The Limits of Interpreting the qEEG and the Issue of Probability
What You Will Learn in This Chapter
Every test has limits, and the qEEG is no exception. The temptation in this field is to treat a colorful topographic map as though it settled a question, when in fact it has only contributed one line of evidence to it. Knowing where your method stops being informative is not a weakness in your practice. It is the thing that makes the rest of your practice defensible.
This unit examines two kinds of limits. The first is the limit imposed by your choice of normative database, because a database is not a neutral yardstick. It embodies decisions about who counts as normal, which features were normed, how the data were recorded and filtered, and which statistical tests were applied. Choose a database that cannot image coherence and you will not see a connectivity problem, however clearly it is there.
The second limit is conceptual, and it trips up experienced clinicians as often as new ones. A finding can be statistically significant and clinically meaningless, or clinically important and statistically unremarkable. You will learn to hold statistical probability and clinical probability apart, to say which one you are talking about, and to resist the pull of a two-standard-deviation deviation that makes no difference to the person in front of you.
IQCB Blueprint Coverage: This unit addresses IX. Clinical Practice/Forensic, specifically A. Knowledge regarding the limits of interpreting the qEEG regarding the choice of reference databases and recognizing statistical probability versus clinical probability.
Learning Objectives
After completing this section, you will be able to:
Explain how the choice of reference database shapes the interpretation of qEEG data, including connectivity patterns and network properties.
Describe the factors to weigh when selecting a normative database for a specific referral question.
State why a clear referral question must precede the choice of method and database.
Identify the conditions for which qEEG findings can be integrated with data from other sources.
Explain why qEEG findings cannot serve as the sole basis for assessment, diagnosis, treatment formulation, or outcome evaluation.
Compare the temporal and spatial resolution of the qEEG and describe what it cannot image.
Distinguish statistical probability from clinical probability and explain why one does not imply the other.
Click the button below to hear the lecture over the Clinical Practice and Forensic unit.
Choosing a Reference Database
The choice of reference database can significantly impact the interpretation of qEEG data, influencing connectivity patterns and network properties. Each type of normative database poses its own special requirements and details, which can affect the interpretation of qEEG data (Thatcher & Lubar, 2023).
What should one consider when choosing a normative database? These facts are also reviewed in qEEG Tutor's section on normative databases. In addition to those factors, it is important to consider the referral question and client-specific details when selecting a normative database.
For example, when assessing a client with autism spectrum disorder, where connectivity is often an issue, it is important to select a normative database that includes measures of coherence. When assessing clients who have posttraumatic stress disorder, where the default mode network may show dysrhythmias, it is important to select a database that images that network, if not the structures involved.
Critical issues include the concept of normalcy, norming of qEEG features, and validation of clinical findings, which are essential for accurate interpretation. Technical aspects such as recording methods, filter use, frequency resolution, and statistical testing are crucial for the construction and use of these databases (Johnstone et al., 2005).
The choice of reference database and electrode significantly impacts qEEG interpretation, affecting both connectivity patterns and network properties. Consistent reference schemes and careful consideration of normative database features are essential for reliable and accurate qEEG evaluations.
A referral arrives asking whether a 12-year-old on the autism spectrum shows evidence of atypical connectivity. You could run the assessment against a database that norms absolute and relative power beautifully and reports nothing unusual, then tell the referring pediatrician that the qEEG was unremarkable. That would be a true statement about the wrong question. If the database you selected does not norm coherence, it cannot answer a connectivity question, and its silence is not evidence of absence. The database has to be chosen after the referral question is clear, never before.
What the qEEG Can and Cannot Answer
All tests have limits, as does the qEEG. Further, the various normative databases used for qEEG each have particular strengths and limitations that can affect the interpretation of qEEG data (Thatcher & Lubar, 2023). Understanding these strengths and limitations depends on formulating a clear referral question.
With a clear referral question, the qEEG provider can then decide whether qEEG methods are generally appropriate and whether a given normative database suits the specific needs of addressing the referral question. Suppose qEEG is a valid method for assessing the type of client who has been referred. In that case, the practitioner then essentially asks which of the available normative databases is best suited to the details of the referral question, given their respective strengths and limitations.
Understanding the limitations of interpreting qEEG findings and the questions it can answer is crucial. Livint Popa and coauthors (2020) offer a comprehensive review of diagnoses where qEEG findings can be integrated with data from other sources. For instance, qEEG findings can be instrumental in diagnosing conditions like epilepsy, stroke, dementia, traumatic brain injury, and mental health disorders such as ADHD, autism spectrum disorder, anxiety, and depression. Koberda (2021) further explores the benefits of qEEG in the assessment of dementia and traumatic brain injury.
Other uses of qEEG are reviewed in qEEG Tutor's normative databases section. However, qEEG is limited in that it cannot serve as the sole type of information on which to base an assessment, diagnosis, treatment formulation, or outcome evaluation.
Rather, qEEG findings should be integrated with information from other sources. Doing so helps to guard against assigning large importance to isolated findings that are statistically significant without meaningfully correlated findings of clinical significance.
Although the qEEG's strength is its temporal resolution, its spatial resolution is weak compared to MRI. Further, there are CNS structures, particularly subcortical structures, that qEEG cannot image. Other strengths of qEEG are its ability to image connectivity, including effective connectivity, in many CNS networks, and its ability to demonstrate activity outside well-defined norms with respect to multiple metrics.
The referral question comes first, the method second, and the database third. A normative database is a set of decisions about normalcy, recording, filtering, frequency resolution, and statistics, and those decisions determine what your assessment is capable of detecting. The qEEG has excellent temporal resolution, weak spatial resolution, and no view at all of many subcortical structures. It is never sufficient on its own for assessment, diagnosis, treatment formulation, or outcome evaluation.
Statistical Versus Clinical Probability
A common dilemma for clinicians who rely on scientifically informed findings is understanding the difference between the statistical significance of a finding and its clinical significance. One does not imply the other, and both must be considered when making reports or other statements and clinical decisions. Statistical and clinical significance can also be reframed as statistical and clinical probability.
Statistical probability is fundamentally concerned with the likelihood of an event occurring based on mathematical principles and data analysis. It is often used to determine the significance of research findings through hypothesis testing and the calculation of p-values (Stratton, 2018).
Statistical probability involves hypothesis testing, where a null hypothesis of no effect is compared against an alternative hypothesis in which an effect is present. The p-value indicates the probability that the observed data would occur if the null hypothesis were true. A smaller p-value, typically at or below 0.05, suggests that the observed effect is statistically significant, meaning it is unlikely to have occurred by chance.
Another key aspect of statistical probability is the use of confidence intervals, which provide a range within which the true value of an effect is likely to lie. A 95% confidence interval, for example, means there is a 95% probability that the true effect size falls within the specified range. This helps in understanding the precision and reliability of the estimated effect.
Clinical probability, on the other hand, is more focused on the practical implications of research findings in a healthcare setting. It considers the real-world impact of an intervention or treatment on patient outcomes.
The clinical approach is often deterministic and causal, emphasizing diagnosis and treatment over mere prediction. It is less concerned with the statistical significance of findings and more with their practical relevance and applicability in clinical practice (Einhorn, 1986).
Clinical probability assesses the magnitude and importance of an effect in a real-world context. An effect that is statistically significant may not always be clinically significant. For instance, a small effect size that is statistically significant might not have meaningful implications for patient care. Conversely, a clinically significant finding may not always achieve statistical significance but could still be crucial for treatment decisions (Stratton, 2018).
Statistical probability answers the question of whether a difference is likely to be chance. Clinical probability answers the question of whether the difference matters to this person. A two-standard-deviation z-score is a statistical statement, and by itself it licenses no clinical conclusion. Report which kind of claim you are making, and never let the arithmetic of a database stand in for judgment about the client.
Check Your Understanding
- Explain how the choice of normative database can determine whether a connectivity abnormality is detected at all.
- Why must the referral question be formulated before the database is selected rather than after?
- List the technical aspects of database construction that affect interpretation, and explain why each matters.
- Contrast the temporal and spatial resolution of the qEEG, and name a class of structures it cannot image.
- Give an example of a finding that is statistically significant but clinically insignificant, and one that is clinically important but statistically unremarkable.
Cutting-Edge Topics in qEEG Research
Normative Databases Are Being Held to Published Standards
Thatcher and Lubar (2023) traced the history of scientific standards for qEEG normative databases and made the criteria explicit: how normalcy is defined, how features are normed, how the recording and filtering were done, and how the statistics were validated. The practical consequence for you is that a database is now something you can evaluate rather than simply trust. Expect referral sources, and opposing counsel, to start asking which database you used and why.
Reviews Are Mapping Where the Method Actually Contributes
Livint Popa and colleagues (2020) reviewed the role of qEEG across neuropsychiatric diagnoses, and Koberda (2021) examined its contribution to early cognitive change in dementia and traumatic brain injury. Work of this kind is gradually replacing the older habit of claiming the qEEG is useful in general. The direction of travel is toward condition-specific statements about what the method adds, which is exactly the form a defensible clinical claim needs to take.
The Statistical-to-Clinical Translation Problem
Stratton (2018) restated for clinicians a problem the wider health sciences have been wrestling with for years: significance testing was never designed to tell you whether a finding matters. Effect sizes, confidence intervals, and minimal clinically important differences are steadily displacing the bare p-value in medical reporting. The qEEG literature is beginning to follow, and reports that give a z-score without a statement of clinical relevance are starting to look dated.
Assignment
Now that you have completed this unit, describe a referral question you might realistically receive, and explain which normative database features would be necessary to answer it. Then describe a qEEG finding that would be statistically significant in that client but clinically insignificant, and explain how you would report it.
Glossary
clinical probability: the practical relevance and applicability of a finding in clinical practice, assessed in terms of the magnitude and importance of an effect in a real-world context rather than its statistical significance.
confidence interval: a range of values so defined that there is a specified probability that the value of a parameter lies within it.
hypothesis testing: a method of statistical inference used to decide whether the data at hand sufficiently support a particular hypothesis.
null hypothesis: the hypothesis that there is no significant difference between specified populations, any observed difference being due to sampling or experimental error.
p-value: a statistical measurement used to validate a hypothesis against observed data. A p-value measures the probability of obtaining the observed results, assuming the null hypothesis is true. The lower the p-value, the greater the statistical significance of the observed difference.
statistical probability: the likelihood of an event occurring based on mathematical principles and data analysis, determined through hypothesis testing and the calculation of p-values.
Test Yourself on ClassMarker
Click the button below to take a 10-question exam over this entire unit. There is no password.
Review Flash Cards on Quizlet
Click the button below to review our chapter flash cards.
References
Einhorn, H. (1986). Accepting error to make less error. Journal of Personality Assessment, 50(3), 387-395. https://doi.org/10.1207/s15327752jpa5003_8
Johnstone, J., Gunkelman, J., & Lunt, J. (2005). Clinical database development: Characterization of EEG phenotypes. Clinical EEG and Neuroscience, 36(2), 99-107. https://doi.org/10.1177/155005940503600209
Koberda, J. L. (2021). QEEG as a useful tool for the evaluation of early cognitive changes in dementia and traumatic brain injury. Clinical EEG and Neuroscience, 52(2), 119-125. https://doi.org/10.1177/1550059420914816
Livint Popa, L., Dragos, H., Pantelemon, C., Verisezan Rosu, O., & Strilciuc, S. (2020). The role of quantitative EEG in the diagnosis of neuropsychiatric disorders. Journal of Medicine and Life, 13(1), 8-15. https://doi.org/10.25122/jml-2019-0085
Stratton, S. (2018). Significance: Statistical or clinical? Prehospital and Disaster Medicine, 33(4), 347-348. https://doi.org/10.1017/S1049023X18000663
Thatcher, R., & Lubar, J. (2023). History of the scientific standards of QEEG normative databases. In Introduction to quantitative EEG and neurofeedback (3rd ed., pp. 121-141). Academic Press.