An Overview of the Small Area Health Insurance Estimates Program

1
An Overview of the Small Area Health Insurance
Estimates Program

• Joanna Turner, Robin Fisher,
• and David Waddington
• U.S. Census Bureau
• September 23, 2004

2
Motivation for Estimates of the Uninsured

• Broad interest in health insurance coverage
  issues
• Not a question on the decennial census or the
  American Community Survey (ACS)
• State Childrens Health Insurance Program (SCHIP)

3
Small Area Income and Poverty Estimates (SAIPE)

• Model-based estimates of median household income
  and poverty for states, counties, and school
  districts
• Uses survey data, administrative records, and
  decennial census data
• Estimates evaluated favorably by National Academy
  of Sciences NAS panel
• Estimates used by Department of Education and
  Health and Human Services

4
Small Area Health Insurance Estimates (SAHIE)

• Extend SAIPE knowledge and methodologies to the
  area of health insurance coverage
• Estimates for all states and counties
• Estimates for various age groups
• Estimates of Mean Squared Error

5
Experimental Estimates

• State
• Number of uninsured children 0-18 in households
  with income 200 of the Federal Poverty
  Threshold (SCHIP)
• County
• Total number of uninsured

6
ModelCounty Total Insured

• Hierarchical Bayesian linear model

7
Model (2)County Total Insured

• Response Variable
• Insured Rate Insured/Population
• Annual Social and Economic Supplement (ASEC) to
  the Current Population Survey (CPS)
• 3 year average
• Log scale

8
Model (3)County Total Insured

• is the random effects term
• is the sampling error term
• True Log Insured Rate is represented by

9
Model (4)County Total Insured

• Predictor Variables
• Administrative Records
• Tax data
• Medicaid Statistical Information System (MSIS)
  enrollment
• Food Stamp Program number of participants
• Census 2000 demographic variables

10
Variable SelectionCounty Total Insured

• IRS proportions between multiples of FPT
• (100, 130)
• (200, 300)

• Proportion eligible for Medicaid
• Children ages 0-18
• Adults ages 35-64
• Hispanics
• Blacks

11
Variable Selection (2)County Total Insured

• Census 2000
• Total Population
• Proportion Hispanic

• Food Stamps
• Number of participants

12
ResultsCounty Total Insured

• Overall Fit is good
• Examination of p-values and plots
• Mean posterior coefficient of variation is about
  6.0

13
Conclusions

• Research is promising
• Refine state and county models
• Model forms and data
• In the meantime, sum counties.
• We expect to produce experimental estimates for
  public examination soon

14
Future Plans

• Include other groups
• Age groups
• SCHIP 0-18, 200 Federal Poverty Threshold
• Model validation

15
Questions for SDC

• Is there interest in these estimates?
• Would you like a chance to comment on the
  estimates before we move to producing them on a
  regular basis?

16
Questions for SDC (2)

• Are you aware of data appropriate for validation
  of these estimates?
• What if there are administrative records
  available for some areas but not others?

17
Further Reading

• Health Insurance Estimates for States by Robin
  Fisher and Jennifer Campbell
• Health Insurance Estimates for Counties by
  Robin Fisher and Joanna Turner

18
Further Reading (2)

• Small Area Estimation of Health Insurance
  Coverage from the Current Population Surveys
  Annual Social and Economic Supplement and the
  Survey of Income and Program Participation by
  Robin Fisher and Joanna Turner
• Available in November, 2004

19
SAIPE website

• http//www.census.gov/hhes/www/saipe.html
• SAHIE website coming in Spring 2005

20
Contact Information

• [phone]
• hhes.sahie_at_census.gov