The Hidden Architecture of Clinical Judgment

How Health Care Systems Shape What Physicians See Before They Decide

Timothy Lesaca, MD

Psychiatrist | Pittsburgh, Pennsylvania

ABSTRACT

Clinical judgment is commonly represented as an internal sequence: the clinician gathers information, interprets it, weighs alternatives, and acts. That model is necessary but incomplete. It places the health care environment outside the act of judgment, as though the system merely helps or obstructs a decision after the physician has already made it. This conceptual essay proposes the architecture of clinical judgment: the total arrangement of temporal, informational, administrative, economic, technological, legal, and cultural conditions that shapes what a clinician notices, which possibilities become available for thought, what actions appear feasible, and what patterns are eventually experienced as intuition. The architecture operates through three linked functions: salience, friction, and internalization. Within it, invisible triage filters possibilities before conscious comparison, so an option may be absent from thought rather than explicitly rejected. Repeated exposure can then convert external limits into tacit professional expectations. This framework does not erase clinical agency or individual accountability. It explains why agency is exercised within a field whose boundaries have often been drawn elsewhere. The theory yields testable propositions and practical implications for patient safety, medical education, health information technology, artificial intelligence, payment policy, equity, and clinician well-being. Improving clinical judgment therefore requires more than educating better reasoners. It requires constructing environments worthy of the judgment medicine asks physicians to exercise.

Keywords: Clinical judgment; clinical reasoning; health care systems; invisible triage; diagnostic error; artificial intelligence; moral injury

Before the first thought

Before a physician reasons, the room has already been arranged.

The appointment has a length. The chart has a shape. A triage note has named the problem before the patient begins to speak. The electronic record places some facts in red and leaves others several screens away. A prior diagnosis waits at the top of the problem list, carrying the authority of repetition. The formulary has made one treatment easy and another conditional. The schedule is already running late. Somewhere beyond the room, a utilization rule, a quality metric, a staffing decision, and a malpractice fear have taken their seats.

We usually describe these conditions as context. The physician, in this familiar account, first observes, reasons, and decides; the system then facilitates or obstructs what follows. The denial comes after the prescription. The documentation burden comes after the visit. The unavailable referral comes after the plan. The environment appears to surround judgment without entering it.

But that sequence is often false.

A clinician does not ask every possible question and then run out of time. The questions themselves are shaped by the time believed to be available. A physician does not calmly rank every medically plausible treatment and only afterward discover which one is difficult to obtain. Years of prior authorizations, unavailable specialists, rejected referrals, and institutional scrutiny have already altered which treatments feel realistic enough to consider. The chart does not merely store the patient's story. It arranges the story into a hierarchy before the clinician reads it.

None of this requires a malicious actor. No administrator needs to instruct the physician to overlook complexity. No insurer needs to enter the examination room. No algorithm needs to issue a command. The architecture works most powerfully when it looks like the ordinary arrangement of things: the template, the queue, the default, the familiar label, the extra click, the anticipated denial, the sentence there is no time to ask.

This essay proposes that clinical judgment is not simply exercised within health care systems. It is partly formed by them. The physician's mind is one of the places where institutional structures do their work.

The inherited picture of clinical judgment

Clinical judgment is one of medicine's oldest practices and one of its least settled concepts. It is often described as a sequence involving the collection of clinical information, interpretation and reasoning, and decision making.1 This sequence captures something essential. A physician listens, examines, tests hypotheses, integrates evidence, and chooses a course of action. The National Academies has likewise emphasized that diagnosis is a complex, collaborative process involving information gathering, reasoning, communication, and the work system in which care occurs.2

Yet much of the literature and teaching on judgment still places the decisive action inside the clinician. Expertise, knowledge, pattern recognition, uncertainty tolerance, heuristics, and cognitive bias occupy the foreground. When judgment fails, we ask what the physician did not know, what cue was missed, which bias intervened, or why the differential diagnosis closed too early. These questions matter. Cognitive errors can arise from failures of perception, distorted heuristics, and predictable biases, and efforts to recognize them have made medicine safer.3

The difficulty is that this internal model can make the environment appear secondary. The system becomes the setting in which reasoning happens rather than one of the forces that determines the contents of reasoning. Time pressure is treated as an obstacle to an otherwise intact cognitive process. The electronic record is treated as a container for information. Prior authorization is treated as an implementation problem. Protocols are treated as guides applied after a clinician has recognized the relevant possibilities. Artificial intelligence is evaluated by the accuracy of its answer, as though the more consequential question were not also how it frames the problem.

Herbert Simon's account of bounded rationality helped dislodge the fantasy of unlimited decision making. Human beings reason with finite information, finite time, and finite computational capacity.4 In clinical medicine, this means that no physician can generate every hypothesis, retrieve every fact, or calculate every possible consequence. But bounded rationality is still commonly understood as a limit within the decision maker: the mind cannot fully process the world before it.

The theory proposed here moves the boundary upstream. The relevant limits do not begin only when a physician starts to think. They are also built into the environment that decides what will be presented, what will be hidden, what will be easy, what will be costly, and what will feel possible. The clinical problem is not merely that a finite mind must choose among too many options. It is that the option set has already been constructed.

A theory of the architecture of clinical judgment

The architecture of clinical judgment is the total arrangement of temporal, informational, administrative, economic, technological, legal, and cultural conditions that shapes what a clinician notices, which possibilities become available for thought, what actions appear feasible, and what patterns are eventually experienced as intuition.

Architecture is more than a metaphor here. A building does not dictate every movement within it, but it makes some movements natural, others awkward, and some impossible. Doors create passage and exclusion at the same time. Hallways direct traffic without issuing instructions. A staircase quietly assumes who can climb it. The people inside retain agency, but they exercise that agency within a field of structured possibility.

Clinical environments work in the same way. Appointment length directs attention. Staffing determines whether collateral history can be obtained. The order of information in a record creates a path through the case. Coding requirements make some descriptions legible and others administratively weightless. Protocols reduce variation while making deviation costly. Formularies, network adequacy, and prior authorization shape feasibility. Metrics create visibility for what can be counted. Legal expectations alter thresholds for testing, disclosure, and documentation. Digital systems rank, summarize, and recommend.

No single element constitutes the architecture. Its effect lies in the arrangement. A brief appointment may be manageable when the record is coherent, the patient is known, the team is stable, and referrals are accessible. The same appointment becomes cognitively dangerous when the chart is fragmented, interruptions are constant, staffing is thin, and every nonstandard treatment requires an appeal. Architecture is relational: the load placed on one component depends on the support or weakness of the others.

The theory is not deterministic. Physicians resist defaults, notice what systems obscure, and sometimes create time where none has been allotted. Patients interrupt the expected story. Colleagues rescue one another's attention. Good architecture can also protect judgment by surfacing allergies, standardizing dangerous calculations, prompting follow-up, and making rare but consequential possibilities easier to see. The claim is not that architecture always corrupts judgment. It is that architecture is never absent from it.

               Three linked operations explain how this occurs: salience, friction, and internalization.

Salience: what the system teaches the clinician to notice

Clinical information does not arrive as an undifferentiated field. It arrives arranged.

A red laboratory value is more visible than a sentence describing loneliness. A diagnosis repeated in the problem list acquires more authority than a question buried in an old note. A templated review of systems can appear complete while the patient's central fear remains unasked. A referral reason may compress a complicated life into two words. By the time the physician encounters the case, someone or something has already decided what deserves the foreground.

Time intensifies this hierarchy. Time-motion studies have shown how much ambulatory practice is divided between direct patient care and electronic or desk work, with substantial work continuing after clinic hours.5,6 Primary care has been described as an environment of information chaos, including overload, underload, scatter, conflict, and erroneous information.7 Interruptions and time pressure can alter the conditions under which diagnostic reasoning occurs.8 The point is not simply that clinicians become busy. Busyness changes which signals can compete successfully for attention.

Electronic records are often discussed as repositories, but every repository is also an editor. Interface design, data placement, copied text, alert behavior, and navigation determine what can be found without unusual effort. Usability and safety studies have documented wide variation and persistent risks in electronic health record design.9 The clinically important fact that requires six screens does not remain cognitively equivalent to the fact displayed on the opening page.

Salience is therefore not the same as importance. It is the probability that something will become available to attention under actual conditions of practice. A system can preserve a fact perfectly and still make it functionally invisible. What is recorded but not findable, measurable but not interpretable, or present but not given time may never become part of the clinical reality from which judgment proceeds.

This is the first architectural operation: the system does not merely supply information. It assigns visibility.

Friction: how feasibility enters the differential

The second operation is friction: the effort, delay, exposure, uncertainty, or conflict attached to a possible action.

Some friction is necessary. Controlled substances should not be prescribed without safeguards. Dangerous procedures should require deliberation. Expensive or scarce interventions may require review. The problem begins when friction is treated as though it affects only implementation. In practice, clinicians learn the resistance of the world before they choose what to attempt.

A treatment that requires six phone calls, a peer-to-peer review, an uncertain appeal, and a delay of several weeks does not remain psychologically equivalent to a treatment available with one click. A specialist who appears in a directory but never accepts a referral gradually ceases to function as an option. A test that predictably triggers scrutiny acquires a different threshold from one embedded in a routine order set. A recommendation that will create hours of unpaid documentation is weighed differently, even when the physician sincerely intends to reason only from clinical need.

Research on prior authorization makes this pre-decisional effect visible. In a national survey, clinicians reported changing treatment decisions, avoiding newer medications, or modifying diagnoses in relation to prior authorization requirements and associated burdens.10 Administrative burden is therefore not merely the cost paid after a clinical choice. It can become part of the process that selects the choice.

This produces a crucial coupling between feasibility and plausibility. The more difficult an option is to carry out, the less readily it may arise as a serious option at all. The clinician may still know, in an abstract sense, that the treatment exists. But abstract knowledge is not the same as cognitive availability in a crowded visit. The path of greatest institutional resistance gradually becomes the path of least mental presence.

Protocols and guidelines create a related form of friction. They can protect patients and reduce unwarranted variation, yet single-disease recommendations may fit poorly when patients have multiple conditions, conflicting priorities, or treatment burdens.11,12 Deviation requires explanation, time, and professional confidence. Following the supported path is often effortless; leaving it may require the clinician to become advocate, exception writer, and risk bearer at once.

Friction does not order the physician to abandon judgment. It changes the price of exercising it.

Internalization: when the institution becomes intuition

The third operation is internalization. Architecture begins outside the clinician, but it does not remain there.

At first, the constraint may be conscious. The physician thinks: this medication will be denied; that consultant is unavailable; this appeal will consume an hour; this concern will not survive the handoff; there is no room in today's schedule for the full story. Repetition changes the form of the knowledge. What was once an explicit obstacle becomes a tacit expectation. Eventually, the mind stops generating options that experience has taught it cannot use.

This adaptation is not evidence of indifference. It is one of the ways skilled people preserve function in environments of excess demand. Clinicians learn patterns, reduce unnecessary search, and anticipate downstream consequences. Such efficiency is indispensable. The danger is that practical wisdom and institutional conditioning can become indistinguishable from the inside. What feels like seasoned intuition may partly be the sediment of repeated denials, compressed visits, inaccessible services, and defensive routines.

The system has then moved from being an obstacle in front of the physician to a boundary inside the physician.

This helps explain why some forms of professional distress are not captured by exhaustion alone. Administrative burden, poorly designed electronic records, and loss of control are associated with physician dissatisfaction and burnout.13-15 Yet the deeper experience may be a narrowing of authorship. The physician continues to work, often competently, while recognizing fewer parts of the role as genuinely his or her own. Judgment is exercised, but within limits that have become so familiar they no longer appear external.

Internalization also explains why removing a rule may not immediately restore the thinking that existed before it. Once a referral has disappeared from the practical imagination, once a treatment has become synonymous with futility, or once a clinician has learned that curiosity creates unfinishable work, the architecture can persist after the wall is gone.

Invisible triage: the upstream filter

Within this larger architecture sits a specific mechanism I have called invisible triage.16

Ordinary triage is explicit. Patients, needs, or treatments are ranked because time and resources are limited. Invisible triage occurs earlier and without a declared decision. It is the preconscious filtering process through which some possibilities enter the field of clinical awareness and others do not. The missing option is not judged and rejected. It never becomes sufficiently present to be judged.

This distinguishes invisible triage from cognitive bias. Anchoring, availability, confirmation bias, and premature closure operate on information or hypotheses that have reached awareness. Invisible triage helps determine what reaches awareness in the first place. A bias may lead a physician to choose wrongly among visible explanations. Invisible triage may prevent a relevant explanation from appearing on the list.

It also differs from the usual idea of choice architecture. A default or nudge rearranges visible options to influence selection. Invisible triage acts on the construction of the option set itself. The question is not only which door the clinician chooses. It is which doors the building permits the clinician to see.

Because invisible triage is distributed, it often leaves no identifiable decision maker. The scheduler did not decide against a trauma history. The record designer did not decide against a malignancy workup. The insurer did not explicitly instruct the physician never to discuss the preferred medication. The algorithm did not command the clinician to ignore an unranked possibility. Each component may be defensible in isolation. Their arrangement nevertheless produces a field in which certain lines of thought are less likely to occur.

This is why omissions generated by architecture can be so difficult to detect. There is no rejected order, canceled referral, or documented disagreement. The record contains no evidence of the option that never appeared. The absence leaves no footprint.

Four clinical rooms

The theory becomes clearer when viewed at the level of ordinary care. The following situations are hypothetical, but their structure is familiar.

The prewritten follow-up. A patient returns with fatigue and mild anemia. The visit is labeled an iron-deficiency follow-up. The opening screen foregrounds a low ferritin from the prior year. The template asks about supplement adherence. The appointment is brief, the clinician is interrupted twice, and the most accessible action is to renew iron and repeat familiar laboratory tests. An underlying malignancy is not weighed and judged unlikely. The possibility never gathers enough presence to be considered. Prior labeling, interface design, workflow, and time have already organized the case.

The compressed psychiatric evaluation. A frightened patient arrives in an emergency department with mistrust, withdrawal, poor sleep, and disorganized fragments of a story. A prior note uses the word psychosis. Collateral information is unavailable. The unit is crowded, the disposition decision is urgent, and the psychosis pathway is already open in the record. A careful trauma history, cultural formulation, longitudinal mood history, and assessment of substance effects would require time the environment has not preserved. Bias may be present, but architecture magnifies it by withholding the conditions under which an initial impression might be corrected.

The treatment never discussed. A physician knows that a particular medication is the best fit for a patient's prior response and comorbidities. The same medication has repeatedly required prior authorization, step therapy, and appeals. The patient has limited transportation and cannot tolerate a prolonged delay. Before entering the room, the physician selects the easier second-line option. The preferred treatment is not denied because it is never ordered. From the payer's data, no barrier occurred. From the patient's perspective, the choice appears purely clinical.

The algorithmic summary. A clinical system summarizes a long record and highlights heart failure, prior admission, and elevated risk. The summary is accurate as far as it goes. It does not foreground the patient's recent bereavement, medication confusion, or inability to obtain food. A recommended order set appears beside the summary. The clinician remains free to disagree, but the system has already established the center of gravity. Automation bias and verification burden can make algorithmic suggestions difficult to resist, particularly when checking them is more laborious than accepting them.17 Algorithms may also reproduce inequity when the variables chosen as proxies reflect unequal access or expenditure rather than underlying need.18

These are not stories of bad physicians. Nor are they excuses for careless care. They show that the visible decision is the end of a longer sequence of arrangements. By the time the clinician acts, much of the field has already been built.

The decision has many authors

Medicine concentrates accountability at the bedside because the bedside is where the decision becomes visible. The physician signs the order, explains the plan, and faces the patient when the outcome is poor. This concentration has ethical value. Patients need a responsible professional, not a maze of unanswerable institutions.

Yet the authorship of the decision may be widely distributed.

The appointment length was set elsewhere. The available data were shaped by prior documentation and interface design. The differential diagnosis was influenced by labels inherited from other clinicians. The feasible treatment set was altered by coverage rules, network access, staffing, and local protocols. The legal threshold was shaped by previous claims and institutional counsel. The algorithm was trained on choices made under earlier systems. The physician remains an author, but not the only one.

This creates an asymmetry: authorship is distributed while accountability is concentrated.

The asymmetry distorts both patient safety and professional life. After a missed diagnosis, review may focus on why the clinician failed to order a test without asking how the case was framed, what information was visible, how many interruptions occurred, or which alternatives the workflow made costly. When treatment is delayed, the physician may carry the patient's anger even when the delay was produced by a sequence of approvals no one person controls. When clinicians repeatedly feel responsible for outcomes they lack authority to prevent, moral distress can become a rational response to misaligned responsibility.

The architectural account does not dissolve personal responsibility into the system. A physician still has duties that cannot be delegated: to remain curious, to recognize uncertainty, to resist unsafe defaults, to advocate when the ordinary path is wrong, and to acknowledge error. But accountability should be causally honest. It should trace not only the final act but the conditions that formed the available choices.

Responsibility in complex medicine is neither absent nor indivisible. It is layered. The task is not to decide whether the clinician or the system is responsible. It is to understand how responsibility travels through both.

Five propositions of the theory

A useful theory must do more than rename familiar frustration. It should make claims that can be examined, challenged, and refined. The architecture of clinical judgment yields five propositions.

1. Clinical judgment is situated. Holding the patient and clinician constant while changing the decision environment should change what is noticed, which hypotheses are generated, and what action is selected. Judgment is therefore a property of the clinician-in-environment, not of the clinician alone.

2. Architecture acts before deliberation. Environmental conditions influence not only the ranking of known options but the content of the initial option set. Studies should therefore measure the questions asked, hypotheses generated, and alternatives mentioned, not only the final decision.

3. Friction and salience are coupled. As the effort, delay, or professional exposure attached to an option rises, that option will become less likely to appear early in reasoning, even when its clinical merit is unchanged. Administrative policies should be evaluated as cognitive interventions, not merely operational requirements.

4. Architectural effects persist through internalization. Repeated exposure to constraints will create habits and expectations that may continue after the original barrier is removed. Natural experiments involving formulary changes, improved access, or redesigned workflows should therefore examine whether clinical behavior changes immediately or lags behind structural reform.

5. Misalignment between authorship and accountability has psychological consequences. The greater the distance between responsibility for outcomes and authority over the conditions producing them, the greater the risk of moral distress, professional alienation, and withdrawal of discretionary effort.

These propositions are testable. Researchers could randomize the order and visibility of information in otherwise identical records; vary appointment time or interruption load in standardized cases; add or remove prior-authorization friction while holding clinical evidence constant; use think-aloud protocols to identify when an option first appears; compare clinician reasoning before and after network or formulary changes; and pair electronic event logs with interviews about perceived possibility. The crucial outcome is not simply whether the clinician chose A or B. It is whether B ever became a live option.

The framework also suggests a new unit of analysis. Traditional decision research often studies the individual encounter. Architectural research must also study the history of encounters through which clinicians learn what their systems permit. A single denial may be an inconvenience. A hundred denials may become a way of seeing.

What changes if the framework is true

If clinical judgment has an architecture, several familiar approaches to improvement become necessary but insufficient.

Patient safety. Debiasing, checklists, and diagnostic education remain valuable, but they cannot correct information the environment never makes available or options it has trained clinicians not to generate. Root-cause analysis should ask architectural questions: What did the interface foreground? What required unusual effort to discover? Which pathways were frictionless? What assumptions were embedded in the visit type, template, or handoff? What did the organization repeatedly teach clinicians was futile? The inquiry must extend from why the physician missed the diagnosis to how the diagnosis was made difficult to see.

Medical education. Trainees are taught to build differentials, recognize bias, and defend decisions. They should also learn to inspect the environments shaping those decisions. A mature clinician must be able to ask: Is this truly my intuition, or have I learned the institution's limitations so well that they feel like clinical facts? Would I view this treatment differently if it were easy to obtain? What part of the patient's story has the record rendered peripheral? Education should cultivate not only metacognition, but architectural awareness.

Health information technology and artificial intelligence. Clinical tools should be evaluated not only for accuracy, speed, or user satisfaction, but for how they redistribute salience and friction. Does a summary preserve uncertainty or compress it prematurely? Does a recommendation make dissent easy or merely technically possible? Does the tool identify missing variables, reveal alternative explanations, and know when not to answer? Can the clinician inspect the source of a conclusion without prohibitive effort? A highly accurate model can still create unsafe architecture if it narrows attention too early or makes verification harder than acceptance.

Payment and utilization policy. Prior authorization, step therapy, network design, and documentation requirements are often justified as controls on spending or variation. Whatever their purpose, they also alter cognition. Their effects cannot be measured only by approval rates, turnaround times, or completed appeals. Systems should ask how often clinicians avoid proposing an option because they anticipate the burden. The most consequential denial may be the one that never enters the data because the treatment was never requested.

Equity. Architecture does not act evenly across patients. Fragmented records, inaccessible interpreters, poorly represented symptoms, cost-based proxies, and unequal access to specialists can make some forms of suffering less visible and some remedies less feasible. Bias is not replaced by architecture; it can be amplified through it. A prejudgment becomes harder to correct when the system withholds time, context, or alternative pathways. Equity work must therefore examine not only attitudes and outcomes, but the structure of clinical visibility.

Clinician well-being. Burnout interventions often focus on the individual's capacity to recover. The architectural account asks whether the work still allows physicians to exercise judgment in ways that remain continuous with their professional identity. Rest can restore energy. It cannot by itself restore authorship. Resilience can help a clinician endure friction. It cannot decide whether the friction is justified. When professional distress arises from the repeated internalization of constraints, healing requires changes in the conditions that are being internalized.

Organizational design. Every new policy, template, metric, and digital tool should be treated as a clinical intervention because it changes the field in which care is conceived. The relevant review question is not only, Will this improve throughput or compliance? It is also, What will this make easier to notice, harder to imagine, safer to say, or more costly to pursue?

An architectural audit of judgment

The theory points toward a practical method: an architectural audit of judgment.

Such an audit would begin with a real clinical decision and move backward. It would reconstruct not only what the clinician knew, but how the case became knowable. It would examine the first label attached to the encounter, the order in which information appeared, the time and interruption structure, the defaults and prompts, the options excluded by access or coverage, the documentation required for deviation, and the consequences attached to delay, dissent, or uncertainty.

The audit would then ask three questions corresponding to the theory's central operations.

What was made salient? Which facts, risks, diagnoses, and patient goals were placed in the foreground, and which were buried, fragmented, or absent?

Where was friction placed? Which actions were available immediately, and which required extra time, justification, money, negotiation, or professional risk?

What had already been internalized? What did experienced clinicians assume would fail, be denied, take too long, or create trouble, and how had those expectations shaped the option set before this particular encounter began?

This method would not replace conventional quality review. It would deepen it. Human beings still make mistakes, and some failures remain plainly individual. But an architectural audit can reveal why the same kinds of mistakes recur among different competent clinicians. Repetition across people is often a clue that the environment is doing more than hosting the problem.

The audit should also search for protective architecture. Where did the system create the pause that prevented harm? Which colleague, alert, checklist, pharmacist, or workflow made the overlooked possibility visible? Which forms of friction appropriately slowed a dangerous action? The purpose is not to make every path effortless. It is to place attention and effort where clinical value justifies them.

Boundaries and limitations

This is a conceptual framework, not a completed empirical theory. Its constructs will require clearer operational definitions, validated measures, and testing across specialties and settings. Salience, friction, internalization, and invisible triage overlap in practice, and separating their effects may be difficult.

The framework also risks becoming too expansive. If every feature of the environment is called architecture, the concept could explain everything and therefore predict little. Its proper use requires specificity: identifying the concrete arrangement, the mechanism through which it alters perception or feasibility, and the observable change in reasoning that should follow.

Nor should architecture become a refuge from accountability. Clinicians can invoke system pressure too easily when a duty was clear and the opportunity to act was present. The theory does not claim that physicians are passive instruments or that all constraints are unjust. Scarcity is real. Standards, protocols, review processes, and boundaries can be ethically necessary. Good architecture often limits action precisely to protect patients.

The more modest claim is also the more consequential one: clinical agency is real, but it is structured. A fair account of judgment must attend both to the person who decides and to the environment that shaped what decision could become thinkable.

Designing environments worthy of judgment

Medicine has spent generations trying to produce better clinicians. We select for intelligence, teach vast bodies of knowledge, supervise decisions, examine competence, study cognitive error, and urge lifelong reflection. All of this is necessary. Yet we have paid less attention to whether the environments we build allow those capacities to appear when patients need them.

A physician cannot attend indefinitely. No system can make every option visible, every treatment accessible, or every decision free of trade-offs. Architecture is unavoidable because care must have pathways, thresholds, schedules, records, and limits. The ethical question is not whether medicine will have an architecture. It is what that architecture will teach its occupants to see.

Clinical judgment remains human, but it is not private. It arises in the encounter between a mind, a patient, and a constructed world. The physician brings knowledge, experience, character, and responsibility. The patient brings a life that exceeds every category. The system brings an arrangement of visibility, effort, permission, and consequence. Judgment emerges from all three.

This perspective changes the direction of reform. We should continue asking physicians to reason carefully, resist bias, and remain curious. We should also stop building environments that punish the time curiosity requires, bury the information it needs, and make the clinically appropriate path the least available one. We should evaluate policies not only by what they command, but by what they make disappear. We should judge technology not only by the answers it gives, but by the questions it prevents us from asking.

Every system teaches its occupants how to see. Over time, physicians cease merely to practice within institutions; they begin to perceive through them. The future of medicine therefore depends not only on educating wiser clinicians, but on constructing environments worthy of the judgment we hope they will exercise.

Before the physician enters, the room has already been arranged. We cannot ask doctors to see clearly while building rooms that teach them not to look.

References

1. Fava GA, Sonino N, Guidi J. Clinical judgment: an essential method in medicine. Postgrad Med J. Published online March 23, 2026. doi:10.1093/postmj/qgag030.

2. National Academies of Sciences, Engineering, and Medicine. Improving Diagnosis in Health Care. Washington, DC: National Academies Press; 2015. doi:10.17226/21794.

3. Croskerry P. The importance of cognitive errors in diagnosis and strategies to minimize them. Acad Med. 2003;78(8):775-780. doi:10.1097/00001888-200308000-00003.

4. Simon HA. A behavioral model of rational choice. Q J Econ. 1955;69(1):99-118. doi:10.2307/1884852.

5. Sinsky C, Colligan L, Li L, et al. Allocation of physician time in ambulatory practice: a time and motion study in 4 specialties. Ann Intern Med. 2016;165(11):753-760. doi:10.7326/M16-0961.

6. Arndt BG, Beasley JW, Watkinson MD, et al. Tethered to the EHR: primary care physician workload assessment using EHR event log data and time-motion observations. Ann Fam Med. 2017;15(5):419-426. doi:10.1370/afm.2121.

7. Beasley JW, Wetterneck TB, Temte J, et al. Information chaos in primary care: implications for physician performance and patient safety. J Am Board Fam Med. 2011;24(6):745-751. doi:10.3122/jabfm.2011.06.100255.

8. Monteiro SD, Sherbino JD, Ilgen JS, et al. Disrupting diagnostic reasoning: do interruptions, instructions, and experience affect the diagnostic accuracy and response time of residents and emergency physicians? Acad Med. 2015;90(4):511-517. doi:10.1097/ACM.0000000000000614.

9. Ratwani RM, Savage E, Will A, et al. A usability and safety analysis of electronic health records: a multi-center study. J Am Med Inform Assoc. 2018;25(9):1197-1201. doi:10.1093/jamia/ocy088.

10. Salzbrenner SG, Lydiatt M, Helding B, et al. Influence of prior authorization requirements on provider clinical decision-making. Am J Manag Care. 2023;29(7):331-337. doi:10.37765/ajmc.2023.89394.

11. Boyd CM, Darer J, Boult C, Fried LP, Boult L, Wu AW. Clinical practice guidelines and quality of care for older patients with multiple comorbid diseases: implications for pay for performance. JAMA. 2005;294(6):716-724. doi:10.1001/jama.294.6.716.

12. Wyatt KD, Stuart LM, Brito JP, et al. Out of context: clinical practice guidelines and patients with multiple chronic conditions: a systematic review. Med Care. 2014;52(suppl 3):S92-S100. doi:10.1097/MLR.0b013e3182a51b3d.

13. Rao SK, Kimball AB, Lehrhoff SR, et al. The impact of administrative burden on academic physicians: results of a hospital-wide physician survey. Acad Med. 2017;92(2):237-243. doi:10.1097/ACM.0000000000001461.

14. Melnick ER, Dyrbye LN, Sinsky CA, et al. The association between perceived electronic health record usability and professional burnout among US physicians. Mayo Clin Proc. 2020;95(3):476-487. doi:10.1016/j.mayocp.2019.09.024.

15. National Academies of Sciences, Engineering, and Medicine. Taking Action Against Clinician Burnout: A Systems Approach to Professional Well-Being. Washington, DC: National Academies Press; 2019. doi:10.17226/25521.

16. Lesaca T. What hidden constraints shape clinical decisions? KevinMD. May 6, 2026. https://kevinmd.com/2026/05/what-hidden-constraints-shape-clinical-decisions.html.

17. Lyell D, Coiera E. Automation bias and verification complexity: a systematic review. J Am Med Inform Assoc. 2017;24(2):423-431. doi:10.1093/jamia/ocw105.

18. Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447-453. doi:10.1126/science.aax2342.