Diagnostic uncertainty in primary care: what is known about its communication, and what are the associated ethical issues?

Cox, Caitríona; Miller, Benjamin M.; Kuhn, Isla; Fritz, Zoë

doi:10.1093/fampra/cmab023

Cited by 27 publications

(24 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This implies that sharing uncertainty may be more common than reported, a noteworthy finding especially in this time-pressured acute care setting. In our study uncertainty expressions can be depicted as a communication strategy and a way to explain diagnostic reasoning to patients, something suggested in previous studies [ 27 , 28 ].…”

Section: Discussionmentioning

confidence: 95%

Anatomy of diagnosis in a clinical encounter: how clinicians discuss uncertainty with patients

et al. 2022

View full text Add to dashboard Cite

Background Studies consider the clinical encounter as linear, comprising six phases (opening, problem presentation, history-taking, physical examination, diagnosis, treatment and closing). This study utilizes formal conversation analysis to explore patient-physician interactions and understanding diagnostic utterances during these phases. Methods This study is a qualitative sub-analysis that explores how the diagnosis process, along with diagnostic uncertainty, are addressed during 28 urgent care visits. We analyzed physicians’ hypothesis-generation process by focusing on: location of diagnostic utterances during the encounter; whether certain/uncertain diagnostic utterances were revised throughout the encounter; and how physicians tested their hypothesis-generation and managed uncertainty. We recruited 7 primary care physicians (PCPs) and their 28 patients from Brigham and Women’s Hospital (BWH) in 3 urgent care settings. Encounters were audiotaped, transcribed, and coded using NVivo12 qualitative data analysis software. Data were analyzed inductively and deductively, using formal content and conversation analysis. Results We identified 62 diagnostic communication utterances in 12 different clinical situations. In most (24/28, 86%) encounters, the diagnosis process was initiated before the diagnosis phase (57% during history taking and 64% during physical examination). In 17 encounters (61%), a distinct diagnosis phase was not observed. Findings show that the diagnosis process is nonlinear in two ways. First, nonlinearity was observed when diagnostic utterances occurred throughout the encounter, with the six encounter phases overlapping, integrating elements of one phase with another. Second, nonlinearity was noted with respect to the resolution of diagnostic uncertainty, with physicians acknowledging uncertainty when explaining their diagnostic reasoning, even during brief encounters. Conclusions Diagnosis is often more interactive and nonlinear, and expressions of diagnostic assessments can occur at any point during an encounter, allowing more flexible and potentially more patient-centered communication. These findings are relevant for physicians’ training programs and helping clinicians improve their communication skills in managing uncertain diagnoses.

show abstract

Section: Discussionmentioning

confidence: 95%

Anatomy of diagnosis in a clinical encounter: how clinicians discuss uncertainty with patients

et al. 2022

View full text Add to dashboard Cite

show abstract

“…There were limited findings regarding uncertainty, compared with those concerning risk. Uncertainty in medicine is an under-researched area3 54–56 thus eRATs are influencing aspects of the consultation which are already poorly understood.…”

Section: Discussionmentioning

confidence: 99%

How do electronic risk assessment tools affect the communication and understanding of diagnostic uncertainty in the primary care consultation? A systematic review and thematic synthesis

et al. 2022

View full text Add to dashboard Cite

ObjectivesTo conduct a systematic review and synthesise qualitative research of electronic risk assessment tools (eRATs) in primary care, examining how they affect the communication and understanding of diagnostic risk and uncertainty. eRATs are computer-based algorithms designed to help clinicians avoid missing important diagnoses, pick up possible symptoms early and facilitate shared decision-making.DesignSystematic search, using predefined criteria of the published literature and synthesis of the qualitative data, using Thematic Synthesis. Database searches on 27 November 2019 were of MEDLINE, Embase, CINAHL and Web of Science, and a secondary search of the references of included articles. Included studies were those involving electronic risk assessment or decision support, pertaining to diagnosis in primary care, where qualitative data were presented. Non-empirical studies and non-English language studies were excluded. 5971 unique studies were identified of which 441 underwent full-text review. 26 studies were included for data extraction. A further two were found from citation searches. Quality appraisal was via the CASP (Critical Appraisal Skills Program) tool. Data extraction was via line by line coding. A thematic synthesis was performed.SettingPrimary care.ResultseRATs included differential diagnosis suggestion tools, tools which produce a future risk of disease development or recurrence or calculate a risk of current undiagnosed disease. Analytical themes were developed to describe separate aspects of the clinical consultation where risk and uncertainty are both central and altered via the use of an eRAT: ‘Novel risk’, ‘Risk refinement’, ‘Autonomy’, ‘Communication’, ‘Fear’ and ‘Mistrust’.ConclusioneRATs may improve the understanding and communication of risk in the primary care consultation. The themes of ‘Fear’ and ‘Mistrust’ could represent potential challenges with eRATs.Trial registration numberCRD219446.

show abstract

“…Indeed, reasoning under uncertainty is a linchpin of human cognition (Lake et al 2017) and has been shown to be a central component of "good" decision-making (Laidlaw and Russell 2021;Bhatt et al 2021). Fields which have traditionally de-emphasized uncertainty, such as in the medical profession, are now encouraging trainees to recognize and thoroughly consider when there is uncertainty in a diagnosis or treatment plan in an effort to improve medical safety (Hall 2002;Platts-Mills, Nagurney, and Melnick 2020;Cox et al 2021). Hence, annotation schemes which enable the expression of probability judgements, particularly in datasets wherein the underlying "ground truth" or the "true" label is unknown, are sensible and desirable for improving ma- chine safety, trustworthiness, and efficacy.…”

Section: Discussionmentioning

confidence: 99%

Eliciting and Learning with Soft Labels from Every Annotator

Collins¹,

Bhatt²,

Weller³

2022

Preprint

View full text Add to dashboard Cite

The labels used to train machine learning (ML) models are of paramount importance. Typically for ML classification tasks, datasets contain hard labels, yet learning using soft labels has been shown to yield benefits for model generalization, robustness, and calibration. Earlier work found success in forming soft labels from multiple annotators' hard labels; however, this approach may not converge to the best labels and necessitates many annotators, which can be expensive and inefficient. We focus on efficiently eliciting soft labels from individual annotators. We collect and release a dataset of soft labels for CIFAR-10 via a crowdsourcing study (N = 248). We demonstrate that learning with our labels achieves comparable model performance to prior approaches while requiring far fewer annotators. Our elicitation methodology therefore shows promise towards enabling practitioners to enjoy the benefits of improved model performance and reliability with fewer annotators, and serves as a guide for future dataset curators on the benefits of leveraging richer information, such as categorical uncertainty, from individual annotators.* Both authors contributed equally.setting, specifically CIFAR-10 (Hinton et al. 2012). Fig. 1 illustrates how our method compares to that of Peterson et al. (2019). We highlight the following contributions:• We introduce an efficient approach to elicit soft labels from individual annotators and will release the code for our elicitation interface. • We release our dataset of 6,200 soft labels over 1,000 image datapoints from CIFAR-10. We call this new dataset of soft labels CIFAR-10S. • We show that models trained with CIFAR-10S obtain similar performance (in terms of accuracy, robustness, and calibration) to models trained on CIFAR-10H with approximately 8.5x fewer annotators.

show abstract

Diagnostic uncertainty in primary care: what is known about its communication, and what are the associated ethical issues?

Cited by 27 publications

References 43 publications

Anatomy of diagnosis in a clinical encounter: how clinicians discuss uncertainty with patients

Anatomy of diagnosis in a clinical encounter: how clinicians discuss uncertainty with patients

How do electronic risk assessment tools affect the communication and understanding of diagnostic uncertainty in the primary care consultation? A systematic review and thematic synthesis

Eliciting and Learning with Soft Labels from Every Annotator

Contact Info

Product

Resources

About