Errors in diagnostic reasoning can occur when clinicians use cognitive shortcuts, relying on intuition and past experience to reach conclusions or make decisions without subjecting their intuitive thought processes to analytical scrutiny. As clinicians gain experience, diagnostic error more often occurs due to flawed reasoning processes rather than knowledge gaps.1 Knowing the sources of biases and being aware of how biases can impact clinical encounters is an important step in counteracting them.
CASE
Office Visit
A 32-year-old woman presents to her clinician’s office with a nonproductive cough, dyspnea, and left-sided chest pain that began 3 days ago. The pain is dull and persistent, but is relieved somewhat when she sits up and aggravated during coughing or deep breathing. She has a history of asthma, allergic rhinitis, recurrent sinusitis, and obesity (body mass index, 34.9 kg/m2), and was last seen in the office 2 weeks ago with ear and facial pain and nasal fullness and was diagnosed with acute sinusitis. At that visit, she was prescribed a 5-day course of amoxicillin-clavulanate and instructed to contact the office if there was minimal or no improvement in her symptoms after 3 or 4 days of therapy. Her asthma has been well controlled with a long-acting beta agonist/inhaled corticosteroid combination. Other medications include a combined oral contraceptive (COC) containing drospirenone and ethinyl estradiol, which she started taking 18 months ago, stopped taking after 9 months, and then restarted 4 months ago. She works as a computer programmer, and for the past month has been working on a project with an upcoming deadline that has required her to work sitting at a desk 12 to 15 hours daily. She does not drink alcohol. She currently smokes 1 pack of cigarettes daily, and has a 10 pack-year history.
On physical exam, blood pressure is 132/90 mm Hg; pulse rate, 103 beats/min; respiratory rate, 29 breaths/min; pulse oximetry, 95%; and temperature, 100.4°F (38.3°C). Auscultation reveals coarse breathing sounds in the left lower lung, but no other abnormalities are detected on physical examination.
The clinician orders a chest x-ray, which shows infiltrates in the left lower lung. The clinician makes a diagnosis of pneumonia, prescribes clarithromycin 500 mg twice daily, and tells the patient to contact the office if her symptoms do not improve.
Emergency Department Visit
Two days later, the patient faints while walking to her mailbox and is taken to the emergency department by ambulance. At the ED, she reports continued chest pain, cough, and dyspnea. Blood pressure is 129/82 mm Hg; pulse rate, 105 beats/min; respiratory rate, 25 breaths/min; pulse oximetry, 95%; and temperature, 99.8°F (37.6 °C). Physical exam does not reveal any skin color changes, edema, or leg asymmetry. Laboratory testing reveals an elevated plasma D-dimer level at 2.49 μg/mL (normal, <0.5 μg/mL). The patient undergoes computed tomography (CT) pulmonary angiography, which reveals thrombus of the posterior basal segmental artery of the left lung. Treatment is initiated with subcutaneous low molecular weight heparin for 5 days, and the patient is maintained on oral rivaroxaban 20 mg daily for 2 months. The patient is advised to stop taking the COC. In addition, her primary care provider counsels the patient about alternative birth control methods, advises her to stop smoking, and provides counseling and quit strategies to help her achieve this goal.2
DISCUSSION
The patient was misdiagnosed with pneumonia even though she had several risk factors for pulmonary embolism, including obesity, recent initiation of COC therapy containing a progestin that can be associated with a higher risk for venous thromboembolism than other progestin-containing COCs, smoking, and long periods of immobility related to her job as a computer programmer. The cognitive biases involved in this missed diagnosis were availability bias, in which the clinician overestimates the likelihood of a diagnosis based on the ease with which it comes to mind, and premature closure, in which a clinician stops considering or exploring important alternative diagnoses once a clear reason for the patient’s symptoms has been found. In this case, the clinician had seen the patient just 2 weeks earlier for an upper respiratory infection, and linked her prior presentation with certain features of the current visit, excluding other possibilities. The diagnosis of “respiratory infection” was readily “available” to the clinician’s mind, given the recency of the patient’s visit. Another example of availability bias is making a diagnosis based on recall of a previous patient with similar symptoms. A subtype of availability bias is significant case bias, where a rare diagnosis, a particularly astute diagnosis, or a diagnosis that had a significant impact on the clinician (such as one that led to litigation) comes to mind before other, more common diagnoses. Referring back to the case, having missed a diagnosis that could have had dire consequences for their patient, this clinician could over estimate the likelihood of pulmonary embolism in patients with cough, chest pain, and dyspnea in future encounters.
One study that explored availability bias in first- and second-year internal medicine residents showed that recent experience with a particular diagnosis increased the chance that these clinicians would pursue that same diagnosis over others when they relied only on non-analytical reasoning.3 However, the study authors note that availability bias was reduced by reflective reasoning, or considering the information and data informing any given decision and determining whether that is sufficient or whether additional data need to be gathered. In other words, when making clinical decisions or drawing conclusions, clinicians must use explicit, deliberate reasoning to override the rapid but error-prone intuitive, unexamined thought processes humans employ most of the time.1
