Pulmonary embolism (PE) diagnosis presents a significant challenge for emergency department (ED) physicians, as both missed or delayed diagnosis and overtesting can have serious consequences for patients. The implementation of health information technology, such as clinical decision support systems, has the potential to mitigate diagnostic errors and enhance the overall diagnostic process. However, to achieve this, the technology must be practical, user-friendly, and seamlessly integrate into clinical workflows. This calls for a sociotechnical systems approach to understand the cues involved in the PE diagnosis process and how they relate to the information available in electronic health records (EHRs).
In this study, we sought to comprehend the cues in the PE diagnosis process within the ED sociotechnical system and compare them to the information found in the EHR. The objective was to establish design requirements for clinical decision support for PE diagnosis in the ED.
Pulmonary embolus (PE) is among the most challenging diagnoses made in the emergency department (ED). While missed or delayed diagnosis of PE is a major problem in the ED1, overtesting, which subjects patients to harm from radiation, overdiagnosis, and increased cost, is also a concern. Health information technology, such as clinical decision support, has the potential to reduce diagnostic errors and support the diagnostic process. However, this requires that the technology be useful and usable, and fit within the clinical workflow, providing justification for a sociotechnical systems approach. The purpose of this study is to understand cues in the PE diagnosis process in the ED sociotechnical system and to compare these cues to the information available in the EHR. This will help in defining design requirements for a clinical decision support for PE diagnosis in the ED. Using the Critical Decision Method, we interviewed 16 attending physicians and residents in three EDs of two academic medical centers and one community hospital. The total duration of the interviews was over 12 hours. Using an iterative qualitative content analysis, we identified 4 categories of cues: (1) patient signs and symptoms (e.g., leg swelling, chest pain), (2) patient risk factors (e.g., immobilization, surgery or trauma, cancer), (3) explicit risk scoring (e.g., PERC), and (4) clinical judgment. We then mapped these cues to information available in the EHR at one of the participating hospitals. About 80-90% of the cues may be available in the EHR; many of them rely on the physical exam and information obtained by talking to the patient. This finding underlines the need to identify the various roles involved in obtaining, documenting and reviewing the information that informs the PE diagnostic process. The PE diagnostic process in the ED is distributed across multiple roles, individuals and technologies in a sometimes chaotic and often busy physical and organizational environment.