Design

We applied the well-established, systematic qualitative research methodology, grounded theory [21]. All study procedures were approved by the Cancer Institute NSW and the University of Sydney Human Research Ethics Committee [#15245]. GPs had an opportunity to discuss the study, and gave written consent, prior to participation.

Participants and Setting

We recruited a sample of 69 GPs (40 Australian, 29 UK) for this study. In Australia we advertised via the newsletters and email lists of GP organisations (Medicare Locals) in Sydney, in mass and social media, and in medical journals. Rural GPs were accessed by phoning practice managers, through colleagues, and advertising with rural Medicare Locals [22].

We included GPs from the United Kingdom to also explore PSA testing reasoning and practice in a jurisdiction with comparatively lower rates of prostate cancer screening than Australia. We subsequently recruited 29 GPs throughout England (n = 24) and Scotland (n = 5). Our initial sample of GPs responded to an invitation distributed by academic colleagues through professional networks. We then broadened our sample by advertising via email to members of the Royal College of General Practitioners (RCGP), primary health care departments, university academic departments, and general practice and research mail lists. We also advertised via newsletter including the Society for Academic Primary Care (SAPC) and RCGP Scotland’s eBulletin.

GPs were invited to contact KP if they were interested and willing to participate. Participating GPs were of varying ages, clinical experience, gender, and patient populations. All GPs who expressed interest in participating were included. GPs were compensated financially for their time.

Interviews / Data Collection

We generated data via in-depth interviews. The semi-structured interview schedule covered a broad range of topics, including GPs’ recent clinical encounters involving PSA testing decisions; communicating information; screening pathways; and overdiagnosis. The schedule was modified between interviews, informed by the developing analysis. Uncertainty was not specifically included as a topic for discussion in the schedule; rather it was a recurring concept that was identified during data analysis. Interviews with Australian GPs took place between March 2013 and June 2014 and with UK GPs between September and December 2014. They were all conducted by KP, mostly by telephone and Skype, and ranged in duration from 18 to 70 minutes. All interviews were audio-recorded and transcribed verbatim.

Data Coding and Analysis

The analysis was led by KP, who coded the transcripts and wrote detailed memos which were regularly reviewed and discussed by the authors in analysis meetings. A subset of transcripts was read and coded by three authors independently; this coding was compared and discussed to inform the development of the central concepts in the study.

A longstanding point of contention in grounded theory methodology is the relation between the theory being produced, which is ‘grounded’ in the data collected, and existing relevant theory. While early expressions of grounded theory methodology [23] strongly emphasised the development of new theory as opposed to the testing of existing theory, contemporary mainstream grounded theorists strongly concur that qualitative empirical work must be conducted in the context of existing knowledge [24]. As uncertainty was identified as a core category in our data analysis, we turned to the literature to develop a better understanding of the concept, and identified Han’s taxonomy of uncertainty in health care [14]. This taxonomy resonated with our interpretation of the data and suggested face validity for our early analysis of the sources and issues of uncertainty. We used Han’s taxonomy to develop our analysis of GPs’ experiences with uncertainty and categorised our data according to the ‘sources’ and ‘issues’ of uncertainty as described in the framework. In addition, we developed a new set of concepts related to how GPs respond to uncertainty in PSA testing, an issue that was not included in Han’s typology.

Where Does GPs’ Uncertainty Come from and What Is the Uncertainty About?

Table 1 outlines sources of GPs’ uncertainty about PSA testing. We have included Han’s definition of each source of uncertainty followed by a summary of how GPs described this type of uncertainty manifesting in their practice.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sources of uncertainty (where is GP uncertainty coming from?).

https://doi.org/10.1371/journal.pone.0153299.t001

Table 2 captures the issues of uncertainty, again presented with Han’s definition of the issues followed by specific examples in each cell from our data. Han’s framework characterises ‘personal uncertainty’ as patient-centred. However because this study focused on the perspective of GPs, our data also includes personal uncertainty with a locus in GPs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Issues of uncertainty (what is the uncertainty ABOUT?).

https://doi.org/10.1371/journal.pone.0153299.t002

Our data indicate that GPs may experience a diverse range of uncertainties with respect to PSA testing. There were important differences however, between Australian and UK GPs’ descriptions regarding what their uncertainty was about. Australian GPs’ uncertainty was related to all three of the issues described above: scientific, practical, and personal. UK GP experiences of uncertainty were mostly about personal issues, because (a) GPs were clear about procedural expectations coming from government and medical bodies about PSA testing, and scientific uncertainty was dealt with via clear guidelines and norms; and (b) UK GPs expressed a sense that the medical profession was collectively managing the uncertainty so individual GPs were facing less practical uncertainty because the UK system has processes in place to help them manage it. Thus the UK GPs’ uncertainty was predominantly patient-centred; they were mostly concerned that when their patients sought or asked about PSA testing, they were then burdened with uncertainty due to the uncertain nature and quality of the available information. UK GPs did not feel uncertain themselves but worried that their ‘patients have to make well-informed decisions and I suppose that’s where it’s such a minefield of uncertainty, it must be very difficult for people to say that they’ve made a well-informed decision’ (UKGP26).

Below we outline strategies GPs described using to handle uncertainty in the context of PSA testing of asymptomatic men. The information below draws primarily from the Australian data; where observations are based on UK data this is noted.

What Strategies Do GPs Use to Manage Uncertainty?

We identified three main approaches GPs used when faced with uncertainty around PSA testing, specifically about decision-making and responsibility:

1. Sometimes GPs took charge of uncertainty, considered it a usual feature of their daily practice, and managed (at least some) uncertainties on their own
2. Sometimes GPs engaged others in managing uncertainty: they accepted some uncertainty as a challenge and used it as an opportunity to engage the medical profession, colleagues, and patients about how the uncertainty would be handled, to enable them to better support patients and find shared solutions
3. Sometimes GPs sought to transfer to other parties the responsibility for reducing or managing some uncertainties.

Some GPs tended to manage all types of uncertainty about PSA testing in a relatively consistent way, adopting one of the three approaches described above (i.e. some GPs usually took charge, usually accepted and engaged, or usually transferred, although this was never absolute). Other GPs were more variable in their approach, applying different management strategies in different situations. Particular sources of uncertainty also tended to elicit particular types of responses. So GPs who tended to use different strategies to manage different types of uncertainties may call upon one or all of the techniques depending on the type of uncertainty, the particular situation, and the GPs individual interpretation of it and level of tolerance.

We describe the three categories of GPs approaches below.

How Do GPs Respond to Different Sources of Uncertainty?

Probabilistic uncertainty was identified as a major source of uncertainty by the GPs, yet they tended to speak of it as being more tolerable than uncertainties arising from ambiguity and complexity. Overall, GPs accepted responsibility for probabilistic uncertainty and shared their knowledge about it; it was not as common for GPs to attempt to transfer responsibility for this source of uncertainty. Many GPs described handling probabilistic uncertainty reasonably comfortably on their own. They saw it as central to the GP role and to clinical judgment, which necessarily involves interpreting scientific evidence and probabilities. Individuals or organisations were not blamed or held directly responsible for probabilistic uncertainty. GPs described probabilistic uncertainty as challenging (e.g. indeterminate medical outcomes), but had mostly found a relatively settled way of dealing with it because they had few alternatives.

The two other types of uncertainty invoked different responses: uncertainty arising from ambiguity—for example, varied recommendations—and complexity—the vagueness of clinical descriptors, or the difficulty in judging patient understanding. Many GPs preferred to negotiate or transfer responsibility for these uncertainties. For example, a GP experiencing uncertainty in a testing decision might tell the patient about discordant recommendations to justify their uncertainty and immediately refer to specialists for further opinion. Some GPs appeared to hold that responsibility for ambiguity or complexity should be collective and others should be involved in negotiating decisions, or they preferred to transfer those decisions to specialists. Ambiguity and complexity commonly led to practical issues for GPs arising from this uncertainty–how to engage in ‘good’ practice and appropriate communication based on complex evidence and ambiguous guidance. Ambiguity and complexity were also a foundation for GPs’ sense of personal burden, as some GPs expressed anxiety about their capacity to make ‘good’ decisions without good evidence, and ability to judge what level of information was suitable to support consent. Theoretically these sources of uncertainty could be modified or improved with system change (for example, consistent recommendations, established consent protocols) but GPs felt limited in their capacity to make changes at the clinical level.

Sources, Issues, and Management of Uncertainty

GPs in our study appeared to manage uncertainty arising from probabilistic information with reasonable confidence when compared to the other sources of uncertainty. One possible explanation for this finding is that GPs are familiar with probabilities and are trained to interpret and manage probabilistic information in the clinical setting. Numerous resources are available presenting probabilistic and statistical information in multiple formats (numerically, graphically) to assist GPs and to support the decision making capacity of their patients. While probabilistic uncertainty is particularly challenging in relation to PSA testing and cannot be readily resolved with leaflets or information sheets, we propose that such resources ‘normalise’ the probabilistic uncertainty inherent to this context. It is possible that GPs feel more comfortable taking charge or engaging with others about probabilistic uncertainty because they are better able to get a handle on the uncertainty they are dealing with because it is an ongoing and familiar type of uncertainty. Probabilistic information about outcomes is also the typical kind of information assumed to be shared in processes of shared decision-making.

It is clear from our results that uncertainty from ambiguity on the other hand was extremely frustrating for GPs; GPs were uncertain about guidance from medical authorities and were unclear of their clinical and legal obligations. These uncertainties are grounded in ambiguity, which, as the decision psychology literature has shown, people generally prefer to avoid (e.g. [25]. Ambiguity is more directly related to uncertainties about what to do in practice than probabilistic uncertainty, and perhaps leaves more scope for GPs to take a “wrong” course of action. Aspects of ambiguity in this context are potentially modifiable, which could reduce some uncertainty for GPs. For example, clearer and more consistent expert guidance may assist. Yet even if more ‘certainty’ was implemented at the system-level via clear guidelines or consistent advice from authoritative sources, clinicians will inevitably still experience some personal uncertainty stemming from the complexity of this information, and because of diverse GP and patient value systems in the clinical context (e.g. questioning appropriateness of recommendations).

Complexity was also experienced as particularly uncomfortable for GPs; some GPs described feeling uncertain about their own clinical judgment. Complexity is essentially a source of uncertainty amenable to individual value judgment, interpretation, and assumptions. It is therefore difficult to offer training or advice to GPs about ideal management of complexity uncertainty in relation to PSA testing. While clearer expert opinion may lessen ambiguity, GPs are left to deal with complexity at the clinical level with the individual patient. Our data from UK GPs illustrates that even with system-level guidance and clear ‘best practice’ formulations, GPs continue to experience personal uncertainty because every patient and consultation involving PSA testing is different and involves difficult value judgments. Clinicians will invariably face ongoing uncertainty about the nature of medical evidence and individual and distributive health care [26].

The Burden of Uncertainty

Ordering PSA tests for screening purposes is paradoxical in relation to uncertainty. Testing is a response to uncertainty: an attempt to better predict a man’s risk of developing prostate cancer. Yet while PSA testing aims to reduce uncertainty, the characteristics of the test when used as a screening tool mean that it tends to introduce more scientific, practical, and personal uncertainty than it eliminates. Faced with a PSA test result, the clinical significance of which is intrinsically uncertain, some GPs actively tried to lessen uncertainty by looking to guidelines on which to base decisions, or by sharing their knowledge with patients. However these strategies also tended to compound uncertainty, piling up knowledge about the uncertainties regarding outcomes, and GPs felt further burdened as a result. This is the great paradox of the PSA. If GPs chase certainty via more and more testing and investigation, this may create more uncertainty. Testing can undoubtedly create certainty for some GPs and patients at the individual level, but it comes at a cost of lots of uncertainty, including GPs’ top line uncertainty about using the PSA test at all. This, as our research has demonstrated, is what inflicts considerable burden on GPs: because they have burdened patients with information which cannot provide any clear answer, because their efforts to locate certainty have gone unrewarded, and because of remaining uncertainty about what counts as good practice.

Uncertainty and the associated struggle is not just a burden to be borne by individual practitioners. There is an expectation that GPs will practice in line with professional standards and evidence-based medicine. If GPs are expected to act according to these standards, it would be reasonable to suppose that the health care system might owe GPs some reciprocal supports to make it possible for them to do so.

Ethical Considerations

There are judgments to be made about how best to involve men in PSA testing decisions. Our data touches on uncertainties about whether, when, and in what forms communicating about uncertainty is considered appropriate. Communicating uncertainty is not simply about probabilities; ambiguity and complexity are also key sources of GP uncertainty and subsequent burden. Any attempt to guide communication about uncertainty in PSA testing practice may be more effective if it addresses all three sources of uncertainty.

Ethical questions remain in relation to presenting uncertainty to patients who wish to avoid making medical decisions; coping with uncertainty is known to be a source of substantial stress and anxiety for many patients (e.g. [27]. Manson and O’Neill argue that a trusting and responsive relationship might be more valued by patients than detailed information exchange [28]. Perhaps in some situations, a “taking charge” approach, whereby GPs call on their professional experience and apply ‘rules of thumb’ to direct their practice might be appropriate, while in more complex situations, GPs may need to discuss and reach agreement with others regarding how decisions based on uncertainty will be allocated and resolved.

More work is needed to clarify under what circumstances uncertainty should be communicated to patients, and if so what aspect of the uncertainty construct should be addressed. Parascandola et al argue that respectful interaction with patients requires disclosure about uncertainty even when it is not offered to gain consent or in the service of patient decision making; as long as patients understand the general decision making context [29]. Ideally, doctors need to feel supported in their dealings with uncertainty. Research suggests that when doctors are comfortable with uncertainty and collaborate with patients in their medical care, patient trust and satisfaction are high [29].

Limitations

This analysis was inductive rather than commencing with research questions about uncertainty. Further research could test our findings and explore the impacts of uncertainty in more depth.