The bias of the question posed: a diagnostic “invisible gorilla”

Introduction

Diagnostic errors often result from physician’s susceptibility to cognitive biases [1–3]. Our understanding of biases largely derives from principles established by cognitive psychologists Daniel Kahneman and Amos Tversky [3]. These principles are applicable to reasoning and decision-making processes in many fields, including medicine. A large number of cognitive biases have been noted to lead to diagnostic errors in medical practice [2]. Self-reflection on potential biases (so-called “debiasing”) is advocated as part of the diagnostic process, in order to identify and compensate for biases before they lead to diagnostic errors [2, 4–6]. Familiarity with common, clinically-relevant biases is therefore a necessary step in avoiding them. I would like to describe and demonstrate a common, potent cognitive bias which may distract the physician’s attention from important diagnostic information.

The bias is introduced by the diagnostic question posed to the diagnostician by another person: a referring colleague or the patient. Such questions sometimes focus attention on certain items, “blinding” the doctor to other, often more important items. I will refer to this as Bias of the Question Posed (BQP) which is a specific case of a more general cognitive phenomenon referred to as “inattentional blindness” [7].

Report of a case

The following clinical case demonstrates the potent effect of BQP: A 49-year-old patient was referred to a tertiary referral eye hospital by a specialist ophthalmologist for the diagnostic assessment of a pigmented lesion on the sclera of the left eye. The referral was made in order to “rule out malignant melanoma”. The patient was otherwise healthy and had a history of a left eye penetrating injury from a high-velocity metallic object 10 years earlier. He was assessed by a team of ophthalmology sub-specialists at the hospital. Several clinic visits took place. The written question posed to the clinicians in the patient’s record at each visit was “rule out malignant melanoma”. Clinical and radiological assessments of the lesion (Figures 1–3) were performed. As clinically suspected, a CT scan showed that the lesion in question was in fact an epi-bulbar metallic foreign body. The patient then consented to attend the hospital’s weekly grand rounds in which patients are routinely examined by a team of both specialist ophthalmologists and trainees, in order for their diagnostic or treatment dilemmas to be discussed in a larger forum. The patient was presented with the referral question and the available work-up, and advice was sought about the recommended management. He was examined by 12 specialist ophthalmologists, each spending a few minutes talking to the patient and examining him with the help of a slit lamp. A group discussion of the case followed, concluding that the dark lesion on the sclera was not a melanoma, but rather a metallic foreign body situated external to the globe, and therefore not requiring removal. However, at that stage, one ophthalmologist commented that the patient’s iris was obviously darker on the affected eye (heterochromia). This detail had escaped the attention of the other clinicians, and required re-examination of the patient to confirm (Figure 4). In the context of an ocular metallic foreign body, heterochromia is a classical clinical sign of siderosis, namely a late sight-threatening iron toxicity to the intraocular tissues, which colours the affected iris brown. Siderosis is a serious condition which needs to be anticipated after penetrating injuries from iron-containing foreign bodies. It is a gradual and slow process, which in this patient’s case had resulted in heterochromia and mild, diffuse retinal function loss as evident by diffuse reduction in sensitivity of the visual field. The patient’s subjective vision was still satisfactory at that stage, but siderosis is likely to lead to eventual blindness from retinal toxicity and glaucoma unless treated promptly [8]. It therefore requires immediate surgical removal of any such foreign bodies in an attempt to stop the process from progressing further.

Figure 1

The left eye showing evidence of prior penetrating eye injury, with a corneal scar at 5 o’clock and a traumatic iris defect at 5–7 o’clock.

The presumed trajectory of the foreign body is: entry at 5 o’clock at the corneal limbus, and passage through the iris and lens at 5–7 o’clock, passage through the ciliary body and anterior choroid at 8 o’clock and eventual lodgement in sclera at 9 o’clock.

Figure 2

A dark sub-conjunctival mass, originally suspected to be a malignant melanoma.

Figure 3

A metallic foreign body is seen in the corresponding area on axial orbital computerised tomography.

Figure 4

“The gorilla”.

Comparative examination of both eyes reveals an obvious heterochromia, with the injured left eye showing a hyperpigmented, brown iris. The diagnosis is left ocular siderosis due to a retained metallic foreign body.

Conceptually, the question posed to the clinicians therefore changed from “Is this a melanoma?” to “What is/are this patient’s ocular diagnosis/es” and, subsequently, “Does this patient require urgent surgical treatment for his sight-threatening siderosis?” The consensus then changed from “Do nothing, no concern, as this is not a melanoma” to “The patient has ocular siderosis of the left eye, urgent removal of the metal foreign body is required to avoid blindness”.

Discussion

As seen in Figure 4, the patient’s heterochromia would probably be obvious to most lay observers after a casual conversation. In fact, the patient himself noticed heterochromia developing after the injury, but thought it was irrelevant. Therefore, I suggest that it takes a powerful cognitive bias for multiple competent clinicians to repeatedly miss such an obvious sign.

The above (flawed) cognitive pathway was introduced by the question “Is this a melanoma?” This conceivably focused the examiners’ attention so strongly on the potentially life-threatening item in question that they did not notice the obvious and significant finding of heterochromia.

The same psychological phenomenon (termed “inattentional blindness”) has been famously demonstrated by Simons in the “Invisible gorilla” experiment [7, 9].

In the experiment, a video of a basketball game is shown. The question posed to the viewers (“how many passes were made by the black team?” or “how many bounce passes vs. aerial passes have taken place?”) focuses their attention so effectively that it distracts and “blinds” half of the viewers to the obvious presence of the gorilla crossing the basketball court in the middle of the video.

Any question posed to a clinician prior to making an independent diagnostic assessment may potentially act in the same way. This may be regarded as one type of a “context error”, defined by Ely et al. as “the clinical signal is distorted by the background against which it is perceived” [10, 11]. The term “framing effect” is also related, describing “How diagnosticians see things may be strongly influenced by the way in which the problem is framed” [12]. There is therefore some overlap between “the bias of the question posed”, “context error” and “framing effect”. Salient distracting features have been identified as a potent diagnostic bias, especially if introduced early in the diagnostic process [13]. Many radiologists and pathologists therefore routinely attempt to avoid this phenomenon, preferring to examine the radiological or pathological information before reading the questions posed by the referring clinician, as “it forces one to scan the entire image or images, since the specific area of interest in not yet known. It promotes unexpected observations” [14].

It is therefore advisable to start clinical diagnostic assessments with as little knowledge of previous questions, opinions and data as possible, in order to preserve the opportunity for an unbiased initial assessment. This is especially relevant to clinical situations of high-level (secondary and tertiary) referrals from doctor to doctor, where the opportunity for a truly “second” opinion may be lost if the second clinician is biased by the opinion of the first one.

Conclusions

In self-reflecting during a diagnostic process, one should therefore ask oneself not only “Have I correctly answered the question I was asked?” but also “Is the question posed to me really the (only) important question about this patient’s problem?” In doing so, the “invisible gorillas” would hopefully receive the attention they deserve.