The rational diagnostician: Understanding probability, risk and communicating uncertainty
•Download as PPT, PDF•
0 likes•309 views
The document discusses the uncertainty inherent in medical diagnosis and treatment. It notes that diagnoses assign probabilities rather than certainties, and that tests have varying sensitivity and specificity levels. A key challenge is determining acceptable risk levels, such as how many missed diagnoses are tolerable. The document advocates considering pre-test probability, test accuracy metrics, and residual risk to determine appropriate diagnostic or discharge pathways. Overall, it emphasizes communicating risk to patients and balancing risk versus benefit in clinical decision making.
Recommended
Recommended
More Related Content
What's hot
What's hot (18)
Viewers also liked
Similar to The rational diagnostician: Understanding probability, risk and communicating uncertainty
Similar to The rational diagnostician: Understanding probability, risk and communicating uncertainty (20)
More from Simon Carley
More from Simon Carley (20)
Recently uploaded
Recently uploaded (20)
The rational diagnostician: Understanding probability, risk and communicating uncertainty
- 2. 3 Things • Probability • Risk • Me
- 3. When is a diagnosis useful?
- 5. How sure are you? • 98-100% • 95-98% • 90-95% • 85-90% • <85%
- 7. • Medicine is a science of uncertainty and an art of probability.
- 8. Truth table and calculations Sensitivity=a/a+c Specificity=d/d+b
- 9. H-FAB in AMI Sensitivity 97% Specificity 54% HFABP 125 261 4 315
- 10. What answers do we really get from a Truth table? What does a +ve test result really mean? What does a -ve test result mean? HFABP 125 261 4 315
- 11. H-FAB testing for AMI • A +ve test means they have the diagnosis • A +ve test means they probably have the diagnosis • A +ve test means they are more likely to have the diagnosis than before I did the test HFABP 125 261 4 315
- 12. Myth: Reality vs. research •Researchers live in the clouds •Clinicians have their feet on the ground
- 13. Probability • Diagnosis often assigns a probability • We will miss some people with disease • We will mislabel (and treat) people without disease
- 14. A problem of risk
- 15. PE
- 16. • Patients with pleuritic type chest pain in the ED.
- 18. • Angiography? • CTPA? • 3/12 follow up?
- 19. • Time • Expense • Danger • Misdiagnosis?
- 21. Which screening test? • D-dimer rules out • Wells score rules out • PERC score rules out • Wells and PERC rules out • Wells + D-dimer rules out • Other (e.g. dopplers, flip a coin, astrology etc)
- 22. What’s an acceptable level of risk?
- 23. 1:50 1:25 1:10
- 24. Example - PERC rule • • • • • • • • Age>49 Pulse>100 Pulse Ox<95% Current haemoptysis Oestrogen use Previous venous thromboembolism Recent surgery Unilateral leg swelling PERC Rule sens = 97.4% spec = 21.9%
- 25. 1:50 1:25 1:10
- 26. Pre test probability for PE 6.9 % pre test probability of PE.
- 27. 21.9% negative 78.1% don’t know
- 28. 21.9% negative
- 30. PERC positive 6.7 of 6.9 PE patients PERC negative 0.2 of 6.9 PE patients
- 31. PERC positive need further testing PERC negative patients sent home
- 32. Overall risk is... • Gestalt plus PERC rule will fail to spot two PEs for every thousand patients investigated.
- 33. Happy? Yes No
- 36. QuickTime™ and a decompressor are needed to see this picture.
- 37. Diagnosis missed It gets worse They don’t come back The treatment does not work Harm takes place Diagnosis made They get better anyway They do come back ..but the diagnosis is missed again! The treatment works Significant harm avoided
- 39. Multi-step sequence needed for harm to take place
- 40. Me
- 41. • The only statistics you can trust are the ones you falsified yourself.
- 43. Patient confidence Physician certainty Patient concern Physician discomfort Uncertainty, responsibility and the evolution of the physician/patient relationship Henry MS JMedEthics 2006;32:321-323
- 44. I don’t have a PE You don’t have a PE I might have a small risk of PE That patient might have a PE and sue me Uncertainty, responsibility and the evolution of the physician/patient relationship Henry MS JMedEthics 2006;32:321-323
- 45. I don’t have a PE You don’t have a PE Is honesty is good for doctor and patient? Uncertainty, responsibility and the evolution of the physician/patient relationship Henry MS JMedEthics 2006;32:321-323
- 46. Take away • Do I make diagnoses or probabilities? • Is some risk acceptable in my practice? • How do I communicate risk to patients?
- 47. Thank you
- 48. Risk vs Benefit • Benefits to patients • Access to therapy • Knowledge • Treatment benefit • Cost • Risks to patients • Denied therapy • Risk of therapy • Risk of investigation • Cost
Editor's Notes
- General question as ice breaker Common reasons likely to be labels, prognosis, therapy. Ultimately because it is useful to know. BUT the key to a rational diagnostician is that it should be beneficial for the patient and not just the physician
- 55 year old man with 2 hour hx of chest pain. no risk factors. previously fit and well
- Digivote slide Answer is 90-95%
- Howard Atwood Kelly (February 20, 1858 – January 12, 1943) was an American gynecologist. He was one of the "Big Four" founding professors at Johns Hopkins Hospital.[1] (The "Big Four" were William Osler, Professor of Medicine; William Stewart Halsted, Professor of Surgery; Howard A. Kelly, Professor of Gynecology; and William H. Welch, Professor of Pathology.) Kelly is credited with establishing gynecology as a true specialty.[2] Harvey Williams Cushing, M.D. (April 8, 1869 - October 7, 1939), was an American neurosurgeon and a pioneer of brain surgery, and the first to describe Cushing's syndrome.[1] He is often called the "father of modern neurosurgery. Thayer prof medicine at John Hoplins
- Objective: Many Emergency Departments (EDs) utilise ‘triple marker’ testing with CK-MB, myoglobin and troponin I (cTnI) to exclude acute myocardial infarction (AMI) within hours of presentation. We evaluated the ability of 8 biomarkers to rapidly exclude AMI at the point of presentation and investigated whether ‘triple marker’ testing represents the optimal multimarker strategy. Methods: We recruited patients who presented to the ED with suspected cardiac chest pain occurring within 24 h. Blood was drawn at the time of presentation. Diagnostic value was assessed by calculating the area under the ROC curve (AUC) and a multivariate model was constructed by logistic regression. The primary outcome was a diagnosis of AMI, established by ≥12-h troponin testing in all patients. Results: 705 included patients underwent venepuncture a median of 3.5 h after symptom onset. Heart fatty acid binding protein (H-FABP) had an AUC of 0.86 (95% CI 0.82–0.90), which was significantly higher than any other biomarker including cTnI. While no single biomarker could enable exclusion of AMI, multivariate analysis identified cTnI and H-FABP as the optimal biomarker combination. Combined with clinical risk stratification, this strategy had a sensitivity of 96.9%, specificity of 54.7%, PPV 32.4% and NPV 98.8%. Conclusions: We have derived an algorithm that would enable AMI to be immediately excluded in 315 (44.7%) patients at the cost of missing 6 AMIs per 1000 patients treated. While the risk is likely to be unacceptable for clinical implementation, we have highlighted an area for future development using serial testing and increasingly sensitive assays.
- Digivote LR +ve 2.138 LR -ve 0.057 +ve test is not a diagnosis They still probably don’t have a diagnosis. Pretest probability was 129/705 = 18.3% After testing probability is = 33.2%
- So as a clinician we have no idea whether our patient has the right diagnosis or not. We just know that they probably do or they probably don’t. Even with tests that have a high sensitivity.
- Probability So if we accept that we have a probability issue then we have a problem of risk.
- Digivote D-dimer sensitivity 90% Wells score misses 3.4% PERC score rules out 2.6% Wells and PERC rules out 1% (but you only rule out a small number) Wells + D-dimer rules out 2% (rules out about 50%) Other (e.g. dopplers, flip a coin, astrology etc)
- chest pain tests.
- Digivote
- 1666 ?PE patients in 13 US and NZ EDs Gestalt plus PERC gives the following PERC Rule sens = 97.4% spec = 21.9% Designed with acceptable miss rate of 2% Even in this study PEs were missed. ^ Kline, JA; Courtney, DM; Kabrhel, C; Moore, CL; Smithline, HA; Plewa, MC; Richman, PB; O'Neil, BJ et al (2008). "Prospective multicenter evaluation of the pulmonary embolism rule-out criteria". Journal of Thrombosis and Haemostasis 6 (5): 772–780. doi:10.1111/j.1538-7836.2008.02944.x. PMID 18318689. J Thromb Haemost. 2008 May;6(5):772-80. Epub 2008 Mar 3. Prospective multicenter evaluation of the pulmonary embolism rule-out criteria. Kline JA, Courtney DM, Kabrhel C, Moore CL, Smithline HA, Plewa MC, Richman PB, O'Neil BJ, Nordenholz K. Source Department of Emergency Medicine, Carolinas Medical Center, Charlotte, NC 28323-2861, USA. jkli[email_address].org Abstract BACKGROUND: Over-investigation of low-risk patients with suspected pulmonary embolism (PE) represents a growing problem. The combination of gestalt estimate of low suspicion for PE, together with the PE rule-out criteria [PERC(-): age < 50 years, pulse < 100 beats min(-1), SaO(2) >or= 95%, no hemoptysis, no estrogen use, no surgery/trauma requiring hospitalization within 4 weeks, no prior venous thromboembolism (VTE), and no unilateral leg swelling], may reduce speculative testing for PE. We hypothesized that low suspicion and PERC(-) would predict a post-test probability of VTE(+) or death below 2.0%. METHODS: We enrolled outpatients with suspected PE in 13 emergency departments. Clinicians completed a 72-field, web-based data form at the time of test order. Low suspicion required a gestalt pretest probability estimate of <15%. The main outcome was the composite of image-proven VTE(+) or death from any cause within 45 days. RESULTS: We enrolled 8138 patients, 85% of whom had a chief complaint of either dyspnea or chest pain. Clinicians reported a low suspicion for PE, together with PERC(-), in 1666 patients (20%). At initial testing and within 45 days, 561 patients (6.9%, 95% confidence interval 6.5-7.6) were VTE(+), and 56 others died. Among the low suspicion and PERC(-) patients, 15 were VTE(+) and one other patient died, yielding a false-negative rate of 16/1666 (1.0%, 0.6-1.6%). As a diagnostic test, low suspicion and PERC(-) had a sensitivity of 97.4% (95.8-98.5%) and a specificity of 21.9% (21.0-22.9%). CONCLUSIONS: The combination of gestalt estimate of low suspicion for PE and PERC(-) reduces the probability of VTE to below 2% in about 20% of outpatients with suspected PE.
- Digivote
- 6.9 % pre test probability of PE. Kline study multicentre 8138 patients gestalt plus PERC
- 3500
- Digivote slide
- No guarantee that failure to diagnose = harm No guarantee that diagnosis = success.
- Sequence has to be completefor harm to occur IN appendicitis. Miss diagnosis, not come back early, not get recognised, not get to theatre, not respond to treatment, not survive.
- Sequence has to be complete for harm to occur IN appendicitis. Miss diagnosis, not come back early, not get recognised, not get to theatre, not respond to treatment, not survive.