Beth Hartman Ellis, PhD

1
QualityNet ConferenceSeptember 21, 2006
Medicare Health Outcomes Survey Health Status
Disparities in Medicare Managed Care
Beneficiaries

• Beth Hartman Ellis, PhD
• MaryAnne D. Hope, MS
• Health Services Advisory Group
• Phoenix, AZ

2
Medicare Health Outcomes SurveyBackground

• The Medicare Health Outcomes Survey (HOS)
• Assesses each Medicare Advantage (MA) health
  plans ability to maintain or improve the
  physical and mental health functioning of its
  Medicare beneficiaries over a two-year period
• Is sponsored by CMS
• Launched in 1998
• First Medicare managed care outcomes measure
• More than 1.8 million Medicare beneficiaries
  surveyed to date

3
Medicare Health Outcomes SurveyMethodology

• MA members are surveyed at baseline, and
  respondents are resurveyed two years later
• A cohort comprises respondents from one baseline
  and associated follow up
• Baseline cohort of 1,000 beneficiaries randomly
  sampled from each participating plan
• In plans with less than 1,000, all MA
  beneficiaries are sampled
• Survey mailed to baseline sample
• Telephone follow up of non-respondents

4
Medicare Health Outcomes SurveyPopulation

• Beneficiaries included in the HOS
• Community dwelling
• Nursing home
• Institution
• Disabled under 65
• End stage renal disease patients excluded

5
Survey Content
6
Research Goal for Current Study

• To examine physical health status after a
  two-year interval for living and deceased
  Medicare managed care beneficiaries

7
Analytic Sample for Current Study

• Medicare HOS 2002 2004 Cohort 5 Baseline and
  Follow Up data
• 60,317 beneficiaries
• 65 and over, physical component summary
• (PCS) score at Baseline
• 6,993 of these beneficiaries were deceased
• at follow up and included in the analyses

8
Excluded Groups at Follow Up

• Excluded Groups at follow up
• 1. Invalid survey at follow up (n781)
• Beneficiaries not enrolled in the plan, bad
  address and non-working/unlisted phone number
• 2. Voluntarily disenrolled at follow up
  (n18,603)
• Beneficiaries who left their plan between
  baseline and follow up
• 3. Involuntarily disenrolled at follow up
  (n8,111)
• Beneficiaries whose plans were no longer
  available at follow up
• 4. Non-respondents at follow up (n12,733)
• Beneficiaries who did not respond to the survey
  at follow up

9
Analytic Strategy for Current Study

• We employed the methodology by Diehr and
  colleagues (2001) for including the deceased in
  health outcomes research
• Healthy at follow up defined as a response of
  excellent, very good, or good to the question,
  In general, would you say your health is..

10
Analytic Strategy for Current Study, contd

• Logistic regression used to obtain the
  probability of being healthy at follow up,
  estimated from the baseline PCS score
• Deceased assigned a value of zero
• Clustering among health plans assessed with the
  intraclass correlation coefficient - found to be
  0.02, suggesting clustering (Cohen et al., 2003)
• Solution multilevel model
• SAS PROC MIXED

11
Analytic Strategy for CurrentStudy, contd
Reference groups for regression models

• Race - White
• Income of 50,000 and over
• College graduate
• Male
• Married

• Not a Medicaid recipient
• Self-respondent
• Non-smoker
• No chronic conditions
• Negative response to 3 depression screening
  questions

12
Analytic Strategy for Current Study, contd

• Two multilevel models constructed
• Demographics only
• Demographics and health risks
• Smoker
• Positive depression screen
• Sum of an individuals chronic conditions

13
Specific Predictors

• Demographics
• Race African American, Hispanic, Asian/Pacific
  Islander, American Indian/Alaskan Native, Other
  Race
• Household Income
• Less than 10,000
• 10,000 to 19,999
• 20,000 - 29,999
• 30,000 - 49,999
• Missing income

14
Specific Predictors, contd

• Demographics, continued
• Educational level
• 8th grade or less
• Some high school
• High school graduate/GED
• Some college/2 year degree
• Gender
• Female
• Age
• Proxy respondent

15
Specific Predictors, contd

• Demographics, continued
• Marital Status
• Divorced/separated
• Widowed
• Never married
• Medicaid Status
• Dually eligible (Medicaid Medicare)
• Smoking Status
• Smoker (every day/some days/smoked 100 cigarettes
  in your life)

16
Specific Predictors, contd

• Positive depression screen
• Positive response to any of the 3 depression
  screening questions in the HOS
• Comorbidities
• Individuals sum of 9 chronic conditions

17
Demographics Model
18
Demographics and Health Risks Model
19
Excluded Groups Comparison at Baseline

• Effect sizes for proportions (Cohen, 1988) and
  Hedges g for means (Rosenthal Rosnow, 1991)
  used to assess significance of findings

20
Excluded Groups Comparison at Baseline, contd

• The invalid survey group had significantly
• More Hispanics
• Less Whites
• More with 8th grade education or less
• More with less than 10,000 household income
• Small effect size gt 0.20 lt 0.50
• Medium effect size gt 0.50 - lt 0.80
• Large effect size gt 0.80

21
Excluded Groups Comparison at Baseline, contd

• The invalid survey group had significantly
• Less homeowners
• More dually eligible
• More who had a positive depression screen
• Older
• Lower PCS and MCS scores
• More impaired ADLs
• Small effect size gt 0.20 lt 0.50
• Medium effect size gt 0.50 - lt 0.80
• Large effect size gt 0.80

22
Conclusions

• Probability of not being healthy at follow up
  related to
• Low socioeconomic status
• Low educational level
• Female
• Proxy respondent
• Medicaid recipient (dually eligible)
• Positive depression screen
• Chronic conditions
• Advanced age

23
Conclusions, contd

• Demographics and health risks model
• Better overall fit compared to the demographics
  only model
• Socioeconomic disparities exist in Medicare
  managed care for enrollees in this sample

24
Conclusions, contd

• Medicare managed care plans and QIOs should
  consider targeting beneficiaries with low income
  and low educational levels, depression, and
  comorbidities for disease management programs

25
Medicare HOS Webinars

• Getting the Most out of Your Medicare HOS
  Reports held September 14, 2006
• Upcoming Webinars
• A Guide for Researchers
• October 18, 2006
• Mining Your HOS Data A Toolkit
• November 14, 2006
• Check the HOS Website for information
• about specific dates

26
Contact Information

• Beth Hartman Ellis, PhD Bellis_at_azqio.sdps.org
• 602.665.6133
• MaryAnne D. Hope, MS Mhope_at_azqio.sdps.org
• 602.745.6211
• HOS Web Site www.hosonline.org
• HOS Technical Support
• Medicare HOS Information and Technical Support
  Telephone Line
• 1-888-880-0077
• E-Mail
• hos_at_azqio.sdps.org

27
References

• Agency for Healthcare Research and Quality
  (2005). National Healthcare Disparities Report.
  Available at www.ahrq.gov/qual/nhdr05/nhdr05htm.
  
• Cohen, J. (1988). Statistical Power Analysis for
  the Behavioral Sciences (2nd ed). Hillsdale, NJ
  Lawrence Erlbaum Associates.
• Cohen, J., Cohen, P., West, S.G., Aiken, L.S.
  (2003). Applied multiple regression/correlation
  analysis for the behavioral sciences (3rd ed).
  Mahwah, NJ Lawrence Erlbaum Associates.
• Diehr, P., Patrick, D.L., Spertus, J., et al.
  (2001). Transforming self-rated health and the
  SF-36 scales to include death and improve
  interpretability. Medical Care 39 (7) 670-680.
• Menard, S. (1995). Applied logistic regression
  analysis. Sage Series Quantitative Applications
  in the Social Sciences. Thousand Oaks, CA Sage
  Publications.
• Rosenthal, R. Rosnow, R. L. (1991). Essentials
  of behavioral research methods and data analysis
  (2nd ed). Columbus, OH McGraw-Hill.
• Singer, J. (1998). Using SAS PROC MIXED to fit
  multilevel models, hierarchial models, and
  individual growth models. Journal of Educational
  and Behavioral Statistics, 24(4), 323-355.