Using&Medicare&Part&D&Data& - txcercit.org · Using&Medicare&Part&D&Data& HollyM.Holmes,MD&...
Transcript of Using&Medicare&Part&D&Data& - txcercit.org · Using&Medicare&Part&D&Data& HollyM.Holmes,MD&...
Objectives
• Understand the contents of the Medicare Part D files
• Discuss the strengths and limitations in using Medicare Part D data in research
Medicare Part D – what is it? • Established by the Medicare Prescription Drug, Improvement, and Modernization Act of 2003
• Covers prescription medications – Not OTC, not supplements
• Available to all 43 million Medicare beneficiaries
• Started January 1, 2006
PART D QUIZ
True or False:
Enrollment in Part D is optional.
True or False:
Once you choose a Part D plan, you cannot
switch.
Part D enrollment
• 60% of Medicare beneficiaries were enrolled
in Part D in 2011
• Open enrollment for 2012: 10/15/11-‐12/7/11
• 6% switch every year (therefore 94% do not
switch)
PART D QUIZ
True or False:
Medicare Advantage enrollees who have Part D
coverage are enrolled in stand-‐alone Part D
plans (PDP).
Part D Enrollment in 2010
Source: ResDAC
If you are studying A+B+D, you really only have 37% of Medicare beneficiaries. And since ½ of PDP is Low Income Subsidy, you only have 18% of “middle-‐class”-‐ish benes. If you want to study all of the Low Income Subsidy beneficiaries, you need both PDP and MAPD.
PART D QUIZ Part D plans can offer different
A. Deductibles
B. Copays
C. Premiums
D. Gap Coverage
E. All of the above
This goes to 0 by 2020 (fed govt picks up rest)
Source: Kaiser Family Foundation.
These are determined based on true out-‐of-‐pocket cost (TROOP)
This goes to 0 by 2020 (fed govt picks up rest)
Source: Kaiser Family Foundation.
These are determined based on true out-‐of-‐pocket cost (TROOP)
Beneficiary Phase Variable identifies where the bene is when the prescription is filled.
Part D QUIZ
True or False:
Part D data for Medicare Advantage (HMO)
beneficiaries is available to researchers.
What types of Part D data are available?
• Public use (landscape) files
– Not linkable to beneficiary-‐level files
• Beneficiary-‐Level Part D Data
– Final Part D Rule (5/28/2008): keeps sensitive data
encrypted and cost data aggregated
– Cannot identify prescriber, pharmacy, or plan
What types of Part D data are available?
• Public use (landscape) files
– Not linkable to beneficiary-‐level files
• Beneficiary-‐Level Part D Data
– Final Part D Rule (5/28/2008): keeps sensitive data
encrypted and cost data aggregated
– Cannot identify prescriber, pharmacy, or plan
Part D files are linkable to A + B only through the beneficiary ID, which is encrypted and the same as the bene ID from A + B. This will not change until the Part D Rule changes.
Part D Denominator File
• Called the Master Beneficiary Summary File
(MBSF)
• Contents – Beneficiary Summary File (BSF)
– Chronic Conditions (26)
– Cost & Utilization – annual summary
– Death Information
Part D Denominator File
• Called the Master Beneficiary Summary File
(MBSF)
• Contents – Beneficiary Summary File (BSF)
– Chronic Conditions (26)
– Cost & Utilization – annual summary
– Death Information
The BSF is available by summer of year 20xx+1. The Part D denominator part of the BSF is added in March of 20xx+2. You may need 2 pieces of the BSF.
Demographic information known to CMS overwrites all demographic information from a claim or Part D event in the MBSF. That means agreement between MBSF and claims/PDE unless there is a change.
Part D Event Files • Bene_ID
• Drug info: from NDC linkage with First Data Bank
– Generic/brand name
– Dosage (strength and unit of measure, eg mg)
– Form (tablet, capsule, cream, spray, patch)
• Claim info:
– Quantity dispensed
– Date dispensed
– Days supply
Part D Event Files • Bene_ID
• Drug info: from NDC linkage with First Data Bank
– Generic/brand name
– Dosage (strength and unit of measure, eg mg)
– Form (tablet, capsule, cream, spray, patch)
• Claim info:
– Quantity dispensed
– Date dispensed
– Days supply
The directions on the prescription are not part of the PDE files. Frequency of dosing can be calculated by quantity dispensed / days supply.
Dosing Frequency Example
Drug % of Claims with Dosing Frequency
<1/day 1-‐<2/day 2-‐<3/day >3/day
Amlodipine 6.4 89.3 4.1 0.22
Warfarin 15.5 69.7 9.8 5.0
Part D Appended Files
• Prescriber
– Encrypted ID, type, specialty, subspecialty, and state
• Plan
– Encrypted ID, benefits, premiums, tiers, service areas
• Pharmacy
– Encrypted ID, type of pharmacy, state
Part D Appended Files
• Prescriber
– Encrypted ID, type, specialty, subspecialty, and state
• Plan
– Encrypted ID, benefits, premiums, tiers, service areas
• Pharmacy
– Encrypted ID, type of pharmacy, state
The encrypted ID for prescribers, plans, and pharmacies does not link with A or B files or with any other beneficiary-‐level files.
PART D QUIZ
True or False:
Medicare beneficiaries with Medicaid (dual
eligibles) continue to get their prescriptions
through Medicaid.
Identifying Low Income Beneficiaries
• State Buy-‐In: A state paid for the beneficiary’s Part B coverage through Medicaid or a savings program
• Low Income Subsidy: benes with help paying
premiums, deductibles, and/or copay, no gap, no late
enrollment
• Dual Eligible status: traditional Medicaid and other
Medicare savings programs
Identifying Low Income Beneficiaries
• State Buy-‐In: A state paid for the beneficiary’s Part B coverage through Medicaid or a savings program
• Low Income Subsidy: benes with help paying
premiums, deductibles, and/or copay, no gap, no late
enrollment
• Dual Eligible status: traditional Medicaid and other
Medicare savings programs
How to identify: State Buy-‐In: State buy-‐in variable in BSF LIS: Cost share group variable Dual Eligible: State reported dual eligible status code
PART D QUIZ What is the minimum number of Part D plans a
beneficiary could choose from for 2012?
A. 5
B. 15
C. 25
D. 30
What drugs are covered?
• No OTCs
• Protected Drugs:
– Immunosuppressant
– Antidepressant
– Antipsychotic
– Anticonvulsant
– Antiretroviral
– Antineoplastic
What drugs are covered?
• No OTCs
• Protected Drugs:
– Immunosuppressant
– Antidepressant
– Antipsychotic
– Anticonvulsant
– Antiretroviral
– Antineoplastic
Benzodiazepines and barbiturates are not covered under Medicare Part D, but they may be covered by some states’ LIS programs
Medication-‐Centered Research Ideas with Part D
• Polypharmacy
– Overutilization
– Use of inappropriate medications
• Suboptimal prescribing
– Drug interactions
– Underutilization
• Adherence
• Prescribing patterns
Medication Adherence
1. Determine predictors of adherence in Medicare Part D beneficiaries with hypertension.
2. Quan<fy the extent to which con<nuity of care with a provider affects adherence.
3. Es<mate the amount of nonadherence aBributable to the provider.
Methods Study Popula<on
• Medicare claims and Part D event files for 5% sample of
Medicare beneficiaries
• 66 and older on January 1, 2007
• Coverage – 24 months A and B without HMO 2006-‐2007
– 12 months Part D in 2007
– PDE files in 2006 and 2007
Methods
• Stable an<hypertensive users – Uncomplicated Hypertension (401.xx)
– PDE files for an<hypertensive medica<on in 2006 and 2007
– Mul<ple med classes, categorized as BP meds by JNC7
– No dose changes
• Excluded hospitalized and ins<tu<onalized persons
Medication Possession Ratio: (MPR)
# days’ supply dispensed between 1st and last fill date # days between 1st and last fill date
Exclude days’ supply on last fill date
Measure of adherence: MPR
Medica'ons Thiazides
Beta blockers Calcium channel blockers
ACEIs/ARBs Vasodilators Clonidine
Renin inhibitors Alpha blockers
Potassium sparing diure<cs Loop diure<cs
Measures • Medica<on Possession Ra<o (MPR)
– Sum of days’ supply between 1st and last fill / total days between 1st and last fill
• Main outcome = % adherent
– Adherent = average MPR 80% or greater
• Possible predictors
– Demographics
– Socioeconomic status
– Comorbidity
– Medica<on use
– Number of prescribers
Age 66 on 1/1/07 n=382939
24 months of Part A and B n=319359
Part D events for blood pressure medica<ons n=276443
Uncomplicated HTN n=236158
Medicare beneficiaries with HTN n=471190
1,698684 benes without hypertension
88,251 younger than 66 on 1/1/07
63,580 with HMO
14,900 no BP meds 27,810 with <2 claims 206 with dose change
Medicare enrollees with PDE files in 2006 – 2007 n=2,169,874
38,125 comp. HTN 2160 with MPR >1.43
Uncomplicated HTN N=168522
Study Popula<on N=168522
56,651 hospitalized 2007 9,550 in nursing home 324 in Territories or unknown
Table 1. Characteristics of Study Population (n=168,522)
Variable Category Total (Column %) Percent Adherent
AGE GROUP
66-‐69 36468 (21.6%) 78.9% 70-‐74 42178 (25%) 79.3% 75-‐79 37936 (22.5%) 79.7% 80-‐84 28488 (16.9%) 79.2% 85+ 23452 (13.9%) 80.5%
SEX Female 116942 (69.4%) 79.4% Male 51580 (30.6%) 79.6%
RACE/ETHNICITY
Non-‐Hispanic white 137981 (81.9%) 81.5% Black 14249 (8.5%) 67.8% Hispanic 9656 (5.7%) 69.3% Amer. Indian/Alaskan 563 (0.3%) 67.7% Asian/pacific island 4950 (2.9%) 78.4% Other 882 (0.5%) 74.7% Unknown 241 (0.1%) 80.1%
Table 1. Characteris'cs of Study Popula'on (n=168,522)
Variable Category Total
(Column %) Percent Adherent
DIVISION
East North Central 27970 (16.6%) 82% East South Central 13688 (8.1%) 77.3% Middle Atlantic 20407 (12.1%) 80.4% Mountain 7189 (4.3%) 78.2% New England 10079 (6%) 83% Pacific 17964 (10.7%) 76.7% South Atlantic 35746 (21.2%) 78% West North Central 17192 (10.2%) 84.2% West South Central 18287 (10.9%) 75.6%
LOW-‐INCOME SUBSIDY
No 127781 (75.8%) 80.5% Yes 40741 (24.2%) 76.1%
MEDIAN INCOME IN CENSUS TRACT
0 -‐ 31,000 36229 (22.1%) 76.1% 31,000-‐38,000 42412 (25.9%) 79.8% 38,000-‐49,000 42033 (25.7%) 80.6% 49,000+ 43187 (26.4%) 81%
% IN CENSUS TRACT WITH <12 YEARS EDUCATION
0 -‐ 12.2 41473 (25.3%) 81.7% 12.2-‐18.6 40877 (24.9%) 81.6% 18.6-‐27.1 40786 (24.9%) 79.5% 27.1+ 40712 (24.8%) 75.2%
Table 1. Characteristics of Study Population (n=168,522)
Variable Category Total
(Column %) Percent Adherent
DEPRESSION No 155027 (92%) 79.6% Yes 13495 (8%) 78.2%
DEMENTIA No 160071 (95%) 79.5% Yes 8451 (5%) 79%
NUMBER OF ELIXHAUSER’S CONDITIONS
0-‐1 comorbidity 74943 (44.5%) 80.4% 2-‐3 comorbidity 62119 (36.9%) 79.3% 4+ comorbidity 31460 (18.7%) 77.6%
IN COVERAGE GAP IN 2007
No 108046 (64.1%) 77.2% Yes 60476 (35.9%) 83.5%
NUMBER OF MEDS 8.9 (+/-‐4.8)
NUMBER OF BP MEDS 2.1 (+/-‐1.1)
Logistic Regression Model for Adherence (n=168,522)
Characteristic Odds Ratio (95% CI)
Low Income Subsidy 1.14 (1.10-‐1.18)
% in Census Tract <12 yrs Education (>18.6% vs. 0-‐18.6%)
0.93 (0.90-‐0.95)
Depression 0.94 (0.90-‐0.99)
Comorbidity (0-‐1 conditions as reference)
2 to 3 conditions 0.93 (0.90-‐0.95)
4 or more conditions 0.85 (0.82-‐0.88)
Number of Medications 0.97 (0.97-‐0.98)
In the Coverage Gap in 2007 1.65 (1.60-‐1.70)
Total Copay ($1000 incr.) 1.23 (1.20-‐1.25)
Number of Unique Prescribers 0.98 (0.97-‐0.98)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 8 15
22
29
36
43
50
57
64
71
78
85
92
99
106
113
120
127
134
141
148
155
162
169
176
183
190
197
204
211
218
225
232
239
246
253
260
267
274
281
288
295
302
Percent of Population in HRR with ≥80% Adherence
Regional Differences in Adherence
% ADHERENT 0.757 0.793 0.836 percentile 20th 50th 80th
Regional Differences • Model with HRR (mull): ICC = 1.8%
• Model with HRR level variables: ICC = 1.4%
– Medicare enrollees in HRR
– Total Part B expenditure
– Primary Care Physicians in HRR
• Model with Pa<ent level variables: ICC = 1.0%
• Model with HRR and Pa<ent Factors: ICC = 0.87%
Conclusions • Significant differences in an<hypertensive medica<on adherence among different racial/
ethnic groups and in persons with higher
levels of comorbidity.
• Differences by region, medica<on use, number
of prescribers.
Medication-‐Centered Research Ideas with Part D
• Polypharmacy
– Overutilization
– Use of inappropriate medications
• Suboptimal prescribing
– Drug interactions
– Underutilization
• Adherence
• Prescribing patterns
Inappropriate Medication Use
1. Investigate the utility of Medicare Part D data to describe prescriber-level variation in medication use
2. Evaluate the variation in PIM use in Medicare Part D beneficiaries at the prescriber level, controlling for patient characteristics associated with getting a PIM
Design and Methods • 100% Texas Medicare claims and Part D event files for 2007
and 2008
• Enrollees 66 and older in 2008 with 12 months of A, B, and D, without HMO in 2008
• Prescribers who were physicians, with 10 or more beneficiaries per prescriber
• PIMs defined according to Beers 2003 list
• List of medications/drug classes only (did not include drug-disease combinations)
• Unable to assess over-the-counter meds
Design and Methods Variables Data Source
Patient Age, sex, race/ethnicity, state buy-‐in PDE denominator
Comorbidities (Elixhauser’s Index) 2007 carrier file and MEDPAR
Hospitalization in 2007 MEDPAR
PIM in 2008 according to Beers list PDE files
Prescriber Credentials, specialty PDE Prescriber Characteristics File
Analysis Plan Patient and prescriber characteristics associated with PIM use by patients
• Bivariate • Multivariable model for patient factors • Multilevel model for prescriber, controlling for patient level
Data Elements
Texas Medicare Part D Beneficiaries Age 66 in 2008
N = 2,261,766
12 months of A, B, and D coverage and no HMO all of 2008
N = 760,703
Had Part D claims for any drug in 2008
N = 716,930
10 or more beneficiaries per physician prescriber
N = 677580 (24,561 MD/DO)
Results: Study Flow Chart
Results: Number of Beneficiaries Per Prescriber Distribution of BID_PCT_PROVIDER & CLM_PCT_PROVIDER
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 76 80 84 880
1
2
3
4
5
6
7
8
9
Percent
BID_PCT_PROVIDERNumber of beneficiaries per prescriber
Percen
t of p
rescribe
rs
Results: Prevalence of PIM Use
• Overall, 216,364 (31.9%) of 677,580
Texas Part D beneficiaries who filled
prescriptions received a PIM in 2008
• 85% of the 24,561 prescribers prescribed
at least 1 PIM
Table 1: Characteristics of 677,580 Beneficiaries
• PIM use increased with increasing age, and differed between sexes and categories of race/ethnicity
Characteristic Category Number % Getting a PIM
Age 66-‐69 157,530 29.6
70-‐74 171,984 30.9
75-‐79 142,225 32.7
80-‐84 107,999 33.8
85+ 97,842 34.4
Sex Female 441,657 35.0
Male 235,923 26.2
Race/Ethnicity White 465,680 32.2
Black 52,611 34.2
Hispanic 139,223 31.3
Asian 16,797 22.9
Other 3,269 28.3
PIM – poten<ally inappropriate medica<on
Table 1 (cont’d): Characteristics of 677,580 Beneficiaries
Characteristic Number % Getting a PIM
State Buy-‐in in 2008 YES 206,113 35.0
NO 471,467 30.6
Hospitalization in 2007 143,741 41.5
Comorbidities Heart Failure 108,703 42.5
Uncomplicated DM 213,993 35.9
Complicated DM 81,271 40.3
Hypertension 523,380 34.2
Pulmonary Disease 150,589 38.5
PVD 133,021 39.7
Depression 70,283 42.4
Cancer 72,654 32.7
Psychoses 43,211 40.1
Neurologic Disorder 91,586 37.7
Total Number of Medication Claims (SD)
39.2 (+/-‐ 32.0) 52.2 (+/-‐ 35.9)
PIM – poten<ally inappropriate medica<on DM – Diabetes mellitus PVD – peripheral vascular disease
Table 2: PIM Use According to Number of Prescribers
• 73% of beneficiaries had >1 prescriber for all prescriptions • PIM use increased considerably with increased numbers of
unique prescribers
Number of Unique Prescribers
Number of Beneficiaries
% of Beneficiaries Getting a PIM
1 182,884 19.2
2 178,487 26.6
3 130,779 34.1
4+ 185,430 48.1
Table 3: Prescriber Characteristics and PIM Use
• 14.4% of all beneficiaries who got a prescription from an MD got a PIM from that MD
Characteristic of Prescriber Number of Prescriptions
% of Beneficiaries Getting PIMs
Credentials MD 1,753,953 14.4
DO 133,790 18.5
Specialty Gen. Internal Medicine 355,262 19.0
Family Medicine 438,185 19.3
General Practice 20,730 19.7
Internal Medicine Specialty
364,597 8.0
Geriatrics 30,767 18.7
Gynecology 27,021 11.4
Table 4: 10 Most Commonly Prescribed PIMs
PIM Name Number of
Beneficiaries
Propoxyphene 83,415
Nitrofurantoin 37,908
Clonidine 28,496
Cyclobenzaprine 27,893
Amitriptyline 19,390
Doxazosin 11,941
Amiodarone 10,906
Dicyclomine 9753
Carisoprodol 8475
Methocarbamol 7958
Table 5: Multivariable Model for Odds of PIM Use
Characteristic Odds Ratio 95% CI
Age 66-‐69 1.0 Ref
70-‐74 1.0 0.98-‐1.01
75-‐79 0.99 0.97-‐1.01
80-‐84 0.98 0.96-‐1.00
85+ 0.97 0.95-‐0.99
Gender Female 1.37 1.35-‐1.38
Male 1.0 Ref
State Buy-‐in Yes 1.11 1.09-‐1.12
No 1.0 Ref
Race/Ethnicity White 1.0 Ref
Black 1.04 1.02-‐1.07
Hispanic 0.94 0.92-‐0.95
Asian 0.74 0.71-‐0.77
Other 0.92 0.85-‐1.00
Table 5 (cont’d): Multivariable Model for Odds of PIM Use
• Adjusted for other patient factors, sex and hospitalization were still significant, and most comorbidities were not statistically or clinically significant predictors of getting a PIM.
Characteristic Odds Ratio 95% CI
Hospitalization in 2007 1.11 1.10-‐1.13
Heart Failure 0.98 0.96-‐0.99
Uncomplicated DM 0.92 0.91-‐0.94
Complicated DM 0.96 0.94-‐0.98
Hypertension 0.93 0.92-‐0.95
Pulmonary Disease 1.03 1.02-‐1.05
PVD 1.05 1.03-‐1.06
Depression 1.08 1.06-‐1.10
Cancer 0.97 0.95-‐0.99
Psychoses 0.84 0.82-‐0.86
Neurologic Disorder 0.85 0.84-‐0.87
Table 5 (cont’d): Multivariable Model for Odds of PIM Use
• Increasing number of unique prescribers remained a strong
independent predictor of PIM use in the multivariable model for
patient factors.
Number of Unique Prescribers
Odds Ratio 95% CI
1 1.0 Ref
2 1.42 1.40-‐1.44
3 1.90 1.87-‐1.94
4+ 2.92 2.87-‐2.96
!
Results: Adjusted % of Beneficiaries on PIMs Across All Prescribers Adjusted % of b
enefi
ciaries geing a PIM
Prescribers (N=10747)
1113 (10.4%)
607 (5.7%)
10th percen'le = 13.4% 90th percen'le = 30.0%
mean = 21.1%
Opportunities Provided by Part D Data
• Link with other data sources
– MCBS data
– SEER, Texas Cancer Registry
– Medicare A and B
• Compare PDP and MAPD enrollees
What can you do with Part D Data?
• Toxicities
• Cost
• Policy
• Access
• Disparities
• Adjustment for medication use