Background: Diagnostic error in acute care represents an unresolved safety issue: error rates range from 4.8 to 49.8%. If the diagnosis is delayed or incorrect, the patient may not get correct treatment in a timely manner. Underlying contributing factors include system flaws (e.g., communication barriers) and cognitive errors (e.g., anchoring), factors that are often overlooked by workflow and technological solutions. New frameworks developed for addressing diagnostic error and recent work identifying the types of events often captured in the electronic health record (EHR) have yet to be used to operationalize key “e-triggers” that could proactively identify patients at risk for preventable harm.
Purpose: Our Patient Safety Learning Laboratory, a collaboration between Brigham and Women’s Hospital and Northeastern University Healthcare Systems Engineering Institute, is designing digital health tools that seek to use e-triggers from the EHR to identify hospitalized patients at risk for diagnostic error in real-time. We describe our approach to addressing diagnostic errors in acute care.
Description: Our team is engaged in activities (Table 1) that utilize a variety of systems engineering and human factors methods to analyze cases, validate potential risk factors and previously described e-triggers, and understand stakeholders’ perspectives of actual diagnostic error. We are identifying representative cases of diagnostic error for 14 conditions (e.g., pneumonia) and undifferentiated symptoms (e.g., abdominal pain) from our safety reporting system, hospital medicine quality assurance committee, and patient and family relations department. We are using Ishikawa and fault-tree analyses to understand the types of system and cognitive factors that led to harm for these cases. Simultaneously, we are retrospectively identifying cases of potential diagnostic error by applying various screening criteria to previously hospitalized patients. As part of our quantitative activities, we will determine the positive predictive value of these screening criteria by confirming the presence or absence of diagnostic error via chart review and an adjudication process. As part of our qualitative activities, we are conducting structured interviews with key stakeholders about their perceptions of diagnostic error in general and for representative cases. Finally, we are using a user-centered design approach to identify requirements for our technology-based solution according to key diagnostic process dimensions. We plan to use the knowledge generated from these analyses to operationalize selected e-triggers in context of our digital health intervention in order to prospectively identify hospitalized patients at-risk for diagnostic error. In general, we will design and develop enhancements to a suite of EHR-integrated tools (Figure 1) that serve to engage patients and care teams with the goal of ensuring clinical trajectories match the anticipated course for working diagnoses or symptoms, as well as alignment with patient and clinician expectations.
Conclusions: Our experience to date underscores the complexity and interdisciplinary expertise required to design digital health enabled strategies that leverage patient-reported and EHR data in real-time to enable front-line clinicians to proactively identify patients at risk for diagnostic error. During implementation and evaluation phases, our goal is to demonstrate a reduction in the rate of diagnostic error for patients admitted to general medicine units.
.png)
