There are a number of factors that make us fundamentally human. We eat. We sleep. We experience emotions. And unfortunately at some point, we all become sick. Where we tend to diverge is how we process and treat our sicknesses. Some people run immediately to the doctor’s office, or if you are anything like me, you think NyQuil cures all ailments of the body, and therefore, you self-treat.

Regardless of our methods of treatment, one question that remains is how do we develop explanations and diagnoses for sickness based on the presented symptoms?

Diagnosing illness can be quite tricky (just try entering your symptoms into an online symptom checker – Fair warning, symptom checkers have low accuracy, with the correct diagnosis being listed within the top 20 options only 58% of the time) .

We learn to employ different strategies for diagnosing our illnesses. We may rely on the order in which symptoms present themselves. For example, if a stomach ache is my first symptom followed by fatigue and a headache, I might believe I have food poisoning before I think I have the flu.

Alternatively, we may use prior familiarity or environmental prevalence of certain illnesses to reason about diagnoses. For instance, if I have a runny nose and red eyes during the first few weeks of spring, I may quickly deduce that I have allergies as opposed to a cold.

Regardless of the method we choose, a correct diagnosis is of the utmost importance, especially when illnesses are on the rise, like this 2018 flu season. You cannot treat an illness unless you know what it is.

According to a recent online article posted on CNN.com, disease specialists with the Center of Disease Control (CDC) and the US National Institute of Allergy and infectious Diseases have reported that this flu season rivals the worst in recent years. Especially in the U.S. where there has been a seasonal total of over 126,000 confirmed cases of the flu, a sharp increase in pediatric death caused by influenza-related symptoms, and significant antiviral shortages, have been observed.

Shocking, isn’t it? This should be even more incentive to get our diagnosing correct.

This, however, begs the question, what factors play a role in developing diagnoses for symptoms?

In a recent study published in the Psychonomic Bulletin and Review, Agnes Scholz and colleagues used eye tracking to evaluate the role of “memory indexing” (a novel method‒observing eye movements while individuals solve memory-based cognitive tasks) and eye gaze in symptom processing. The goal of the study was to explore memory processes during diagnostic reasoning involving ambiguous symptoms. Ambiguity results when symptoms can support two or more diagnostic hypotheses (e.g. vomiting, dizziness, and fatigue can support both the flu and food poisoning).

The study was carried out in two phases: a learning phase and a reasoning phase. During the learning phase, participants saw four squares of information containing symptom classes that corresponded to each of 4 chemicals (see the left panel of the figure below that explains the task) and learned how symptoms were assigned to symptom classes, and then how symptoms related to chemicals. For example, when participants heard the symptom ‘sting’, they learned to associate it with the symptom class ‘pain’ which was caused by chemical A (see task figure) which appeared in the top left quadrant.

During the reasoning phase, the four quadrants appeared blank (right panel of the figure below). While viewing the emptied screen, participants were presented with four symptoms heard audibly (one at a time).  Participants’ task was to indicate which chemical most likely caused the symptoms by denoting the quadrant for that chemical (remember, during training they learned to associate each chemical with a specific quadrant based on the symptom classes presented on the screen). In other words, participants needed to remember which symptom classes belonged to which chemical and in which quadrant the chemical was represented.

Based on this methodology, Scholz and colleagues made several predictions: 1. If eye gaze reflects memory retrieval in ambiguous diagnostic scenarios, eye movements should correspond to the outcome of the reasoning process. In other words, participants should look longer and have more eye movements directed toward the quadrant of the chemical (based on memory) they believe is most likely to have caused the symptoms. 2. If participants are strongly influenced by the presentation of the first symptom, they will fixate longer than expected by chance alone on the chemical that can cause the first symptom. 3. If participants change their hypothesis, they will fixate most often on the alternative hypothesis relative to the hypothesis supported by the symptom sequence. 4. The proportion of fixations toward a participant’s final choice will increase toward the end of the symptom reasoning trials and will be highest during the response trial.

Scholz and colleagues report a number of noteworthy results:

1. Participants were relatively good at selecting (and remembering) the hypothesis that was most supported by the presented symptoms.
2. When multiple hypotheses were supported, participants chose the option with at least two symptoms presenting from the same class instead of alternative options where the symptoms belonged to classes associated with two or more chemicals.
3. The proportion of gaze fixations toward a hypothesis (e.g., chemical A) was greater when symptoms supporting it were presented earlier on the symptom list. For instance, suppose chemical A was a known skin irritant, the earlier that symptoms supporting a skin irritant appeared on the symptom list, such as rash, redness, itchiness, the longer people gazed at the location for chemical A. However, if there were other symptoms on the list (presented later on) that strongly favored a different hypothesis (e.g., symptom favoring a digestive irritant), fixations toward the new hypothesis increased by the end of the symptom interval. In this way, gaze data can predict the leading hypothesis as well as a hypothesis change.

In a nutshell, gaze data can predict diagnostic responses as evidenced by the finding that the hypothesis that has garnered the most fixations by the end of the trial is most likely the diagnostic response. Gaze data can also predict the formulation of a leading hypothesis, as people gaze longer at the hypothesis that is supported by the symptoms presented earliest on the list and if there is a sudden change in hypothesis due to new symptoms, people start to gaze more at the new leading hypothesis. A preference for gazing at one hypothesis over the others based on the amount of support given by the symptoms that appear earliest demonstrates a bias in symptom processing toward earlier symptoms.

So how do people cope with difficult diagnostic reasoning? They strive to find the most coherent explanation even if biased processing (up-prioritizing the presentation order of some symptoms) is necessary.

I bid you good luck in your diagnostic reasoning this flu season, you’re going to need it!

Featured Psychonomic Society article:

Scholz, A., Krems, J. F., Jahn, G. (2017). Watching diagnoses develop: Eye movements reveal symptom processing during diagnostic reasoning. Psychonomic Bulletin & Review, DOI 10.3758/s13423-017-1294-8.