The global crisis of antimicrobial resistance (AMR) is frequently framed as a failure of pharmacology or a lack of new drug development. The cognitive bias of the prescribing physician is often ignored. However, a growing body of evidence suggests that the persistence of suboptimal prescribing is not merely a scientific challenge, but a deeply psychological one. Despite the proliferation of clinical practice guidelines and educational initiatives, inappropriate antibiotic use remains rampant. To advance antimicrobial stewardship (AMS), we must look beyond the pathogen and the molecule to examine the complex machinery of the prescriber’s mind. By understanding the cognitive biases and psychosocial drivers that influence medical decision-making, we can design interventions that work with, rather than against, human nature. To understand why clinicians overprescribe, we must first understand how humans think. As described by the Nobel laureate Daniel Kahneman, human decision-making is governed by dual process theory.1 Type 1 thinking is fast, intuitive and heuristic-based – a survival mechanism that allows humans to process the thousands of decisions daily with minimal cognitive load. Type 2 is slow, deliberate and analytical. In the high-pressure environment of healthcare, where clinicians make the thousands of decisions per shift, reliance on type 1 thinking is necessary for efficiency.2 However, this intuitive mode is prone to systematic errors known as cognitive biases.2 While these mental shortcuts are often useful, in the context of antimicrobial prescribing, they can lead to catastrophic long-term consequences.3 The application of type 1 thinking to infectious diseases often results in predictable patterns of error. One of the most pervasive is diagnostic momentum. Consider the common scenario of an elderly resident transferred from a long-term care facility with confusion and foul-smelling urine. The initial label of ‘suspected urinary tract infection UTI’ from the nursing home, reinforced by family concern, creates a momentum that is difficult to arrest. Even when clinical evidence (such as the absence of dysuria or fever) points to asymptomatic bacteriuria – a condition that does not require treatment – the initial anchor drags the clinician towards prescribing. This is compounded by confirmation bias, where the clinician subconsciously seeks evidence to support the initial diagnosis (e.g., a positive urine culture) while ignoring disconfirming evidence (e.g., the patient is dehydrated, not infected).2 Furthermore, the decision to prescribe is heavily influenced by commission bias, the innate tendency to favour action over inaction. In medicine, the regret associated with ‘doing nothing’ and missing an infection is psychologically heavier than the regret of prescribing an unnecessary drug. This is inextricably linked to hyperbolic discounting, where immediate, tangible benefits (patient satisfaction and the feeling of safety) are prioritised over distant, abstract harms (Clostridioides difficile infection and future resistance).2 Fear is a potent driver of these biases. The fear of adverse outcomes and a low tolerance for uncertainty drive clinicians to ‘cover’ patients with broad-spectrum antibiotics. This emotional response is often reinforced by the negativity effect, where memories of rare, adverse events (e.g., a patient who deteriorated rapidly) remain more vivid and accessible than the thousands of cases where conservative management was successful.4 Prescribing does not occur in a vacuum; it is a social act. It has been emphasised that individual cognitive biases are amplified or mitigated by the social environment. Hierarchical structures in hospitals can suppress appropriate prescribing; junior doctors may fear contradicting a senior physician’s incorrect plan due to established social norms. Similarly, the pressure to maintain a good relationship with patients or families often overrides the clinical guidelines.4 These psycho-social factors create a gap between knowledge and behaviour. A clinician may possess the correct knowledge (competence) but fails to apply it due to social pressures or emotional fatigue (performance). This disconnect explains why passive education alone is rarely sufficient to change behaviour. Telling a doctor that antibiotics are unnecessary for a viral infection is ineffective if the prescription is serving a social function – such as validating the patient’s illness or managing their anxiety.5 Recognising these flaws in human reasoning allows us to re-engineer stewardship interventions. The goal is not to eliminate type 1 thinking – which is impossible – but to implement ‘de-biasing’ strategies that trigger type 2 reflection at critical moments. The most direct counter to rapid, error-prone thinking is to force a pause. The ‘antibiotic time-out’, typically occurring 48–72 h after initiation, is a structural mechanism to engage type 2 thinking. This forces the prescriber to move from the ‘empirical’ moment of decision-making to the ‘re-assessment’ moment.6 By mandating a review of culture data and clinical progress, the time-out disrupts diagnostic momentum and encourages de-escalation. However, for these time-outs to be effective, they must be integrated into the workflow, potentially through checklists or prompts during rounds.2 Human beings place a disproportionately high value on products they have helped create – a phenomenon known as the IKEA effect.7 The IKEA effect was named after the Swedish furniture retailer IKEA, whose business model requires customers to assemble their own furniture and refers to a cognitive bias in which consumers place a disproportionately high value on products they partially created or assembled themselves, even when the results are imperfect; in other words, people place a higher value on things because they had put work into them.7 In the context of AMS, this suggests that guidelines and protocols should not be imposed from above but co-designed with the end-users. When prescribers are involved in the development of stewardship pathways, their sense of ownership and satisfaction increases, leading to higher adherence.2 A participatory approach where stakeholders analyse their own barriers and co-design interventions, arguing that this empowerment reduces resistance to change has been advocated.4 Clinicians must be trained in metacognition – thinking about their thinking. Educational initiatives should move beyond microbiology to include the psychology of error. Techniques such as ‘considering the opposite’ can help mitigate anchoring and confirmation bias. For example, when diagnosing a confused elderly patient, a clinician might be trained to explicitly ask, ‘What else could cause this confusion if it is not an infection’?8 In addition, providing feedback on prescribing performance compared to peers can trigger self-reflection and recalibrate optimism bias (the belief that one’s own prescribing is better than average).2 Other cognitive biases prevalent in the field of medicine that can subtly affect clinical judgment and decision-making include the following.9-11Attribution bias refers to the tendency to discover reasons for clinical observations, which may sometimes be inaccurate. Search-satisficing bias occurs when practitioners assume that the information currently available is sufficient, leading them to cease the search for alternative explanations. False consensus bias involves overestimating the extent to which others share similar views, and blind spot bias is characterised by the belief that one is less susceptible to bias than one’s peers.9 Finally, because cognitive effort is a finite resource, to make taking the ‘right’ decision the easier one, ‘nudging’ or altering the choice architecture is required. For instance, selective reporting by microbiology laboratories – where susceptibility results for broad-spectrum agents are withheld when narrower options are effective – forces the clinician towards the preferred therapy without requiring a cognitive battle.10 This utilises the prescriber’s natural tendency to follow the path of least resistance. The integration of behavioural science into antimicrobial stewardship is no longer optional; it is essential. The AMS programmes should include behavioural experts and follow a structured four-phase approach: assessment of specific context needs, design of interventions using behavioural frameworks, implementation with participatory communication and rigorous follow-up.4 Antimicrobial stewardship (AMS) is critical for clinicians today as antimicrobial resistance (AMR) has evolved into a top public health threat, directly causing an estimated 1.3 million deaths annually.11 For clinicians, AMS provides an essential framework to combat inappropriate prescribing practices driven by diagnostic uncertainty and patient demand. Evidence indicates that effective stewardship programmes significantly improve patient safety by optimising dosage, reducing hospital lengths of stay and lowering the prevalence of multidrug-resistant organisms.12 Ultimately, AMS empowers clinicians to preserve the efficacy of life-saving drugs while ensuring superior clinical outcomes. Recent evidence from systematic reviews13-15 suggest that artificial intelligence (AI) and machine learning (ML) offer substantial utility in enhancing AMS programmes. Evidence demonstrates that AI algorithms – particularly random forests, decision trees and XGBoost – achieve high diagnostic accuracy and sensitivity in predicting antimicrobial resistance and optimising empirical therapy. By analysing the clinical data faster than traditional culture methods, AI tools facilitate timely de-escalation to narrow-spectrum antibiotics and personalised dosage adjustments, particularly for drugs such as vancomycin. Consequently, AI-driven approaches successfully reduce inappropriate prescribing, lower mortality rates and improve guideline adherence, although current implementation is largely concentrated in the high-income healthcare settings. It must be acknowledged that prescribers are not purely rational. They are humans operating under stress, driven by fear of error, influenced by social hierarchy and reliant on mental shortcuts. By illuminating the ‘black box’ of cognitive bias – from the initial anchoring on a diagnosis to the hyperbolic discounting of future resistance – we can design smarter, more empathetic and ultimately more effective stewardship interventions. The future of antimicrobial preservation, thus, depends not only on discovering new molecules but on mastering the psychology of the clinician who prescribes. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.
Mohan et al. (Wed,) studied this question.