Diagnostic uncertainty: strategies to decide well

The patient arrived with diffuse abdominal pain, the initial tests came back normal, and the preceptor asks what your diagnosis is. You want to say "I don't know yet" and ask for one more test, because deciding without certainty seems wrong. A review on the topic describes exactly this pattern: students and residents tend to treat certainty as a condition for acting, and that is why they struggle in situations where it does not come (Ilgen et al., 2019).

NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.

This text explains what diagnostic uncertainty is, why a doctor's confidence and diagnostic accuracy do not always go together, and which strategies help you decide with incomplete information. At the end, we show how to train this skill with short cases.

Key points

• Diagnostic uncertainty is the perception of not being able to explain the patient's problem precisely, and it changes throughout the encounter.
• Confidence and accuracy go together only weakly, so it is worth checking your own confidence with objective questions.
• Decision thresholds help answer whether it is time to observe, investigate further or treat.
• Planned reassessment gives uncertainty a script: alarm signs, when to return and what to do with the results.
• Tolerating uncertainty is associated with greater well-being in students and doctors, and practice with feedback helps calibrate confidence.

Summary

• What diagnostic uncertainty is
• Confidence and accuracy go separately
• Tolerating uncertainty is also learned
• Strategies for deciding without certainty
• Planned reassessment: uncertainty becomes a script
• How to train uncertainty with cases
• Frequently asked questions
• Checklist for deciding with uncertainty

What diagnostic uncertainty is

Every diagnosis starts uncertain. The main complaint suggests a few hypotheses, each new piece of data changes their order, and at some point you decide with what you have. The difficulty is that the word "uncertainty" covers different things, and each one calls for a different response.

A definition proposed by the literature

A systematic review looked at how diagnostic uncertainty was defined and measured and found 123 articles for full reading, none of them with a definition of the term (Bhise et al., 2018). From what they found, the authors proposed defining diagnostic uncertainty as the "subjective perception of an inability to provide an accurate explanation for the patient's health problem" (Bhise et al., 2018).

Three characteristics that matter

The same review identified three attributes useful for measuring uncertainty (Bhise et al., 2018):

• It is subjective. The clinician is the one who feels the uncertainty, and two doctors facing the same case may feel it differently.
• It affects the investigation. Poorly managed, it can delay the diagnosis.
• It is dynamic. It changes over time, as new data arrive.

The third point is the most practical. The uncertainty of the first minute of the consultation is not the same as at the end of the shift, and your task is to track how it changes, not to eliminate it all at once.

Different sources, different responses

A taxonomy published in a medical decision journal organizes uncertainty in health into three dimensions: the source, the type of question and the place where it is (Han et al., 2011). The authors argue that separating the types of uncertainty helps choose suitable strategies for each one (Han et al., 2011).

In diagnostic practice, it helps to ask three questions:

Question	Example	Most useful response
Is information about this patient missing?	The test that separates the hypotheses has not been done yet	Decide whether the test changes management
Is the available knowledge limited?	The evidence on that finding is weak or conflicting	Recognize the limit and use the best available data
Is the situation too complex for a single answer?	Several diseases at the same time, overlapping symptoms	Separate the problems and prioritize the most serious one

Table source: questions prepared by the Editorial team from the idea of separating sources and types of uncertainty in Han et al., 2011; the examples are illustrative.

Confidence and accuracy go separately

Feeling confident is not the same as being right. Studies on calibration, the relationship between how often you are right and how often you think you are right, show that this difference appears even in experienced doctors.

What calibration studies show

Vignette studies compared how often physicians and students get it right with how confident they are in their own answer, in easy and difficult cases. The numbers from these studies are gathered in the "Feedback and calibration" section of the text about clinical reasoning games: in short, confidence and accuracy go together only weakly, and overconfidence appears among the causes of diagnostic error.

Why this matters for uncertainty

The editorial reading of these studies is useful: the feeling of uncertainty is an imperfect signal. Sometimes you feel confident in a difficult case and you are wrong; sometimes you feel unsure and you are right. So it is worth having a way to check confidence that does not depend only on the feeling, as we will see in the strategies.

The opposite side: too much uncertainty

Overconfidence has a mirror: the difficulty of acting while doubt exists. The review on comfort with uncertainty describes that comfort is not feeling certain about the diagnosis, but trusting your own ability to act, or not to act, safely in that situation (Ilgen et al., 2019). According to the authors, this comfort relies on self-monitoring and on planning what comes next (Ilgen et al., 2019).

Tolerating uncertainty is also learned

Tolerance of uncertainty is the ability to deal with insufficient, ambiguous or unreliable information without freezing. It has been studied as an individual trait, and the results matter to those in training.

What studies associate with tolerance

A systematic review gathered 67 studies on tolerance of uncertainty and outcomes in health and in health education (Strout et al., 2018). Tolerance was associated with cognitive, emotional and behavioral outcomes of students, professionals and patients, and the most consistent association was with emotional well-being (Strout et al., 2018).

Another review, focused on medical education, included 11 studies, and all reported an association between lower tolerance of ambiguity or uncertainty and lower psychological well-being (Hancock and Mattick, 2020).

The limits of this evidence

Both reviews make the same caveat. The studies were mostly cross-sectional, heterogeneous in how they measured and of low methodological quality, which limits strong conclusions about cause and effect (Strout et al., 2018; Hancock and Mattick, 2020). What can be said is that tolerating uncertainty and being well go together; it is not yet known with confidence which comes first.

What can be trained

If tolerance is partly a skill, it develops with exposure to uncertain situations and with feedback on the decisions made in them. The review on comfort with uncertainty proposes precisely using this concept to guide teaching and research programs (Ilgen et al., 2019). For the student, this means practicing decisions with incomplete information in an environment where making a mistake teaches and harms no one.

Strategies for deciding without certainty

The strategies below organize reasoning when the answer is not yet clear. They do not replace knowledge of diseases, and they work best when you already have reasonable hypotheses.

Name the uncertainty and its source

Tell yourself, or write down, what exactly is uncertain. "I don't know if it is appendicitis or mesenteric adenitis" is different from "I don't know what this imaging finding means" and from "I don't know if she will come back for reassessment". Each sentence points to a different next step, as in the table of sources of uncertainty.

Think in thresholds, not in certainty

The threshold model describes two probability points that guide the decision: below the testing threshold, do not treat and do not test; above the test and treatment threshold, treat without further exams; between the two, do the exam and decide by the result (Pauker and Kassirer, 1980). The authors show that these thresholds depend on the reliability and risks of the exam and on the benefits and risks of treatment (Pauker and Kassirer, 1980).

The lesson for uncertainty is direct: you rarely need certainty to act. You need to know whether the probability of the disease has already passed the point where action pays off. The text on pretest probability shows how to estimate that probability, and the one on clinical decision-making goes deeper into choosing the test.

Work with one hypothesis and an open list

Choose a working hypothesis and keep visible what might be behind it. The question "what can I not let slip by?" ensures that a serious and treatable cause stays on the list until it is ruled out. This habit also protects against premature closure, explained in the text on differential diagnosis.

Check confidence with questions, not with a feeling

Since confidence and accuracy go poorly together, as the section "Feedback and calibration" of the text on clinical reasoning games shows, it is worth using objective questions before closing:

• Do all the important findings fit my hypothesis?
• Is there a finding that contradicts the hypothesis and that I set aside?
• If my hypothesis is wrong, what would the second one be, and what would separate it from the first?
• Is this case similar to one I saw recently and that may be pulling my answer?

Ask for help for the right reason

Practical suggestion: since the feeling of doubt is an imperfect signal, use objective criteria to ask for a second opinion, such as a finding the hypothesis does not explain, a sign of severity or a case that falls outside what you usually see. In the internship, presenting the case to the preceptor with your uncertainty made explicit is part of the training, as the text on priority and hypothesis in the internship shows.

Record what you know

When the visit ends without a diagnosis, record the most specific symptom or finding, instead of a diagnosis that has not yet been made. The guide on how to find the right code in ICD-10 explains how to use the symptoms chapter for this. An honest record helps whoever will review the patient know where to start.

Common mistakes in the face of uncertainty

• Ordering a test to relieve anxiety. Before ordering it, ask which result would change your management. If none would, the test does not resolve the uncertainty that matters.
• Hiding the doubt in the case presentation. Telling the preceptor "my hypothesis is this, with moderate confidence, and what would make me change is that finding" shows reasoning, not weakness.
• Closing early to end the discomfort. The urge to end the doubt pushes you toward the first plausible hypothesis. An open list, reviewed with each piece of data, protects against this.
• Treating all uncertainty as the same. Lack of information, the limit of knowledge and the complexity of the case call for different responses, as the table of sources shows.
• Forgetting the afterward. A decision made under uncertainty without a reassessment plan leaves the patient not knowing what to do if the picture changes.

Planned reassessment: uncertainty becomes a plan

One of the most concrete ways to deal with uncertainty is to plan what happens after the visit. In the primary care literature, this has a name: safety netting, which can be translated as rede de segurança or planned reassessment.

What the review found

A literature review with 47 studies defined planned reassessment as a consultation technique to communicate uncertainty, inform the patient about alarm signs and plan future visits that ensure reassessment at the right time (Jones et al., 2019). The guidance may include the expected course of the disease, the symptoms that should worry the patient and how and when to seek help (Jones et al., 2019).

The authors point out that planned reassessment goes beyond the conversation in the consultation. It includes following up on test results and referrals, and it was considered especially important in the care of children, people with an acute picture, patients with several diseases and people with mental health problems (Jones et al., 2019).

The four elements of a good plan

Based on the definition from the review, a reassessment plan has four parts:

Element	Question it answers	Example sentence to the patient
Communicate the uncertainty	What do we know and what do we not know yet?	"Based on today's examination, the most likely cause is this, but it is not yet possible to be sure."
Alarm signs	What changes the plan?	"If this symptom appears, seek care the same day."
When to reassess	How soon do we look again?	"If it doesn't improve by such and such a day, come back so we can reassess."
Results and referrals	Who checks the test and the specialist's follow-up?	"I'll look at the result and get in touch."

Source of the table: elements of the definition by Jones et al., 2019; the sentences are examples from the Editorial and do not replace the clinical guidance for each case.

Time as a tool

When the probability of a serious disease is low and the patient is stable, watching how things evolve can be part of the investigation. Planned reassessment is what makes this choice safe: time passes with a script, and not with the patient alone. The decision to observe, investigate or treat remains with the clinician, based on the picture and the guidelines; the script only makes clear what to do if the picture changes.

A commented example

Go back to the patient from the opening, with diffuse abdominal pain and normal initial tests. Applying the strategies from this text, the reasoning goes like this:

1. Name the uncertainty: the main doubt is between a benign, self-limited cause and a surgical cause in an early stage, which has not yet given clear signs.
2. Think about the threshold: with the patient stable and without signs of severity, the probability of a serious cause may be below the point at which immediate treatment would be worthwhile, but not low enough to close the matter.
3. Working hypothesis and open list: the working hypothesis is the benign cause; the surgical cause remains on the list as the one that cannot be missed.
4. Reassessment script: explain to the patient what is known and what is not known, which symptoms call for returning the same day, how soon to reassess and who checks the pending tests.
5. Record: note the symptom as the diagnosis for the visit and leave the script written for whoever sees the patient later.

The example is educational and does not replace clinical assessment or the guidelines for each situation. What it shows is the structure: uncertainty still exists, but each step has a reason and a next step.

What this teaches the student

For those who are learning, planned reassessment shows that uncertainty does not need to be resolved during the consultation. It needs to be managed. A good practice is to end each study case with the question: if I am wrong, how and when will I find out?

How to train uncertainty with cases

Tolerance and calibration improve with practice: many decisions with incomplete information, each followed by feedback. Short cases, with data arriving little by little, reproduce this situation well.

A script for each case

1. After the first piece of data, write the working hypothesis and your confidence, from 0 to 10.
2. With each new piece of data, update the hypothesis and the confidence, and note what changed.
3. Before answering, say what the source of the remaining uncertainty is.
4. When you see the answer, compare the confidence you gave with the result.
5. After a few cases, look at the pattern: do you get it wrong more often when you are sure or when you are unsure?

This simple record is a direct way to check your own calibration, the topic of the section "Feedback and calibration" in the text about clinical reasoning games.

Training on NEXO

On NEXO, the case arrives in up to six clues, one at a time, and you decide when to make your guess. Each wrong guess shows the proximity in ICD-10 across five levels, from different chapter to exact ICD, and opens the next clue. It is a practical way to exercise the script above: decide with incomplete information, see how close you got and adjust the hypothesis.

The Daily brings one case per day, good for keeping the habit of recording your confidence. The Archive gathers previous daily cases for training in volume. Adrenaline adds time, which brings the training closer to the pressure of deciding quickly. Multimedia brings cases with images, audio and other tests, and Continents changes the underlying epidemiology. How each mode is unlocked is on the NEXO economy page.

Keeping what uncertainty taught

When a case shows that you were too sure, or unsure for no reason, record the finding that should have changed your confidence. The Anki button creates the card straight from NEXO and opens it in the Anki app. The error notebook method helps separate errors of knowledge, reasoning and attention.

Discussing in a group, researching and reviewing

In the Group tab of the ranking, invited colleagues share their own ranking. With the class playing the same Daily case, it is worth comparing which clue each person felt confident enough to answer on and who got it right. If you coordinate a medical school, a league or a residency program and want to use cases like this in teaching, check out NEXO Closer. Researchers interested in confidence calibration and diagnostic reasoning can propose studies through NEXO Researcher, and physicians who want to review cases in their specialty can take part through NEXO Curator, and those who produce medical content can check out NEXO Creators.

Frequently asked questions

What is diagnostic uncertainty?

It is the perception of not being able to explain the patient's health problem precisely. A systematic review proposed this definition and described uncertainty as subjective, capable of affecting investigation and dynamics over time (Bhise et al., 2018).

How do you deal with diagnostic uncertainty in practice?

Name the source of the uncertainty, consider whether the probability of the disease has already crossed the threshold to act, keep a working hypothesis with an open list and plan reassessment. The threshold model shows that the decision depends on the probability and on the risks and benefits of the test and the treatment (Pauker and Kassirer, 1980).

Is feeling confident in the diagnosis a good sign?

It is an imperfect sign. The calibration studies summarized in the "Feedback and calibration" section of the text about clinical reasoning games show that confidence and accuracy go together only weakly. Use objective questions to check the hypothesis before closing.

What is safety netting?

It is planned reassessment: communicating the uncertainty, explaining the alarm signs and agreeing on when and how to reassess, as well as following up on tests and referrals (Jones et al., 2019).

Is tolerance of uncertainty related to well-being?

Studies indicate that it is: two systematic reviews found an association between greater tolerance and more well-being, with the caveat that the quality of the studies was low (Strout et al., 2018; Hancock and Mattick, 2020).

Does ordering more tests resolve uncertainty?

Sometimes. The test helps when the probability of the disease is between the testing and treatment thresholds, that is, when the result can change the decision (Pauker and Kassirer, 1980). Outside that range, the test adds little to management.

How can a student train decisions under uncertainty?

Practice cases in which the data arrive gradually, record your confidence at each step and compare it with the outcome at the end. The review on comfort with uncertainty points to self-monitoring and planning as the basis of that comfort (Ilgen et al., 2019).

How does NEXO help train uncertainty?

In NEXO, you decide your guess with up to six clues and receive the ICD-10 proximity with each attempt. This allows you to train decision-making with incomplete information and compare your confidence with the outcome, case by case. To learn about the game, read what NEXO is.

Checklist for deciding under uncertainty

1. Say what is uncertain and where the uncertainty comes from.
2. Choose a working hypothesis and keep an open list.
3. Include the serious cause that cannot be missed until it is ruled out.
4. Think about the threshold: observe, investigate or treat.
5. Ask whether the test changes management before ordering it.
6. Check your confidence with questions, not just with the feeling.
7. Use the difficulty of the case as a trigger to ask for help.
8. Plan reassessment: alarm signs, timeframe and who checks the results.
9. Record what you know, and keep the suspicion recorded as a suspicion.
10. Train with short cases and compare confidence with accuracy.

Want to train decisions with incomplete information today? Play the case of the day on NEXO.

References

1. Bhise V, Rajan SS, Sittig DF, Morgan RO, Chaudhary P, Singh H. Defining and Measuring Diagnostic Uncertainty in Medicine: A Systematic Review. J Gen Intern Med. 2018. https://pubmed.ncbi.nlm.nih.gov/28936618/
2. Han PK, Klein WM, Arora NK. Varieties of uncertainty in health care: a conceptual taxonomy. Med Decis Making. 2011. https://pubmed.ncbi.nlm.nih.gov/22067431/
3. Ilgen JS, Eva KW, de Bruin A, Cook DA, Regehr G. Comfort with uncertainty: reframing our conceptions of how clinicians navigate complex clinical situations. Adv Health Sci Educ Theory Pract. 2019. https://pubmed.ncbi.nlm.nih.gov/30390181/
4. Strout TD, Hillen M, Gutheil C, et al. Tolerance of uncertainty: A systematic review of health and healthcare-related outcomes. Patient Educ Couns. 2018. https://pubmed.ncbi.nlm.nih.gov/29655876/
5. Hancock J, Mattick K. Tolerance of ambiguity and psychological well-being in medical training: A systematic review. Med Educ. 2020. https://pubmed.ncbi.nlm.nih.gov/31867801/
6. Jones D, Dunn L, Watt I, Macleod U. Safety netting for primary care: evidence from a literature review. Br J Gen Pract. 2019. https://pubmed.ncbi.nlm.nih.gov/30510099/
7. Pauker SG, Kassirer JP. The threshold approach to clinical decision making. N Engl J Med. 1980. https://pubmed.ncbi.nlm.nih.gov/7366635/