Authors: Sonia Twigg, Andy Little, Moises Gallegos
Editor: Sara Krzyzaniak, MD

A Case

It’s a nightshift in your busy ED. The ED is pumping and Sarah, a senior resident, is running the pod. She has seen sick patient after sick patient and knows there are still long waits for patients in the waiting room. She is going as fast as she can and juggling as many patients as possible. 

She is seeing Mr. Smith, her fifth patient with chest pain this shift. Mr Smith is in his 60s and has a typical presentation for ischemic chest pain: heavy central chest pressure for the past 30 minutes. He is hemodynamically stable and his pain improves after aspirin and sublingual nitro. His ECG shows ST depressions in V1-V3. Sarah makes sure she follows the department’s chest pain protocol including orders for serial troponins and ECGs, as well as sublingual nitro as needed. She expects Mr Smith has had a NSTEMI. 

Another critical patient arrives, so she drops what she is doing to go and care for that patient. 30 minutes later she is called to Mr. Smith’s bedside by his nurse who is concerned that he is looking worse. Mr. Smith is diaphoretic, tachycardic, and has become hypotensive. Reviewing the initial ECG, her attending points out that the ST depressions in the anterior leads are concerning for a posterior STEMI. The attending orders a posterior lead ECG which demonstrates ST elevations in V7-V9, confirming her suspicion. 

Sarah is upset about this missed diagnosis and the delay in care that it created. She wonders if she would have been able to arrive at the correct diagnosis if she had not been distracted by her other patients and frequent interruptions.

Background

Cognitive Load Theory builds on Atkinson and Shiffrin’s model for human memory.³ Sensory inputs enter into our working memory, and we organize (“chunk”) this information into “schemas” that are then stored in long term memory. Working memory is limited. At any given time we can hold 7+/- 2 items in our working memory for only 30 seconds. Long term memory is infinite, but we have to retrieve the schemas back into working memory when we need them. Working memory becomes the bottleneck for learning. Over time, and with repeated experiences, specific schemas become more complex, organized and eventually automated — this is expertise.⁴ A single schema, however complex, counts as one item in working memory. The premise of CLT is that the processes of learning may be different from the processes required to complete a task. If working memory is spent on tasks that do not contribute to the development of schemas from working memory to long-term memory, learning cannot occur.

Intrinsic load describes the cognitive demand of the task itself.² Intrinsic load is affected by task complexity and the prior knowledge of the learner. Complex tasks, with a high number of elements or highly interactive elements, impose a higher intrinsic load than simple tasks.³ Let’s use the example of analyzing an ECG. Rare or subtle ECG findings are harder to pick up than obvious ones. Novice learners experience a higher intrinsic load than experienced learners doing the same task. An experienced senior resident would be expected to analyze the same ECG faster and more accurately than a medical student.

Extraneous load is the extra information we experience that is not necessary to complete the task yet uses up part of our precious working memory.² It’s what gets in the way of doing the task: the buzzer that goes off while you are reading the ECG, the interruption with a question, the poor print out, and the time pressure of the busy ED. Each of these work to increase your cognitive load and make reading the ECG harder. But it’s not always bad! Sometimes the extra information helps you elaborate the schema. For example a quick review of STEMI mimics from an online reference before reading the ECG may help you make the correct interpretation of the ECG.

Germane load describes the effort associated with learning. It is the load we experience from processing the information in our working memory into (or out of) the schemas we store in long term memory.³ It is closely linked with intrinsic load, but it relates more to the attention devoted to learning rather than the effort dedicated to actually performing the task (i.e. intrinsic load). 

Modern takes on this Theory

In 2010 Van Merrienboer and Sweller offered principles to guide instructional design in medical education.⁴ Fraser et al. elaborate on them in their 2015 review of how CLT applies to healthcare simulation.⁵

Strategies to manage intrinsic load:

• Sequencing effect: present simple concepts first followed by more complex ones.
• Segmenting effect: divide a task into manageable “chunks” and once schemas have been formed for each, combine them to perform the whole task (i.e. create one schema).
• Pre-training effect: teach components of a task before the whole task. For example in a healthcare simulation, do a pre-brief on how the mannequin and simulation monitors work.
• Low to High Fidelity effect: start with a low fidelity environment and build up.

Strategies to decrease extraneous load:

• Goal free effect: specify tasks without a single answer, e.g. rather than asking “What is this patient’s diagnosis?” one should ask “List as many differential diagnoses you can for this presentation.”
• Worked example effect: Demonstrate to the learner how to perform a task.
• Completion effect: offer partially completed tasks, e.g. “Please finish this management plan.”
• Split attention effect: don’t split the learner’s attention, e.g. don’t use a slide graphic whose key is on another slide or on a handout
• Modality effect: use both visual and auditory modes when presenting information.
• Redundancy effect: don’t give redundant info. Replace multiple sources with a single source of information.

Strategies to optimize germane load:

• Variability in practice effect: offer tasks that provide variation on a theme, e.g. read many different types of STEMI ECGs.
• Contextual interference effect: give a series of tasks that use different skills (high contextual interference).
• Self-explanation effect: prompt the learner to “self-explain”, e.g. “Write down why you came to this diagnosis.”

Cognitive load & instructional design:

Leppink describes three dimensions that help guide instructional design⁶:

• Task fidelity
• Task complexity
• Instructional support

He recommends that fidelity and complexity should be gradually increased while ensuring adequate instructional support, and then ultimately fade the instructional support when it is no longer needed.

These principles help explain the “expertise-reversal effect”. As the learner becomes more advanced, techniques that previously augmented learning for the novice actually start to interfere. For example, asking an expert to use an unfamiliar mnemonic for a task they already do well provides too much instructional support which increases extraneous load.

A fundamental characteristic of health professions is the collaborative nature of care. Research has begun to explore the application of CLT to group-based learning and teamwork. Kirschner et al have promoted the “interaction hypothesis” to explain the differences in learning efficiency for groups and individuals.^7,8 They discuss the impact of utilizing a combined, or collective, working-memory, in the development of schema. In a group setting, cognitive load can be distributed to the working memory of various team members for processing. Large quantities of information are serially distilled into workable inputs that team members can then use to form schema. This can be useful when dealing with high complexity tasks that would otherwise overload an individual. However, with low complexity tasks, distribution of cognitive load may leave individuals without enough effortful processing of information to allow for appropriate schema development. It becomes “a trade-off between the group’s advantage of dividing information processing amongst the collective working memories of the group members and its disadvantage in terms of associated costs of information communication and action coordination.”⁷

Fraser et al have begun to look at how emotion affects cognitive load.⁵ They point out that the activation of emotion is inevitable in healthcare simulation, among other situations in medicine. Emotion can increase extraneous load, but it is sometimes an important component of intrinsic load, e.g. when learning to break bad news. Positive emotions appear to improve learning, perhaps through improving motivation and attention or through improving problem solving and creativity. Stress appears to have varying effects; it can focus attention and improve schema formation for the task at hand, but it also decreases working memory and can impair retrieval from long term memory. Healthcare teams have been shown to increase performance when feeling psychologically safe, so if psychological safety is present, “academic emotions,” such as performance anxiety and motivation to learn, can be optimized.

Other Examples of Where this Theory Might Apply

The argument can be made that the Flipped Classroom model for teaching benefits from application of CLT by creating protected learning time through removing it both temporally and spatially from preparatory time.⁹ The inherent separation of preparatory work from active application creates a focus on germane load during in-person sessions and off-loads intrinsic and extraneous processes. The design principles proposed by Van Merrienboer and Sweller to improve the balance of cognitive load in health professions education can be applied to optimize the preparatory material and in-person sessions that create a flipped learning experience.⁴

Simulation-based learning forms an important tenet of skill development and critical action decision making. However the complexity of cases and skills taught through simulation may result in too high a cognitive load. CLT can be applied to simulation curricula for improved results.⁵ While running a simulated pediatric resuscitation, a trainee may become overwhelmed trying to remember medication dosages and lose track of the experiential goals for the session. Providing a Broselow tape, or better yet a confederate such as a pharmacist, would allow the learner to focus on building skills as a team leader.

CLT has been applied to performance in medicine as well as learning. Sewell examined cognitive load as it applies to performing a colonoscopy.¹⁰ The same group has also looked at how cognitive load impacts patient handovers.

Limitations of this Theory

De Jong criticized CLT by pointing out that while it describes a cognitive basis for instructional design principles, it is “impossible to falsify” because CLT relies on post-hoc assumptions.¹² If the load imposed by the task interferes with our schema construction, we think it’s bad and label it as extraneous load. If it helps us construct schema, then we think it’s good and label it as germane load. In other words, we make the evidence fit the theory.

Perhaps we could justify this labeling if we could directly or accurately measure cognitive load. Current measures include:

Psychometric testing. Paas developed and validated a single-item scale, which is the most commonly used measure.¹³ In contrast the NASA-Task Load Index (NASA-TLX) is a multi-item scale.¹⁴ Unfortunately these are self reported measures that occur after the task has finished. They may not reflect the load during the task, and they only measure overall cognitive load.

Secondary task response. The speed and accuracy of performing a secondary task, for example giving a medication order while reading the ECG. While you are concentrating on that complex ECG, you might start mumbling or pausing while giving the medication order.

Physiologic measures. Heart rate or respiratory rate variability, EEG or EMG monitoring, eye tracking, pupillary diameter, blink frequency, serum adrenaline levels, brain imaging and skin conductance have all been examined and validated. Their supposed advantage is that they can measure cognitive load instantaneously and continuously over time.

Performance on the task.³ Examining how well is the task performed, e.g. number of errors made during task, can provide insight into how much cognitive load was involved in the performance.

It is arguable whether these are valid or accurate measures. What are they really measuring? While they are general measures of cognitive load, most do not measure its constituent parts of intrinsic, extraneous and germane load. In other words, they can tell us that cognitive load affects performance but not how the different parts interact.

In 2013, Leppink developed a psychometric instrument to differentiate between different types of cognitive load.¹⁵ However in a subsequent review, it was noted that some studies did not support the instrument described.⁶ The authors of the review suggested that we go back to a two factor framework and think of cognitive load as consisting of only intrinsic and extraneous load along with a “subjective judgement of learning.” Young and Sewell disagree and have derived instruments to measure the different types of load in medical settings including colonoscopy¹⁰ and patient handovers.¹⁶ In 2016, Naismith developed the Cognitive Load Component (CLC) psychometric test and compared it with the Paas and NASA TLX tests.¹⁷ This interesting but small study showed little agreement on total cognitive load between the three tests. However intrinsic load seemed consistently measured by all three. Intrinsic load on the CLC correlated with the NASA TLX subscales of mental demand and frustration.

CLT makes intuitive sense, but we have more research to do to demonstrate how it applies to instructional design and workplace performance.

Returning to the case…

After her whirlwind shift, Sarah is finally able to catch her breath and reflect on Mr. Smith’s case. Luckily he was taken to the cath lab 20 minutes after discovering the posterior MI where emergent PCI and stenting of a 90% occluded RCA occurred without complications and he was transferred to the CCU in stable condition.

She recognizes that as a senior resident she should be more capable of identifying this common ECG pattern. She plans to spend some time on her day off studying ECGs (decrease intrinsic load) and will set a systematic approach for when she is handed ECGs on shift making sure she pauses for each (decrease extraneous load).

References

1. Sweller J. Cognitive load during problem solving: Effects on learning. Cognitive Science. 1988;12:257-285.
2. Sewell JL, Maggio L, ten Cate O et al. Cognitive load theory for training health professionals in the workplace: A BEME review of studies among diverse profession: BEME guide No. 53. Medical Teacher. 2019;41(3):256-270.
3. Young JQ, Van Merrienboer J, Durning S et al. Cognitive load theory: implications for medical education: AMEE Guide No. 86. Medical Teacher. 2014;36:371-384.
4. Van Merrienboer JJ and Sweller J. Cognitive load theory in health professions education: Design principles and strategies. Med Educ. 2010;44:85-93.
5. Fraser KL, Ayres P, Sweller J. Cognitive load theory for the design of medical simulations. Sim in Healthcare. 2015;10(5):295-307.
6. Leppink J and Van den Heuvel A. The evolution of cognitive load theory and its application to medical education. Perspect Med Educ. 2015;4:119-127.
7. Kirschner F, Paas F, Kirschner PA. Individual and group-based learning from complex cognitive tasks: effects on retention and transfer efficiency. Computer Hum Behav. 2009;25:306-314.
8. Kirschner F, Paas F, Kirschner PA. Task complexity as a driver for collaborative learning efficiency: The collective working-memory effect. Appl Cogn Psych. 2011;25: 615-624.
9. Abeysekera L and Dawson P. Motivation and cognitive load in the flipped classroom: Definition, rationale and a call for research. Higher Ed Res and Dev. 2015;34(1):1-14.
10. 10. Sewell JL, Bocardin CK, Young JQ. Measuring cognitive load during procedures skill    training with colonoscopy as an exemplar. Med Educ. 2016;50:682-692.
11. 11. Kirschner, PA, Sweller J, Kirschner F, Zambrano R J. From cognitive load theory to collaborative cognitive load theory. Int J Comp Supported Collab Learn. 2018;13(2):213-233.
12. 12. De Jong T. Cognitive load theory, educational research, and instructional design: Some food for thought. Instr Sci. 2010;38:105-134.
13. 13. Paas F. Training strategies for attaining transfer of problem-solving skill in statistics: A cognitive-load approach. Journal of Educational Psychology. 1992;84:429-434.
14. 14. Hart SG, Staveland LE. Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. Advances in Psychology. 1998;52:139-183.
15. 15. Leppink J, Paas F, Van der Vleuten C et al. Development of an instrument for measuring different types of cognitive load. Behav Res. 2013;45:1058-1072.
16. 16. Young JQ and Sewell JL. Applying cognitive load theory to medical education: Construct and measurement challenges. Perspect Med Educ. 2015;4:107-109.
17. 17. Naismith LM, Cheung JJH, Ringsted C, Cavalcanti R. Limitations of subjective cognitive load measures in simulation based procedural training. Med Educ. 2015;49:805-814.