reviewed e.g. a platelet count is critically low but appears adequate on a smear.
Upgrading hematology instrumentation is a sensible strategy for reducing common cause of variation, but microscopic work remains a highly variable, manual process. To reduce common cause microscopic variation, simplify. Here are suggestions:
- Eliminate needless work. Unnecessary microscopic reviews add to cumulative variation and increase decision fatigue on the bench. Consider adopting the International Society for Laboratory Hematology (ISLH) Consensus Rules7 to optimize work at the scope.
- Upgrade instrumentation. Modern instruments that perform a six-part differential count, immature platelet fraction, nucleated red blood cell count, and other advanced parameters can enhance or eliminate the need for a smear review.
- Utilize instrumentation. Manual differential counts are inherently less precise than automated counts that enumerate thousands of cells. Consider reporting only abnormal populations with automated differential counts e.g. an automated ANC along with a raw percentage of metamyelocytes seen on the smear.
- Standardize a microscopic field of view. It’s easy to over- or underestimate a field size of evenly dispersed red cells; consider standardizing this estimation by counting cells along a portion of the circumference (the “area” of this circle in RBCs = circumference2 / 4?).
- Simplify criteria. Grading criteria that are too confusing, specific, or irrelevant to physicians add work and variation; consider reducing categories or using a single cutoff for clinically significant numbers of cells e.g. greater than 10 per field.
Upgrading your process will help reduce common causes of variation. While cognitive bias cannot be eliminated, discussing this on the bench with techs will help raise awareness to its effects. And discussing these changes with physicians adds value to your reports, leading to better patient care.
- ScienceDaily. Cognitive bias. 2014. Available at: http://www.sciencedaily.com/articles/c/cognitive_bias.htm. Accessed October 26, 2014.
- Cia.gov. Chapter 9 – Central Intelligence Agency. 2007. Available at: https://www.cia.gov/library/center-for-the-study-of-intelligence/csi-publications/books-and-monographs/psychology-of-intelligence-analysis/art12.html. Accessed October 26, 2014.
- Rationalwiki.org. List of cognitive biases – RationalWiki. 2014. Available at: http://rationalwiki.org/wiki/List_of_cognitive_biases. Accessed October 26, 2014.
- Collinsdictionary.com. English Translation of “déformation professionnelle” | Collins French English Dictionary. Available at: http://www.collinsdictionary.com/dictionary/french-english/d%C3%A9formation-professionnelle. Accessed October 26, 2014.
- HRZone. Available at: http://www.hrzone.com/hr-glossary/deformation-professionnelle-definition. Accessed October 26, 2014.
- Micquality.com. SIX SIGMA Glossary: Common and Special Cause Variation. Available at: http://www.micquality.com/six_sigma_glossary/common_cause_variation.htm. Accessed October 26, 2014.
- Islh.org. International Society for Laboratory Hematology. 2012. Available at: http://www.islh.org/web/consensus_rules.php. Accessed October 26, 2014.