Advanced Care Management Task Force Outcomes Research Study: Scope, Methodology, Results

Transcription

1 Advanced Care Management Task Force Outcomes Research Study: Scope, Methodology, Results Presented by Cheri Lattimer, RN, EVP Health Integrated Michael Terpening, VP Systems Analysis, Health Integrated Joan O Connell, Ph.D, Health Care Consultant

2 NASCHIP Challenges Historical Themes: Funding Adequacy and Stability Cost Containment Efforts Access Issues Affordability Issues Sustainability Issues Congruent Themes: Fragmentation of care Growing cost of chronic care Rising costs of drugs Consumer and provider pressures Regulatory / government pressures Premium increases, medical loss ratios and health plan assessments

3 Population Profile Comparison to % of Total Cost Well 75% High Risk Insurance Pool vs. Commercial Commercial Population Profile 56% are well/worried well 20% 30% %of Total Cost Other Acute 43% High Cost Cond 45% CHIP Profile Non-Chronic Acute 30% 50% % of Total Costs % High Cost Conditions 3% Chronic Dx/ Acute Care 22% 15% have Sx 27% have chronic disease 50% Chronic Dx/ Acute Care 12% 20%

4 Leading a new direction Advanced Care Management Model Problem Identification, Education and Logistical Support Care Advocacy, Empowerment and Health DM Integration UM CM Reassessing and Adjusting Care Assessing and Planning Care Optimum Health Care Facilitation and Preemptive Interventions

5 Advanced Care Management Task Force (ACMTF) Cooperative Outcomes Research Project* Rationale For A High Risk Pool Study Goals for the Claims Study Benchmarking Comparative Analysis Predictive Model Concept and Outcomes Task Force Commitments Provide source claims data in consistent format Invest time and resources to the process Collegiality in sharing results *CO, KS, OK, WA, AR

6 Study Objectives Compare CHIP populations to commercial Benchmark = 2.5 Million life MEDSTAT claim set Compare risk profiles, illness burdens and prevalence, costs, co-morbidity Compare CHIP populations to each other Explore implications of results for member identification and care management. Explore DxCG prospective (predictive) models Does the DxCG models executed on this sicker population predict future claims costs significantly better than an age/sex and prior year claims model.

7 What is DxCG Predictive Model licensing company in Boston, MA. Diagnostic Grouper Groups ICD-9 into condition sets ( groups) Assigns Risk Score based on conditions and comorbidity Predicts future spending in Chronic populations Explains high yearly spending in Acute Episodes

8 Study Data Specifics 5 State Comprehensive Health Insurance Pools (Arkansas, Kansas, Oklahoma, Colorado, Washington) 2 years of Medical and Pharmacy claims October 2000 to September 2001 Study Year 1 October 2001 to September 2002 Study Year 2 Over 95% claims closure in final month Members eligible at anytime in 2 year period Approximately 15,000 members WA. Medicare members were excluded

9 Benefit Plans Plan Option ARK COL KAN OKL WAS 300 Ded X 500 Ded X X X X 750 Ded X 1000 Ded X X X X 1500 Ded X X 2000 Ded X X 2500 Ded X 3000 Ded X 5000 Ded X X X X 7500 Ded X X Ded X Pre-ex Waiver Yes Yes Yes Yes Pre-ex Period 6 mo 6 mo. 3 mo. 12 mo 6 mo

10 Benefit Trend Towards High Deductibles Oklahoma 12% year 2 increase enrollment in $2000 Deductible or higher Colorado 11% year 2 increase in $2000 Deductible or higher Washington 6% year 2 increase in $1000 Deductible Arkansas with its higher average deductibles remained constant (approx 70% in the $1000 Deductible)

11 Eligibility Trends Eligibility by Study Quarters Members ARK COL KAN OKL WAS qtr1 qtr2 qtr3 qtr4 qtr5 qtr6 qtr7 qtr8 Study Period: 10/1/00-9/30/02

12 New Members and Plan Contributions 1st Quarter of Member's Eligibility Allowable + Pre-Ex Patient Paid State Deductible+Coinsurance+Paid Deductible+ Coinsurance Pre-Existing Condition Denial Plan Paid ARK $ 8,054,619 19% 4% 77% COL $ 11,091,777 23% 2% 75% KAN $ 4,718,321 25% 6% 69% OKL $ 7,513,894 21% 2% 77% WAS $ 5,376,741 11% 9% 80% 4th Quarter of Member's Eligibility ARK $ 4,931,928 16% 0% 84% COL $ 3,345,577 22% 0% 78% KAN $ 2,273,567 22% 0% 78% OKL $ 4,534,881 19% 1% 80% WAS $ 3,567,924 9% 0% 91% * Percent of allowable PLUS Pre-Existing

13 Utilization Totals Study Year # 1 Total Average Admit per Days per State Members Members ARK COL KAN OKL WAS Totals Study Year #2 Total Average Admit per Days per State Members Members ARK COL KAN OKL WAS Totals

14 Utilization Mix Study Year #1 Percent Percent Percent State Inpatient Outpatient Pharmacy PMPM ARK 38% 36% 19% $ 518 COL 31% 35% 26% $ 576 KAN 39% 40% 14% $ 614 OKL 36% 32% 26% $ 781 WAS 26% 35% 30% $ 916 Totals 34% 35% 23% $ 657 Study Year #2 Percent Percent Percent State Inpatient Outpatient Pharmacy PMPM ARK 42% 35% 18% $ 404 COL 37% 31% 25% $ 651 KAN 39% 40% 14% $ 724 OKL 31% 33% 29% $ 844 WAS 29% 31% 33% $ 1,254 Totals 35% 34% 25% $ 694 *Statistics are percent of Allowed Paid+Deductible+Coinsurance *Percents exclude Home Health, DME etc

15 State Pools Differ from Commercial Population Condition Category MEDSTAT 5 State Difference Populations 10,000 10, % Malignant Neoplasm % Diabetes 267 1, % Liver % Hematological % Mental 600 1, % Neurological % Heart 1,019 2, % Vascular % Lung 1,089 1, % Transplants, Openings, Other % Numbers shown are prevalence per 10,000 members MEDSTAT reflects 2.5 Million commercial lives

16 Diagnostic Condition Group Member Illness Profiles DxCG Company Diagnostic Condition Groups. Condition Category Members % Comorbidity * Cost per Member Heart % $ 10,846 Lung % $ 16,692 Diabetes % $ 9,152 Hematological % $ 27,556 Vascular % $ 27,466 % Comorbidity over the 30 Condition Category Sets Above conditions in top 10 Prevalence, Cost, Comorbidity

17 Number of Members Co-Morbid in Vascular (out of 636 matched) : Nutritional 2:Gastrointestinal 3:Musculoskeletal 4:Hem atologic 5:Cardio-resp. 6: Lung 7:Ears 8:Urinary 9: Skin * 9 of the 30 possible Co-morbid matches are shown

18 Comparison of CHIP to Commercial Population Member Distribution by Spending Level Percentage Commercial State Pools Lowest Level 2 Level 3 Level 4 Highest

19 CHIP Members Compare to Top 7% of Spenders in Benchmark Average Prospective Risk Scores CHIP members: 2.68 Benchmark: All members: 1.00 Top 7% of Spenders: 2.81

20 Compare Top 7% Co-Morbidity Example: Lung Top 7% total 34,370 State Pools: 1754 State Pools Average Comorbidity over 30 Conditions 25% compared to 23% Condition Group MEDSTAT Top 7% State Pools Screening/History 48% 55% Musculoskeletal and Connective Tis. 55% 49% Heart 42% 48% Nutritional 30% 42% Genital 26% 21% Diabetes 15% 18%

21 Diagnostic Condition Group Methodology Some categories are for chronic conditions which will occur during the next period Examples: heart conditions, diabetes Risk of spending/utilization high in YR 1 & YR 2 Some categories are for acute conditions which are not expected to occur again Examples: pneumonia, accidents Risk of spending/utilization high in YR1, not YR2

22 Diagnostic Condition Group Methodology Based on claims and membership data, each member is assigned: 1. Age/gender risk score 2. Concurrent risk score based on condition categories which explain current spending and utilization (Acute and Chronic Conditions) 3. Prospective risk score based on condition categories which predict future spending and utilization (Future Chronic Conditions)

23 DCG Risk Scores Year 1 Data Age/Gender Concurrent Prospective Arkansas Colorado Kansas Oklahoma Washington Average Benchmark

24 Predicting Health Care Spending Specific Study Goals: 1. Assess ability of Diagnostic Groups and Risk Scores to improve predictions of health care spending during following year 2. Use Diagnostic Groups and Risk Scores to conduct more rigorous and reliable ROI evaluations

25 Predicting Health Care Spending Method: Regression analysis of PMPM YR 2 Characteristics: 1. Age, gender 2. Year 1 (current year) spending (include by type of service and by quarter) 3. Risk Score, Condition Categories Evaluate prediction: R-square (percent of spending variation explained)

26 Predicting Health Care Spending Regression Results: R-square Without With Diagnostic Model Diag. info Information Eligibility > 1 month 30.1% 32.5% Eligibility > 3 months 35.0% 36.9% Sample sizes: 8174 for > 1 month, 6916 for > 3 months.

27 Overall Study Conclusions Prospective risk scores are only 2 3% more accurate in predicting future costs in this group, over age/sex and prior claims, due to the constancy/volume of claim data in the population Diagnostic Groups provide a viable means to identify and stratify high risk members in this population by condition and comorbidity

28 Managing Health Care Utilization and Spending Methods used to identify persons for care management programs: 1. Traditional utilization management: high cost or adverse event (example: hospital stay) 2. Condition identification for disease management (targeted conditions: diabetes, heart disease) 3. Risk identification with or without a targeted condition (example: Diagnostic Condition Group high cost prediction)

29 Managing Health Care: Risk identification with a targeted condition Use Condition Category to identify high risk members: Conditions: Diabetes, Lung, Heart, Liver Cancer, Hematological, Cerebro-Vascular Prospective Risk Score >= 2.75 Population identified accounts for: Percent of Members: 28.1% Percent of Total Spending: 55.2%

30 Managing Health Care: Risk identification independent of targeted condition Use Condition Categories to identify high risk members: Prospective Risk Score >= 2.75 Any other conditions AND COMORBIDITY Population identified accounts for: Percent of Members: 4.2% Average annual spending: $13,791 Opportunity: Identify through drug claims due to timely submission before high cost events occur

31 Next Steps for ACM Task Force Use Diagnostic Group and Risk Scores to identify care management candidates proactively and profile Co-morbidity in the risk pool database Investigate the use of Diagnostic and Risk modeling to identify similar members for a Matched Control Group in a Pre and Post intervention study in Colorado Study impacts of Advanced Care Management interventions using the risk pool database benchmarks

32 Contacts: For further information Q and A Contact: Michael Terpening Mterpening@healthintegrated.com