What Decision Fatigue Is Supposed to Change
This chapter examines healthcare professionals making repeated medical judgments and asks what, specifically, cumulative effort is supposed to change in those judgments. The proposed mechanism is not a vague loss of ability: it is a shift in how information is acquired and used, potentially moving a nurse toward a familiar default or a more urgent, simpler choice. Making that pathway explicit lets readers distinguish a testable prediction from a story that can be fitted to any outcome.
From a medical call to a costly information problem
A medical call is an information problem before it becomes a rating or recommendation. The professional must listen to the patient’s account, ask questions, interpret the answers, connect them to possible conditions, weigh trade-offs, and decide what level of care is appropriate. Some calls will be straightforward; others will contain uncertainty that can be reduced only by asking another question, checking another possibility, or mentally comparing several possible outcomes.
The work is effortful not because every call is equally difficult, but because a sequence combines many small judgments under time pressure. The professional has to decide what to attend to, which answer needs clarification, which possibility deserves more thought, and when the available information is enough to act. The sequence can include documentation and consultation as well as the conversation itself. What matters to the theory is the repeated demand to process information and choose an action, not simply the number printed on a queue or clock.
The proposed decision-fatigue mechanism begins with the effort required for that work. Repeated judgments are not just a series of button presses. They involve controlled attention and reasoning, and they may become subjectively tiring, boring, or aversive. In the field study’s account, prolonged mental effort raises the cost of further information processing. The claim is not that the nurse suddenly loses medical knowledge. It is that using that knowledge in full may feel more costly, which could alter the balance between gathering more information and reaching a decision with what is already available.
One way to formalize this idea comes from rational inattention. The phrase sounds technical, but the underlying trade-off is familiar: information can improve a decision, yet finding and processing information takes time and mental effort. A decision-maker must allocate limited attention between the benefit of learning more and the cost of doing so. If the cost increases, the person may rationally stop searching sooner and rely more on prior beliefs or on a choice that is easier to make.
Imagine two otherwise similar calls that both leave some uncertainty. In the first, the nurse has more willingness to ask a further question or consider an alternative interpretation. In the second, the cost of one more round of processing is higher. The theory says that the second judgment may be more sensitive to what the nurse already expects, not necessarily because the nurse has forgotten anything, but because the marginal value of additional information no longer seems worth the effort. This is a proposed change in information allocation. It is not a diagnosis of the nurse, and it is not a claim that the second judgment must be wrong.
In a healthcare setting, “prior belief” does not mean an unsupported hunch. It can mean the professional’s accumulated expectation about a type of patient or symptom pattern, shaped by training and repeated experience. A nurse who has seen many similar calls may have a usual urgency judgment for that type of case. Relying on that usual judgment more often could be efficient when the case is familiar. It could also become a concern if it occurs because additional, relevant information is being left unexamined. The mechanism alone does not tell us which interpretation is correct; it predicts a change in the balance between information gathering and reliance on what is already known.
This is an important distinction from the everyday idea that fatigue makes someone “worse.” A person can preserve the correctness of many decisions while changing the route used to reach them. They may ask fewer questions, spend less time exploring an unusual possibility, or prefer an option that reduces the downside of uncertainty. If those shifts are consistent and linked to cumulative effort, they could matter even when no single decision is obviously an error. Conversely, a faster or more conservative judgment could reflect expertise, a clear guideline, or a different case mix rather than fatigue.
The proposed pathway therefore has several stages:
- demanding medical work requires information processing and judgment;
- repeated effort may increase the subjective or cognitive cost of further processing;
- the higher cost may reduce willingness to seek or weigh additional information;
- the professional may rely more on prior beliefs, familiar defaults, or simpler options;
- the downstream medical judgment may show a systematic change.
Each arrow is a claim that could fail. The work may not produce enough fatigue in the relevant interval. A decision-support system may reduce the marginal cost of retrieving information. A professional may compensate through motivation, training, or an explicit safety rule. Or the outcome may change for reasons that have nothing to do with the proposed pathway. Treating the mechanism as a sequence of possible links, rather than as one indivisible intuition, makes those alternatives visible.
The last line of the diagram is not a minor footnote. Case characteristics and organizational context can influence the final judgment through a path outside the fatigue mechanism. A convincing test must therefore examine not only whether the outcome follows the predicted direction, but whether the comparison makes those outside paths less likely to explain the difference.
Why defaults and urgency are the specific predictions
The field study turns the general mechanism into two observable predictions. The first concerns personal defaults. For a particular kind of patient, a nurse may have a most common urgency rating. The study asks whether a decision made under a plausibly higher level of cumulative effort is more likely to match that nurse’s usual choice for that type of call.
This is a subtle prediction. It does not say that the nurse will always choose the safest option, always choose the most urgent option, or make a random mistake. It says that the distribution of choices may move toward the individual’s own modal judgment because less information is being acquired or processed at the margin. The comparison is within the professional’s pattern of judgment, which makes the prediction more specific than simply asking whether average performance declines.
That individual reference point also changes how the result should be interpreted. A default is not a universal correct answer, and moving toward a personal default is not automatically a sign of reduced quality. The result would be evidence about consistency with a prior pattern under a high-effort condition. To decide whether the pattern is beneficial, harmful, or neutral, one would need additional evidence about patient outcomes and appropriateness. The registered report’s question is narrower: does cumulative effort alter the way the nurse makes the judgment in the predicted direction?
The second prediction concerns urgency ratings. The study proposes that urgency might rise when cumulative effort is higher. One rationale is that assigning a more urgent level can be a simpler or safer response when a severe condition seems unlikely but cannot be ruled out without additional inquiry. If acquiring more information is costly, the decision-maker might prefer a choice that protects against the more serious possibility.
Higher urgency is not automatically a sign of poor care. It can be clinically appropriate, and in some contexts a conservative decision may protect a patient. That is precisely why it is useful as a specified outcome rather than as a moral label. The relevant question is whether urgency shifts systematically under the high-effort condition after the other important features of the comparison have been addressed. A change in urgency can be operationally meaningful without proving that the professional was careless or that the patient was harmed.
The two outcomes also test different parts of the proposed pathway. More default-consistent choices concern reliance on an individual’s prior pattern. Higher urgency concerns the aggregate direction of judgments. If both move as predicted, the result would be more consistent with a common information-processing account than either outcome alone. If they move in different directions, the evidence would be harder to interpret as one simple fatigue mechanism. If neither moves, the tested version of the mechanism loses support, even though other effects might remain possible.
A prediction can also be compatible with more than one mechanism. Suppose urgency rises under a particular condition. That could reflect greater reliance on a conservative prior, a change in the kinds of patients being seen, a policy applied at a certain time, or a fatigue-related information shortcut. The outcome is informative only when the design and the theory jointly reduce those alternatives. This is why the field study treats time of day as a risky proxy and seeks comparison points where cumulative workload differs more clearly than the surrounding context.
That structure illustrates why a causal hypothesis must specify direction, outcome, and comparison point. “Decision fatigue changes judgment” is too broad to decide what would count as support. “Higher cumulative effort increases alignment with personal defaults” has a direction. “Higher cumulative effort increases urgency ratings” has a direction. The researcher can then state what data would favor those claims, what would favor no difference, and what would leave the result inconclusive.
Pre-specification matters because sequential data invite flexible explanations. Suppose later calls receive higher urgency ratings. One could say that the nurse became fatigued, that the patients became more severe, that another service closed, or that the time of day changed the kind of advice patients sought. Suppose urgency falls instead. One could say the nurse became disengaged, more experienced, or less willing to escalate. A broad theory can be made compatible with all of these outcomes unless the expected direction and the relevant comparison are set before the result is visible.
The field study’s predictions also avoid a common category error. They do not treat a proxy for fatigue as the outcome. Time on shift, call order, or being before a break is an attempt to identify a condition in which cumulative effort may differ. Personal defaults and urgency are the behaviors that are supposed to change. Keeping the presumed cause and the measured consequence separate is essential: otherwise the same variable can be used to define fatigue and then “confirm” it.
What the mechanism still cannot tell us
A clear mechanism is not a universal law. The field report presents rational inattention as a coherent account for its setting, while acknowledging that it may place too much emphasis on conscious rationality. People do not necessarily calculate the cost and benefit of every question. Habits, guidelines, professional norms, sleepiness, motivation, boredom, and the social context of work may all shape how effort is experienced and how a decision is made.
Nor does the mechanism specify a universal threshold. It does not tell us that a certain number of calls or hours must produce a behavioral change in every nurse. The effect could be weaker than expected, vary between people, appear only for some kinds of cases, or be buffered by a support system. A null result for defaults and urgency at the tested comparison points would therefore be strongest against that specified, practically relevant version of the hypothesis—not against every possible consequence of fatigue.
This boundary is part of the mechanism’s value, not a defect to hide. A model becomes more useful when it says what it predicts and where the prediction may stop. It can then be improved by asking which link failed. Did repeated work not raise the relevant effort cost? Did professionals compensate? Were the chosen outcomes too distant from the underlying process? Did the comparison fail to isolate cumulative effort? Or did another pathway change the judgment in the opposite direction?
Those questions specify what new evidence would change the conclusion. A direct measure of subjective fatigue could test whether the assumed internal state actually changed. A different clinical setting could reveal whether the effect depends on work organization or decision-support tools. Other outcomes could test whether fatigue appears in the amount of information gathered, consultation behavior, or errors rather than in default alignment or urgency. Such studies would not automatically prove a broad effect, but they would make the remaining uncertainty more informative.
These boundaries also explain why time of day cannot serve as a substitute for the mechanism. The clock may correlate with cumulative effort, but it is not the cognitive process itself. If a later-day difference appears, the next task is to determine whether it reflects the proposed change in information use or a different composition of patients and work conditions. A mechanism gives the investigation a target; it does not grant the target evidence in advance.
The practical lesson is restrained. It is reasonable to care about workload and the experience of fatigue without assuming that every schedule-related difference in medical judgment is caused by decision fatigue. It is also reasonable to demand a behavioral signature before designing a policy around the mechanism. The next question, then, is how much of a later-day pattern survives when clock time and case context are brought closer together—and when the comparison is chosen to challenge the explanation rather than merely illustrate it.