Coping with uncertainty in clinical practice: a narrative review

In the words of renowned American surgeon, writer and public health researcher Dr Atul Gawande, “The core predicament of medicine — the thing that makes being a patient so wrenching, being a doctor so difficult, and being a part of society that pays the bills they run up so vexing — is uncertainty … Medicine's ground state is uncertainty. And wisdom — for both the patients and doctors — is defined by how one copes with it”.1

The ever‐increasing novelty, complexity and, sometimes, insolubility of modern medicine brings decisional uncertainty for clinicians (Box 1). Although diagnostic conundrums attract considerable attention, uncertainty pervades many other areas of practice, such as how to deal with incidental or ambiguous findings from an ever‐increasing array of laboratory investigations,2 what treatments to prescribe for conditions for which there are multiple options,3 how to predict illness trajectories and navigate the care of patients through a complex health system,4 and how to ethically decide what care to provide while reconciling patient wishes with likelihood of benefit and limited resource availability.5 In addition, scientific evidence is non‐existent, conflicting, inconclusive or not applicable for many clinical questions, so deciding what constitutes best care in a particular set of circumstances remains uncertain (epistemic uncertainty).6 Equally, even with high quality evidence, it can be difficult to predict the effects of interventions in individual patients (aleatory uncertainty).7 Contextual factors can also inject more uncertainty into clinical encounters by disrupting reasoning processes, these being clinician‐related (eg, fatigue, hunger), patient‐related (eg, poor English proficiency, presentation complexity) or environment‐related (eg, noise, distractions, time pressures).8

Despite the ubiquity of uncertainty in medicine, clinical culture too often fails to acknowledge it, explore how it affects individual clinicians, or consider what can be done to mitigate it.9 Professional training and socialisation have traditionally placed value on certainty over uncertainty,10 and how to minimise or eliminate uncertainty rather than in how to tolerate or manage it. Clinicians respond to uncertainty in various ways through the interplay of a series of cognitive, emotional and ethical reactions.11 Stress from uncertainty is increasingly recognised as a likely driver of professional burnout in health care,12 reducing clinician engagement and productivity, posing risks for patient safety as a result of more errors,13 and exacerbating workforce shortages by inducing early retirement or change of occupation.14 Acquiring the adaptive ability to understand and tolerate clinical uncertainty and guide patients in trusting relationships amid such uncertainty are important challenges facing clinicians,1,15 and are now recognised as core clinical competencies for medical graduates and trainees.16 Modern computerised decision support systems currently lack intelligent inference mechanisms for handling uncertainties in scientific knowledge and its application to specific clinical scenarios.17 Evidence‐based clinical decision rules do not necessarily improve care processes or patient outcomes,18 and although precision medicine using artificial intelligence, machine learning and big data seeks to lend greater certainty in personalising care, the information explosion that it brings will likely increase rather than decrease levels of uncertainty.19

The aims of this narrative review are to: i) characterise the various types of uncertainty; ii) define the adverse effects, clinician characteristics and clinical contexts associated with intolerance of uncertainty; iii) identify means for assessing intolerance of uncertainty at an individual level; and iv) present strategies that clinicians and their patients can use to cope with uncertainty after doing what can be reasonably done to minimise it. This review is based on original articles and systematic or narrative reviews obtained from a search of MEDLINE, PsycINFO, SocINDEX, CINAHL, Web of Science and Google Scholar published between 1 January 1995 and 30 September 2022 using search terms “uncertainty”, “ambiguity”, “tolerance”, “confidence” and “coping”. We selected articles relating to clinicians providing direct patient care and excluded those relating to more indirect care provided by clinicians in disciplines such as pathology and radiology.

Characterising clinical uncertainty

As there is no universal definition of clinical uncertainty, we will define it simply as the state clinicians are in when they are unsure of the diagnosis, the care to be recommended and delivered, or their understanding of the patient, their problem, or its trajectory. Clinical uncertainty is common, with one study of doctors interacting with older male patients verbally expressing uncertainty in up to 70% of all encounters.20 In primary care, one British study of 50 new patients presenting with a diagnostic issue revealed a certain diagnosis was forthcoming by the end of the first consultation in less than half of the cases.21 Among 592 patients presenting acutely with dyspnoea to one emergency department in the United States, almost a third of cases remained undiagnosed after a full standard work‐up.22

Not surprisingly, clinical uncertainty can take various forms according to the decisional context and goals, as denoted by various conceptual frameworks and taxonomies of uncertainty. An initial framework proposed three types of informational uncertainty: technical (due to limited scientific data or practical skill), personal (arising from gaps in understanding patients’ wishes), and conceptual (resulting from the application of abstract criteria, such as guideline recommendations or protocols, to concrete clinical scenarios).23 Another framework states uncertainty can be about the research evidence, the patient's story, what best to do for a specific patient with a particular set of circumstances, and the interactions between humans and between humans and technology.24 A third framework classifies uncertainty as being centred on probability (ie, inability to predict future outcomes), ambiguity (ie, lack of information or conflicting evidence or opinion) or complexity (ie, multiplicity of causal factors, relationships and interpretations).25 The same authors also identify loci of uncertainty as being disease‐centred (diagnosis, prognosis, treatments), system‐centred (structures and processes of care, networking, teamwork) and patient‐centred (psychosocial and existential effects of illness and its treatment on patients’ perceived meaning of life and personal relationships). A fourth framework similarly categorises uncertainty as arising primarily from deficits in scientific knowledge about disease, insufficient practical understanding of how systems of care operate (eg, lack of clarity about expected skills, how to access or implement different forms of care, what policies and procedures to follow), and inability to predict the aforementioned effects of illness and treatments on patients.26 However, this framework also includes ethical aspects of clinicians reconciling their personal values with those of their patients and with the sociocultural and practice codes of the craft groups and institutions within which they work.

More recently, different researchers have formulated frameworks that not only identify the types and sources of uncertainty, but also include factors that influence how much an individual clinician perceives uncertainty (eg, the clinical context or situation, personal moderators such as personality, cognitive capacity, level of experience, sociocultural influences), and categorise how individuals respond to uncertainty in cognitive, emotional and behavioural terms, both positive and negative.27,28,29 In Box 2, we have constructed a schema that integrates these various frameworks, while noting that dimensions of probability, ambiguity and complexity can be difficult to disentangle, and boundaries between source domains of scientific, practical and personal, and between response domains of cognitive, emotional and behavioural are blurry and often intertwined.

Effects of uncertainty on clinicians and patient care

Like all humans, clinicians will often think and do things in an attempt to extract themselves from situations of uncertainty which are experienced as being uncomfortable or even stressful. When confronted with uncertainty in decision making, clinicians may deny it and apply accepted rituals, simply adopt the practice of their peers, or rely on heuristic reasoning (use of mental shortcuts or rules of thumb).30 In other instances, acknowledging uncertainty may serve as a self‐protective or motivational force, resulting in increased information seeking or considered selection of preventive, diagnostic and therapeutic actions.

Unfortunately, maladaptive reactions induced by an intolerance of uncertainty and desire for decisional comfort can have adverse consequences for patient care. Clinicians with higher intolerance of uncertainty may be more likely to avoid certain kinds of patients with complex needs, such as substance users, the poor, older patients and the underserved, contributing to health inequities.31 Studies show that intolerance of uncertainty can impair clinical decision‐making skills,32 manifesting as misperceptions of clinical goals of care, greater vulnerability to reasoning biases (eg, premature closure in diagnostic reasoning) and lower diagnostic performance,33 overuse and misuse of diagnostic tests,34 prescribing of inappropriate interventions,35 more unnecessary referrals,36 and reduced engagement with colleagues and patients in decision making.32,37 This can all result in compromised doctor–patient relationships and decreased patient satisfaction, confidence and trust in the medical system.38

Intolerance of uncertainty is also associated with more psychological distress and burnout,8,39 loss of self‐compassion,40 career dissatisfaction and disengagement at work, more discomfort dealing with death and grief,41 more concerns about malpractice risk, less propensity to adopt new and effective clinical interventions,32 and more limited leadership abilities.42 Clinicians with higher intolerance of uncertainty may also avoid certain career paths, such as primary care, emergency medicine, psychiatry or geriatric medicine, as these are perceived as disciplines likely to evoke more decisional uncertainty.14,32,43

Personal and contextual factors associated with intolerance of uncertainty

Evidence of any associations between intolerance of uncertainty and clinician and contextual factors is, to date, inconsistent, although some studies suggest a higher propensity towards intolerance of uncertainty among female clinicians,37 general practitioners,37 and surgeons.44 Clinicians with more experience or who are less risk‐averse tolerate uncertainty more than less experienced37,45 or more risk‐averse46 colleagues, and lacking a trusted advisor is also associated with greater intolerance of uncertainty.37

Regarding contextual influences, clinicians are less tolerant of uncertainty when dealing with acute undifferentiated clinical scenarios, especially in settings where access to advice is reduced,46,47 or when interviewing more educated persons who show a greater desire for information,48 or with patients who present with mental health or psychosocial issues rather than biomedical ones.49 Other patient characteristics such as age, gender or race appear to exert no influence.

Assessing individual tolerance of uncertainty and response

In measuring a clinician's level of intolerance of uncertainty and response, the 1995 version of the Physician's Reactions to Uncertainty (PRU) scale is the most widely cited and validated tool (Box 3).50 This 15‐item, Likert‐scale questionnaire has two parts, the first assessing the level of stress arising from uncertainty in terms of anxiety and concerns about bad outcomes, and the second determining the extent to which uncertainty results in reluctance to disclose uncertainty to patients or disclose mistakes to other clinicians. The points for each item are summed to give a final score, with a higher score indicating greater intolerance of uncertainty. The PRU scale is limited in that it addresses only emotional reactions to uncertainty, rather than those of a cognitive and ethical nature.9,37

The PRU scale helps to identify different clinician phenotypes in terms of affective and maladaptive response to intolerance of uncertainty. In one cross‐sectional analysis of 1209 clinical encounters involving 594 Australian general practice trainees to whom the PRU scale was applied,51 the anxiety and concern subscales were associated with female gender, less experience in hospital before commencing general practice training, and graduation overseas. On the other hand, the reluctance to disclose subscale was associated with urban practice, health qualifications before studying medicine, practice in an area of higher socio‐economic status, and being trained in Australia.

Strategies for managing uncertainty

As tolerance of uncertainty is thought, at least partly, to be determined by situational factors, it may be amenable to change through educational and experiential processes. Much of the research on such processes involves novice clinicians (medical students, residents, trainees or newly qualified specialists), although the insights gained have application to all grades of experience. At the outset, novice clinicians should expect uncertainty to be part of their routine work, and manage the expectations of all parties involved in delivering care, including themselves, about how much they are capable of eliminating it.52 Where uncertainty results from limited knowledge and experience, information‐seeking actions can be constructive, but, at some point, clinicians have to accept they will never know everything they need to know and no amount of study or decision support will eliminate uncertainty.

Evidence is emerging of the value of incorporating formal teaching in how to characterise uncertainty and how to cope with it into undergraduate and postgraduate curricula, best integrated with courses in clinical reasoning.53 Various educational formats have been shown to have favourable impact, including clinical debriefs, small group exercises, role plays, simulations, standardised cases, chart‐stimulated recall interviews, peer‐to‐peer conversations, and reflective learning and assessment, including narratives from mentoring clinicians which express and normalise their own vulnerability to uncertainty.46,54,55,56 Discussing uncertainty with members of multidisciplinary teams allows others to offer advice and perspectives that can decrease errors and cognitive overload.57 Commitment to lifelong learning and intellectual curiosity are also strengths, as researching a patient's illness and possibly becoming aware of new case reports or new treatments may benefit existing or future patients.

However, others argue the acquisition of appropriate reactions to uncertainty will mostly occur over time with more experience.58 In reality, a complex balance of education and experience is most likely, while recognising that fixed personality traits and deeply imprinted prior experiences may predispose individuals to specific psychological responses.59 Of concern is the observation in some studies of medical students becoming more intolerant of ambiguity, losing empathy and seeking decisional perfectionism as they progress through their course.43,60,61

Acting with confidence while simultaneously remaining uncertain in dealing with complex, ill‐defined problems epitomises expert practice. In such situations, clinicians continuously reconstruct and redefine their understanding of the problem, even as they are trying to solve it.62 This goal of becoming comfortable with uncertainty and remaining able to make decisions is different to being certain but uncomfortable, where one perceives a clear understanding of the problem but realises one is incapable of managing it. Clinicians with intolerance of uncertainty often perceive uncertainty as a threat to self‐identity, and in reducing such existential distress, cognitive dissonance provoked by uncertainty must be transformed into self‐directed learning moments, whereby clinicians become aware of, and are shown how to lower, perceived threats while improving resilience.63 Recent articles offer various strategies for coping with uncertainty in routine practice,64,65,66,67 as summarised in Box 4.

Communicating uncertainty with patients

In an era of shared decision making, whether to share uncertainty with patients is not the issue, rather the question is how best to communicate it without causing undue anxiety or loss of trust, which both become more likely the longer uncertainty persists.68 Clinicians who perceive their patients as more likely to have negative reactions to uncertainty may make unilateral and premature decisions about preferred care and withhold interventions where there is uncertainty about benefits and harms.69 In contrast, disclosing and managing uncertainty may actually strengthen the therapeutic clinician–patient relationship depending on how uncertainty is communicated, and what strategies are used to prioritise continuity of care, patient‐centred communication, and trust.70 Acknowledging uncertainty and exercising appropriate safety netting build patient confidence in their clinician's care, contrary to beliefs that such admissions may imperil clinician credibility and authority.15 Instructing a patient to call if any change or clinical worsening occurs and being specific about what the patient should look for and how to seek medical attention build patient trust and confidence. Similarly, sharing uncertainty with colleagues applies when patients are being discharged from hospital. The clinician who will assume follow‐up (likely the patient's general practitioner) should be contacted directly by the inpatient team, and any uncertainty regarding the patient's diagnosis or plans of care should be made clear. The same principle applies during a hospitalisation, as care is transitioned from one shift to another or from one clinician to another.

Despite the growth in studies exploring multiple sources and types of uncertainty, most continue to focus on identifying, testing and evaluating strategies to communicate probabilistic risk.70 This may not always be appropriate due to the increasing realisation of limitations in scientific evidence and a multiplicity of alternative care options. Consequently, more focus needs to be given to improving communication of uncertainty inherent in ambiguous and complex scenarios.70 Box 5 provides a structured approach to communicating uncertainty to patients based on findings from recent reviews.71,72,73,74 However, successfully enacting this approach requires both clinicians and patients to be free of time pressures, distractions and competing priorities.

Conclusion

Uncertainty is intrinsic to clinical practice, affecting both trainees and experienced clinicians. As Sir William Osler wrote: “Medicine is the science of uncertainty and the art of probability”.75 Limitations in knowledge, complexities of care, and variation in patient preferences contribute to uncertainty. Personal factors such as personality traits and resilience influence one's tolerance of uncertainty and responses to it. Clinicians need to be able to recognise and manage uncertainty in its various forms, communicate uncertainty to patients, and normalise and openly discuss clinical uncertainty with their peers. At the same time, the current design and funding of health care favour investigations and procedures over potentially lengthy and cognitively demanding discussions between clinicians and patients and shared decision making in the context of irreducible clinical uncertainty. Clinicians and managers must together advocate for system of care reforms that support the recognition, acceptance and management of uncertainty, and the means for minimising its potentially harmful effects on patient care and clinician wellbeing.

Box 2 – An integrated theoretical framework of clinical uncertainty

---