Communicating uncertainties when disclosing diagnostic test results for (Alzheimer's) dementia in the memory clinic: The ABIDE project

Abstract Background The development of novel diagnostics enables increasingly earlier diagnosis of Alzheimer's disease (AD). Timely diagnosis may benefit patients by reducing their uncertainty regarding the cause of symptoms, yet does not always provide patients with the desired certainty. Objective To examine, using both quantitative and qualitative methods, uncertainty communicated by memory clinic clinicians in post‐diagnostic testing consultations with patients and their caregivers. Methods First, we identified all uncertainty expressions of 22 clinicians in audiotaped post‐diagnostic testing consultations with 78 patients. Second, we statistically explored relationships between patient/clinician characteristics and uncertainty expressions. Third, the transcribed uncertainty expressions were qualitatively analysed, determining the topic to which they pertained, their source and initiator/elicitor (clinicians/patients/caregivers). Results Within 57/78 (73%) consultations, clinicians expressed in total 115 uncertainties, of which 37% elicited by the patient or caregiver. No apparent relationships were found between patient/clinician characteristics and whether or not, and how often clinicians expressed uncertainty. Uncertainty expressions pertained to ten different topics, most frequently patient's diagnosis and symptom progression. Expressed uncertainty was mostly related to the unpredictability of the future and limits to available knowledge. Discussion and conclusions The majority of clinicians openly discussed the limits of scientific knowledge and diagnostic testing with patients and caregivers in the dementia context. Noticeably, clinicians did not discuss uncertainty in about one quarter of consultations. More evidence is needed on the beneficial and/or harmful effects on patients of discussing uncertainty with them. This knowledge can be used to support clinicians to optimally convey uncertainty and facilitate patients' uncertainty management.


| INTRODUC TI ON
About 50 million people are currently living with dementia worldwide, and this number is rapidly increasing due to the ageing population. 1 Alzheimer's disease (AD) is the most common cause of dementia, with no existing cure yet. 2 Many aspects of this disease are still unknown. Clinical progression is, for example, difficult to predict, impairing patients' and their caregivers' ability to plan ahead and prepare for the future. Although uncertainty has always been inherent to the AD and dementia context, two recent developments have expanded the scope of uncertainty with which patients, caregivers and clinicians are confronted. First, additional uncertainty has resulted from the introduction of 'pre-dementia' diagnostic categories in memory clinic practice, such as 'mild cognitive impairment (MCI)'. 3,4 Roughly half of MCI patients develop dementia in the course of 3 years, while the other half remains stable or improves. 5 Whether, when and if so how MCI patients will develop dementia is still difficult to predict. 6,7 Second, novel diagnostic measures are developed to enable earlier and a more accurate diagnosis of AD. 8,9 Examples are tests detecting biomarkers through imaging techniques or in cerebrospinal fluid (CSF). 10,11 Earlier diagnostic testing might benefit some patients and their caregivers by reducing uncertainty regarding the cause and the course of the patient's symptoms. [12][13][14][15][16] However, these tests do not always provide the desired certainty. 17 The interpretation of diagnostic test results is often complicated, for instance when they are borderline abnormal or conflicting. 9 In addition, for individuals without dementia, results of AD-biomarker diagnostic tests only yield a risk indication of developing dementia within the next years. 13 Concluding, the introduction of both 'MCI' as a diagnostic category and (early) biomarker testing, may have heightened uncertainty for patients and caregivers in the diagnostic trajectory for (AD) dementia.
The phenomenon of uncertainty in illness can be described as someone's 'subjective consciousness or awareness of one's lack of knowledge'. [18][19][20][21][22] Uncertainty may relate to a wide variety of topics, varying from disease progression to test reliability. In addition, several distinct types or 'sources' of uncertainty can be distinguished, 18 that is (a) probability, (b) ambiguity and (c) complexity. 'Probability' refers to uncertainty caused by an inability to predict the future. Probability most frequently takes on the form of a risk, such as a 50% chance to develop dementia after an MCI diagnosis. 'Ambiguity' refers to uncertainty resulting from the lack of reliable, credible or adequate information about a phenomenon.
An example is the uncertainty resulting from conflicting test results. Finally, 'complexity' refers to uncertainty due to difficulties in comprehending aspects of the phenomenon itself-for example uncertainty about a definitive diagnosis resulting from the interplay between a multitude of complex factors. 18 Besides conceptual work on uncertainty, previous research so far has focused on how clinicians cope with uncertainty. Evidence outside of the explored relationships between patient/clinician characteristics and uncertainty expressions. Third, the transcribed uncertainty expressions were qualitatively analysed, determining the topic to which they pertained, their source and initiator/elicitor (clinicians/patients/caregivers).
Results: Within 57/78 (73%) consultations, clinicians expressed in total 115 uncertainties, of which 37% elicited by the patient or caregiver. No apparent relationships were found between patient/clinician characteristics and whether or not, and how often clinicians expressed uncertainty. Uncertainty expressions pertained to ten different topics, most frequently patient's diagnosis and symptom progression. Expressed uncertainty was mostly related to the unpredictability of the future and limits to available knowledge.

Discussion and conclusions:
The majority of clinicians openly discussed the limits of scientific knowledge and diagnostic testing with patients and caregivers in the dementia context. Noticeably, clinicians did not discuss uncertainty in about one quarter of consultations. More evidence is needed on the beneficial and/or harmful effects on patients of discussing uncertainty with them. This knowledge can be used to support clinicians to optimally convey uncertainty and facilitate patients' uncertainty management.

K E Y W O R D S
Alzheimer's disease, dementia, diagnostic work-up, memory clinic, physician-patient communication, uncertainty context of dementia suggests that clinicians vary in their recognition and acknowledgement of uncertainty, and are often hesitant to discuss uncertainty with patients. [23][24][25][26][27] A recent observational study in the field of cancer genetic counselling indicates that clinicians do express and address uncertainty, but to a widely varying degree. 28 Several factors may explain such variation, first of which is clinicians' general tendency in responding to uncertainty-their tolerance for uncertainty. 29 Clinicians' tolerance for uncertainty has been suggested to influence their openness in sharing uncertainty with patients. Second, clinicians may adapt the amount of uncertainty they convey depending on patient characteristics such as education level or diagnosis. Evidence supporting these hypotheses is, however, scarce. 30,31 We do not know yet to what extent and how clinicians in the (AD) dementia diagnostic context discuss uncertainty with patients and caregivers, or which characteristics predict uncertainty communication. Communicating uncertainty is challenging, but often cannot be avoided, especially when patients and/or caregivers are actively involved in the medical interaction by asking questions. To enable the development of strategies for clinicians to optimally address uncertainty and support patients' and caregivers' coping with uncertainty, 32 we need to first establish the spectrum of uncertainties communicated by clinicians within this context.
In this study, we aimed to examine if and how often clinicians communicate uncertainty in disclosure consultations in the memory clinic setting, to what topics the uncertainty relates and what its sources are. We also examined who elicited clinicians' expressions of uncertainty, that is whether clinicians expressed uncertainties on their own initiative or in response to remarks or questions by patients or caregivers. Additionally, we explored whether the observed amount of uncertainty expressions was related to clinician and/or patient characteristics, such as clinicians' tolerance of uncertainty and patients' diagnosis.

| Design and study context
In an observational design, we combined qualitative, descriptive and exploratory quantitative analysis of clinician-patient consultations in a memory clinic setting in which results of diagnostic testing for AD and dementia were discussed. This study was part of a larger research project entitled Alzheimer's Biomarkers in Daily Practice (ABIDE). 9 ABIDE addresses the value and use of diagnostic tests for AD to daily practice in memory clinics, including communication about test results. 9 ABIDE encompasses an observational study of clinical encounters at eight Dutch memory clinics. 33 During the routine diagnostic work-up for dementia at those clinics, we audiotaped clinician-patient consultations, that is the clinical encounter(s) prior to and after diagnostic testing. The current study focused on the latter encounters only. The board of the Medical Ethics Committee of the Academic Medical Centre (AMC) Amsterdam reviewed and approved of this study.

| Sample and procedure
All clinicians at the eight memory clinics, that is neurologists and geriatricians, were invited to participate in this study. They were eligible if they were involved in patient consultations during the diagnostic process and if they were willing to participate. Their newly referred patients and patients' informal caregivers who accompanied them to the memory clinic were invited to participate prior to their first visit at the memory clinic. Only patients with sufficient comprehension of the Dutch language who were willing and able to sign informed consent were included (based on clinicians' evaluation). In the current analyses, we included fully recorded consultations (June 2016-July 2017) in the Dutch language.

| Measures
Prior to audiotaping their consultations, clinicians completed a questionnaire assessing sociodemographic and work-related characteristics, that is their age, gender, specialty (neurology or geriatrics) and level of experience (years of experience and estimated number of new patients per month). In addition, clinicians' tolerance of uncertainty was assessed with nine items from the Physicians' Reaction to Uncertainty Scale (PRUS). 34 We used the subscales 'Anxiety due to uncertainty' and 'Reluctance to disclose uncertainty to patients', excluding one item on the use of treatment (6-point Likert scale ranging from 1: strongly disagree to 6: strongly agree). Item scores were summed to a total score, with a higher score indicating lower tolerance for uncertainty (range 9-54). Patients and their caregivers completed a questionnaire assessing sociodemographic characteristics, that is age, gender, educational level and (only for caregivers) their relation to the patient.
Patients also completed an adapted version of the PRUS to assess their tolerance of uncertainty. We used the subscales 'Anxiety due to uncertainty', 'Reluctance to disclose uncertainty to patients' and one item from the subscale 'Concern about bad outcomes' (six-point Likert scale strongly disagree to strongly agree).

| Analyses
Two trained research assistants (BA, AH) listened to all consultations twice and marked any instances in which they perceived the clinician to express uncertainty. They interpreted the content of the conversation and marked all verbal expressions reflecting the clinician's awareness of uncertainty or a lack of certainty (eg, 'that's uncertain' or 'I am not sure about…'), a chance, risk or probability (eg, 'there is a chance that…' or 'the risk of developing dementia is 50%'), or missing, ambiguous or indefinite knowledge or information (eg, 'we do not know that yet', 'that could still mean two things', or 'we don't know what will happen in the future'). 18 An 'expression' could be one or several sentences long. An expression was considered finished when the focus was no longer on uncertainty. All consultations in which no uncertainty expressions had been marked were double-checked by the first author (SP) for uncertainty expressions.
If in doubt, BA, AH and SP were instructed to mark the expression, so that it could be checked for inclusion by MH, an expert on uncertainty communication. Next, all sentences in which uncertainty expressions were identified were transcribed verbatim. We also transcribed the clinician-patient interaction immediately preceding the uncertainty expression (all sentences necessary to understand the uncertainty expression and the initiation of this topic).
We used descriptive statistics to report characteristics of clinicians and patients/caregivers and to report how often clinicians expressed uncertainty. To explore relationships between participant characteristics and uncertainty expressions (ie, the number of expressions and a dichotomous variable indicating if uncertainties were expressed 'yes' or 'no'), we used correlations, t tests, Kruskal-Wallis and Mann-Whitney U tests, and the Chi-square statistic, depending on the normality and type of data. Significance testing was done two-sided at an alpha of .05.
All quantitative analyses were performed using IBM SPSS Statistics 24.
Next, MAXQDA 12 software 35 was used to assist in further organizing and analysing the transcribed uncertainty expressions. All transcripts were independently read and double coded, that is by SP and MH or LV, who have backgrounds in psychology and are trained in qualitative data analysis. Codes were subsequently compared and discussed until consensus was reached. First, we identified to which topic the uncertainty expressions pertained, using inductive analysis aimed at generating coding categories from the data. 36 Second, we assessed whether three conceptually distinguished main sources of uncertainty, that is probability, ambiguity and complexity, and associated subcategories (see Table 2) were reflected in the current data. To that end, we used deductive analysis, originating from predefined codes. 18,37 Third, all clinicians' uncertainty expressions were categorized based on who initiated or elicited the uncertainty utterance: the clinician, the patient or the caregiver. Table 1 displays the characteristics of the 22 clinicians and 78 patients participating. Of these patients, 73 were accompanied by a caregiver, most often their partner (64%). Clinicians each recorded post-testing consultations with one to eight patients (Mdn = 3) with a mean duration of 18 minutes (SD = 8, range 3-39 minutes). We found no relationships between clinician or patient characteristics (all characteristics displayed in Table 1 Median test Chi-square = 6.26, P = .044). Pairwise comparisons indicated that clinicians expressed uncertainty more frequently when disclosing an MCI diagnosis, compared to a cognitively normal 'diagnosis' (Mann-Whitney U = 71.00, P = .038). This is visualized in

| Topics of uncertainty
Clinicians' uncertainty expressions related to ten different topics (see Table 2 In other cases, uncertainty related for example to conflicting test results, the effects of medication or heredity of the disease.

| Sources of uncertainty
For about half of the uncertainty expressions, the clinician acknowledged his/her inability to predict the future; that is, the uncertainty was due to probability (see Table 3, including representative quotes).
For example, the clinician addressed the risk of developing dementia (Extract 1), or conveyed uncertainty about predicting the course of the symptoms or underlying disease (Extract 2).
The other half of uncertainty expressions was caused by ambiguity, that is related to limits in the quality of information. We could further identify six subtypes of ambiguity, in line with previous conceptual work 29 (Table 3). Clinicians most frequently expressed uncertainty caused by: (a) limits in (the currently available) knowledge (incompleteness), for example when not all test results are yet known (Extract 4); (b) their inability to provide a single solution (indefinitiveness), for example the lack of a single diagnosis (Extract 5); or (c) their inability to provide a definitive answer (tentativeness), for example about the cause of the patient's symptoms (Extract 6). Other subcategories of ambiguity were found only occasionally; in these cases, the uncertainty was caused by inconsistency between test results (Extract 7), the existence of multiple possible meanings, or polysemousness (Extract 8) or by the impossibility to systematically test the potential effects of medication, that is insolubility (Extract 9).
Complexity was only occasionally the cause of the uncertainty expressed by clinicians in these consultations, for example uncertainty caused by the complexity of disease development and brain functions (Extract 10). Table 2 illustrates that different uncertainty topics were explained by different sources of uncertainty. Particularly, probability, that is the inability to predict the future, was the source of uncertainty in almost all instances in which clinicians expressed uncertainty about disease/symptom progression, the risk of developing a dementia or the effects of medication. For other topics, ambiguity was the primary source of uncertainty. For example, when clinicians expressed uncertainty about diagnostic labels or limitations of diagnostic tests, their uncertainty was almost exclusively explained by ambiguity.

| Elicitation of clinicians' uncertainty expressions
Of the 115 uncertainty utterances expressed by the clinicians, 72 (63%) were initiated by the clinician (Table 4)

| Discussion
This is the first study examining whether and to what extent clinicians discuss with patients the uncertainty associated with results of diagnostic testing at a memory clinic. Discussion of uncertainty is prevalent in these memory clinic consultations; in three-quarter of the consultations, uncertainty was addressed at least once. These uncertainties pertained to various topics, such as diagnosis, disease progression and heredity. Roughly half of them were related to clinicians' inability to predict the future ('probability') and the other half to limitations (eg, incompleteness, unreliability) to the knowledge available ('ambiguity'). This means that, if clinicians expressed uncertainty, they emphasized not only their inability to predict symptom and disease progression, but also the limits to scientific knowledge and to the reliability of diagnostic tests and test results. In contrast, requests for more certainty about disease progression or symptoms.
Future research should establish whether the advantages of early testing outweigh the disadvantage of having to deal with uncertainty. 42 If possible, patients should be given the choice between whether or not they want to undergo early diagnostic testing. 43,44 To facilitate that choice and help them manage their expectations, they need to be aware of possible uncertainties prior to testing. 45 A recent Delphi consensus study by our group indicated that clinicians, patients and caregivers agree about the importance of discussing the potentially uncertain result of diagnostic testing during pre-testing clinician-patient consultations. 46 Noticeably, in about one quarter of disclosure consultations, cli-  59 A useful three-step approach to provide support has been recently proposed. 60 First, clinicians can normalize uncertainty, by acknowledging patients' wish for more certainty while explaining that uncertainty is unfortunately inherent to the situation. Second, they can acknowledge and address patients' and caregivers' emotions regarding uncertainty by acknowledging that it is unpleasant not to know things. Third, they can help patients and caregivers cope with uncertainty by stimulating them to focus on living in the here and now instead of dwelling on the uncertainty. The effectiveness of these and other proposed strategies remains to be investigated.

| Strengths and limitations
Among the strengths of this study are its multicentre design, in which we sampled 78 consultations at eight different Dutch memory clinics. Moreover, we were able to gain rich insight into the data by Dementia Mild cognitive impairment Cognitively normal

Number of uncertainty expressions per consultation/patient
Frequency also be an interesting characteristic to examine more thoroughly.
Results from a recent survey study indicate that clinicians who are less tolerant for uncertainty prefer a more paternalistic approach in medical decision making. 61 If clinicians' tolerance for uncertainty is an important factor for the extent to which they share uncertain information with patients, interventions aimed at enhancing tolerance for uncertainty may eventually contribute to a more adequate discussion of uncertainty. Second, response bias might be a limitation in that sense that clinicians who are more comfortable with communication, including communication of uncertainty, may have been more likely to participate in this study.  'What I'm trying to explain is that it is possible we are dealing with someone who had an accident in the past and is now having normal aging in the brain. The effects from the past are more noticeable, because the reserve capacity of the brain is decreasing, because you're getting older. That is one possibility. The other possibility is that we do have to think about a disease of the brain, that is causing early aging of the brain. So, indeed a form of dementia.' (p6103; Male; 69 y; dementia) Insolubility Impossibility of something to be worked out or explained Extract 9: 'The effect of the medication is difficult to predict. Some experience more benefits than others. What we sometimes notice is that some people become more alert. It is a comparison you cannot make, because you cannot compare yourself with how you were without the medication, that is the whole point.' (p1106; Male; 73 y; dementia)

| Future research directions
Complexityone relevant subcategory: Features of information that limit its understanding Complexity Intricateness, multidimensionality or multifacetedness of information

| CON CLUS ION
In addition to communicating their inability to predict symptom and disease progression, many clinicians openly communicated the limits of scientific knowledge and diagnostic testing with patients.
However, in approximately one quarter of consultations concerning the disclosure of results of diagnostic testing in the context of dementia no uncertainties were communicated. More research is warranted to investigate the beneficial and/or harmful impact of uncertainty communication, to establish how clinicians can optimally discuss uncertainty with patients, that is in a way that would benefit and not harm patients. Eventually, such research can result in useful advice and/or communication tools for memory clinic clinicians, thereby improving the quality of care provided to patients and their caregivers.

ACK N OWLED G EM ENTS
We thank all patients, caregivers and clinicians who so generously contributed to this study. Furthermore, we thank Bahar Azizi and Anneke Hellinga for their assistance during data coding. All funding is paid to the institution.

DATA AVA I L A B I L I T Y S TAT E M E N T
Research data are not shared (for privacy reasons).