Why Clinicians Need a New Way to Study in the Age of AI. Introducing ReviewBytes
Read more
A practical guide to 10 high-stakes decisions for new attendings: discharge, escalation, supervision, uncertainty, handoffs, and safer independent practice.
A new attending’s first high-stakes decisions usually involve admission or discharge, escalation of care, consultation, supervision, uncertainty, complications, conflict, prioritization, signing off, and knowing when to say no.
That list may sound obvious. What changes when you become an attending is that the decision eventually stops somewhere—and increasingly, it stops with you.
As a resident or fellow, there was usually another layer available. You could formulate the plan, defend it, and act on it, but an attending remained somewhere in the chain. Becoming an attending does not mean that consultation or collaboration suddenly disappears. In fact, some of the safest and most experienced physicians are remarkably comfortable asking for help. What changes is that you are now responsible for recognizing when help is needed, when action cannot wait, and when the available plan is simply not safe enough.
This field guide focuses on ten decisions that repeatedly test that transition.
The first months of independent practice are not mainly a test of whether you can remember enough medicine. They are a test of whether you can turn clinical knowledge into safe decisions while managing uncertainty, other people, competing priorities, and imperfect systems.
A useful way to think about these decisions is through three questions:
A 2025 qualitative study of new-to-practice attendings identified decision-making, supervision, workload, and time management among the struggles encountered during the transition to independent practice (PMID: 39980956).
Before finishing this guide, you should be able to ask yourself:
These are useful questions not only for new attendings, but also for senior residents, fellows, physician assistants, and nurse practitioners preparing for greater clinical independence.
We often talk about being “ready for independent practice” as if there is a clear point at which supervision ends and independent judgment begins.
Real clinical work is not that neat.
A physician can be very strong in medical knowledge and still struggle with the transition because independent practice adds several things at once: a larger workload, greater responsibility for supervision, new administrative expectations, more competing demands, and the need to make decisions without routinely presenting them to someone else first.
That is one reason traditional board prep, exam prep, ABIM preparation, and in-training exams can only measure part of readiness. They remain important, but knowing the correct answer to a question is different from deciding what to do when the patient is deteriorating, the consultant has not called back, the family disagrees, and three other patients also need your attention.
The important skill is not certainty. It is being able to act responsibly despite incomplete certainty.
Most high-stakes attending decisions can be simplified into four steps.
What is happening now?
Look beyond one isolated value. Consider the patient’s trajectory, new symptoms, vital signs, examination, laboratory and imaging information, nursing concerns, family concerns, and what has changed.
What needs to happen now?
The answer may be treatment, observation, consultation, a procedure, transfer, discharge, or simply more information before committing to any of these.
What happens if the plan does not work?
This is where a decent plan becomes a safer plan.
A patient may reasonably be discharged, for example, but only if there is a clear plan for worsening symptoms. A patient may remain on the medical floor, but only if deterioration can be detected and treated quickly.
Who owns the next action?
Many problems in medicine occur not because nobody knew what should happen, but because everyone assumed somebody else would make it happen.
Structured handoff research supports explicit communication and contingency planning, although it does not validate this particular four-step framework (PMID: 25372088).
Readiness question: “Can this patient safely receive the next phase of care in the proposed setting?”
Discharging someone is not simply the opposite of admitting them.
A patient may look considerably better than when they arrived and still not be ready to go home. Conversely, an abnormal laboratory value does not automatically require admission if the patient can safely be managed outside the hospital.
The attending has to connect the medical plan with the real-world environment into which the patient is being sent.
Before discharge, consider:
A useful final question is:
“Tell me what you will do if you feel worse tonight.”
That short conversation can reveal considerably more than asking, “Do you understand?”
A randomized trial of a coordinated discharge program demonstrated lower 30-day hospital utilization compared with usual care, supporting the importance of organized discharge planning rather than relying on the examination alone (PMID: 19189907).
Readiness question: “Can this setting recognize and respond to the deterioration I am worried about?”
One of the most consequential attending decisions is not whether a patient is sick, but how sick the patient may become before the current environment can no longer safely manage them.
Sometimes the change is obvious. At other times, the patient is still talking, the blood pressure has not collapsed, and no single value looks catastrophic.
Yet the trajectory is wrong.
A rising oxygen requirement, progressive work of breathing, worsening mental status, repeated hypotension, increasing lactate, or escalating nursing concern should not be dismissed simply because the patient has not yet crossed some dramatic threshold.
The practical approach is:
Current critical care guidance supports explicit rapid-response activation criteria and incorporating patient and family concerns when assessing deterioration (PMID: 38240510).
Readiness question: “What specific question needs specialist input, and how urgently?”
Calling a consultant is not an admission that you do not know enough medicine.
It is often evidence that you understand where your responsibility intersects with somebody else’s expertise.
The less useful call is:
“Can you see this patient?”
The more useful call is something like:
“I am concerned about X because of Y and Z. We have already done A and B. I need your help deciding whether this patient needs C tonight.”
A good consultation request should usually include:
There also needs to be a closed loop.
If the consultant will see the patient in three hours, what happens if the patient worsens in thirty minutes?
Clear communication about the reason for consultation has long been recognized as important to consultation effectiveness (PMID: 6615097).
Readiness question: “For this patient and this task, what can this trainee do safely, and what must I directly observe?”
New attendings sometimes respond to responsibility in one of two ways.
Some supervise too little because they want residents or fellows to experience autonomy. Others supervise almost everything because they remain uncomfortable with their own new responsibility.
Neither extreme is particularly useful.
The better question is not, “Is this a senior resident?”
It is:
What can this particular trainee safely do in this particular situation?
Consider:
A resident who independently manages routine decompensated heart failure may still need direct involvement for an unfamiliar procedure or an unusual diagnostic problem.
One randomized trial found that increased attending presence during rounds did not significantly reduce medical errors. That finding does not mean that supervision is unnecessary or that high-risk tasks should occur without appropriate oversight (PMID: 29868877).
Readiness question: “Have I explained the uncertainty and the plan for managing it?”
Doctors sometimes fear that saying “I do not know” will undermine confidence.
Patients usually have a more practical concern: What happens next?
There is a major difference between:
“We don’t know.”
and:
“The most likely explanation is X. We are also considering Y because of these findings. The current testing has not completely excluded it, so this is what we are doing next.”
Good communication of uncertainty generally includes:
The evidence around the best wording remains heterogeneous. A systematic review found that clinicians communicate diagnostic uncertainty in many ways and that patient responses vary, although patient-centered approaches were generally viewed positively (PMID: 36127538).
The goal is not false certainty.
The goal is structured uncertainty.
Readiness question: “What needs immediate treatment, what do we know, and who needs to hear it?”
Complications are among the moments when attending responsibility feels most real.
Something has happened. The patient may be worse. The family wants answers. The trainee may be distressed. The team may be looking to you.
The first task is patient care.
Only then should the conversation broaden.
It is also important to distinguish a complication from a preventable error. An adverse outcome does not automatically mean somebody made a mistake.
At the same time, uncertainty about cause should not become an excuse for silence.
Patients generally want to understand:
When an error has occurred, literature on disclosure indicates that patients also value an honest explanation and genuine apology (PMID: 17664451).
Avoid speculation about blame before the facts are established. Involve appropriate institutional patient-safety, disclosure, or risk-management resources when necessary.
Readiness question: “Are we disagreeing about facts, goals, burdens, or the patient’s ability to decide?”
A patient who refuses your recommendation is not necessarily being “difficult.”
Sometimes the disagreement is about medical information.
Sometimes it is about values.
Sometimes there are practical issues the medical team has not recognized.
A patient declining admission may be worried about a child at home. A family demanding continued treatment may not understand the prognosis. A patient declining a procedure may understand the risks perfectly well but value those risks differently than you do.
Before escalating the conflict, determine what the conflict is actually about.
Consider:
When decision-making capacity is genuinely uncertain, the standard framework includes the ability to communicate a choice, understand information, appreciate the situation and consequences, and reason about the options (PMID: 30896380).
A decision you consider unwise is not, by itself, evidence that the patient lacks capacity.
Readiness question: “Where could delay cause the most harm, and what must be reassigned?”
The new attending quickly discovers that clinical medicine is not a sequence of neatly separated decisions.
You may simultaneously have:
Everything may matter.
Everything does not matter equally right now.
The safest approach is to prioritize according to the consequence of delay.
Ask:
An observational study of hospitalist workload found associations between increasing workload and aspects of care efficiency, although it did not establish a universal safe patient census (PMID: 24686924).
The lesson is not that there is one perfect number.
The lesson is that excessive workload should not be treated as a personal endurance contest when it begins interfering with safe reassessment or follow-up.
Readiness question: “Have I verified what I am signing and assigned every consequential unfinished task?”
Signing a note feels final.
Clinical responsibility often is not.
A culture may still be pending. A biopsy may return tomorrow. A patient discharged today may have an imaging result finalized later. A consultant recommendation may require somebody to act after the shift changes.
That unfinished work needs an owner.
Before signing off:
This issue is not theoretical. A study of results returning after hospital discharge found that potentially actionable results were sometimes not known to the relevant physicians (PMID: 16027454).
Structured handoff programs have also demonstrated improvements in communication and reductions in medical errors (PMID: 25372088).
Responsibility has not been transferred simply because your shift ended.
Readiness question: “What am I declining, why, and what safe alternative will I offer?”
One of the less discussed transitions into attending practice is realizing that leadership sometimes means saying no.
A patient may request a treatment you believe is inappropriate.
A family may request escalating interventions that do not align with the clinical situation.
A colleague may ask you to accept a patient into a setting that cannot provide the necessary monitoring.
An administrator may push for a disposition that you believe is unsafe.
A trainee may want to perform something independently before you believe they are ready.
“No” should not mean, “This conversation is over.”
A better formulation is:
“I cannot recommend that approach because of this concern. Here is what I think we can safely do instead.”
For persistent high-stakes disagreement, use appropriate ethics, institutional, and conflict-resolution processes rather than relying on authority alone.
Critical care consensus guidance addressing requests for potentially inappropriate treatment similarly emphasizes communication and fair processes for resolving disagreement (PMID: 25978438).
How to interpret this table: The safest decision depends on the clinical context. These examples show what changes the decision, rather than providing fixed rules.
| Decision | Tempting shortcut | Better readiness question | Evidence note |
|---|---|---|---|
| Admission vs discharge | “They look better.” | Can the next phase of care actually occur safely outside the hospital? | Coordinated discharge planning; PMID: 19189907 |
| Escalation | “The numbers are not that bad yet.” | Can this setting recognize and rescue deterioration? | Clinical deterioration guidance; PMID: 38240510 |
| Consultation | “I’ll just ask them to see the patient.” | What specific decision needs their expertise, and how urgently? | Consultation communication; PMID: 6615097 |
| Supervision | “They’re a senior resident.” | What can this trainee safely do for this patient and task? | Supervision RCT; PMID: 29868877 |
| Uncertainty | “I should sound confident.” | Have I explained what is known, unknown, and next? | Uncertainty communication; PMID: 36127538 |
| Complication | “I need to explain why this happened.” | What needs treatment now, and what facts are actually established? | Error disclosure; PMID: 17664451 |
| Conflict | “The patient is refusing reasonable care.” | What are we actually disagreeing about? | Capacity framework; PMID: 30896380 |
| Prioritization | “I should clear the list in order.” | Where could delay cause the greatest harm? | Workload study; PMID: 24686924 |
| Sign-off | “The note is signed.” | Who owns every unfinished consequential task? | Pending results/handoffs; PMIDs: 16027454, 25372088 |
| Saying no | “I am the attending, so the answer is no.” | Why am I declining this, and what safe alternative can I offer? | ICU consensus; PMID: 25978438 |
Medicine becomes considerably easier when we recognize which things are genuinely context-dependent.
A patient who might safely remain on the floor in a hospital with rapid nursing reassessment and immediate respiratory support may require a different plan somewhere without those resources.
The disease has not changed.
The rescue environment has.
More treatment is not automatically better treatment.
Escalation should remain consistent with the patient’s goals, prognosis, and previously expressed wishes. Critical care guidance specifically recognizes the importance of incorporating goals of care into response planning (PMID: 38240510).
A postgraduate year is a useful shorthand, but it is not a complete description of competence.
The attending needs to know what the trainee has actually demonstrated.
If you remember nothing else, remember these:
There is a useful difference between studying a diagnosis and rehearsing a decision.
Consider this case:
Near handoff, a patient who was expected to go home develops increasing shortness of breath. The resident asks whether discharge can proceed. At the same time, your clinic inbox contains an unresolved abnormal result, and a consultant is returning your call about another patient.
Pause.
Do not ask only, “What is the diagnosis?”
Instead ask:
Then change one variable.
The consultant cannot come for three hours.
Run the decision again.
Now the patient lives alone.
Run it again.
Now the trainee is an experienced fellow rather than an intern.
Run it again.
This type of decision rehearsal can become a useful bridge between traditional medical education and real-world attending readiness.
For a Bytes+ case series, each of the ten decisions could become a separate interactive case:
That format is particularly useful for transition-focused upskilling because it moves beyond “What is the correct answer?” toward “What do you actually do next?”
Simulation research supports improvements in knowledge, skills, and behaviors compared with no intervention, although it does not prove that this particular exercise—or a Bytes+ case series—will improve patient outcomes (PMID: 21900138).
The same framework can be adapted for residents, fellows, medical students, physician assistants, and nurse practitioners, with expectations adjusted to the individual’s training, supervision requirements, and scope of practice.
Becoming an attending does not require you to suddenly know everything.
It requires something more realistic and, in many ways, more difficult: recognizing which decisions belong to you, which decisions require additional help, what cannot safely wait, and what must happen after you leave the room.
That transition can feel uncomfortable because residency and fellowship often reward getting the answer right. Independent practice requires another layer.
You have to make the answer work.
Sometimes that means admitting the patient.
Sometimes it means sending the patient home with a carefully constructed plan.
Sometimes it means calling the ICU, calling a consultant, standing beside a trainee, telling a patient that the diagnosis remains uncertain, or acknowledging that a complication has occurred.
And sometimes it means saying:
“I do not think this is safe.”
That is not failure of independence.
It is one of the clearest signs that clinical judgment is beginning to mature.
Educational only, not personalized medical advice. Individual concerns require clinician guidance; local emergency, supervision, credentialing, and institutional policies remain essential.
Yes. Becoming an attending does not mean that consultation suddenly becomes inappropriate. If another clinician’s expertise, experience, or resources could materially affect a consequential decision, asking for help is part of good clinical judgment. Emergency escalation should not be delayed simply because the diagnosis remains uncertain.
Not automatically. The decision depends on the clinical risk, trajectory, ability to follow up, available support, and whether deterioration could be recognized and managed safely outside the hospital.
Supervision should match the trainee’s demonstrated ability, the complexity of the task, the acuity of the patient, and applicable institutional requirements. Training year alone should not determine how much oversight a particular situation requires.
Explain what you think is most likely, what important possibilities remain, what the available testing can and cannot establish, and what happens next. Uncertainty is usually easier for patients to understand when it comes with a plan.
No. Important pending results need a clearly identified person responsible for reviewing and acting on them. Handoffs should make unfinished work explicit rather than assuming somebody else will notice it.
No. Decision rehearsal can supplement training and help clinicians practice difficult judgment calls, but it does not establish independent competence, confer credentials, or replace appropriate patient-specific supervision.
The name ReviewBytes reflects our belief that medical learning should be clear, focused, and built for the modern learner. Review speaks to scientifically grounded learning methods that improve retention and recall, while Bytes reflects both bite-sized learning and a technology-forward educational experience.
We chose ReviewBytes because it captures the way we think learning should work: evidence-based, efficient, and thoughtfully designed. The name brings together proven review methods with microlearning and AI-powered innovation.
Yes. Many learners search for Review Bytes as a variation of ReviewBytes, and both refer to the same brand and mission.
Absolutely. The “Bytes” in ReviewBytes is a nod to bite-sized learning—breaking complex medical concepts into smaller, easier-to-review pieces—while also reflecting our tech-forward approach.
“Bytes” reflects two ideas: bite-sized learning and a modern, technology-forward approach to education. It captures both accessibility and innovation.
Yes—some people hear or search for ReviewBytes as “review bites.” While the spelling is different, the meaning aligns closely with our mission of smarter, more focused medical learning.
⚠️ Educational disclaimer: This article is intended for education and professional development. It is not personalized medical advice and does not replace institutional policies, credentialing requirements, scope-of-practice rules, or direct clinical supervision.
.webp)
.webp)