Tolerance of ambiguity and psychological well‐being in medical training: A systematic review

Abstract Context The prevalence of stress, burnout and mental health disorders in medical students and doctors is high. It has been proposed that there may be an association between levels of tolerance of ambiguity (ie an ability to tolerate a lack of reliable, credible or adequate information) in clinical work and psychological well‐being within this population. The aims of this systematic review were: (i) to assess the nature and extent of the literature available, in order to determine if there is an association, and (ii) to develop a conceptual model proposing possible mechanisms to underpin any association, in order to inform subsequent research. Methods MEDLINE, Cumulative Index to Nursing and Allied Health Literature (CINAHL) and PsycINFO databases were searched for articles published from inception to September 2018. Additional literature was identified by searching the reference lists of included articles, forward searches of included articles, hand searches of key journals and a grey literature search. Of the 671 studies identified, 11 met the inclusion criteria. A qualitative synthesis of included studies was performed. Results All 11 included studies reported an association between a lower level of tolerance of ambiguity or uncertainty and reduced psychological well‐being. Included studies were heterogeneous in terms of population and measurement approach, and were often of low methodological quality. Subsets of items from previously developed scales were often used without sufficient consideration of the impact of new combinations of items on scale validity. Similar scales were also scored inconsistently between studies, making comparison difficult. Conclusions There appears to be an association between tolerance of ambiguity and psychological well‐being. This provides new opportunities to understand and prevent the development of stress, burnout and mental health disorders in medical students and doctors. The conceptual model developed provides a framework for future research, which we hope will prevent wasted research effort through duplication and promote higher methodological quality.


| INTRODUC TI ON
The prevalence of stress, burnout and mental health disorders, such as depression and anxiety, in doctors worldwide is alarmingly high. [1][2][3][4][5][6] Similarly, there is a high prevalence of depression and anxiety in medical students, and by the end of undergraduate training levels of 'psychological distress' are higher than in the age-matched general population. 7 This may lead to absenteeism (where doctors or students miss work or study due to their mental ill-health), presenteeism (where doctors or students come to work or study when unwell) and loss of staff from the workforce (where doctors or students leave the profession of medicine altogether). 8  Although the problems of stress, burnout and mental health disorders in doctors are well described, less is known about individual, team, organisation or societal factors that increase the risk of doctors developing these problems. 7 It is likely that multiple factors contribute towards their increased prevalence in this population. A number of studies have proposed a link between intolerance of ambiguity or uncertainty in clinical practice and a range of outcomes, which could be considered under the broader term reduced psychological well-being, including psychological distress, burnout and mental health disorders. [9][10][11] Although it is widely acknowledged that ambiguity is inherent within the practice of medicine, 12 a greater understanding of the implications of this has been slow to develop, partly due to the conceptual complexity.
Varying definitions of tolerance of ambiguity and uncertainty have been proposed and used to underpin measures of these constructs in medical undergraduate and qualified doctor populations. Recently, a conceptual model for understanding tolerance of ambiguity and uncertainty was proposed based on a review of 18 existing measures of tolerance of uncertainty and ambiguity. 13 This review outlined some of the challenges with existing measures, such as their poor conceptual clarity (eg using the terms uncertainty and ambiguity interchangeably) or inconsistent use of definitions for these constructs. In response, Hillen et al set out their definition of tolerance of uncertainty as 'the set of negative and positive psychological responses-cognitive, emotional and behavioral-provoked by the conscious awareness of ignorance about particular aspects of the world.' They state that uncertainty is the response to either ambiguity, probability or complexity and define ambiguity as a 'lack of reliable, credible or adequate information. ' We adopt these definitions for the current study as they precisely and explicitly distinguish between ambiguity and uncertainty, based on the current literature. They also set out a clear definition for what it means to 'tolerate' these two closely related constructs. 13 We hypothesise that intolerance of ambiguity in medical students and doctors could place an individual at increased risk of experiencing reduced psychological well-being. By this, we mean stress, burnout or a more persistent pathological state such as anxiety or depression, which are themselves considered to be a mental health disorder. Examining the broader concept of psychological well-being would aid understanding of the role that these potentially more transient states (stress and burnout) could play in the development of a mental health disorder. However, the terms stress and burnout themselves are not without controversy. For example, although the term 'burnout' is commonly used there remains no universally accepted definition for this; thus, the validity of burnout scales is unclear. 14 Given that the set and extent of pressures that doctors and medical students encounter may be unique to this population, 8 it seems reasonable to examine this potential association within this population without extending our hypothesis to include other health care professionals at this stage. By synthesising the existing literature on this topic, we hope to draw conclusions about the potential association between tolerance of ambiguity and psychological well-being and offer a conceptual model that can be tested through subsequent research. This is particularly important given that many of the included concepts appear to be inconsistently defined and overlapping. A conceptual model would help to advance the research field and save unproductive research effort. It would also take us closer to designing evidence-based interventions that might support doctors in coping with their often inherently ambiguous medical work.

| ME THODS
The aims of this systematic review were: (a) to assess the nature and extent of the literature available, in order to determine if there is an association between levels of tolerance of ambiguity and psychological well-being within medical students and doctors, and (b) to develop a conceptual model proposing possible mechanisms to underpin any association, in order to inform subsequent research. The preferred reporting items for systematic reviews and meta-analyses (PRISMA) framework has been used to help guide and ensure the high quality of this systematic review. 15,16

| Information Sources and Search Strategy
The search protocol was developed by undertaking pilot database searches in MEDLINE and PubMed and discussions between the authors (JH and KM) and others (see Acknowledgements), to identify and refine search terms. We included search terms that would identify studies that evaluated levels of 'stress' or 'burnout', and studies that assessed evidence of a mental health disorder, based on the author's definitions of these terms. Given the challenges with existing measures of tolerance of ambiguity and uncertainty, and the evidence that these measures often overlap substantially, 13,17 we decided to include search terms that would identify studies evaluating levels of tolerance of ambiguity or uncertainty. It is important to note that although we include studies evaluating tolerance of both ambiguity and uncertainty in this review this does not mean that we are using these terms interchangeably, just that we are aware that terminology has not always been used precisely in the published literature. The search terms and strategy were then further refined and finalised through discussion with an information specialist (AB).

Searches of PsycINFO, Cumulative Index to Nursing and Allied
Health Literature (CINAHL) and MEDLINE databases for articles published from inception to 3 September 2018 were conducted.
These databases were selected to ensure that published mental health and qualitative literature were considered in addition to more traditional biomedical articles. Relevant Medical Subject Headings or subject terms were explored and included. No search limitations were applied at this stage. The full search strategies for each database are included in Table S1.

| Study selection
In total 669 papers were identified through the initial search of the three databases (see Figure 1). There were 71 duplicates, meaning that 598 papers were screened using the inclusion and exclusion criteria. Studies were included if they: (a) were an empirical study; (b) used any defined measure of ambiguity or uncertainty tolerance, (c) used a measure of psychological well-being (stress, burnout or evidence of a mental health disorder); and (d) were conducted within the undergraduate medical student or postgraduate doctor population. These are defined in more detail in Table 1. JH and KM independently screened 75 (12.5%) of the titles and abstracts to determine if they met the inclusion criteria for further analysis or inclusion in the review. There was agreement on 74 out of 75 and in the one case where there was disagreement this was resolved through discussion. The remaining 523 papers were screened by JH alone, with further discussion with KM where needed.
Of the 598 papers that were screened, 571 were excluded following a review of the title and abstract as they did not meet the inclusion criteria for the review. Full papers were requested and reviewed for 27 studies. Of these 27 full papers, 17 did not meet the inclusion criteria, meaning that 10 papers in total were included.  Table S2 for full details of all supplementary and grey literature searches.

| Supplementary and grey literature searches
F I G U R E 1 PRISMA (preferred reporting items for systematic reviews and meta-analyses) flow diagram Through this process, two additional studies were identified.
One was excluded because, although physicians were included in the study population, it was not possible to evaluate their scores independently of other health care professionals. One study met the inclusion criteria and was included in our final review.

| Data extraction and summary of findings
Relevant information, including country, sample, study design, measure of tolerance of ambiguity or uncertainty and measure of psychological well-being and outcomes, was extracted from each included article by JH.

| Quality assessment
Study quality was appraised using the Medical Education Research Study Quality Instrument (MERSQI). 18 The MERSQI evaluates study quality based on study design, sampling, type of data, validity of evaluation instrument, data analysis and outcomes. Each item is scored on a scale of 1-3 and summed to determine a total score.
The maximum score for each domain was 3, therefore the maximum MERSQI score is 18 with a potential range of 5-18. The total MERSQI score was calculated as the percentage of total achievable points (accounting for 'non-applicable' responses) and then adjusted to a standard denominator of 18 to allow for comparison of scores across studies.

| RE SULTS
The 11 included studies, along with the quality assessment score, are summarised in Table 2 In all 11 included studies there was a reported association between a higher level of intolerance of ambiguity or uncertainty and reduced psychological well-being. We now present results for the measurement tools used for tolerance of ambiguity or uncertainty, the measurement tools used to evaluate psychological well-being and the associations identified between tolerance of ambiguity and measures of psychological well-being.

| Tolerance of ambiguity or uncertainty
A number of measurement approaches were used to assess levels of tolerance of ambiguity or uncertainty in the included studies (Table   S3). Tolerance of uncertainty was assessed in eight studies and tolerance of ambiguity was assessed in five studies. Two studies assessed levels of tolerance of both ambiguity and uncertainty. 19,20 In 10/11 studies, previously validated scales (either the complete scales or a selected component) were used and in the remaining study a new ad hoc single-item self-reported questionnaire was used. 21 The Physicians' Reactions to Uncertainty (PRU) Scale, which evaluates physicians' 'affective' response to uncertainty, was most frequently used (six studies). The 1990 version of this scale 11 was used in two studies 19,22 and the 1995 version 23 was used in four studies. [24][25][26][27] The Intolerance of Uncertainty Scale (IUS-12) 28 was used in two studies. 20,24 The Tolerance for Ambiguity Scale (Geller) 29 was used in two studies. 19,30 The original Tolerance for Ambiguity Scale (Budner) 31,32 was used in one study, 33 whereas a modified version of this scale was used in another study. 20 The Ambiguous Scenario Task (AST-D) was used in one study. 34,35 This differed from the other

Inclusion criteria Exclusion criteria
Empirical study (peer-reviewed article that presents quantitative and/or qualitative data)

Not published in English
Describes any quantitative or qualitative methodology used to evaluate tolerance of ambiguity or uncertainty Contained medical students or doctors within the population but did not include a subgroup analysis that allowed these populations to be evaluated independently of other health care professionals Describes methodology used to assess for evidence of psychological well-being. This could include the presence of a mental health diagnosis (self-rated or clinician assessed), any measure of psychological distress (self-rated or clinician assessed), stress or burnout Conducted in either undergraduate medical students or postgraduate doctors scales used as it evaluated an individual's interpretation bias (dichotomised into either positive or negative) in response to ambiguity.
In eight of the studies, level of tolerance of ambiguity or uncertainty was treated as a continuous variable, with a 'score' for tolerance of ambiguity or uncertainty being calculated based on scale responses. In three of the studies, level of tolerance of ambiguity or uncertainty was treated as an ordinal variable. 21,25,34 In one of these studies 34 scores above the midpoint were considered to represent a positive interpretation bias towards ambiguous situations, whereas in another 25 participants were only considered to be tolerant of uncertainty if participants scored above the 75th centile.

| Evaluation of psychological well-being
Psychological well-being was assessed using a range of different measurement approaches (Table S4). Self-reported burnout was as- used in the other study. 30 The General Health Questionniare-12, 44 a measure of psychiatric morbidity, was used in one study. 20

| Associations between tolerance of ambiguity and psychological well-being
In all 11 included studies there was a reported association between a lower level of tolerance of ambiguity or uncertainty and reduced psychological well-being.
In all studies where self-reported burnout was assessed 21,22,[24][25][26][27]33 an association was demonstrated with lower levels of tolerance of ambiguity or uncertainty. Only one study measuring burnout evaluated levels of tolerance of ambiguity. 33 The 1995 version of the PRU Scale 23 claims to measure several different emotional or behavioural responses to uncertainty, including 'anxiety caused by uncertainty,' 'concern about bad outcomes,' 'reluctance to disclose uncertainty to patients' and 'reluctance to disclose mistakes to physicians.' Different studies demonstrated different relationships between the type of intolerance of uncertainty and self-reported burnout. One study (n = 193) showed that for emergency physicians in the USA the 'anxiety due to uncertainty' and 'concerns about bad outcomes' components of the PRU scale appeared to be linked with burnout, but not 'reluctance to disclose uncertainty' or 'reluctance to disclose mistakes to physicians.' 25 Another study (n = 128) showed that for GP registrars in Australia 'anxiety due to uncertainty' and 'reluctance to disclose uncertainty to patients' appeared linked to burnout, but not 'concern about bad outcomes' or 'reluctance to disclose mistakes to physicians.' 24 Two studies attempted to identify if participants had evidence of depression. One showed that depressed paediatric residents in North America were more likely to have increased 'stress from uncertainty' than residents without depression. 26 In this study, depression was defined as a score of ≥ 9 on the Harvard National Depression scale). 11 One study showed that in matriculating medical students those expressing higher levels of stress over the past month on the Perceived Stress Scale 43 reported lower tolerance of ambiguity on Geller's Tolerance for Ambiguity Scale. 29,30 This was the largest study identified within this review (n = 13 314).
One study showed that students with psychological distress, evidenced by a General Health Q uestionnaire (GHQ -12) 44 score > 3, had a higher mean intolerance of uncertainty score compared with those without psychological distress using the Intolerance of Uncertainty

| Additional variables
In addition to evaluating tolerance of ambiguity and levels of psychological well-being, additional variables were assessed in a number of the included studies. 'Resilience' was assessed in three studies. 24,26,34 In two of these studies 24 TA B L E 2 Continued through evaluating five individual characteristics: (i) purpose; (ii) perseverance; (iii) self-reliance; (iv) equanimity, and (v) existential aloneness. 45 In one of these studies 26 doctors with depression, identified on the Harvard National Depression Screening Day Scale, 40 were found to have lower resilience scores. The final study showed that trait resilience was associated with positive interpretation bias in ambiguous situations (itself associated with a reduced risk of depression). This study was the only study that was longitudinal in design with interpretation bias in response to ambiguity and resilience being measured at baseline and evidence of depression being evaluated at 6 months. 26 One paper 27 looked at other factors that may be associated with burnout within emergency medicine residents in the USA.
In addition to describing a significant correlation between intolerance of uncertainty and burnout, this paper also identified that those residents with a significant other or spouse had a higher prevalence of burnout compared to single residents (60% vs 40%, P = .002), and that other features such as lack of administrative autonomy and lack of clinical autonomy were also correlated with risk of burnout. Another paper 21 reported that 'feeling alone at work' was associated with burnout (emotional exhaustion). However, in both of these papers these additional variables were not found to be associated with levels of tolerance of ambiguity or uncertainty.

| Quality of studies
The overall methodological quality of studies was low. Total adjusted MERSQI scores amongst the 11 studies ranged from 8.4 to 10.2, with a mean (standard deviation) of 9.2 (0.62). All studies were observational and used self-reported questionnaires to assess outcomes, with only one of the included studies being longitudinal in design.

| D ISCUSS I ON
The The model builds on work by Hillen et al, 13 and incorporates the included studies from this systematic review and our a priori knowledge as researchers (KM and JH) and a consultant psychiatrist (JH). Ambiguity is defined as one of the potential causes of uncertainty. 13,46 Our systematic review found that there appears to be an association between intolerance of ambiguity and uncertainty and reduced psychological well-being (stress, burnout or a mental health disorder) in medical students and doctors. This is indicated in the model with unidirectional arrows, reflecting our hypothesis that intolerance of ambiguity could be a factor in the development of reduced psychological well-being, which needs testing through further research.
Stress was defined differently across studies 19,30 and its definition is more wide ranging than the more consistently defined patho- which evaluates level of tolerance of ambiguity in medical students and postgraduate trainees, 46 and the PRU Scale, which evaluates a physicians' 'affective' response to uncertainty. 11 We have also highlighted some established measures of mental health disorders, such as the PHQ-9 (for depression) and the GHQ-12 (to identify common psychiatric conditions) and suggest that they may be included in future studies.

| Future research
To build upon and complement existing research, future studies that are longitudinal in design, multicentre, include medical students and doctors in training or are based on a power calculation are now needed. These might be hypothesis led and test a specific component of the conceptual model proposed in Figure 2. Finally, we recommend that future studies use measurement tools with more careful consideration of their validity for the population studied.

| CON CLUS IONS
There appears to be an association between intolerance of ambiguity and reduced psychological well-being in medical students and doctors. However, the strength and direction of this relationship is unclear. This is hampered by the small number of studies completed to date, the cross-sectional nature of studies, the small sample sizes of studies and the wide range of measurement approaches used.
This is particularly the case when evaluating levels of tolerance of ambiguity, when subcomponents of previous validated tools are often used and scoring is inconsistent.
Subsequently, the research field to date is patchy and fragmented, rather than programmatic and additive with one study building on the next. Our proposed conceptual model, although based on this limited evidence, does provide researchers with a number of testable hypotheses that could be explored through subsequent research. Our hope is that this can advance this field of research through saving unproductive research efforts. Ultimately, we hope that this will take us closer to designing evidence-based interventions that might support doctors in coping with their often inherently ambiguous medical work, and may reduce the risk of them developing their own problems with stress, burnout or mental health disorders.

ACK N OWLED G EM ENTS
We would like to thank Alison Bethel, Information Specialist with the Evidence Synthesis Team of PenCLAHRC, Exeter, UK, for her advice and input regarding the final search terms and strategy. Funding for the open access of this paper has been supplied by the University of Exeter.

DATA ACCE SS I B I LIT Y S TATE M E NT
The research data supporting this publication are provided within this paper.

AUTH O R CO NTR I B UTI O N S
JH: designed the research, carried out data acquisition, screening, extraction and synthesis, quality appraised the data, drafted and critically revised the manuscript, including revisions, and approved the final version for publication. KM: designed the research, carried out data screening and synthesis, drafted and critically revised the manuscript, provided feedback on revisions and approved the final version for publication. Both authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

CO N FLI C T S O F I NTE R E S T
None.

E TH I C A L A PPROVA L
No human subjects were involved. This was a systematic review.