Niels Peek Professor of Health Informatics The University of Manchester Clinical prediction methods CAVEAT Why do we need prognostic models Prevention is more effective than cure ischemia ID: 1012289
Download Presentation The PPT/PDF document "Why we need causal prognostic reasoning ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Why we need causal prognostic reasoning in medicineNiels PeekProfessor of Health InformaticsThe University of Manchester
2. Clinical prediction methodsCAVEAT
3. Why do we need prognostic models? Prevention is more effective than cureischemiaanginanecrosistime
4. Take home messageCAVEAT Existing prognostic models fail to properly support clinical decision making, because they do not support causal reasoning
5. Patient Joe
6. GP Brian Patient Joe
7. We need to treat 20 people with statins, on average, to save one lifeFor every person they do help (blue), the ten highest grossing drugs in the US fail to improve the conditions of between 3 and 24 people (red). Taken from:Schork NJ. Nature 2015;520:609-611.Skeletal formula of rosuvastatin
8. Another example: Preventing emergency hospital admissionsCAVEAT Introduction of a predictive risk stratification model (PRISM; AUC=0.749) into primary care resulted in more emergency hospital admissions and use of other NHS services without evidence of benefits to patients or the NHS.Snooks H, et al. BMJ Qual Saf 2019;28:697–705.What GPs need is not a tool that tells them which patients are at high risk, but which patients are likely to benefit from preventive treatment – which is causal prediction.
9. Third (and final) example: Covid-19CAVEAT Many prognostic models are currently being developed for Covid-19 patients.Most models focus on expected duration of ventilation and mortality risk.Wynants L et al. BMJ2020;369:m1328low mortality riskshort length of ICU staylow mortality risklong length of ICU stayhigh mortality risklong length of ICU stayhigh mortality riskshort length of ICU stay
10. ConclusionCAVEAT Clinicians need tools that can tell them whether a course of action is likely to benefit their patient. This requires causal reasoning about the consequences of actions.Existing prognostic tools focus exclusively on risk, and therefore fail to properly support clinical decision making.
11. Thank youNiels PeekCentre for Health InformaticsThe University of Manchester, UK niels.peek@manchester.ac.uk @NielsPeek