Hindsight bias and outcome bias in judging error
Table of contents
In the morning discussion, someone projects the CT scan that confirmed the diagnosis two days after the first consultation. "It was obvious," says a colleague. On the screen, with the outcome known, the signs seem to shout. At the first consultation, with one patient among many and no CT scan, they were whispers among other whispers.
NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.
This gap between what was known before and what is known after creates two biases that weigh in whenever an error is judged: hindsight bias and outcome bias. This guide explains each one, shows what studies found when doctors assessed colleagues' care knowing the outcome, and proposes a fairer way to look at a past decision.
Key points
- Hindsight bias is the tendency to think, after knowing the outcome, that it was more predictable than it seemed beforehand.
- Outcome bias is judging the quality of a decision by the outcome, and not by the information available at the time.
- In a classic study, anesthesiologists judged the same care as less appropriate when the outcome described was permanent.
- Doctors who read vignettes with an error outcome identified about twice as many cognitive biases in the colleague's reasoning.
- Assessing care without revealing the outcome and reconstructing what was known at each moment are the most cited defenses.
What the two biases are
Hindsight bias
Hindsight bias, or hindsight bias, is the "I knew it" feeling after the outcome appears. Anyone who knows the ending reorganizes the memory of the data: the ones that pointed to the diagnosis stand out, and the ones that pointed elsewhere disappear. A 1981 study already investigated this effect in doctors estimating the probability of diagnoses (Arkes et al., 1981).
Outcome bias
Outcome bias, or outcome bias, is a close relative. It judges the decision by what happened afterward: the same conduct seems correct if the patient improved and careless if the patient got worse. The decision is the same; what changed was luck.
Before and after
Both are distinct from calibration and metacognition, which involve assessing one's own reasoning before knowing the result, as shown in the guides on confidence calibration in diagnosis and metacognition in medicine. Here the problem is the judgment that comes afterward, almost always about someone else's decision.
What the studies show
The outcome changes the grade for the care
In a classic study, 112 practicing anesthesiologists judged the appropriateness of care in situations with adverse anesthesia outcomes (Caplan et al., 1991). For each original situation, the authors created an identical version except for the outcome, temporary or permanent, and randomly assigned which version each evaluator received (Caplan et al., 1991).
There was an inverse relationship between the severity of the outcome and the judgment of appropriate care in 15 of the 21 pairs, or 71% (Caplan et al., 1991). The proportion of assessments of appropriate care fell 31 percentage points when the outcome changed from temporary to permanent and rose 28 points in the opposite direction (Caplan et al., 1991).
The same effect in emergency vignettes
A more recent cross-sectional study, with 191 emergency and family medicine doctors in the Netherlands, randomly assigned the outcome reported in acute care vignettes (Plaum et al., 2024). In one of the vignettes, the care was considered appropriate by 44% of the doctors when the outcome was bad, versus 88% when it was good and 84% when no outcome was reported (Plaum et al., 2024).
The doctors who received the bad outcome also estimated a higher probability of that outcome happening (Plaum et al., 2024). It is hindsight in action: knowing the ending makes the ending seem likely.
The outcome changes the biases you see
In a study with vignettes, 37 doctors read eight cases and marked the cognitive biases present in the reasoning described; in half, the outcome indicated a correct diagnosis and, in the other half, an incorrect one (Zwaan et al., 2017). With a correct outcome, the doctors identified an average of 1.75 biases; with an incorrect outcome, 3.45 (Zwaan et al., 2017).
Agreement among physicians on each bias was nearly nil, with kappa between 0.000 and 0.044 (Zwaan et al., 2017). The authors concluded that judging biases is strongly influenced by hindsight, which complicates error reduction strategies based on identifying biases after the fact (Zwaan et al., 2017).
Risk and negligence
In a study with 235 psychiatrists, half received vignettes with the information that a serious outcome had happened shortly after discharge (LeBourgeois et al., 2007). Hindsight bias appeared in the risk assessment but was absent in the judgment of negligence (LeBourgeois et al., 2007). The effect, therefore, varies with the question asked of the evaluator.
In other areas
A scoping review on cognitive biases in the interpretation of radiology images included 24 publications, and hindsight bias was one of the five main themes (Chen et al., 2023). The results were mixed, and strategies to reduce these biases remained largely untested (Chen et al., 2023).
.jpg){kind=link}
Why this matters in judging error
In case discussions
Case discussion and morbidity and mortality sessions almost always start with the outcome. This is unavoidable, but it creates the ideal ground for both biases. Good moderation asks that the presentation follow the timeline and that the audience reason with what was known at each moment.
In peer review and expert assessment
The 1991 study was motivated by this problem: those who review care knowing of a permanent harm tend to judge it more strictly (Caplan et al., 1991). This applies to internal committees, audits and expert assessments.
In the physician
The bias does not affect only those judging from outside. The physician involved in a bad outcome also revisits their own decision with hindsight, and may blame themselves for having ignored a sign that, at the time, was ambiguous. The emotional impact of error on the professional, called the second victim, and the conversation with the patient about the error have their own guides in this series.
In learning
If every error analysis overestimates what was predictable, the lesson drawn becomes "pay more attention", and not a concrete change of process. The result is blame without learning.
How to judge a decision fairly
Hide the outcome when you can
The most direct defense is to evaluate care without knowing the outcome. Vignette studies show the difference this makes: in the Dutch vignette, the assessment without outcome came close to the assessment with a good outcome (Plaum et al., 2024). In quality reviews, an evaluator can read the record up to the point of the decision before knowing what came after.
Rebuild the timeline
For each decision, note three things: what was known, what was reasonable to ask and what the alternatives were. The guide on how to organize clinical reasoning in the medical record shows why a clear record of hypotheses makes this reconstruction easier later.
Separate decision from outcome
A good decision can have a bad outcome, and a bad decision can turn out well. Evaluating the two in separate columns helps: was the decision reasonable with the information at the time? Why was the outcome bad? The guide on diagnostic uncertainty deals with deciding well when the result is not guaranteed.
Distrust the list of biases made afterwards
The 2017 study showed that, knowing of an error, physicians see twice as many biases and disagree among themselves about which ones they are (Zwaan et al., 2017). Naming biases helps to teach, as in the guide on examples of heuristics, but the label assigned after the outcome is a hypothesis, and not a diagnosis of someone else's reasoning.
Ask about the system
When the question changes from "who made the error?" to "what made this error more likely?", hindsight loses force, because the analysis moves away from the person and toward working conditions.
For the student: learning from someone else's case
Stop before the end
When reading a case report or watching a discussion, try to stop before the reveal and write your hypothesis with the data from that moment. Only then read the end. The habit protects against "it was obvious" and shows how difficult the case really was.
Compare your estimate with the outcome
Also note how confident you were in the hypothesis. The gap between confidence and outcome, accumulated over many cases, teaches more than any single case, and it is a topic in the error log.
Be generous with whoever made the decision
When a colleague made a mistake, ask what they knew and what you would have done in their place, with the same tiredness and the same queue. The guide to feedback in clinical cases covers the tone of these conversations.
See the case clue by clue in NEXO
In NEXO, the case of the day arrives in up to six clues, and the player decides without knowing the ending. After the game, looking back and asking at which clue the diagnosis actually became visible is an exercise against hindsight: often the answer is later than it seemed. The distribution of correct answers by clue shows when you usually get there.
With each guess, the proximity in ICD-10 across five levels shows how far off the hypothesis was. The cases are fictional with a real clinical presentation, and the guide what NEXO is explains the whole game. How the modes are unlocked is on the NEXO economy page.
If you coordinate a medical school, a student league or a residency and want to discuss cases in a group clue by clue, see NEXO Closer. Researchers in medical education can propose studies through NEXO Researcher. Doctors who review cases take part through NEXO Curator, and those who create health content on social media can take part through NEXO Creators.
Frequently asked questions
What is hindsight bias?
It is thinking, after knowing the outcome, that it was more predictable than it seemed beforehand.
What is outcome bias?
It is judging the quality of a decision by the outcome, and not by the information available at the moment it was made.
Do doctors have these biases?
Yes. Anesthesiologists judged the same care as less appropriate when the outcome was permanent (Caplan et al., 1991).
Does the outcome change the analysis of biases?
In one study, doctors pointed out about twice as many biases when the outcome indicated error (Zwaan et al., 2017).
How can hindsight bias be reduced?
By assessing the care without knowing the outcome when possible, reconstructing the timeline and separating decision from result.
What is the difference from calibration?
Calibration looks at your own confidence before the outcome; hindsight bias distorts judgment after it.
Checklist for judging without hindsight
- Read up to the decision before knowing the outcome.
- Note what was known at each moment.
- List the reasonable alternatives of that moment.
- Separate decision from outcome in different columns.
- Treat the list of biases as a hypothesis.
- Ask about the system, and not only about the person.
- Write your hypothesis before reading the end of a case.
Want to train your decision-making without knowing the ending? Play the case of the day in NEXO.
References
- Arkes HR, Wortmann RL, Saville PD, Harkness AR. Hindsight bias among physicians weighing the likelihood of diagnoses. J Appl Psychol. 1981. https://pubmed.ncbi.nlm.nih.gov/7240061/
- Caplan RA, Posner KL, Cheney FW. Effect of outcome on physician judgments of appropriateness of care. JAMA. 1991. https://pubmed.ncbi.nlm.nih.gov/2008024/
- Plaum P, Visser LN, de Groot B, Morsink MEB, Duijst WLJM, Candel BGJ. Using case vignettes to study the presence of outcome, hindsight, and implicit bias in acute unplanned medical care: a cross-sectional study. Eur J Emerg Med. 2024. https://pubmed.ncbi.nlm.nih.gov/38364049/
- Zwaan L, Monteiro S, Sherbino J, Ilgen J, Howey B, Norman G. Is bias in the eye of the beholder? A vignette study to assess recognition of cognitive biases in clinical case workups. BMJ Qual Saf. 2017. https://pubmed.ncbi.nlm.nih.gov/26825476/
- LeBourgeois HW 3rd, Pinals DA, Williams V, Appelbaum PS. Hindsight bias among psychiatrists. J Am Acad Psychiatry Law. 2007. https://pubmed.ncbi.nlm.nih.gov/17389347/
- Chen J, Gandomkar Z, Reed WM. Investigating the impact of cognitive biases in radiologists' image interpretation: A scoping review. Eur J Radiol. 2023. https://pubmed.ncbi.nlm.nih.gov/37541180/