Please note! This essay has been submitted by a student.
This report seeks to analyze the phenomenon of confirmation bias in medical decision-making, exploring the implications from patient to doctor. Confirmation bias connotes the seeking or interpreting of evidence in ways that are partial to existing beliefs, expectations, or a hypothesis in hand (Nickerson, 1998). Put another way, confirmation bias describes our tendency to apply more weight to confirming evidence than disconfirming evidence, to substantiate less evidence to confirm an initial idea and to interpret vague or ambiguous information in ways that align with our preexisting beliefs. There is a multitude of examples where confirmation bias supplies explanatory value. Facebook’s newsfeed, for example, is curated with information that aligns with users’ preexisting views. Confirmation bias also helps explain the proliferation of fake news on social media, as users are more likely to share false stories that are congruent to their views while discarding false stories that stand in opposition to these beliefs.
There are several reasons why this topic is relevant and important. In our everyday lives, we employ mechanisms – either implicitly or explicitly – to search for meaning and truth. Naturally, this makes us susceptible to cognitive biases, namely confirmation bias. Confirmation bias is ingrained in our cognition, and when determining the voraciousness of ideas, it can significantly shape the process of arriving at a conclusion and subsequent decision-making. This is particularly consequential in medical-decision making, as it could be the deciding factor on whether to see a doctor – the potential difference between either returning to good health or deteriorating to a crippling condition. It could also lead to overly exhaustive use of medical resources, because self-diagnosed patients may assume the worse and thus in constant need of medical attention, albeit only having a minor illness. This essay explores these potential implications by discussing relevant literature and how my study fits within this research.
Although confirmation bias can be found in a variety of sources, I chose to focus on its impact on medical decision-making, particularly how it impacts an individual’s decision to seek medical attention. This analysis will further explore how the effects of confirmation bias can be potentially compounded from patient to doctor. This approach differs from previous research conducted on confirmation bias in medicine, as the literature primarily focuses on confirmation bias from the perspective of the care provider, rather than from the patient’s perspective. This essay seeks to integrate both findings to paint a holistic picture of how confirmation bias influences modern medicine.
My first hypothesis was that people presented with a preconceived notion would be more likely to exhibit confirmation bias. My second hypothesis was that participants, who exhibited confirmation bias, would be less likely to seek medical attention. To test these hypotheses, I created two surveys – one for an experimental group and another for the control. For the experimental survey, I provided a list of symptoms and stated, “You think you have a sinus infection.” The next page then provided a list of WebMD results, offering the participant the option to choose which sickness they would like to obtain more information on. The next page then asked, “After reviewing the information provided, do you feel the need to see a doctor or other medical professional?” Participants then had to choose whether they were comfortable with the self-diagnoses or would like to see a doctor. The structure of the survey was the same for the control group, except instead of telling the participant they thought they had a sinus infection, I stated: “You aren’t sure what illness or disease may be affecting you.” In the experimental group, I supplied participants with the preconceived notion that they have a sinus infection. Consequently, I was able to examine how this affected the information they subsequently acquired. In doing so, I was able to capture whether participants sought to confirm their initial hypothesis, and thus determined which participants succumbed to confirmation bias. The last question asking whether they’d see a doctor was important, as it tested my second hypothesis and provided insight into how confirmation bias affected a participant’s subsequent decision to seek further medical attention
The survey had 64 participants, with 36 in the experimental group and 28 in the control group. Out of the 36 participants in the experimental group, 16 (~44%) sought confirmatory evidence, as in they chose to acquire more information on sinus infections. Of these 16 participants, 10 (~63%) were confident in their self-diagnoses and chose not to seek further medical attention. Only 6 participants in the control group sought additional information on sinus infections, highlighting how the initial information supplied affected their subsequent search. A total of 18 experimental participants chose to forego seeing a doctor, and of the 28 participants in the control group, 13 chose to forego seeing a doctor. In total, 31 participants chose not to seek further medical attention (~48% of all participants). My first hypothesis, testing whether the experimental group exhibited confirmation bias, was statistically significant (p-value = .021). My second hypothesis, testing if participants who exhibited confirmation bias were less likely to see a doctor, wasn’t statistically significant (p-value = .147).
My results indicate that participants, supplied with the preconceived notion of their illness, were likely to seek confirming evidence. And although the second hypothesis wasn’t statistically significant, the majority of participants, who exhibited confirmation bias, were confident in their self-diagnosis. For the participants who exhibited confirmation bias, this implies they may be less likely to seek further medical attention, indicating that confirmation bias may affect patient decision-making. More importantly, considering that nearly half of all participants were confident in their self-diagnosis, my study reveals that patients use WebMD as a vehicle to self-diagnose and as a justification to forego seeking medical attention. With the modern proliferation of information services – such as WebMD – my study indicates that patients are now more prone to confirmation bias, self-diagnosing, and ultimately, that these services can significantly influence their decision to see a doctor.
There is extensive literature theorizing how we form beliefs, exploring the process of determining whether ideas are either true or false. Gilbert (1992) proposed that acceptance of an idea is part of the automatic comprehension of the idea and the rejection of an idea occurs after, and more effortfully than, its acceptance (Gilbert, 1992). In other words, he argued that comprehension and acceptance are intertwined processes, and rejection is a proceeding, discrete process. He supported this argument with an experiment that provided participants with nonsensical statements, such as, “A Tarka is a wolf”. Simultaneously, he had participants remember a series of random numbers. He found that participants were susceptible to believing these false statements when they were deliberately remembering numbers. Participants also found it difficult to later disbelieve these false ideas. (Gilbert, Krull, & Malone, 1990). This supports the notion that acceptance is an automatic, seamless process, and when one is attending to other stimuli, it can inhibit one subsequent ability to falsify an idea. This notion is supported by the fact that falsifying propositions is one of the last linguistic capabilities acquired by children (Bloom, 1970). These findings further allude to the idea that apprehension and acceptance of ideas are entwined processes, and they are mutually exclusive from belief negation.
The implications of these ideas are twofold: first, the doctor – who is likely simultaneously dealing with an array of clients, all with different symptoms and diagnoses – may lack the cognitive resources needed to serve patients. Consequently, the doctor may be more susceptible to confirmation bias, as their ability to disbelieve their initial hypothesis could be impaired. This could detrimentally impact their proceeding diagnostic tests, their eventual diagnosis, and ultimately their recommended treatment plan. Second, from a patient’s perspective, potentially someone prone to self-diagnosing based on WebMD results, they could accept their preliminary diagnoses as a byproduct of trying to understand their illness. This could impact their decision on whether to seek medical help. And in the case that they do see a doctor, it may be difficult for them to disbelieve their preliminary self-diagnosis. This may impact their decision to follow through with the doctor’s recommended course of treatment. It also could mean the patient continues to seek out doctors who they think would be more inclined to agree with their self-diagnosis – a potential burden on care providers.
Empirical research has proven that medical decision-makers are not immune to confirmation bias. One study found that psychiatrists conducting a confirmatory search made a wrong diagnosis in 70% of the cases compared to 27% or 47% for a disconfirmation or balanced information search (Mendel, Traut-Mattausch, Jonas, Leucht, Kane, Maino, Kissling, & Hamann, 2011). This study involved both experienced psychiatrists and medical students. In the case of misdiagnoses, participants also recommended the incorrect course of treatment. Overall, 13% of the participants, who were experienced psychiatrists, conducted confirmatory searches and provided subsequent misdiagnoses (Mendel et al., 2011). Evidently, psychiatrists and medical students are prone to confirmation bias. Although only 13% of experts sought to confirm evidence, misdiagnosed, and provided incorrect treatment, one misdiagnosis can be fatal; hence, this number is alarming and problematic.
This study presents another interesting implication of confirmation bias in medical decision-making – the overuse of medical resources. In the process of sourcing confirmatory evidence, doctors are likely to carry out unnecessary laboratory tests to validate their initial hypothesis (Dawson, 2000). Chapman (2003) found that confirmation bias not only explains why physicians order superfluous diagnostic tests but also why doctors tend to impartially view the results (Chapman, 2003). Two doctors can draw diametrically different conclusions from equivalent diagnostic results (Dawson, 2000). Confirmation bias may be the underlying reason for these interpretations. Although the literature primarily focuses on excessive resource deployment from a provider’s perspective, patients may effectuate the same problem. When a self-diagnosing patient experiences a headache, for example, he or she may tenaciously presume they have a brain tumor. Consequently, they may insist that their doctor prescribe them an MRI – a test typically over $2000 (Glover, 2014). In both cases of confirmation bias – from patient to doctor – medical resources may be inefficiently deployed to justify initial hypotheses.
While there is extensive support for the notion that confirmation bias affects doctor’s decision-making, further research needs to explore this bias from a patient’s perspective. The two biases aren’t necessarily mutually exclusive. When seeking medical attention, patients, for example, can filter out and apply varying weight to information they feel is congruent with their self-diagnosis. Doctors then apply their own filter and evidence-weighting to already biased evidence. This may potentially compound the effects of confirmation bias and may have pronounced effects on the diagnosis, subsequent testing, and ultimate treatment plan. Further research needs to examine the interaction of this bias between patient and doctor.
There are several limitations to my study. First, given the inherent nature of surveying, participants’ responses may not be reflective of their actions in a real-world scenario. Secondly, it is difficult to ascertain whether other biases may have contributed to my results. In the experimental group, for example, the words “illness or disease” may have evoked vivid images of a recent friend or family member that was sick, and we may have inadvertently induced the ease of recall bias (Bazerman, 2012). This may have affected both the information they acquired and the subsequent decision to see a doctor. Also, a participant may have recently come across someone with a sinus infection or one of the listed illnesses, which could have affected their subjective probability assessment as well as the subsequent information they acquired.
In a technologically rich landscape, it is easy to validate a false hypothesis and leverage an abundance of information to reaffirm one’s position. This ceaselessly improving technological landscape may exacerbate the effects of confirmation bias at the patient level. My survey indicated that nearly half of my participants found the WebMD information sufficient and were confident in their self-diagnosis. This points to the fact that these information services may be compounding confirmation bias in patients, as they can easily validate their preliminary hypothesis. Evidently, this can impact their decision to seek medical attention, which could result in either untreated patients or overuse of medical resources. Doctors are also liable to falling prey to confirmation bias. Research has shown that confirmation bias can be at the heart of incorrect diagnoses and treatment recommendations. It has also illustrated that confirmation bias can be the precursor to inefficient resource deployment. Further research needs to analyze how the symptoms of confirmation bias may be magnified from patient-doctor interaction.