Disease Burden and Intensity of Antidiabetic Drug Use by Medicare Beneficiaries with Diabetes: Will Part D Make a Difference? AcademyHealth June 25, 2006

1
Disease Burden and Intensity of Antidiabetic
Drug Use by Medicare Beneficiaries with Diabetes
Will Part D Make a Difference?
AcademyHealth June 25, 2006

• Bruce Stuart, Thomas Shaffer, Linda
  Simoni-Wastila, Ilene Zuckerman
• The Peter Lamy Center on Drug Therapy and Aging
• University of Maryland Baltimore

2
Outline

• Sponsor acknowledgment funding provided by The
  Commonwealth Fund under grant Benchmarking the
  Quality of Medication Use by Medicare
  Beneficiaries
• Background the growing controversy over the
  importance of glycemic control for frail
  diabetics
• Study objectives and hypotheses
• Data and study sample
• Measures
• Statistical strategy
• Results
• Discussion and study implications for policy

3
The Importance of Glycemic Control among Frail
Older Persons

• Traditional diabetes guidelines
• Glycemic control is the mainstay of traditional
  guidelines for treatment of diabetes promulgated
  by the ADA, AHRQ, NCQA, AMDA, and other
  organizations.
• Most guidelines recognize the importance of
  comorbidity and frailty in making clinical
  treatment decisions, but offer no quantitative
  guidance
• New thinking
• Current approaches to assessing quality of
  diabetes treatment do not account for
  heterogeneity among older patients
• Less intense targets for glycemic control are
  warranted for older diabetics in poor health and
  low life expectancy

4
Study Objectives and Hypotheses

• Objectives
• Benchmark prevalence and level of use of
  antidiabetic agents (oral hypoglycemic agents and
  insulins) among community-dwelling Medicare
  beneficiaries with diabetes stratified by burden
  of illness
• Identify factors associated with differences in
  antidiabetic use by burden of illness
• Hypotheses
• Factors hypothesized to be associated with higher
  antidiabetic drug use
• Income, prescription drug coverage, severity of
  diabetes, obesity, health system contacts
  (surveillance hypothesis)
• Factors hypothesized to be associated with lower
  antidiabetic drug use
• Increasing disease burden, kidney disease,
  hospitalization, health system contacts
  (competing demands hypothesis)

5
Data and Study Sample

• Data
• 2002 MCBS Cost and Use files (N12,697)
• Study Sample
• Inclusion criteria
• Self report of diabetes and/or ICD-9 diagnosis
  codes 250.xx, 357.2, 362.01, 362.02, 366.41. Two
  or more Dx of diabetes on outpatient and carrier
  claims or one diagnosis on an impatient claim
• Exclusion criteria
• In MCBS facility sample all or part year, Part A
  and/or B part year, any Medicare HMO enrollment,
  incomplete surveys
• Final study sample N1,956

6
Measures

• Overall Burden of Illness
• Stratify study sample into 10 equal sized groups
  (deciles) by cumulative spending for all medical
  services including drugs
• Dependent Variables
• Any antidiabetic drug use in 2002 by type (orals,
  insulin), number of prescription fills per year,
  total pill count for oral hypoglycemics
• Explanatory Variables
• 6 domains (1) decile assignment, (2)
  demographics (age, sex, race, census region), (3)
  economic variables (income, prescription
  coverage), (4) health (self-reported, ADLs, BMI,
  self-reported DM/no claims, diabetes
  complication, chronic kidney disease) (5)
  contacts with health system (inpatient hospital,
  SNF, HHA, hospice, count of physician EM visits,
  count of different physicians seen in office
  visits), (6) denominator days

7
Statistical Strategy

• Descriptive Charts
• Prevalence rates for common comorbidities by
  decile of medical spending
• Prevalence and level of antidiabetic drug use by
  decile of medical spending
• Regression Analysis
• Stepwise logistic and OLS regressions with
  explanatory variables entered in blocks
  representing the 6 domains of explanatory
  variables. Objective is to determine whether
  differences in drug use observed in the
  unadjusted stratification by decile lose
  significance with additional covariates, and if
  so which are responsible
• Digging Deeper
• Within-decile regression analysis

8
Descriptive Results
9
Characteristics of the Study Population (N1,956)
Characteristic Frequency or count (SE)
Demographics
Age () Under 65 65 69 70 74 75 79 80 14.8 17.1 24.7 20.8 22.6
Female sex () 54.8
Nonwhite () 20.7
Geographic residence () East Midwest South West 18.5 25.6 43.3 12.6
10
Characteristics of the Study Population (N1,956)
Economic variables
Annual income () lt 10,000 10,000 20,000 20,001 30,000 gt30,000 25.1 31.4 21.5 22.0
Supplementary medical insurance No coverage () Full year () Part year () Covered months (count) 8.5 88.0 3.5 6.4 (0.5)
Prescription drug coverage No coverage () Full year () Part year () Covered months (count) 25.0 65.1 9.9 6.0 (0.2)
11
Characteristics of the Study Population (N1,956)
Health status and disease severity factors
Self-reported health () Poor Fair Good to excellent 13.0 28.7 57.6
3 or more activity limitations (ADLs) () 5.5
BMI gt 30 () 39.1
Self reported diabetes/no claim diagnosis () 10.5
Diabetes complication (ICD-9 250.1 250.9) () 52.5
Chronic kidney disease (ICD-9 ) () 7.1
Died () 4.6
Denominator days (count) 353.3
12
Characteristics of the Study Population (N1,956)
Health system contact variables
Any hospital admission () 27.8
Any SNF stay () 4.3
Any home health visit () 11.4
Any hospice stay () 1.2
Number of office visits for evaluation and management (EM) (count) 9.1
Number of different physicians seen for EM visits (count) 4.8
13
Figure 1. Prevalence of Selected Diseases among
Medicare Beneficiaries with Diabetes Stratified
by Decile of Annual Medical Spending, 2002
14
Disease Burden and Medication Use the Inverted
U
15
Figure 2. Unadjusted Annual Prevalence of
Antidiabetic Medication Use by Decile of Annual
Medical Spending
16
Figure 3. Unadjusted Mean Annual Prescription
Fills for Oral Antidiabetic Medications by Users
of These Drugs
17
Figure 4. Unadjusted Mean Annual Pill Counts for
Oral Antidiabetic Medications Stratified by Users
of These Drugs
18
Multivariate Results
19
Adjusted Odds Ratios for Any Use of Antidiabetic
Medications by Study Subjects
Decile Decile Only Regression Unadj. Odds (95 CI) Full Model Adj. Odds (95 CI)
1 2 3 4 5 6 7 (reference) 8 9 10 0.40 (0.26-0.60) 0.61 (0.40-0.91) 0.79 (0.52-1.20) 0.82 (0.54-1.24) 0.64 (0.42-0.97) 0.87 (0.57-1.33) 1.00 0.81 (0.53-1.23) 0.83 (0.55-1.27) 0.53 (0.36-0.81) 0.56 (0.34-0.91) 0.71 (0.45-1.13) 0.81 (0.51-1.28) 0.87 (0.56-1.37) 0.66 (0.42-1.03) 0.80 (0.51-1.24) 1.00 0.79 (0.50-1.25) 0.83 (0.52-1.35) 0.69 (0.41-1.16)
p lt.05 p lt.01
20
Factors Affecting Treatment Odds (or not)

• None of the demographic or economic factors was a
  consistent predictor of likelihood of using
  antidiabetic medicine. Prescription coverage
  plays no significant role.
• Significant (p lt.01) health and diabetes severity
  indicators (range of odds ratios in models 5 and
  6) BMI gt30 (1.51 to 1.53), self reported
  diabetes/no claim (0.20), diabetes complications
  (1.75 to 1.80), chronic kidney disease (0.55)
• Only one significant (p lt.05) health system
  contact variables (odds ratios in model 6)
  number of different physician seen in office
  visits (0.95)
• The adjusted treatment odds in the full model are
  16 points higher in each tail of the
  distribution, but the inverted U pattern
  remains

21
Adjusted Coefficients for Annual Prescription
Fills for Study Subjects Using Oral Antidiabetic
Medications
Decile Decile Only Regression Unadj. Coef. (95 CI) Full Model Adj. Coef. (95 CI)
1 2 3 4 5 6 7 (reference) 8 9 10 -3.79 (-5.30 -2.28) -2.30 (-3.74 -0.86) -0.60 (-2.20 1.00) -0.78 (-2.37 0.80) -0.99 (-2.66 0.69) -0.32 (-1.98 1.33) 0.00 -1.30 (-2.95 0.35) -1.11 (-2.81 0.59) -2.77 (-4.36 -1.19) -4.37 (-6.07 -2.67) -3.05 (-4.67 -1.43) -1.32 (-3.01 0.38) -1.58 (-3.27 0.12) -1.60 (-3.32 0.13) -0.94 (-2.59 0.72) 0.00 -0.71 (-2.39 0.98) 0.43 (-1.43 2.30) -0.38 (-2.26 1.50)
p lt.05 p lt.01
22
Factors Affecting Oral Hypoglycemic Treatment
Levels (or not)

• Full year prescription coverage has a small
  effect in increasing antidiabetic drug use in
  models 4 and 5, but loses significance in the
  full model 6. Lower incomes are associated with
  higher drug use but the effect is small and not
  consistently significant
• Significant (p lt.05) health and diabetes severity
  indicators (range of coefficients in models 5 and
  6) BMI gt30 (0.90 to 0.93), self reported
  diabetes/no claim (-1.49 to -1.80),
  hospitalization (-2.37)
• Only two significant health system contact
  variables in model 6 any hospitalization
  (-2.37 plt.01), and number of different
  physicians seen in office visits (0.02 p lt.05)
• Covariate control essentially eliminates the
  upper tail in the inverted U for level of
  antidiabetic drug use, but unexplained variance
  in the lower tail increases significantly

23
Discussion Main Points

• High level of heterogeneity in antidiabetic drug
  use
• Almost 1 in 4 Medicare beneficiaries with
  diabetes took no antidiabetic drugs in 2002
• All 6 variable domains combined could explain
  only a small fraction of the likelihood and level
  of antidiabetic use among beneficiaries
• The inverted U pattern in medication intensity
  by disease burden
• Increasing disease burden is associated with a
  sharp rise in antidiabetic medication intensity
  among the least sick, plateaus, and then falls
  sharply among those with the greatest disease
  burden
• Study limitations
• No measure of blood glucose levels
• Cross-sectional design precludes causal inferences

24
Conclusions Implications for Part D

• Drug coverage makes no significant difference in
  antidiabetic drug use
• Quality of care and the medication intensity
  curve
• All three zones in the inverted U have
  potential clinical significance but require
  additional study
• Does the low level of antidiabetic use among
  diabetics with the least disease burden reflect
  lack of need or underuse?
• Does the plateau in prevalence and level of
  antidiabetic use among those in the middle of the
  range of disease burden represent a meaningful
  target for improving the quality of diabetic
  treatment in the Medicare program?
• Is the sharp drop in medication use in the upper
  tail of the distribution of disease burden
  justified by new thinking about the need for
  glycemic control among frail older people or does
  it represent underuse?

25
Thank You!