Pres jsm2011 aug1_turner
-
Upload
soder145 -
Category
Economy & Finance
-
view
390 -
download
0
description
Transcript of Pres jsm2011 aug1_turner
![Page 1: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/1.jpg)
Modeling Health Insurance Coverage Estimates for Minnesota Counties
Joint Statistical Meetings
Miami Beach, Florida
August 1, 2011
Joanna Turner and Peter GravenState Health Access Data Assistance Center (SHADAC)University of Minnesota, School of Public Health
Supported by a grant from The Robert Wood Johnson Foundation
![Page 2: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/2.jpg)
www.shadac.org
Acknowledgements
• Thanks to the Minnesota Department of Health, Health Economics Program for their support of this work
• www.health.state.mn.us/healtheconomics
2
![Page 3: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/3.jpg)
www.shadac.org3
Outline
• Background
• Research Objective
• Methodology
• Results
![Page 4: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/4.jpg)
www.shadac.org
Background
• Minnesota Health Access Survey (MNHA) – Telephone survey conducted every 2 years– Provides MN and regional estimates, including
estimates for select populous counties and cities
• County level estimates are frequently requested data
4
![Page 5: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/5.jpg)
www.shadac.org
U.S. Census Bureau County Level Uninsurance Estimates
• American Community Survey (ACS)– 1-year estimates: 12 Minnesota counties (Available) – 3-year estimates: 47 Minnesota counties (Fall 2011)– 5-year estimates: 87 Minnesota counties (Fall 2013)
• Small Area Health Insurance Estimates Program (SAHIE)– 2007 estimates (most recent year) for all 87
Minnesota counties– Release of 2008 and 2009 estimates planned for
Fall 2011
5
![Page 6: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/6.jpg)
www.shadac.org
Research Objective
Produce Minnesota uninsurance rates by county for 2009
– Use the Minnesota Health Access Survey (MNHA)– Use other sources of uninsurance estimates– Include estimates of uncertainty – Allow for future input sources– Use methods accessible to Minnesota Department of
Health staff – Use methods that can be applied to other states
6
![Page 7: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/7.jpg)
www.shadac.org
Methodology
• Hierarchical Bayesian Small Area Estimation (SAE) Model – Spatial Conditional Autoregressive (CAR) errors
• ACS county level estimates• Hierarchical Bayesian Simultaneous
Equation Model (SEM)– 2009 MNHA– 2009 ACS– 2008 ACS– 2007 SAHIE
7
![Page 8: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/8.jpg)
www.shadac.org
Methodology Overview
8
Simultaneous Equation Model
2009 MNHA SAE Estimates
2008, 2009 ACS County Model
2008, 2009 ACS 1-year estimates of uninsurance - PUMA level
PUMA-County Crosswalk
2007 SAHIE Estimates
2008, 2009 ACS County Estimates
Final Results: 2009 MN County Estimates of Uninsurance
2009 MNHA SAE Model
Additional Data Sources – county level
2005-2009 ACS 5-year estimates of demographics – county level
2009 MNHA direct estimates of uninsurance – county level
![Page 9: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/9.jpg)
www.shadac.org
MNHA SAE Model (1)
• Telephone survey conducted every 2 years• Primarily collects data on one randomly
selected member of household (target)• Certain information, such as health
insurance coverage, is collected for all household members
• Provides MN and regional estimates, including estimates for select populous counties and cities
• 2009 complete (turned) file of all household members has a sample size of 31,802
9
![Page 10: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/10.jpg)
www.shadac.org
MNHA SAE Model (2)
• Area-level spatial conditional autoregressive (CAR) model
• Covariates X include ACS 5-year data, Census Bureau population estimates, and additional sources including administrative records. For example:– Quarterly Census of Employment and Wages
(QCEW)– Local Area Unemployment Statistics (LAUS)
10
![Page 11: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/11.jpg)
www.shadac.org
MNHA SAE Model (3)
• Variables predicting the direct estimate of uninsurance in the MNHA survey: – Immigration Rate, 2000-2009 (Population estimates)– Percent Employed Working in Retail, 2009 (QCEW)– Percent Moved into State, 2005-2009 (ACS)– Weekly Wage, 2009 (QCEW)– Percent White, 2005-2009 (ACS)– Percent Land in Farms, 2007 (USDA)– Percent 65 and Over, 2005-2009 (ACS)– Average Unemployment Rate, 2009 (LAUS)
11
![Page 12: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/12.jpg)
www.shadac.org
Methodology Overview – ACS County
12
2009 MNHA SAE Model
Simultaneous Equation Model
2009 MNHA SAE Estimates
2009 MNHA direct estimates of uninsurance – county level
2005-2009 ACS 5-year estimates of demographics – county level
Additional Data Sources – county level 2007 SAHIE
Estimates
2008, 2009 ACS County Estimates
Final Results: 2009 MN County Estimates of Uninsurance
2008, 2009 ACS 1-year estimates of uninsurance - PUMA level
2008, 2009 ACS County Model
PUMA-County Crosswalk
![Page 13: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/13.jpg)
13
Public Use Microdata Area (PUMA)
![Page 14: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/14.jpg)
www.shadac.org
ACS County Model
• Two types of estimates– County estimate exists in 1-year ACS (12 counties)
Estimate and SE used directly
– County is a subset of PUMA (75 counties) Estimate from PUMA SE is the PUMA estimate times the ratio of the PUMA
poverty SE divided by the county poverty SE
14
![Page 15: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/15.jpg)
www.shadac.org
Methodology Overview – SAHIE
15
2009 MNHA SAE Model
Simultaneous Equation Model
2009 MNHA SAE Estimates
2009 MNHA direct estimates of uninsurance – county level
2005-2009 ACS 5-year estimates of demographics – county level
Additional Data Sources – county level
2008, 2009 ACS County Model
2008, 2009 ACS 1-year estimates of uninsurance - PUMA level
PUMA-County Crosswalk
2007 SAHIE Estimates
2008, 2009 ACS County Estimates
Final Results: 2009 MN County Estimates of Uninsurance
![Page 16: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/16.jpg)
www.shadac.org
SAHIE Estimates
• Census Bureau's Small Area Health Insurance Estimates (SAHIE) program produces model-based estimates of health insurance coverage– State estimates by age/sex/race/income categories– County estimates by age/sex/income categories
• Estimates are for 0-64 so we need to make a correction to use in our all ages model
16
𝑈𝑛𝑖𝑛𝐴𝑙𝑙𝑆𝐴𝐻𝐼𝐸=𝑈𝑛𝑖𝑛𝑢𝑛𝑑𝑒𝑟 65
𝑆𝐴𝐻𝐼𝐸 −𝑈𝑛𝑖𝑛𝑢𝑛𝑑𝑒𝑟 65𝑆𝐴𝐻𝐼𝐸 ∗𝑈𝑛𝑖𝑛𝑢𝑛𝑑𝑒𝑟 65
𝑆𝐴𝐻𝐼𝐸 +𝑃𝑟𝑜𝑝65𝑜𝑣𝑒𝑟 𝐴𝐶𝑆5 𝑦𝑒𝑎𝑟∗𝑈𝑛𝑖𝑛65𝑜𝑣𝑒𝑟𝐶𝑃𝑆
![Page 17: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/17.jpg)
www.shadac.org
Methodology Overview – SEM Model
17
2009 MNHA SAE Model
Simultaneous Equation Model
2009 MNHA SAE Estimates
2009 MNHA direct estimates of uninsurance – county level
2005-2009 ACS 5-year estimates of demographics – county level
Additional Data Sources – county level
2008, 2009 ACS County Model
2008, 2009 ACS 1-year estimates of uninsurance - PUMA level
PUMA-County Crosswalk
2007 SAHIE Estimates
2008, 2009 ACS County Estimates
Final Results: 2009 MN County Estimates of Uninsurance
![Page 18: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/18.jpg)
www.shadac.org
Simultaneous Equation Model (SEM)
18
Model
Prediction
![Page 19: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/19.jpg)
www.shadac.org
Limitations/Enhancements
• MNHA SAE model could include more advanced variable selection and transformations
• MNHA SAE model could take advantage of information outside the state
• Multi-staged methodology removes non-parametric errors– Integrated model could propagate errors more
accurately19
![Page 20: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/20.jpg)
20
SEM Model Results - Percent Uninsured
![Page 21: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/21.jpg)
www.shadac.org
SEM Model Results - Precision
• Standard Deviation from SEM model ranges between (0.5-1.3)
• Coefficient of Variation (CV) – CV=Standard deviation/Estimate– Used as threshold for release by Census/NCHS
• >30% CV estimates are unreliable
– Ranges between (5.7%-14.4%)
21
![Page 22: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/22.jpg)
22
SEM Model Results - CV
![Page 23: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/23.jpg)
23
SEM Model Results - Source Comparison
0.0
5.0
10.0
15.0
20.0
25.0O
lmst
edW
ashi
ngto
nCa
rver
Nic
olle
tW
asec
aPo
lkD
akot
aSt
earn
sW
ilkin
Bent
onSh
erbu
rne
Dod
geM
ower
Rose
auAn
oka
Penn
ingt
onSc
ott Clay
Swift
Stee
leBr
own
Wab
asha
Blue
Ear
thM
eeke
rPo
peM
artin
Goo
dhue
Kand
iyoh
iCo
tton
woo
dH
oust
onPi
pest
one
Chis
ago
Jack
son
Rock
Mah
nom
enO
tter
Tai
lSi
bley
Big
Ston
eKi
ttso
nIs
anti
Redw
ood
Le S
ueur
Nor
man
Lyon
Renv
ille
Yello
w M
edic
ine
St. L
ouis
Wrig
htW
inon
aFr
eebo
rnW
aden
aD
ougl
asSt
even
sTo
ddG
rant
Chip
pew
aH
enne
pin
Mor
rison
Fillm
ore
Rice
Red
Lake
Lake
Lac
qui P
arle
Mur
ray
Trav
erse
Linc
oln
Ram
sey
Carlt
onM
arsh
all
Kooc
hich
ing
Itasc
aFa
ribau
ltN
oble
sW
aton
wan
McL
eod
Crow
Win
gPi
neCa
ssKa
nabe
cM
ille
Lacs
Cook
Aitk
inLa
ke o
f the
Woo
dsBe
cker
Hub
bard
Beltr
ami
Clea
rwat
er
SEM Rate
2009 MNHA (SAE)
2009 ACS (County)
2008 ACS (County)
2007 SAHIE
![Page 24: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/24.jpg)
www.shadac.org
Conclusions
• Produced uninsurance estimates and estimates of uncertainty using a state survey and multiple outcomes
• Methodology is accessible and can be applied to other states
24
![Page 25: Pres jsm2011 aug1_turner](https://reader033.fdocuments.us/reader033/viewer/2022051514/5481d5f2b479590f0d8b4615/html5/thumbnails/25.jpg)
25
Sign up to receive our newsletter and updates at www.shadac.org
facebook.com/shadac4states
@SHADAC
Joanna TurnerState Health Access Data Assistance Center
University of Minnesota, Minneapolis, MN612-624-4802