Crossing the Last Mile with Predictable Hospital Readmissions
summary
Predictive model scores on a 20,000-patient cohort give a baseline 13% 30‑day readmission rate; the top 10% by score hits ~66% and targeting the top 70% can save about $6.3M assuming $1K intervention cost, $20K avoided loss and 30% intervention efficacy. The delivery is a production app: SQL Server 2008 R2, Analysis Services multidimensional, SSIS for ETL and persisted scores, and an HTML5/C# web service front end that lists high‑risk patients, tracks interventions and is live after a 2.5‑month build.
AI-written summary · about
chapters (13)
- 0:00Predictive analytics in healthcare
- 1:27Why healthcare needs it now
- 5:02Healthcare spending pressures
- 8:51Capitation and penalties
- 12:22Targeting high-risk patients
- 18:01Last mile to real impact
- 19:25Risk modeling and validation
- 27:55Clinical-IT collaboration
- 28:58Deploying in hospital
- 37:32Patient-level interventions
- 41:22Building solution fast
- 49:21Tech architecture
- 53:10Summary & Q&A
related talks
10mTransforming Population Health with Predictive BI
23mTransferring clinical prediction models across hospitals...
23mEngineering AI-Powered Predictive Healthcare Systems
32mSHAP and Beyond
22mBuilding Predictive Analytics Data Warehouses in Healthcare
38mRealtime prediction of C-section risk for laboring mothers
20mML applications in healthcare - Saudi Arabian hospital case study
29mBart Steverink & Laurens Schinkelshoek