No evidence for decision fatigue using large-scale field data from healthcare.

Abstract

Decision fatigue is the idea that making decisions is mentally demanding and eventually leads to deteriorated decision quality. Many studies report results that appear consistent with decision fatigue. However, most of this evidence comes from observed sequential patterns using retrospective designs, without preregistration or external validation and with low precision in how decision fatigue is operationalized. Here we conducted an empirical test of decision fatigue using large-scale, high-resolution data on healthcare professionals' medical judgments at a national telephone triage and medical advice service. This is a suitable setting for testing decision fatigue because the work is both hard and repetitive, yet qualified, and the variation in scheduling produced a setting where level of fatigue could be regarded as near random for some segments of the data. We hypothesized increased use of heuristics, more specifically convergence toward personal defaults in case judgments, and higher assigned urgency ratings with fatigue. We tested these hypotheses using one-sided Bayes Factors computed from underlying Bayesian generalized mixed models with random intercepts. The results consistently showed relative support for the statistical null hypothesis of no difference in decision-making depending on fatigue (BF₀₊ > 22 for all main tests). We thus found no evidence for decision fatigue. Whereas these results don't preclude the existence of a weaker or more nuanced version of decision fatigue or more context-specific effects, they cast serious doubt on the empirical relevance of decision fatigue as a domain general effect for sequential decisions in healthcare and elsewhere.