Presentation is loading. Please wait.

Presentation is loading. Please wait.

1 iBudget Florida Stakeholders’ Meeting December 3, 2009.

Similar presentations


Presentation on theme: "1 iBudget Florida Stakeholders’ Meeting December 3, 2009."— Presentation transcript:

1 1 iBudget Florida Stakeholders’ Meeting December 3, 2009

2 2 Introduction Jim DeBeaugrine APD Director Betty Kay Clements Family Care Council Florida Chairperson

3 3 The Meeting Process Facilitator Introduction –Keep the group focused and on track –Ensure stakeholder input is heard and recognized

4 4 Introductions Workgroup members Staff

5 5 Review of Minutes from Previous Meeting

6 6 Meeting Objectives Share information Receive input on preferred options, potential improvements, implementation considerations, and other relevant issues Build consensus toward the final plan

7 7 Project Background APD Challenges to Overcome –System is complex –More consumer control possible –Managing funding is difficult –Wait list is growing

8 8 Project Background Legislative Direction: Plan by February 2010 –Fair and equitable distribution of resources based on assessment process –Client choice of services and providers –Formulas to predict resource needs –Recommended roles for providers and support coordinators during the assessment process

9 9 Overall Goals Simplify the system Enhance: –Self-direction –Sustainability –Equity

10 10 1.Empower individuals 2.Funding is fair and equitable, yet responsive to individual needs 3.Protect health and safety Detailed Objectives

11 11 4.Transparent process 5.Operate within agency budget 6.Meet federal requirements Detailed Objectives

12 12 Needs met through waiver AND state plan, natural supports and community resources Waiver Support Coordinators freed to help get other resources Service review limited for many consumers, for spending up to individual iBudget maximum amount Requests for additional funding require thorough review iBudget Paradigm

13 13 Proposed Plan Development Timeframe October: Gather stakeholder ideas November and December: Statistical analysis, policy option development January: Stakeholders review draft plan, second round of public meetings February: Plan to Legislature

14 14 Pending feedback from Legislature: –Draft to Legislature: February 1, 2010 –CMS approval sought as soon as possible –Phase-in begins Summer/Fall 2010 –QSI improvements/algorithm refinement: Through Spring/Summer 2011 –Wider phase in begins Summer/Fall 2011 Suggested High-Level Implementation Plan

15 15 Review of Action Items from Previous Meeting

16 16

17 17

18 18 Algorithm What is an algorithm? –Mathematical formula that considers data (consumer characteristics) and determines a budget amount –Captures patterns of spending for similar consumers from previous years –Used in many other contexts: Amazon.com recommendations—data on product views Credit scores—data on accounts, payment history

19 19

20 20 Algorithm An algorithm has two parts: Dependent variable –What we’re trying to predict—in this case, an individual’s funding for waiver services –We choose dependent variable Independent variables –The factors that go into calculating the dependent variable –Determined through trial & error—testing options and ideas

21 21 Algorithm (weight) Independent Variable A + (weight) Independent Variable B + (weight) Independent Variable C + (weight) Independent Variable D = Dependent Variable ***This is for illustrative purposes only and is not a proposed formula.

22 22 Algorithm What makes a good algorithm? High “r-square”—a measure that tells us how well a formula fits its data Higher r²Lower r²

23 23 Algorithm 1.0 is perfect fit to data; but unattainable due to unique circumstances Wide range of “goodness of fit” Louisiana:.46 Georgia:.75 Colorado:.26 &.51 (two waivers) Oregon:.45 Wyoming: about.80 (started at.50)

24 24 Algorithm Examples of other states’ algorithms: Subscales or sections from assessments Overall scores from assessments Individual questions from assessments Diagnosis Age Living situation Services received Chronic conditions Mental health status Community safety risk

25 25 Algorithm What is the dependent variable for our model—what funding patterns we’re trying to pattern after? Fiscal Year 2007-08 funding Prior to tier waiver implementation Are not seeking perfect correlation

26 26 Algorithm What makes a good algorithm? Fewer independent variables Valid and reliable data Useable—can collect data, run it Refined over time (ex: Wyoming)

27 27 Algorithm Statistical Procedures to Create Models Used Box-Cox Power Transformation Family—method used to select best method for ensuring assumptions required for multiple regression modeling are satisfied

28 28 Algorithm Statistical Procedures to Select Most Promising Model Used Generalized Information Criterion—especially helpful in avoiding an over-fitting model (involving too many predictor variables)

29 29 Algorithm Remember: Model development and selection is work in progress! Examples shown are for discussion purposes only and not APD proposals.

30 30 Algorithm What work has APD done so far? Considered the dependent variable—funding patterns during FY07-08—to make it appropriate to use, given policy changes since then and potential iBudget processes –Removed “outliers” –Rate changes –Deleted services –Temporarily put aside: “One-time” purchases—a separate fund Geographic differentials—will add back in Waiver Support Coordinator funding—will add back in

31 31 Algorithm What work has APD done so far? Tested 49 potential independent variables: the factors that will predict a person’s funding –Age –Living setting –QSI subscores –QSI individual questions –Community safety

32 32 Algorithm What work has APD done so far? Tested 49 potential independent variables: the factors that will predict a person’s funding –Community-based care status –Chronic disease status –Mental health treatment status

33 33 Algorithm Best fit model thus far: Age (quadratic equation) Living setting (group home, family home, or supported living) QSI subscores: behavioral, physical, functional Individual QSI questions: Hygiene (Question 20) Mental health status But more variables still to be tested and more work still to be done! For discussion purposes only.

34 34 Portion of Variance Explained by Individual Variables in Draft Model Due to correlation between variables, actual variance as part of model is slightly less For discussion purposes only.

35 35 07-08 Claims by Living Setting Percent of Consumers In Living Settings For group comprising dependent variable: 1 2 3 4 5

36 36 07-08 Claims by Functional Status Percent of Consumers by Functional Status For group comprising dependent variable:

37 37 07-08 Claims by Behavioral Status Percent of Consumers by Behavioral Status For group comprising dependent variable:

38 38 07-08 Claims by Physical Status Percent of Consumers by Physical Status For group comprising dependent variable:

39 39 07-08 Claims by Age Percent of Consumers by Age For group comprising dependent variable:

40 40 07-08 Claims by Age For group comprising dependent variable:

41 41 07-08 Claims by QSI Q20 (Hygiene) Response Percent of Consumers by QSI Q20 (Hygiene) Response For group comprising dependent variable:

42 42 Variable NameWeightsExample Level Example Value (Weight times Example Level) Intercept-71.3841 Age4.7092094.17 Age-Square-0.049620*20-19.836 Living Setting44.59289.179 QSI Behavioral Subscore5.9831 QSI Functional Subscore6.82320.46 QSI Physical Subscore2.13124.261 QSI Question 20 (Hygiene)6.6651 Mental Health Status15.1111 Total in the Square-Root Scale144.61 Predicted Support (requires further adjustment) 20,912 Predicted support will then be uniformly adjusted across the board based on appropriations and system parameters. For discussion purposes only.

43 43 Algorithm Suggestions for APD to Test: Community safety Natural supports –Caregiver age Data from independent assessments Number of children Drug/alcohol abuse Physical/mental abuse

44 44 Algorithm Review of algorithm impact: Ability to meet minimum health and safety needs Percent change from current amount

45 45 Assessment Instruments What makes a good assessment instrument? Reliable—the way it measures will give the same result when results should be the same Valid—it measures what it’s supposed to measure Predictive – assessment information and other factors yield fair and equitable budget amounts

46 46 Assessment Instruments What makes a good assessment instrument? Also need to think about: What consumers assessment is intended for Time it takes to administer Who can administer Training required for administration Nature of assessment Cost of assessment

47 47 Assessment Instruments Tentative plans for enhancing the QSI: Avenues for comment and suggestions Additional information for consumers and families Revisions to administrative guide Rescoring and rescaling Test potential new questions Larger-scale revision using new questions

48 48 Suggestions for new questions for the QSI?


Download ppt "1 iBudget Florida Stakeholders’ Meeting December 3, 2009."

Similar presentations


Ads by Google