Cognitive Task Analysis to Evaluate Resident Physician Decision Making in the Intensive Care Unit.

Background

Heuristics are commonplace among novices and experts making clinical decisions, often guided by clinical intuition when there is diagnostic or therapeutic uncertainty. Compared with more experienced clinicians, trainees may lack the knowledge, insight, and intuition needed to appropriately select and apply heuristics and clinical decision rules. An improved understanding of the mental models and contextual factors that predispose trainees to misapplied heuristics, cognitive biases, and other decision-making errors is needed.

Objectives

To test the use of cognitive task analysis for examining how trainees make high-risk decisions in complex, dynamic, and real-world practice environments.

Methods

We conducted semistructured interviews between September 2019 and March 2020 using a cognitive task analysis technique called the critical decision method. Participants were third-year internal medicine resident physicians rotating in the medical intensive care unit at a major safety-net academic hospital. Interviews focused on fluid-resuscitation decisions for actual patients with septic shock. Data were coded and analyzed using a template approach with the Recognition-Primed Decision model as the guiding framework.

Results

Eleven of 23 eligible residents completed a full interview. The median time from initial sepsis care to interview was 7 days (interquartile range, 6.5-11 d). Seven key domains related to fluid-resuscitation decisions were identified: cues, information, decision making, decision alternatives, analogs, expected outcomes, and goals. In addition to objective clinical data (e.g., serum lactate concentration), fluid-resuscitation decisions were most significantly influenced by clinical intuition, other nonphysiological contextual factors, and volume-based heuristics. For example, residents frequently prescribed fluid dependent on the total volume already administered. They assumed that patients receiving more than 3-5 L would not benefit from additional resuscitation, while using the same heuristic to disregard evidence-based predictors of fluid responsiveness. Evidence of related cognitive biases was also found, including premature closure, confirmation bias, and status quo (or default) bias.

Conclusions

Cognitive task analysis is a promising tool for examining how trainees make high-risk clinical decisions. Better understanding the nature of trainees' heuristics and cognitive biases has implications for designing educational and training strategies that improve their clinical reasoning.