2nd European Conference on Diagnostic Error in Medicine August 30-31, 2018, Bern, Switzerland

David Schwappach, Jason Maude, Juliane E. Kämmer, Laura Zwaan, Maarten ten Berg, Martha Quinn, James Forman, Molly Harrod, Stephan Winter, Kathryn J. Fowler, Sanjay Saint, A. Gupta, Vishal Chopra, Martine Nurek, Miguel A. Vadillo, Olga Kostopoulou, Sandra Monteiro, Jonathan Sherbino, Geoff Norman, Jonathan S. Ilgen, H. Emily Hayden, Francis Ulmer, Veerle Busink, Vincent Ho, A.k.l. Reyners, Egbert F. Smit, Hilde M. Buiting, Donald van TolDiagnosis 2018 · DOI

August 30-31, 2018 | Bern, Switzerland

Abstract

The process of formulating a working diagnosis in the inpatient setting requires that diagnosticians gather, filter, integrate, and interpret substantial amounts of information within complex and hectic hospital environments. This study examined challenges and opportunities for improvement in clinician-to-clinician communication and data sharing during the diagnostic process using a qualitative, focused ethnographic approach. Observations of eight inpatient medicine teams revealed that data flow for each patient is often fragmented over time, with inefficient communication tools like pagers failing to support the dialogue needed for accurate diagnosis. Key recommendations for improvement include replacing pagers with two-way communication technologies, improving EHR design to increase data integration while reducing overload, and increasing face-to-face communication between clinicians. While front-line physicians have identified clear opportunities for improvement, determining the best methods for implementation remains a critical challenge.

Introduction: The Diagnostic Process as a Systemic Frontier

Diagnostic error is a leading cause of preventable patient harm, yet it remains significantly harder to quantify and treat than surgical or medication errors. This conference collection highlights that diagnostic error is not merely a result of individual failure but a systemic crisis driven by a "perfect storm" of cognitive biases, fragmented digital communication, and high-pressure clinical environments.

The core challenges addressed include:

  • Cognitive Load: Electronic Health Records (EHRs) and fragmented communication channels (pagers, emails) create "data overload" that hinders synthesis.
  • Heuristics and Biases: The impact of "availability bias" (being influenced by recent cases) and "information distortion" (subconsciously twisting new data to fit a premature theory).
  • Culture: The persistent "name-blame-shame" environment that prevents clinicians from discussing and learning from near-misses.

Methods and Datasets

Because diagnostic error is multi-faceted, researchers across the conference utilized diverse methodologies to capture its complexity:

  • Focused Ethnography: Researchers shadowed eight inpatient teams for 286 hours to observe real-time communication breakdowns.
  • Randomized Educational Trials: Comparing "high-volume/low-feedback" vs. "low-volume/high-feedback" training for X-ray interpretation.
  • Cognitive Psychology Experiments: Using "think-aloud" protocols and eye-tracking to see exactly where a resident’s attention wavers during a diagnosis.
  • Systematic Audits: Large-scale retrospective reviews of international data, including 22,780 surgical deaths in Australia and nearly 3,000 primary care cases in Sweden.

Key Insights: Cognitive Biases and Training Models

  • The "Atypical" Red Flag: One study found that a physician’s subjective feeling that a patient’s presentation is "atypical" is a statistically significant predictor of actual diagnostic error (OR 2.71).
  • Expertise as a Shield: While availability bias significantly impacts medical students, experienced staff physicians appear more resilient, suggesting that "pattern recognition" eventually matures into a robust defense mechanism.
  • Volume Trumps Depth in Training: For learning to read X-rays, seeing a high volume of examples with minimal feedback was more effective and resulted in faster diagnostic times than seeing fewer examples with deep, extensive feedback.
  • Information Distortion: Physicians often subconsciously "distort" new clinical cues to support their initial (and often incorrect) hypothesis, a primary driver of misdiagnosis.
  • The "Mistake of the Week" (MOTW): Implementing non-punitive forums where department heads lead by disclosing their own errors was shown to successfully shift hospital culture away from blame.

Key Results: Prevalence and Persistence

Diagnostic error was identified in 12.3% of emergency admissions, correlating with a nearly four-fold increase in mortality (OR 3.94). In primary care, missed diagnoses accounted for 42% of all avoidable harm cases, with cancer (specifically colorectal) being the most frequently missed serious diagnosis.

Diagnostic error rates in emergency admissions over a six-year period, demonstrating a lack of significant downward trend (R² = 0.0454).

Furthermore, communication gaps remain a significant hurdle. In specialist-to-PCP transitions, 28.4% of correspondence letters were missing entirely, and 25% were delayed by more than 14 days, directly impacting the continuity of the diagnostic process.

Conclusion: Moving Beyond Hindsight

While the conference identified critical systemic flaws—such as inefficient pagers and fragmented EHRs—the implementation of solutions remains a massive logistical hurdle. Many studies still rely on retrospective audits, which are subject to hindsight bias. Future work must move toward "Predictive Brain" models in AI to help clinicians maintain "cognitive openness" and avoid premature closure. The ultimate goal is to shift from a culture of blame to one of systemic resilience, ensuring that the diagnostic process is supported by integrated data and clear communication channels.

Summary updated

Abstract

The process of formulating a working diagnosis in the inpatient setting requires that diagnosticians working in complex and hectic hospital environments gather, filter, integrate, and interpret substantial amounts of information within a short time. Several years ago, the information needed for diagnosis was gathered primarily at the bedside. More recently and with the advance of new technologies, this information is collected through a series of communication exchanges (e.g., pagers, emails) and interactions with the electronic health record (EHR). In this study, we examined challenges and opportunities for improvement in clinician-to-clinician communication and data sharing during the diagnostic process. Methods: We performed a qualitative, multi-method, focused ethnographic study. Data were gathered between January and May 2016 at two affiliated teaching hospitals. Eight inpatient medicine teams (which included attending physicians, senior residents, interns, and medical students) were observed during morning rounds and in the afternoon on call and non-call days. Focus groups and interviews were then conducted with team members to better understand challenges and opportunities for improvement. Unstructured field notes were taken during observations. All focus groups and interviews were recorded and transcribed. Data were analyzed using qualitative content analysis. Results: Observation data showed that physicians faced a data-gathering and communication environment that made integration and interpretation of information for diagnosis challenging. Notably, data flow and communication for each patient was fragmented over time and diagnostic information was pieced together from multiple sources. Pagers were inefficient and did not support dialogue needed for diagnosis. Suggestions for improvement during interviews and focus groups included: 1) replacing pagers with two-way communication technologies; 2) improving EHR design to support diagnosis by increasing data integration while reducing data overload and information fragmentation; 3) identifying more efficient ways to access the EHR during morning rounds and while in patient rooms; 4) increasing face-to-face communication between clinicians. Conclusion: Teaching hospitals are complex environments. The way patient information is shared and communicated among clinicians has changed with the adoption of electronic health records. Physicians are confronted with data overload, frequent interruptions, fragmented information, and little time to think about diagnosis. Although improvement opportunities suggested by front line physicians for patient diagnosis were identified, how best to implement these ideas remains to be determined.
BibTeX
@inproceedings{schwappach2018nd,
  author = {David Schwappach and Jason Maude and Juliane E. Kämmer and Laura Zwaan and Maarten ten Berg and Martha Quinn and James Forman and Molly Harrod and Stephan Winter and Kathryn J. Fowler and Sanjay Saint and A. Gupta and Vishal Chopra and Martine Nurek and Miguel A. Vadillo and Olga Kostopoulou and Sandra Monteiro and Jonathan Sherbino and Geoff Norman and Jonathan S. Ilgen and H. Emily Hayden and Francis Ulmer and Veerle Busink and Vincent Ho and A.K.L. Reyners and Egbert F. Smit and Hilde M. Buiting and Donald van Tol},
  title = {{2nd European Conference on Diagnostic Error in Medicine August 30-31, 2018, Bern, Switzerland}},
  booktitle = {Diagnosis},
  year = {2018},
  doi = {10.1515/dx-2018-0055},
}
Figure