Medical error remains a common and important problem, with a significant impact on healthcare outcomes. Approximately 1 in 20 patients is exposed to preventable harm,¹ and human error accounts for up to 87% of all medical errors, with varying prevalence stratified by specialty and clinical situation.² There is a lack of good data concerning medical errors in the South African context, but the medical error rate in developing countries has been estimated to occur in approximately 1 in 12 patients.³ Cognitive errors are thought-process errors or thinking mistakes, which lead to incorrect diagnoses, treatment or both.⁴ Cognitive errors in anaesthesia likely contribute significantly to medical error in the perioperative environment, but this is an understudied area. The framing effect is one example of a cognitive bias that can have an impact on medical decision-making. A framing error occurs when subsequent thinking is swayed by the leading aspect of the initial presentation.⁴ This implies that time-sensitive decisions during emergencies might be influenced by the way in which medical information is framed.⁵ To our knowledge, no studies examining the potential impact of the framing effect in anaesthetic medical officers in an emergency scenario have been conducted.

The purpose of this study was to compare the impact of the framing effect on clinical performance in an emergency simulated scenario under two different framing conditions (Control vs Experiment). We hypothesised that framing the patient as an asthmatic would bias participants towards bronchospasm as the primary diagnosis. This anchoring and/or fixating was expected to delay consideration of alternative causes, such as a mucous plug, thereby prolonging the time to correct diagnosis and initiation of appropriate corrective action. The way a clinical simulation scenario is framed will create a significant and clinically relevant bias, resulting in a prolonged time to diagnosis and/or implementation of corrective measures. We further aimed to assess the participants’ perception of the simulation, their performance and their critical thinking.

We conducted a prospective, randomised, double-blinded study on medical officers who had recently passed their Diploma in Anaesthesia (DA) examination, in the Pietermaritzburg Complex (KwaZulu-Natal, South Africa). Ethical approval was obtained from the KwaZulu-Natal Department of Health (KZ_202110_028) and the University of KwaZulu-Natal Biomedical Research Ethics Committee (BREC/00003229/2021). Informed, written consent was obtained from all participants. The study took place at the Grey’s Hospital Simulation Centre.

The participant recruitment process is illustrated in Figure 1. The final analysis included 34 participants, with 17 in each group. No eligible participant refused consent.

The primary outcome was the difference in time to correct diagnosis (bronchial plug). There was no difference in median times (control group 240 [interquartile range {IQR} 195–240] vs experimental group 240 [IQR 162–240] s; z = −0.433, p = 0.6648). Further action-based secondary outcomes are shown in Table 1.

Secondary outcomes relating to diagnostic reasoning were as follows: Participants in the control group entertained the correct diagnosis in 12/17 (70.6%) vs 9/17 (52.9%) in the experimental group (p = 0.48). In the control group, only 1/17 (5.9%) participants were unable to entertain more than one diagnosis, versus 3/17 (17.6%) in the experimental group (p = 0.60).

We used a questionnaire with a rank-order (Likert) response scale for questions not requiring a written answer (1: Bad, 5: Excellent). The participants ranked the quality of the briefing (5, IQR 4–5) and the relevance of the scenario (5, IQR 5–5) as excellent. The quality of the simulation was ranked as 4.5 (IQR 4–5) and the realism of the simulation as a 4 (IQR 4–5). The participants’ experience of the simulation was ranked lower, at 3.5 (IQR 3–4).

We also asked participants to state their leading diagnosis after the scenario: bronchospasm was the diagnosis in 9/17 (52.9%) of the experimental group vs 8/17 (47.1%) in the control group. We further asked if candidates felt that the handover had been problematic in any way, and 12/17 (70.6%) in the experimental group vs 6/17 (35.3%) in the control group felt misled by the scenario.

Our study showed that the framing of a clinical handover did not result in a difference in time to correct diagnosis in a simulated emergency scenario. Despite the cognitive trap, there was very little difference in performance between the two groups. There were also no significant differences in secondary outcomes between groups, which would require a larger study. Both participant and examiner feedback were positive with regard to relevance, realism and quality, and simulation is likely to become a part of the routine DA curriculum.

The psychology of decision-making is becoming increasingly appreciated in the non-medical and medical literature.⁴ Cognitive biases like framing can affect subsequent thinking. The high prevalence of human error in anaesthesia is likely multifactorial in origin. Possible contributing factors include the complexity of theatre environments, the urgency of the crisis, medication errors and psychophysiological variables that are unique to each individual and team.^4,6,7

Framing errors can be divided into two distinct types: equivalence framing and emphasis framing. Equivalence framing is the preference to behave differently depending on whether a decision is viewed as a gain or a loss. This ‘framing effect’ was first characterised by Tversky and Kahneman in 1981.⁸ Emphasis framing is when your subsequent thinking is swayed by the leading aspect of initial presentation. Therefore, decisions involving life-threatening conditions could potentially be influenced by the way in which medical information is framed.⁹ We postulated that the initial framing of a scenario may result in error through a cognitive error called ‘fixation error’. Fixation errors have also been called ‘anchoring errors’ or ‘tunnel vision’ and can be broadly considered as human errors of insight.² Fixation errors are a type of cognitive error in which individuals and teams focus on one aspect of a situation while ignoring more relevant information.^6,7,10,11

We were unable to demonstrate objective evidence for these cognitive errors in our simulation. Examiners noted that participants did seem to fixate on a single diagnosis (like bronchospasm), but that it appeared to occur regardless of the cognitive trap. We did not power this study to examine each secondary outcome, but there were non-significant differences that might inform the design of future studies in this area. For example, the experimental group appeared quicker to auscultate the chest, which makes intuitive sense when bronchospasm is the primary consideration. They also took longer to respond by increasing the FiO₂ and suctioning the ETT with a flexible suction catheter. This would be an understandable reaction when managing acute bronchospasm, as opposed to a mucous plug. Despite the delayed response to suctioning down the ETT, the response time to removing the ETT was similar in both groups. This may possibly highlight the reluctance of the candidates to remove the ETT, irrespective of the clinical scenario. It appears that the experimental group marginally diagnosed bronchospasm more frequently and was less likely to entertain the correct diagnosis. However, this remains speculative in the absence of a larger study. Future studies can use our findings to perform sample size calculations for studies aimed at investigating these important questions. Understanding these errors is important in order to find strategies that minimise human error and ultimately improve patient outcomes. Overall, only 38% of the candidates performed the intervention (removing the ETT tube) that would have rectified the situation and saved the patient’s life. This included seven participants in the control group and six in the experimental group.

The participants rated the experience highly with regard to relevance, realism and quality. This positive experience of the simulation was ranked higher than expected by the investigator and examiners, as it was felt that some participants looked emotionally affected by the experience. This may be partly because of the lack of simulation training and experience in the Pietermaritzburg Complex and may change the way that future simulation scenarios are handled. Performing a similar study in an institution that routinely practices simulations would provide an interesting platform for comparison. During feedback, more people in the experimental group stated that they felt misled by the handover. Importantly, they were able to recognise the cognitive trap after the simulation. However, even though they held bronchospasm as an important consideration, they were able to perform in a similar manner to the control group. This recognition post simulation points to the importance and possible benefit of using similar simulations as cognitive aids to recognise our own errors in crisis situations.

Our study showed that the framing of a clinical handover did not result in a difference in time to correct diagnosis in a simulated emergency scenario. Differences in secondary outcomes between groups require investigation in larger studies. Candidates did poorly overall, with only a third of participants correctly resolving the clinical scenario. Simulation training and research provide a useful tool to evaluate cognitive errors and to provide training in scenarios where these errors may occur.