Transcript Figure 1
Harnessing the Power of Predictive Modeling Future Trends Harnessing the Power of Predictive Modeling Future Trends • Traditional Applications • Recent Applications • Future Trends – Motivation Index – Forecasting Disease Specific Risk – Provider Market • Forecasting Preventable Events Predictive Models Traditional Applications • Risk Stratify the Population for care management – Manage complexly ill members (Inpatient avoidance) – Refine disease management strategies – Manage pharmacy services • Underwrite more accurately • Reimburse based on illness burden • Evaluate physician management strategies Predictive Models Changing Focus • Traditional Application has been to Identify: – High Risk / High Cost members – Inpatient Risk • Recent Applications – Forecasting additional Cost Components • ER Visit Risk • Pharmacy Cost forecasting – Identify Intervenable or Actionable members • Future Trends – Member Motivation – Disease Specific Complications – Preventable Events for the Provider Market Recent Application Identifying Actionable Members • Method A – Query population by multiple filters: • • • • • Disease Cost Risk Inpatient Risk Pharmacy Risk Mover Risk • Method B – Impact Index : Model that identifies members who have the greatest potential for outcome improvement based on guideline compliance Recent Application Multiple filters to identify actionable members Total Population: 216,842 members High Risk Index High Risk + Mover Top 2% 4,362 Members Forecasted Cost: $25,741 Prior Year Cost: $45,006 498 Members Forecasted Cost: $20,084 Prior Year Cost: $8,832 Savings Potential: Savings Potential: -$84,033,930 $5,603,496 Recent Application Impact Index to identify actionable members Total Population: 925,407 Diabetes: 50,847 High-Risk Index High Impact Index Risk Level 4&5 Top 15% 14,250 Members Forecasted Cost: $14,634 Prior Year Cost: $14,527 13,872 Members Forecasted Cost: $8,698 Prior Year Cost: $5,089 Savings Potential: Savings Potential: $1,524,750 $50,064,048 Recent Application Impact Index • These Drivers – – – – Disease Age/Sex Comorbidities Guideline Compliance Patterns • Determine future impactability • Determine potential cost savings Guideline Potential Savings DM ACERx $455 DM HBA1C $449 DM Eye Exam $447 DM LDL $442 Depress RX $360 CHF Rx $302 CVA Coum $290 CHF HTN Ace $280 Asthma Rx $206 Asthma Steroid $186 COPD TheoLvl $167 MI BBlocker $102 CHF InPt-Echo -$14 Future Trends Motivation Index • Identify members – more motivated to ‘self-manage’ – comply with instructions from providers – pursue ways to improve health status • To create index use data sources – – – – Lifestyle Data Health Risk Assessment Demographics Claims Data Future Trends Motivation Index Drivers • Lifestyle Data – – – – – Net Worth Credit History Magazine Subscriptions Hobbies Clubs • Claims Data – Compliance Patterns – Preventive Care – Physician Visit Patterns Claritas US Census Bureau Media Mark Future Trends Motivation Index Variables • Claims Data – Compliance Patterns • To Guidelines • To Psych-Related Drugs • To Maintenance Drugs – Preventive Care • Use of preventive health services • Compliance to Preventive Lab Test • Compliance to standard preventive guidelines – Physician Visit Patterns • Gap/Frequency between Acute Care & Physician visits • Gap/Frequency between Physician visits per disease – Cost Ratios for Inpt / Rehab / Rx / Physician • Demographic / Misc – – – – Age/Sex Obesity / Smoking Drug/Alcohol Dependency Mental Health Future Trends Motivation Index Drivers • Patients with higher motivation scores have – – – – – – Better guideline compliance Older age Higher preventive care use Lower acute-care use Shorter (Time-frames from Inpt discharge to phys-visit) Females 40 t0 65 • Higher mammogram compliance – Asthmatics • Lower ER visits – Hypertension • Higher hypertensive drug use – Depression • Higher depression related drug use Future Trends Motivation Index Drivers Higher Antidepressant Use within Depressed Population correlates with higher motivation 48.0 Avg Motivation Index 45.0 42.0 39.0 36.0 33.0 30.0 1 2 3 4 5 6 7 8 9 10 # Antidepress Rx Filled 11 12 13 14 15 16 Future Trends Motivation Index Drivers With Mammogram Without Mammogram 60 Avg Motivation Index 55 50 45 40 35 30 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Female, Age Mammogram Use in Female Population correlates with higher motivation 61 62 63 64 Future Trends Motivation Index Drivers ER Visits within Asthma Population correlates with lower motivation 40.0 38.0 Average Motivation Index 36.0 34.0 32.0 30.0 28.0 26.0 24.0 22.0 20.0 1 2 3 4 5 6 7 8 9 10 # ER Visits Per Year 11 12 13 14 15 Future Trends Forecasting Disease Specific Outcomes Disease Diabetes Cancer Asthma Complications Drivers Macrovascular Microvascular Events Retinopathy Metabolic Complications Age Infectious Complications Comorbid Condition Cnt Mammography RxFills Inpatient Admit Cancer Severity after Cancer Resp Severity Respiratory Failure Pulmonary Edema Ventilator Pneumonia Pleural Effusion COPD Pneumothorax Rx Cnt Inpt Admit Musculoskeletal $ Positive Complicati Predictive on Rate Value 23% 60% 11% 59% 7% 37% Future Trends Provider Market • Future Trend – Forecasting Preventable Events Pre-discharge • Preventable readmissions • Catheter-Associated UTI • Pressure Ulcers • Vascular Catheter-Associated Infection • Mediastinitis after CABG-Surgical Site Infection • Hospital-Acquired Injuries Future Trends Why Identify Potentially Preventable Readmissions? • • • • Comparing provider performance to enhance quality Developing pay for performance systems Readmission rates provide quality benchmark Costs associated with readmissions are substantial – 30 billion in play for Medicare • Defining Preventable Readmissions – some initial discharges for which subsequent readmits excluded (e.g. LAMA, cystic fibrosis) – Readmissions are for same diagnosis – Readmissions are for related diagnosis Future Trends Data for Forecasting Preventable Events • Electronic Medical Record Data – HL7 Format – Near-real time data outflow • Forecasting Model – Near-real time forecast of Readmission / Decubitus • Probability • Risk Index • Drivers • Data Needed – – – – – – – Vital Signs Lab Results Drug Dosage and Timing Admission Discharge Transfer Data Chief Complaint Prior Discharge Diagnoses Supplies Future Trends Forecasting Decubitus • Drivers to forecasting Risk of Decubitus when a patient is admitted – Vital Signs • Fever • Pulse / BP / Respirations – Lab Results • White Blood Cell Counts • Blood Culture Results – Drug Dosage and Timing • Antibiotic at admission – Chief Complaint – Diarrhea – Admission Source • From SNF – Prior Discharge Diagnoses • Diabetes / CHF / Senility – Demographics – Supplies • Depends Future Trends Hospital Revenue Loss with Preventable Events Worst Case Revenue at Risk by Condition Condition Decubitus Ulcer Falls and Trauma Urinary Track Infection Object Left in Surgery Mediastinitis Air Embolism Blood Incompatibility Adjustment for Multiple Conditions Present Total of Approved Conditions Discharges with Condition Present Discharges with Change in DRG Assignment (Worst Case) 259,356 201,007 8,832 805 111 46 35 117,852 42,943 1,063 174 31 25 5 45% -283,432,250 21% -128,547,128 12% -1,469,338 22% -454,693 28% -252,677 54% -109,681 14% -5,180 5,959 -20,030,826 168,052 36% -434,301,773 -$5,867 464,325 Source:CMS;Advisory Board Analysis Percent of Discharges Total Revenues at Revenue Loss Risk per HA event (Worst Case) -2,405 -2,993 -1,382 -2,613 -8,151 -4,387 -1,036 -22,968 Future Trends ROI from Forecasting Preventable Events Cost Saved by Preventing Hospital DRG Revenue Loss Acquired Event Per Hospital For Avg For Avg HospSystem Per Hospital HospSystem Acquired Preventable Event Acquired Event Annually Event Annually Decubitus Ulcer -$2,405 -$3,366,978 $17,000 $23,800,000 Urinary Track Infection -$1,382 -$276,451 $37,000 $7,400,000 Methicillin Resistant Staff Aureas NA NA $30,000 $6,000,000 Predictive Modeling Applications for Care Management – Paradigm Changes Historical Mbr Educ Demand Concurrent Disease Mgmt Case Mgmt Mgmt UR/UM Current Transformed Predictive Modeling: Used to identify early members who are trending toward high-risk events Personal Health Mgmt Population Risk Mgmt Decision Support UR/ UM Proactive Case Management Disease Management UR/ UM Proactive Case Management Disease Management Future Model Personal Health Mgmt Population Risk Mgmt Actionability Motivation Index Disease Specific Decision Support About the Future “Never let yesterday use up too much of today.” - Will Rogers The best way to predict the future is to create it.