Diagnosis最新文献

Objectives: Evaluate the impact of problem representation (PR) characteristics on Generative Artificial Intelligence (GAI) diagnostic accuracy.

Methods: Internal medicine attendings and residents from two academic medical centers were given a clinical vignette and instructed to write a PR. Deductive content analysis described the characteristics comprising each PR. Individual PRs were input into ChatGPT-4 (OpenAI, September 2023) which was prompted to generate a ranked three-item differential. The ranked differential and the top-ranked diagnosis were scored on a 3-part scale, ranging from incorrect, partially correct, to correct. Logistic regression evaluated individual PR characteristic's impact on ChatGPT accuracy.

Results: For a three-item differential, accuracy was associated with including fewer comorbidities (OR 0.57, p=0.010), fewer past historical items (OR 0.60, p=0.019), and more physical examination items (OR 1.66, p=0.015). For ChatGPT's ability to rank the true diagnosis as the single-best diagnosis, utilizing temporal semantic qualifiers, more semantic qualifiers overall, and adhering to a typical 3-part PR format all correlated with diagnostic accuracy: OR 3.447, p=0.046; OR 1.300, p=0.005; OR 3.577, p=0.020, respectively.

Conclusions: Several distinct PR factors improved ChatGPT diagnostic accuracy. These factors have previously been associated with expertise in creating PR. Future studies should explore how clinical input qualities affect GAI diagnostic accuracy prospectively.

Objectives: The measurement of fasting glucose is a common practice for lowering the risk of hyperglycemia before an oral glucose tolerance test (OGTT). In this study we analyze advantages and limitations of near-patient measurement of capillary fasting glucose with a portable glucometer or blood sampling and measurement of plasma glucose with laboratory instrumentation.

Methods: The final study population consisted of 241 subjects (mean age: 36 ± 8 years; 97.9 % pregnant women) referred to our local phlebotomy center for an OGTT. Fasting glucose was measured in capillary blood using a near-patient glucometer (glucometer-based strategy) and in plasma with laboratory instrumentation using the hexokinase reference assay (laboratory-based strategy).

Results: The mean turnaround time from sample collection to obtaining the glucose value was longer with the laboratory-based strategy (32 min 8 vs. 8 s). The imprecision of the glucometer was higher than that of the laboratory assay (3.4 vs. 0.8 %). A negative bias of -3.3 % in fasting glucose was found with the glucometer compared to the laboratory measurement. The diagnostic accuracy, sensitivity and specificity of the glucometer for detecting fasting glucose values ≥7.0 mmol/L were 99.2 , 50.0 and 100.0 % compared to the laboratory assay. The glucometer-based strategy had an incremental cost of 0.17€ per patient compared to the laboratory-based strategy.

Conclusions: Screening fasting glucose in capillary blood with a near-patient glucometer instead of measuring fasting plasma glucose with laboratory instrumentation allows faster patient management in the phlebotomy center but is associated with higher imprecision, inaccuracy, costs and avoidable finger pricks.

Clinical reasoning is a quintessential aspect of medical training and practice, and is a topic that has been studied and written about extensively over the past few decades. However, the predominant conceptualisation of clinical reasoning has insofar been extrapolated from cognitive psychological theories that have been developed in other areas of human decision-making. Till date, the prevailing model of understanding clinical reasoning has remained as the dual process theory which views cognition as a dichotomous two-system construct, where intuitive thinking is fast, efficient, automatic but error-prone, and analytical thinking is slow, effortful, logical, deliberate and likely more accurate. Nonetheless, we find that the dual process model has significant flaws, not only in its fundamental construct validity, but also in its lack of practicality and applicability in naturistic clinical decision-making. Instead, we herein offer an alternative Bayesian-centric, intuitionist approach to clinical reasoning that we believe is more representative of real-world clinical decision-making, and suggest pedagogical and practice-based strategies to optimise and strengthen clinical thinking in this model to improve its accuracy in actual practice.

Objectives: Diagnostic reconciliation is the collaborative process between patients and clinicians to create and reconcile evidence-based, feasible, and desirable care plans. However, the specific components of this process remain unclear. The objective of this study was to develop the first comprehensive framework to elucidate the diagnostic reconciliation process.

Methods: We followed a multi-step and iterative approach to develop the framework, including a focused systematic review of diagnostic conversations, quantitative evaluation of recordings of real-life clinical visits recordings, and stakeholder engagement (e.g., patients, clinicians, researchers).

Results: We identified 17 potential components to the process of diagnostic reconciliation through literature review and stakeholder engagement. After review of 56 clinical visits and further stakeholder engagement, we developed a final framework including four categories: 1) understanding the need for a test/referral, 2) logistics of test/referral scheduling, 3) test/referral information, and 4) test/referral results.

Conclusions: The proposed framework lays the foundation for evaluation and improvement of diagnostic conversations in practice. Clinicians can enhance patient-centered diagnosis by co-creating diagnostic plans of care in practice and using the components described in the novel diagnostic reconciliation framework.

Objectives: We primarily assessed differences in differential diagnosis (DDx) efficacy of initial and refined top diagnoses (tDDx) and "can't miss" DDx (CMDx) between 3 MMTs (Constellations, Mental CT, and VINDICATES).

Methods: Pediatric clerkship students participated in two 1-h case-based sessions. The case was presented in three aliquots. Students were randomly assigned to MMT groups. Assigned MMTs were used to generate the initial tDDx and CMDx following aliquot 1. tDDx and CMDx were refined following both aliquots 2 and 3. Group DDx responses and student affective data were collected via survey. DDx efficacy was defined using pooled faculty responses and scoring was done by consensus.

Results: There was no significant difference in scores between MMT groups, except the second iteration of CMDx in Case A (Constellations 50 % [interquartile range, IQR, 50-100], Mental CT 50 % [50-100], VINDICATES 0 % [0-50], p=0.02). Students' self-reported confidence in generating (p<0.001) and refining (p<0.001) their DDx significantly increased after the curriculum.

Conclusions: Although prior studies identified a differential effect of MMTs on DDx generation, we did not observe a difference in initial or refined DDx efficacy between MMTs. .

Background: Diagnostic scope is the range of diagnoses found in a clinical setting. Although the diagnostic scope is an essential feature of training and evaluating artificial intelligence (AI) systems to promote diagnostic excellence, its impact on AI systems and the diagnostic process remains under-explored.

Content: We define the concept of diagnostic scope, discuss its nuanced role in building safe and effective AI-based diagnostic decision support systems, review current challenges to measurement and use, and highlight knowledge gaps for future research.

Summary: The diagnostic scope parallels the differential diagnosis although the latter is at the level of an encounter and the former is at the level of a clinical setting. Therefore, diagnostic scope will vary by local characteristics including geography, population, and resources. The true, observed, and considered scope in each setting may also diverge, both posing challenges for clinicians, patients, and AI developers, while also highlighting opportunities to improve safety. Further work is needed to systematically define and measure diagnostic scope in terms that are accurate, equitable, and meaningful at the bedside. AI tools tailored to a particular setting, such as a primary care clinic or intensive care unit, will each require specifying and measuring the appropriate diagnostic scope.

Outlook: AI tools will promote diagnostic excellence if they are aligned with patient and clinician needs and trained on an accurately measured diagnostic scope. A careful understanding and rigorous evaluation of the diagnostic scope in each clinical setting will promote optimal care through human-AI collaborations in the diagnostic process.

Objectives: Time pressure and time constraints have been shown to affect diagnostic accuracy, but how they interact is not clear. The current study aims to investigate the effects of both perceived time pressure (sufficient vs. insufficient time) and actual time constraints (lenient vs. restricted time limit) with regard to diagnostic accuracy.

Methods: Residents from two university-affiliated training programs in the USA participated in this online within-subjects experiment. They diagnosed cases under two perceived time pressure conditions: one where they were told they had sufficient time to diagnose the cases and one where they were told they had insufficient time. The actual time limit was either restricted or lenient (± one standard deviation from the mean time to diagnose). Participants provided their most likely diagnosis and a differential diagnosis for each case, and rated their confidence in their most likely diagnosis.

Results: A restricted time limit was associated with lower accuracy scores (p=0.044) but no effects of perceived time pressure on diagnostic accuracy were found. However, participants self-reported feeling more time pressure when they thought they had insufficient time (p<0.001). In addition, there was an effect of the actual time limit (p=0.012) and perceived time pressure (p=0.048) on confidence.

Conclusions: This study showed that a restricted time limit can negatively affect diagnostic accuracy. Although participants felt more time pressure and were less confident when they thought they had insufficient time, perceived time pressure did not affect diagnostic accuracy. More research is needed to further investigate the effects of time pressure and time limits on diagnostic accuracy.

Nearly a decade after the National Academy of Medicine released the "Improving Diagnosis in Health Care" report, diagnostic errors remain common, often leading to physical, psychological, emotional, and financial harm. Despite a robust body of research on potential solutions and next steps, the translation of these efforts to patient care has been limited. Improvement initiatives are still narrowly focused on selective themes such as diagnostic stewardship, preventing overdiagnosis, and enhancing clinical reasoning without comprehensively addressing vulnerable systems and processes surrounding diagnosis. To close this implementation gap, the US Centers for Disease Control and Prevention (CDC) released the Core Elements of Hospital Diagnostic Excellence programs on September 17, 2024. This initiative aligns with the World Health Organization's (WHO) 2024 World Patient Safety Day focus on improving diagnosis. These Core Elements provide guidance for the formation of hospital programs to improve diagnosis and aim to integrate various disparate efforts in hospitals. By creating a shared mental model of diagnostic excellence, the Core Elements of Diagnostic Excellence supports actions to break down silos, guide hospitals toward multidisciplinary diagnostic excellence teams, and provide a foundation for building diagnostic excellence programs in hospitals.

Objectives: To investigate longitudinal trends in the incidence, preventability, and causes of DAEs (diagnostic adverse events) between 2008 and 2019 and compare DAEs to other AE (adverse event) types.

Methods: This study investigated longitudinal trends of DAEs using combined data from four large Dutch AE record review studies. The original four AE studies included 100-150 randomly selected records of deceased patients from around 20 hospitals in each study, resulting in a total of 10,943 patient records. Nurse reviewers indicated cases with potential AEs using a list of triggers. Subsequently, experienced physician reviewers systematically judged the occurrence of AEs, the clinical process in which these AEs occurred, and the preventability and causes.

Results: The incidences of DAEs, potentially preventable DAEs and potentially preventable DAE-related deaths initially declined between 2008 and 2012 (2.3 vs. 1.2; OR=0.52, 95 % CI: 0.32 to 0.83), after which they stabilized up to 2019. These trends were largely the same for other AE types, although compared to DAEs, the incidence of other AE types increased between 2016 (DAE: 1.0, other AE types: 8.5) and 2019 (DAE: 0.8, other AE types: 13.0; rate ratio=1.88, 95 % CI: 1.12 to 2.13). Furthermore, DAEs were more preventable (p<0.001) and were associated with more potentially preventable deaths (p=0.016) than other AE types. In addition, DAEs had more and different underlying causes than other AE types (p<0.001). The DAE causes remained stable over time, except for patient-related factors, which increased between 2016 and 2019 (29.5 and 58.6 % respectively, OR=3.40, 95 % CI: 1.20 to 9.66).

Conclusions: After initial improvements of DAE incidences in 2012, no further improvement was observed in Dutch hospitals in the last decade. Similar trends were observed for other AEs. The high rate of preventability of DAEs suggest a high potential for improvement, that should be further investigated.