Using State and Federal Data to Analyze and Model State Health Markets: Examples and Lessons Learned

1
Using State and Federal Data to Analyze and Model
State Health Markets Examples and Lessons Learned

• Scott Leitz
• Director, Health Economics Program
• Minnesota Department of Health
• November 10, 2004

2
Overview

• Some background on state and federal data sources
  for analysis and modeling
• A few examples of Minnesota modeling exercises
• Lessons learned and things to consider

3
State versus Federal data sources for analysis
and modeling

• State legislators generally believe their state
  is unique
• Not having state data can be a reason not to do
  something, therefore collection of state-specific
  information is critical
• But not every question asked by state
  policymakers can be answered with state-specific
  data
• Even when it can, the estimates can sometimes
  differ
• Example CPS versus state-specific surveys

4
State versus Federal data sources for analysis
and modeling (II)

• Even where state data may not be available, or is
  limited, national data can be used and
  adjustments made
• Assumptions are important
• National data is a good crosscheck to state data

5
Example 1

• How much uncompensated care might result from a
  proposal to eliminate a state health insurance
  program for very low income people and reduce
  income eligibility for a Medicaid population?

6
The Challenge

• Turning estimates of enrollment loss into
  hospital-specific estimates of uncompensated care
• Multiple steps involved
• How many will end up without coverage?
• How many services will this population seek?
• How will that care get paid for?
• How will behavior change?
• Need for using both state and national data to
  answer these questions

7
A brief overview of methodology

• Estimated number of people who would lose
  coverage under Governors proposal, adjusted for
  take-up (crowd out studies)
• Adjust result to account for differences in
  expenditures between the uninsured and the
  insured
• Uninsured spend approximately half of what the
  insured spend on health care. (MEPS, Hadley
  Holahan 2003, Long Marquis 1994).
• Adjustment to reflect that public program
  enrollees are sicker in general than the
  uninsured (2001 MN Health Access Survey, Holahan
  2001).
• Result estimate uninsured spend 61 of what they
  would have spent if enrolled in a public program.

8
Methodology (II)

• Resulting figure is the estimated use of services
  by the additional uninsured (uninsured costs).
• Uninsured costs can be paid for in two ways
• Out of pocket payments by the uninsured
• Uncompensated care
• Research shows that the uninsured pay around a
  third of their health care costs
• Surprisingly consistent across income levels
• (MEPS, Hadley Holahan 2003).
• Remaining is uncompensated care

9
Methodology (III)

• This uncompensated care figure is divided between
  hospital-based uncompensated care and
  clinic-based uncompensated care.
• UC allocated 34 to clinics and 66 to hospitals
  (Hadley Holahan 2003, 2000 Minnesota-specific
  analysis of uncompensated care).

10
Results Estimated Impact on the Uninsurance Rate

• Percentage of Minnesotans without health coverage
  increases by the following relative to current
  levels, assuming all other things remain
  constant
• Baseline, 2002 5.4
• 2004 6.0
• 2005 6.4
• 2006 6.5
• 2007 6.6
• Additional of approximately 63,000 additonal
  uninsured Minnesotans

11
How Do These Estimated Increases in Uncompensated
Care at Hospitals Compare to Current Levels?
88
80
63
34

12
Lessons learned

• Using state-specific data is important, but it
  likely cant answer every question
• State-specific UC baseline data, uninsured
  characteristics
• Federal/national MEPS, national studies
• Can use both credibly, as long as their
  respective roles are appropriate
• Use national data as crosscheck for
  state-specific data

13
Example 2

• How will an aging population affect use of health
  care services and hospital bed capacity over the
  next 10, 20, and 30 years?

14
Very Brief Background on Example 2

• Minnesota has operated under a hospital inpatient
  bed construction moratorium since 1984
• Bed capacity essentially static for 20 years
• Question how will population demographics affect
  use of services and how will that compare to bed
  capacity?

15
Again The need for both state and federal data

• State Demographic trends and projections,
  average length of stay
• Federal Hospitalization rates by age, average
  length of stay crosscheck

16
Projected Minnesota Population Growth,by Age
Group
Source Minnesota State Demographic Center
17
In Sheer Numbers, How Much Will Minnesotas
Elderly Population Increase?
Source Miinnesota State Demographic Center
18
How Does Use of Health Care Services Vary by
Age? Hospitals
Hospitalization Rates by Age (2000 data)
Baby boomers
Sources National Center for Health Statistics
(2000 National Hospital Discharge Survey) U.S.
Bureau of the Census
19
Projected Growth in Minnesota Hospital Utilization
Source Minnesota Department of Health, Health
Economics Program
20
Sources of Growth in Projected Minnesota
Hospital UtilizationExample Inpatient Days
Source Minnesota Department of Health, Health
Economics Program
21
Projections of Capacity Utilization (as of
total available MN hospital beds)
Baseline 15 increase 15 decrease
2000 57 57 57
2010 66 69 62
2020 77 85 69
2030 91 105 78
Source Minnesota Department of Health, Health
Economics Program
22
Lessons learned

• Questions are sometimes less complicated than
  they seem
• Relatively simple projections can give you
  estimates that are likely as accurate as
  expensive, sophisticated modeling
• Tradeoff timeliness and cost versus perceived
  sophistication and credibility

23
Overall lessons learned and things to consider

• Know what you can answer with state-specific data
  and what you cant, and be prepared to support
  your decision
• Know what to prepare for
• CPS versus state-specific survey findings
• How sophisticated does the analysis need to be?
• Is it important it be an econometric model or
  does simple projection get you just as close?
• Cost/Timeliness/model understanding critical

24
Overall lessons learned and things to consider

• Contracting with experts versus doing your own
  modeling/projection
• Credibility?
• Theres nothing magic or mystical about modeling
  understand assumptions and how the detail was
  arrived at
• Use technical assistance
• SHADAC, SCI, others
• National data can be a critical and important
  crosscheck to state data