UNCERTAINTY + CALIBRATION: WHEN TO PAUSE, TEST OR ESCALATE
ALL YEARS | PAGE 78C - CONFIDENCE -> EVIDENCE -> TEST VALUE -> TIMEOUT -> TEAM / PATIENT INPUT ->
C. UNCERTAINTY CONTROL PANEL
| Signal | Question to ask | Best response |
|---|---|---|
| Confidence > evidence Low probability / high consequence | How much of my certainty comes from objective support vs familiarity? Is the dangerous alternative unlikely but unacceptable to miss? | Lower confidence, reopen alternatives and name missing evidence. Use an appropriate rule-out / escalation strategy rather than probability alone. |
| Test contemplated | Will a normal or abnormal result change diagnosis, treatment, monitoring, location or follow-up? | If neither result changes a decision, the test may have low decision value. |
| Discordant finding Unexpected trajectory | What important feature does my current model fail to explain? Did the patient improve, deteriorate or fail to respond as expected? | Re-rank the differential; do not force the finding into the existing story. Treat trajectory as new diagnostic data and reassess the model. |
| Handoff / inherited label Residual uncertainty at discharge / follow-up | Would I reach the same diagnosis from the raw data? What future change should trigger re-evaluation? | Reconstruct independently before accepting diagnostic momentum. Define follow-up, return precautions and pending-result ownership. |
D. 30-SECOND DIAGNOSTIC TIMEOUT
| Prompt | What it is trying to catch |
|---|---|
| 1. What does not fit? 2. What dangerous diagnosis remains plausible? | Anchoring / confirmation / overfitting. Premature closure / consequence blindness. |
| 3. What alternative would explain the same data differently? 4. What result would actually change my next action? | Framing / representativeness / diagnostic momentum. Shotgun testing / low-value data collection. |
| 5. What will I reassess, and when? | Failure to use trajectory, follow-up and response as diagnostic evidence. |
TOP 10 DIAGNOSTIC-REASONING CONCEPT - 7: CONFIDENCE SHOULD TRACK EVIDENCE - NOT FLUENCY
| DEEP EXPLANATION | WHY STUDENTS MISS IT | EASIEST CLUE |
|---|---|---|
| Pattern recognition can feel effortless and therefore certain. Calibration asks whether confidence is proportional to the | Fluent recognition feels like proof, while uncertainty feels psychologically uncomfortable. | Say the confidence level out loud and name the single piece of evidence most likely to change it. |
| quality, consistency and independence of the evidence. | ||
| TOP 10 DIAGNOSTIC-REASONING CONCEPT - 8: TESTS UPDATE PROBABILITY; THEY DO NOT REPLACE THE CLINICAL MODEL | ||
| DEEP EXPLANATION | WHY STUDENTS MISS IT | EASIEST CLUE |
| A test result has meaning only in context: pretest probability, test characteristics, timing, spectrum and whether the result fits | Numbers and images look objective, so learners may treat them as independent of context or assume any abnormal result | Before ordering or accepting a result, ask what probability or decision it is supposed to change. |
| the clinical picture. A discordant test should trigger interpretation, not automatic obedience. | explains the case. | |
| TOP 10 DIAGNOSTIC-REASONING CONCEPT - 9: DIAGNOSTIC TIMEOUT + FEEDBACK CLOSE THE REASONING LOOP | ||
| DEEP EXPLANATION | WHY STUDENTS MISS IT | EASIEST CLUE |
| A brief pause can interrupt cognitive momentum, but long-term calibration also requires feedback: outcomes, follow-up, test | Learners move to the next case and rarely see whether the diagnosis was ultimately confirmed or revised. | Build one feedback question into case closure: “How will I learn whether this diagnosis was right?” |
| results, supervisor review and case reflection. Without outcome feedback, confidence can remain wrong in either direction. 2026/27 diagnostic-reasoning expanded edition - educational use; follow school requirements, supervising clinicians and current clinical guidance for patient care |