How Carle Health Effectively Integrated Augmented Intelligence
0 likes981 views
The document outlines a webinar on leveraging AI for improving healthcare operations at Carle Health, featuring guest speaker Phillip Rowell. It details the implementation of three AI use cases: forecasting COVID-19 patient needs, analyzing sepsis mortality, and predicting patient outcomes and costs, showcasing how data-driven insights improved operational decision-making. Carle Health aims to continue enhancing its data foundation to support effective AI applications and optimize healthcare outcomes.
How Carle Health Effectively Integrated Augmented Intelligence
- 1. 1 Webinar: Leveraging AI to Accelerate Systemwide Improvements at Carle Health
- 2. 2 Agenda o Introductions o Understanding Augmented Intelligence o Examining AI Use Cases o Q & A
- 3. 3 Guest Speaker Phillip Rowell Vice President, Clinical and Business Intelligence Carle Health
- 4. 4 Our Health System Carle Health provides a broad spectrum of healthcare services to a large and predominantly rural area across all 28 counties in east-central and southern Illinois
- 5. 5 Augmented Intelligence (AI) o Continues to evolve and assists humans in making accurate, insightful decisions o Extracts the most relevant insights from millions of datasets to solve business-critical issues o Requires a strong data foundation to support effective AI and advanced analytics o Is key to achieving sustainable improvements—now and in the future
- 6. 6 AI Use Cases from the Frontlines
- 7. 7 Three AI Use Cases A roadmap addressing pitfalls to avoid and best practices 1 2 3 Forecast COVID-19 patient numbers and staffing needs Identify inflection points within data for sepsis mortality Stratify & Predict patient outcomes and associated costs
- 8. 8 Use Case 1: Forecasting Pandemic Needs o Pandemic pressures required novel solutions to support sustained operations in a challenging operating model o Needed a way to predict the number of COVID-19 patients that would need care during the pandemic o COVID-19 analytic insights allowed Carle to staff appropriately based on New York Times published infection rates integrated into hospitals admission, discharge, and transfer (ADT) data Lacked reporting capabilities to support insights into COVID-19-related admissions
- 9. 9 Use Case 1: Forecasting Approach o COVID-19 Patient and Staff Tracker Tool – patient and staff exposures o COVID-19 Forecasting Analysis – 5-day forecast for inpatient admissions and daily census o Forecasted across levels of care (e.g., non-critical care versus ICU) o Forecasts (and downstream calculations) include confidence limits that represent “best” and “worst” case scenarios o Included calculations for estimating staffing and PPE needs Deployed a suite of predictive capabilities to address pandemic needs
- 10. 10 Use Case 1: Forecasting Results o Promoted AI models to production in Q3 2020 o Forecasted bed utilization and census has matched >4X o Adopted and utilized by senior leaders Combining data and AI fueled real-world clinical and operational improvements “COVID-19 continues to raise awareness about the importance of data in operational decision making... through the thoughtful application of our AI capabilities, we were able to avoid over 1,000 hours of manual work.” Chief Medical Officer, Robert Healy, M.D.
- 11. 11 Use Case 2: Identify Inflection Points to Reduce Sepsis Mortality o Needed to understand sepsis performance at a deeper level: – Have we reduced sepsis rates overtime? – Are the sepsis rates better at one hospital compared to another? – If we set a sepsis improvement goal, is it statistically different from current performance? – What will our sepsis performance be in a year? – What will we change to see expected sepsis results? o Increasing amounts of data made it difficult to understand actual sepsis improvement opportunities Difficulty identifying sepsis improvement signals from noise
- 12. 12 Use Case 2: Identifying a Shift in Sepsis Mortality o AI model identified sepsis outcomes overtime and compared compliance by location o Expected mortality indicates expected mortality has increased approximately at the same time discharges have decreased and unadjusted mortality has increased o AI revealed that a likely reason for the increased, unadjusted mortality rate is the increased patient severity as opposed to alternatives such as care worsening Timeframe: 10/01/2019 - 08/31/2021 EXCLUSIONS: Hospice, COVID Positive, and Outside Hospital Transfers
- 13. 13 Use Case 3: Stratify and Predict Patient Outcomes and Costs 13 Risk Score Overall Disease Burden (ACGs) Age, Gender Medication Patterns Disease Markers (EDCs) Resource Use Population Markers o Needed a way to surface and prioritize high- and risking-risk patients o Used the Johns Hopkins Adjusted Clinical Groups (ACG) Model to stratify patients based on risk: – Disease burden (morbidity and resource use) ACGs – Hospital admission (6,12 mo; ICU; long-LOS (12+d) and injury–related admission) – 30-day unplanned readmission – Cost range (health care, pharmacy) Health Alliance couldn’t predict patient health status and associated cost
- 14. 14 Use Case 3: ACG Model Opportunities Analytic Efficiency Create population analytics across any patients (all ages and plans) Pharmacy Chronic condition medication adherence, polypharmacy review and reduction Care Management Cohort-based and risk-adjusted pre-post spend and utilization Care Coordination Patients engaged with primary care, reduced low value care or duplicative testing and improved medication management Cost Forecasting Actual to expected spend accuracy and intervention-driven change Proactive Care Patients with elevated blood pressure and CV risk factors or prediabetes that progress to hypertension or diabetes Health Alliance leveraged ACG to reveal AI-informed opportunities
- 15. 15 What’s Next for AI at Carle Health? o Continue investment in a strong data foundation to support effective AI and advanced analytics for future success o Identify new opportunities to apply AI and customize algorithms to our specific healthcare dynamics o Support usage and adoption across our healthcare system
- 16. 16 Thank you