Data Mining for Cross System Collaborationcmhtampaconference.com/files/27/presentations/s71.pdfData Mining for Cross System Collaboration Claudia Zundel, M.S.W. Diane Fox, Ph.D. Nancy

3/14/2014

1

Data Mining for Cross

System Collaboration Claudia Zundel, M.S.W.

Diane Fox, Ph.D.

Nancy Johnson Nagel, Ph.D.

Colorado Department of Human Services

The Question…

• Who are the children and adolescents

currently utilizing high-cost intensive

behavioral health services that could

benefit from a System of Care approach?

3/14/2014

2

The Problem….

• How do we understand the array of

intensive services provided to children and

youth when

– Intensive services (residential treatment) are

provided through a variety of agencies and

funding sources

– Data systems are siloed and difficult to

integrate

Finding the Answers… • Step 1:

– How much is being spent by child serving agencies

on residential treatment and hospitalization?

• Step 2:

– How many unique children are served with these

intensive services?

– Are costs equally distributed to the children served?

– What happens to costs when children are involved

with multiple systems?

• Step 3:

– How many children are receiving services from

multiple agencies over the course of their lives?

3/14/2014

3

Step 1: Method for Compiling

Costs of Residential Treatment

• Contacted all agencies that pay for

residential and inpatient services – Child Welfare

– Office of Behavioral Health

• Non-Medicaid eligible

– Division of Youth Corrections

– Medicaid

• Asked for their total expenditures for a

single FY and the number of clients served

Residential and Inpatient

Expenditures for FY2010-11

Funding Agency Number of Children

Agency Expenditure

Additional Medicaid

Contribution Total

Child Welfare 2063 $51,719,376 $5,922,691 $57,642,068

Medicaid Inpatient 1287 $13,938,398 $13,938,398

DYC 577 $12,960,211 $1,495,839 $14,456,050

Medicaid Residential 462 $3,400,666 $3,400,666

Office of Behavioral Health (non-medicaid) 31 $656,148 $147,845.69 $803,993

Total 4420 $82,674,801 $7,566,376 $90,241,177

3/14/2014

4

Step 2: Determining How Many Unique

Children Were Served In the FY

• Went back and asked the same agencies

to provide lists of clients that they served

to comprise these costs

Barriers:

Confidentiality of data!

Success: Able to put in place business associates

agreements to obtain all the necessary data!

Children not identified with a universal identifier

Success: Used a constructed unique identifier in

combination with Medicaid ID to merge data sets

Clients in the Top Third of BHO Medicaid

Spending Accounted for 70% of the Total

Spending

33.3

8.0

33.3

22.0

33.3

70.0

0

10

20

30

40

50

60

70

80

90

100

% of Clients % of Total Medicaid Spending

Per

cen

t

Highest 1/3

Middle 1/3

Lowest 1/3

Cost per client:

$23,398.34

Cost per client:

$7,207.73

Cost per client:

$2,611.57

*Clients were grouped into three equal groups (high, medium, and low utilizers) then their % of total

spending and cost per client were calculated

3/14/2014

5

Unique Clients Served in Residential and

Inpatient Settings in a SINGLE Fiscal Year

• Result: There were 3,888 unduplicated youth

– 488 (12.6%)children/youth were served by multiple systems

87.4

11.5 1.1

0

10

20

30

40

50

60

70

80

90

100

Per

cen

t Three Agencies

Two Agencies

Single Agency

Step 3: Determining how many children are

receiving services in multiple systems

• First looked at a single large urban county

to see the overlap between

– Child Welfare (high utlizers)

– Youth Corrections

– Mental Health Services

– Substance Abuse Services

3/14/2014

6

How this Data was Used

• Provided justification for the creation of a

Care Management Entity Pilot site in the

county.

– Currently serving high needs youth in the

community

Taking That Analyses Statewide

• The Colorado State Division of Child Welfare (DCW) provided data

for the top 20% of children in Colorado who generated the highest

expenditures in Child Welfare in FY2011-2012. The sample was

comprised of 1881 children.

• Historical data that included any case open on July 1, 2006 or later

were obtained from

– Division of Youth Corrections and

– Office of Behavioral Health (mental health and substance data).

• These data were then merged with the data from Child Welfare to

determine the overlap between child welfare, juvenile justice,

substance abuse, and mental health services for these 1881

children.

3/14/2014

7

System Overlap for youth

with high CW costs

CW/MH/DYC, n=854, 45%

of CW has DYC & MH

CW/DYC, n=28, 1.5% of

CW has DYC only

CW/SA, n=6, 0.3% of CW has

SA only

CW/MH/SA, n=100, 5.3%

of CW has MH and SA

CW/MH, n=883, 47%

of CW has MH only

CW/DYC/MH/SA, n=266, 14% of

CW has all

CW/DYC/SA, n=10, 0.5%

of CW has DYC and SA

CW Only n=113, 6.3%

Who are the youth being seen?

Age Distributions by System Involvement

3/14/2014

8

Policy

• IMPLICATIONS

– Need for System Collaboration and Coordinated

Care; System of Care approach

– Streamline systems for efficiency and seamless care

• IMPACTS

– Establishing a Care Management Entity in One

County

– Mental Health Staff housed at Medicaid

– Medicaid Recommendations

Policy

• QUESTIONS!

– Who are the youth served in the various systems and

when is the best time to intervene?

– Is there a common path or trajectory through the

systems ?

– How do changes in one system affect other systems?

3/14/2014

9

Discuss

• What system of care questions might be

addressed with cross system data?

• What systems would be involved?

• What barriers might exist to obtaining

these data?

• Any ideas that you would like to discuss

with the group?

Great Sand Dunes National Park, Colorado