A Value-Based Approach to Clinical Pathology and Informatics
Download as PPTX, PDF1 like201 views
The document discusses a value-based approach to clinical pathology and informatics, emphasizing the need for effective clinical decision support systems (CDSS) to improve patient outcomes and reduce costs. It highlights specific case studies and research findings that demonstrate the impact of context-sensitive interventions and pathologist advice on clinical practice. Additionally, it addresses the challenges of integrating CDSS into workflow and how to enhance the relevance of information delivered to healthcare providers.
![For our patients and our population
Safety of health ICT
• “The dangerous decade” [E. Coiera et al J Am Med
Inform Assoc 2012;19:2]
– “The demands for health system reform are
now so compelling that there appears to be no
choice but to implement complex ICT on a
large, often national, scale”
“If healthcare wants the benefits of ICT then it
must actively manage the risks”](https://image.slidesharecdn.com/edwardscirdanaug17-170907134746/85/A-Value-Based-Approach-to-Clinical-Pathology-and-Informatics-21-320.jpg)
A Value-Based Approach to Clinical Pathology and Informatics
- 1. For our patients and our population A value-based approach to clinical pathology and informatics Dr Glenn Edwards glenn.edwards@sa.gov.au Disclosure Former shareholder, CEO, Medical Director of Pacific Knowledge Systems
- 2. For our patients and our population Version 5
- 3. For our patients and our population Uluru ’93
- 4. For our patients and our population Version 5
- 5. For our patients and our population What happened to “Decision Support?”
- 6. For our patients and our population • Key issues – Most evidence for process outcomes – Remaining challenges • Demonstrate impact on outcomes, cost, users • Means to augment uptake and effectiveness • Integration into workflow • Deployment across diverse settings • Transformation role • “Broad penetration of CDSS will require aggressively seeking a better understanding of what the right information is and when and how it should be delivered to the right person..” Impact of CDSS: 2012 systematic review (Bright et al Ann Int Med 2012;157(1):29)
- 7. For our patients and our population BNP use /1000 patients / PCT Still extremely low use in many areas: • Excess costs • Poor patient experience • Failure to adopt innovation Map from Atlas of Variation
- 8. For our patients and our population Runciman et al MJA 197 (2) · 16 July 2012
- 9. For our patients and our population 9 How would you deal with these results? 39 year old female Cholesterol 5.1 mmol/L Triglyceride 3.5 mmol/L * HDL cholesterol 0.9 mmol/L * LDL cholesterol 2.6 mmol/L
- 10. For our patients and our population Results in context
- 11. For our patients and our population Context-specific opinion “Dyslipidaemic pattern. Note previous results indicating poorly controlled diabetes mellitus, which may account for the lipid disorder. Suggest review glycaemic control (HbA1c to follow) and check urine ACR, which is now overdue. Monitor lipid response to intensified management. Note current statin therapy may be insufficient.”
- 12. For our patients and our population
- 13. For our patients and our population Context-specific interventions • Test ordering interventions – Test additions or deletions • Automatic, • Discretionary – Prompt other interventions – Suggest other links, information • Request error detection • Report opinions • Billing compliance • Report delivery
- 14. For our patients and our population Familial HypercholesterolaemiaMaternal grandmother -South African -died at age 50 Aunt -died at age 50 (heart attack) Aunt -died at age 60 (heart attack) high cholesterol Uncle -died at age 50 (heart attack) -died at age 50 (heart attack) -had a bypass -by age 38 2x bypasses 2x heart attacks -died age 40 2x bypasses Heart attack -by age 48 4x bypasses -age 26 High cholesterol -age 28 High cholesterol -by age 46 High cholesterol 3x bypasses Ms. D (38) High cholesterol (9.2 mmol/L) High cholesterol DNA testing at PathWest, RPH, mutation detected
- 15. For our patients and our population Impact of Pathologists’ advice on LDL cholesterol levels Bell DA et al Clin Chim Acta 2013;422:21-25 Interpretative comment Control Significance Number of individuals 96 100 Repeat LDL-cholesterol Number (%) 63 (71%) 70 (70%) NS Mean reduction in LDL- cholesterol (mmol/L) 3.0 2.3 p<0.005 Specialist referral (whole group) 4 (4%) 1 (1%) p=0.20 Specifically suggesting referral in interpretative comment. 3 26 individuals (11.5%) 1 (1%) p<0.05
- 16. For our patients and our population Impact of context-sensitive interventions • Context-specific intervention to improve specialist referral for at-risk patients • Significant benefit – Controls 8/96 (8%) vs Cases 24/135 (18%) were referred following pathologist advice • First prospective case-control study to demonstrate a positive benefit of pathologist report interpretation R. Bender et al Pathology 2016;48(5):463
- 17. For our patients and our population Tools to manage context • Conventional LIS rules/middleware • Expert systems – Rules – Case-based rules • Ripple down rules • Artificial intelligence – Machine learning – Other ?
- 18. For our patients and our population Context elements • Patient, requestor, location • Previous tests • Other concurrent tests • Clinical information provided • Other clinical information • Other request information – eg billing status, eligibility – special requests eg “attention Dr Edwards” • Information from other sources – eg: EHR, pharmacy, registries
- 19. For our patients and our population Context-specific intervention: Clinical opinions (Interpretation) • Identify “question” if any • Identify other background issues – Non-compliance – Missed tests – Overlooked tests • Formulate opinion – Summarise key history – Summarise key findings – Diagnostic interpretation – Advice
- 20. For our patients and our population Context-definition
- 21. For our patients and our population Safety of health ICT • “The dangerous decade” [E. Coiera et al J Am Med Inform Assoc 2012;19:2] – “The demands for health system reform are now so compelling that there appears to be no choice but to implement complex ICT on a large, often national, scale” “If healthcare wants the benefits of ICT then it must actively manage the risks”
- 22. For our patients and our population Canned comments: Simple knowledge models IF Triglyceride is HIGH AND HDL is LOW AND LDL-C < 2.5 THEN “Common causes of dyslipidaemic pattern include diabetes, alcohol, genetic….” Rules: 1 Conditions: 3 Validation: Straightforward Value: Low
- 23. For our patients and our population
- 24. For our patients and our population Complex knowledge bases: Risk • Error never completely preventable • Can error be formally monitored/measured? • Can impact/outcomes be measured? • Strategy for risk mitigation – “Swiss cheese” model for error detection – Monitoring, audit, reporting – Continuous incremental refinement • Unknowns – Societal tolerance of • Machine error • Uncertainty – Participants – Relationships: “knowledge engineer”, domain experts, quality systems, risk managers, clinicians, consumers, payers..
- 25. For our patients and our population Free text analysis in clinical decision support systems
- 26. For our patients and our population Free text analysis in CDS D. Sittig et al J Biomed Inform 2008;41:387 • Free text (#9 of “10 grand challenges”) • > 50% of key information resides in the free text portions of the EHR • We need methods for accessing and reasoning with free text => enable more specific CDS interventions – highly tailored alerts and reminders, – even condition-specific or patient specific order sets
- 27. For our patients and our population Natural Language Processing • Named Entity Recogniser (NER) – eg: Mayo system (cTAKES) J Am Med Inform Assoc 2010;17:507) – Open source – http://ctakes.apache.org/ • Issues – Conflicts – Training sets – Informality of language – Situated context
- 28. For our patients and our population Version 5
- 29. For our patients and our population UK standards for authorisation and reporting • Comment on all reports: 5% • 42% no policy • 31% consider highlighting “abnormals” to constitute an interpretation of the result Prinsloo P. & Gray T. Ann Clin Biochem 2003;40:149-55
- 30. For our patients and our population CLN August 2014
- 31. For our patients and our population Factory Logistics Pathology Experts Patients Clinicians Clinical setting Healthcare decisions Information Clinical Pathology: Low value model
- 32. For our patients and our population Factory Logistics Pathology Experts Patients Clinicians Clinical setting Healthcare decisions Clinical Pathology: High value model Decision Support Knowledge Bases Clinical Governance
- 33. For our patients and our population Eugenio H. Zabaleta, Ph.D. MedCentral Health System, OH
- 34. For our patients and our population
- 35. For our patients and our population
- 36. For our patients and our population The Challenge for Clinical Pathology • We need to demonstrate, and articulate, the value of pathology – Clinical – Financial • What do stakeholders want? – Doctors, Patients, Community – Payers • Where is the value opportunity? – Value-added clinical pathology interventions – Active management of test utilisation – Clinical decision support – Advocacy, education, policy • The value of the clinical role of pathologists
- 37. For our patients and our population A value-based approach to clinical pathology and informatics Dr Glenn Edwards glenn.edwards@sa.gov.au Disclosure Former shareholder, CEO, Medical Director of Pacific Knowledge Systems
- 38. For our patients and our population Version 5