RISK STRATIFICATION PROJECT UPDATE 6 MARCH 2019 GARETH HUDSON – BUSINESS INTELLIGENCE ANALYST
RISK STRATIFICATION What is Risk Stratification? •
A predictive tool based on pre-determined factors relating to historical patient care and patient demographics
Who uses it? •
United Kingdom - NHS Primary Care Trusts;
•
New Zealand - Compass, ProCare
RISK STRATIFICATION CONTEXT Who should be on the Risk Stratification team? • Economist • Analysts • Clinicians How successful has it been? • United Kingdom - NHS England • New Zealand – ProCare Rheumatic Fever case study
RISK STRATIFICATION CONTEXT • Problem: • Limited awareness of Unplanned Acute Admissions to BOPDHB Hospitals by WBoPPHO patients • Perpetual growth in attendances at Bay of Plenty Emergency Departments
RISK STRATIFICATION CONTEXT • Project Objective: • Construction of a ‘Risk Formula’ to gauge likelihood of patient readmission in Bay of Plenty • Assignment of a Risk score (0-100) to all NHIs within the WBoP PHO register
• Benefits: • Planners have a ‘macro-level’ population view of localities of interest; ‘high-cost’ patient groups or impactible conditions for preventive intervention programmes • PHOs have ‘micro-level’ population views (by Practice sub-populations) of high risk individuals or gaps that may signal sub-optimal care
• Opportunities: • Interactive analysis to better understand patient needs and prompt targeted service delivery • Supplementary information to support patient medical profiles, patient care plans, intervention programmes
RISK STRATIFICATION METHODOLOGY • Process: • Non-identifiable PHO register data submitted to Sapere with 2 years (to 31 Dec 2016) of DHB admissions metrics • Considered Patient demographics; ED/Ward attendance history; diagnosis of diabetes/CVD/COPD; Number of contacts with other DHB services (e.g. District Nursing, Alcohol and Drug, Community Mental Health) • Identify first ED/Admission dates Post census (i.e. > 1 Jan 2017) • Apply results to a regression model to determine weightings of input variables
• Deliverables: • Series of weighted factors that impact probability of a patient’s admission • A risk score between 0 and 100 indicating probability of acute hospital admission in next 12 months • Enhanced understanding in General Practice of the likelihood of acute admission of ‘at risk’ patients
RISK STRATIFICATION METHODOLOGY • Project Outcomes: • A risk register of ‘at risk’ enrolled patients from highest score to lowest in each practice • Patient level data includes demographics, deprivation flags, condition flags, days to last consult, # of GP consults (24 mths), ED Referral Source, # ED attendances (24 mths), Outpatient consults (12 mths)
• Six-monthly refresh of risk scores
RISK STRATIFICATION APPLICATION • Predictive Accuracy reduces for patients lower down the Risk Register • Patients lower than top 5% may be a high relative risk in specific circumstances (e.g. Maori Male 45-64 with a chronic condition flag)
RISK STRATIFICATION APPLICATION • Patients are stratified into Risk Strata: • Very High Relative Risk (0.5%) • High Relative Risk (0.51%-5%) • Moderate Relative Risk (5.1%-20%) • Low Relative Risk (20.1%-31%)
• Preventive interventions are offered to patients in certain strata of risk • Targeted interventions have a higher cost/benefit return
SUMMARY PROFILING – 2017 ADMISSIONS
RISK STRATIFICATION NEXT STEPS • Considerations: • Practice Demography and Patient Locality for service decision making • Monitoring and management of Clinical metrics for at risk patients • Clinical and DHB priorities for illustrating improved health outcomes • Priorities for Phase 2 of Risk implementation? • Frequent Fliers (e.g. Addiction services/Adult Mental Health Services) • Polypharmacy • COPD Patients / Smokers
REFERENCES • Kings Fund & Billings, J et. Al 2006; “PARR - Combined Predictive Model Final Report” • Billings J et. Al 2012; “Development of a predictive model to identify inpatients at risk of re-admission within 30 days of discharge (PARR-30)”; BMJ • Lewis, Dr G 2015; “Next Steps for Risk Stratification in the NHS”