Transcript Frank_SAS_CMS_10172011x

RESEARCH DATA ASSISTANCE CENTER
(RESDAC)
Barbara Frank, MS, MPH
October 17, 2011
CMS SAS Day
RESDAC
The Research Data Assistance Center (ResDAC) is a
CMS contractor that provides free assistance to
academic, government and non-profit researchers
interested in using Medicare and/or Medicaid data
for their research.
ResDAC is staffed by a epidemiologists, public health
specialists, health services researchers,
biostatisticians, and health informatics specialists
from the University of Minnesota.
RESDAC ROLE
 To help CMS increase the number of researchers
skilled in accessing and using CMS databases for
studies of the Medicare and Medicaid programs
and beneficiaries.
 To provide education for researchers interested in
using and obtaining CMS data for Comparative
Effectiveness Research (CER).
RESDAC ASSISTANCE
 ResDAC web site, www.resdac.org
- Information specific to CMS data files and data requests
 What’s new-updates on CMS data release policies
 Available Data
 Data Documentation
 Requesting Data
 FAQs
 Workshops/Education
 Outreach presentations
 ResDAC Technical Register for assistance
RESDAC ASSISTANCE
 ResDAC web site, www.resdac.org
- Tools to support health services research
 Statistical links
 Health services links
 Education – Workshops
-
Introduction to the Use of Medicare data for Research
Conducting Research with Medicaid Claims Data
Intro to the Use of Medicare Part D Data for Research
Using Medicare Data in Comparative Effectiveness
Research (CER)
- Using Cost Report Data for Research
CONTACTING RESDAC ASSISTANCE DESK
 Phone
- Toll free: 888-9ResDAC (888-973-7322 )
 email
- [email protected]
 WEB
- www.resdac.org (information)
- resdac.oit.umn.edu (request assistance)
RESDAC ASSISTANCE
 ResDAC Assistance Desk functions:
- Answer questions regarding Medicare and Medicaid
data: data access and availability, record layouts,
individual variables, location of Medicare and Medicaid
program information, CMS SAS Input Statements
- Work with researchers from first inquiry to submission of
a complete request to CMS for data
- Support ResDAC website
- Tour of ResDAC website www.resdac.org
AVAILABLE CMS DATA
Types of Data for use in Comparative Effectiveness studies
 3 Types of CMS Data
- Non-Identifiable (Public Use Files)
- Limited Data Sets (LDS)
- Research Identifiable
AVAILABLE CMS DATA
Research Identifiable
 Medicare Utilization and Enrollment Data
- Enrollment File – Denominator/Part D Denominator or
CCW Beneficiary Summary File
- Utilization – Institutional






Inpatient
Outpatient
Skilled Nursing Facilities (SNF)
Home Health Agencies
Hospice
MedPAR – Stay record file containing Inpatient and SNF
stays
AVAILABLE CMS DATA
 Medicare Utilization Data
- Utilization – Non-Institutional
 Carrier
 Durable Medical Equipment (DME)
- Utilization – Part D Event Data
 Available 2006 to current
 Additional CCW Characteristic Files

Drug, Plan, Pharmacy, Provider
- Beneficiary Annual Summary File
 Contains CCW Chronic Condition flags, summary utilization
variables, demographic information
AVAILABLE CMS DATA
 Assessment Data
-
Outcome and Assessment Information Set (OASIS)
 Available July 1999 to current
-
Long Term Care Minimum Dataset (MDS)
 Available 1999 to current
-
Inpatient Rehab Facility Patient Assessment Instrument (IRF-PAI)
 Available January 2002 to current
 Medicaid Utilization & Enrollment (MAX)
- Personal Summary , Inpatient, Other Therapy, Long Term Care, and
Prescription Drug
- Available for all 50 States plus DC
AVAILABLE CMS DATA
Limited Data Sets
 Medicare Current Beneficiary Survey (MCBS)
- Rolling Panel Survey of approximately 16,000 per year
 Includes: Aged, Disabled, and Institutionalized Medicare
beneficiaries
 Source of information on socioeconomic and demographic
characteristics, health status and functioning, health care use
and expenditures, and health insurance coverage.
- Access to Care module available 1991 – 2009
 With accompanying Part A & B Claims data
- Cost and Use module available 1992 – 2008
 With accompanying Part A & B Claims data
 Part D data has been integrated into MCBS
CMS DATA TYPE
Non-Identifiable






Basic Stand-alone Public Use Files
Provider of Services File
NPI File
Physician/Supplier Procedure Summary File
Cost Reports
Some files are downloadable or low cost
GOAL FOR REST OF THIS PRESENTATION
 Review key data variables in the Medicare
administrative data available for Comparative
Effectiveness Research
BASIC ELEMENTS OF A CE STUDY
 Basic elements of CER
include:
-
Cohort identification
A “treatment”
Treatment
An outcome
Patient demographics
Measures of patient comorbidity
and severity of illness
- Potential observed
confounders
- Potential unobserved
confounders
- Methods to deal with
confounding/selection
bias
Demographics
Outcome
Observed Co-morbid Conditions
and Other Observed Potential
Confounders
Potential Unobserved
Confounders
COHORT IDENTIFICATION
Numerator and Denominator
 Two general rules are:
- all persons in the denominator must be eligible to have
events
- all persons in the numerator (events) must be eligible to
be in the denominator
 Issues in the Medicare files
- HMO Enrollees in Part A/B and D
- Part D Enrollment
DENOMINATOR INFORMATION
 A patient ID number – may be HIC
- If linking across various types of files be sure how to
identify patient across all files and time




Date of birth
Gender
Race/ethnicity
Place of residence: state, county and zip code
KEY POINTS
 All demographic information in the Medicare
claims data comes from the Enrollment Database
(EDB) maintained at CMS Data Center
 As claims are processed, the demographic
information known to CMS overwrites any
demographic information in the claim with current
information
RACE – ONE COLUMN VARIABLE; HISPANIC
ETHNICITY NOT ASKED NOR CODED SEPARATELY
 Originally, race coded as:
- white, black, other, unknown
 Effective 1994, race codes were expanded
to:
- white, black, Asian, Hispanic, Native American,
other, unknown
 New “RTI Race Code” variable is available in
the Part D Denominator/Beneficiary
Summary File
PERCENTAGE DISTRIBUTION OF MEDICARE ENROLLEES
BY RACE, 2008
ORIGINAL RACE CODE
1.86%
Asian
1.84%
Other
10.12%
Black
2.49%
Hispanic
0.43% N
American
Native
RTI RACE CODE
0.42% American
7.80% Hispanic
Indian/Alaska
Native
2.43% Asian/Pl
1.19% Other
9.77% Black
83.11%
White
77.64% NonHispanic White
SOCIOECONOMIC INFORMATION
 Denominator - “State buy-in” variable: lumped
all Medicare Savings Plan beneficiaries
(Medicaid, QMB, SLMB, QDWI, and QI) into one
variable
 New Part D variables - State Reported Dual Eligible Status: indicates which
of the Medicare Savings Plans the beneficiary is
enrolled in, if any; by month
- Low Income Subsidy (LIS) recipient: Premium
and/or copayment assistance depending on income
and assets; includes persons with higher incomes
and/or assets than those in Medicate Savings Plans
BASIC ELEMENTS OF A CE STUDY
 Basic elements of CER include
Demographics
- A “treatment”
βG
Treatment
Outcome
- An outcome
- Patient demographics
- Measures of patient coObserved Co-morbid Conditions
morbidity
and Other Observed Potential
and severity of illness
Confounders
- Potential observed
confounders
Potential Unobserved
- Potential unobserved
Confounders
confounders
- Methods to deal with
confounding/selection
bias
ELEMENTS OF CER AND HOW TO FIND THEM
IN MEDICARE ADMINISTRATIVE DATA
Why all 3 in RED?
 One researcher’s treatment may be another
researcher’s covariate
 One researcher’s outcome may be another
researcher’s covariate
 So, where do you find this information in the
Medicare data files?
WHERE TO FIND TREATMENT INFORMATION
IN MEDICARE DATA
Data file
Prescription drug
event (PDE)
Medications
Procedures
Devices
DRG codes
ICD-9 Procedure
codes
DRG codes
ICD-9 Procedure
codes
Product Service ID
MedPAR or
Inpatient
Carrier
HCPCS codes
Outpatient hospital
APC codes
HCPCS codes
DME
Home health
Agency
HCPCS codes
APC codes
HCPCS codes
HCPCS codes
Revenue Center
codes
HCPCS codes
NDC: National Drug Code
HCPCS: Healthcare Common Procedure Coding System
APC: Ambulatory Payment Classification
MEDICATION INFORMATION
 Prescription Drug Event (PDE) data file
-
Product Service ID is the variable that = NDC code
Generic name
Brand name
Strength and Drug dosage form
Days supplied
 NO therapeutic drug class – need help
 Medi-Span – Master Drug Database
 First DataBank
 Multum
MEDICATIONS – MAY ALSO BE PAID FOR AS
A PART B SERVICE
 Medicare has paid for
specific drugs under
Part B
- Generally, drugs that
are administered in
physician or other
offices, used as part of
infusion devices
- Some oral drugs used
following organ
transplant.
- Most (40% in 2001)
are oncology drugs
 Identified by HCPCS
codes starting with “J” in
DME claims file
Drug name
Erythropoietin
(anemia)
Lupron (prostate
cancer
Ipratropium bromide
(Asthma)
Zolodex (prostate
cancer)
Albuterol (Asthma)
% of Part B
drug costs in
2001
12.1%
10.4%
7.3%
6.8%
5.5%
PROCEDURES – IN-PATIENT
 Procedures performed in hospital are
incorporated into an institutional claim that
becomes an In-patient file record or MedPAR
stay record
 Identified by ICD-9 Procedure codes
- Up to 6 per claim. First listed is the “primary”
procedure
- Four digits of the form XX.XX with leading zero
BACK-UP FOR INDENTIFYING IN-PATIENT
PROCEDURES
Physician claims in Carrier file
 Surgeon will submit a clam for the procedure
that will appear in the Carrier file with “place
of service” = hospital
PROCEDURES – “OUTPATIENT”
 Services provided in an outpatient clinic or in a
physician’s office
 Defined by HCPCS – next slide - (Outpatient and
Carrier files) or ICD-9 procedure codes (Outpatient
file)
 When billed by physicians or other “noninstitutional” providers, appear in non-institutional
(i.e., Carrier) claims file
 Also, like an in-patient procedure, a physician claim
for work done in an outpatient facility appears in a
Carrier file line item with “place of service” =
hospital outpatient
HEALTHCARE COMMON PROCEDURE
CODING SYSTEM (HCPCS) CODES
 Appear in Outpatient, Home Health Agency, Carrier
(physician claims, ambulatory care center, health
departments, etc.) and Durable Medical Equipment
(DME) claims files
- CMS pays these payment requests (claims or
line items) based on the HCPCS code and its
modifiers
- HIGHEST QUALITY DATA
HCPCS: HEALTHCARE PROCEDURE CODING
SYSTEM CODES – 3 “LEVELS”
 Level 1 - 5 position numeric codes – They are CPT
(Current Procedural Terminology) codes of American
Medical Association
- 52630 Transurethral resection of the prostate
- 99201 Office or other outpatient visit for the evaluation
and management of new patient
 Level 2 - 5 position alpha-numeric codes; national
codes
- J0540 Injection, penicillin G benzathine and penicillin G
procaine, up to 1,200,000 units
 Level 3 - 5 position alpha-numeric codes beginning
with W, X, Y or Z; local codes
OUTCOMES
 Mortality
 Hospital-related
- hospitalization, re-hospitalization
 Diagnoses
 Procedures – see prior slides on locating
and describing treatments and procedures
and information
OUTCOMES - MORTALITY
 Two important fields in Denominator File/BSF
- date of death, and
- death date validation field
 Death dates are missing if the beneficiary is alive
and non-missing if they are deceased
 100% of DEATHS are validated
 96% of death DATES are validated
 Validated death dates are noted with ‘V’
 All files linkable at the beneficiary-level, so can do
survival analysis, 30-day, etc. post-admission
mortality
OUTCOMES - HOSPITAL-RELATED
 Hospitalization
- Yes/no
- Principal diagnosis gives the reason for
hospitalization
 Readmission to hospital
- Dates of admission and discharge are in MedPAR
and In-patient files
OUTCOMES - DIAGNOSES
 All Part A and Part B claims -- ICD-9-CM
diagnoses
- Institutional claims up to 10 codes
- Non-institutional claims up to 8 codes
 Inpatient, Outpatient and Skilled Nursing
Facilities also have admission diagnosis
code
OUTCOME - COSTS
(IF ALLOWED IN CER)
 Amount reimbursed by Medicare for each
service is in each file
MEASURES OF CO-MORBIDITY AND OTHER
OBSERVED CONFOUNDERS
 Charlson Co-morbidity Score, as well as
other co-morbidity scores
 Provider-related confounders
-
High versus low volume hospitals
Teaching versus non-teaching hospitals
Number of physician visits
Treatment by physician specialists
WHERE TO FIND PROVIDER-RELATED
CONFOUNDERS
 Teaching hospital
- Payment adjustment variable in MedPAR or Inpatient file: Indirect Medical Education Amount
 Physician specialty
- Line CMS Provider Specialty code variable in
Carrier file
ADDITIONAL INFORMATION ABOUT PROVIDERS
 Institutional
- Provider ID – link to Provider of Services file for
additional information
 Non-institutional
- National Provider Identification Number (NPI) link to NPI registry for location of provider and
not much else
ADDITIONAL INFORMATION ABOUT PROVIDERS
 Part D PDE File variables
- Provider ID/Pharmacy ID/CCW Encrypted
Pharmacy ID – link to Pharmacy
Characteristics file to obtain additional
information about the pharmacy
- Prescriber ID/CCW Encrypted Prescriber
ID – link to Prescriber Characteristics file
to obtain additional information about the
prescriber
INFORMATION ON OTHER POSSIBLE CONFOUNDERS IN
THE BENEFICIARY ANNUAL SUMMARY FILE (BASF)
 BASF - by calendar year; one record for each
beneficiary
 Number of events and Medicare payments by
Standard Analytic File type
 Information on 21 chronic conditions
-
Based on algorithms supported by literature
During current year of file
By July 1 of current year of file
First ever date of meeting the criteria in the algorithm
OTHER DATA SETS TO CONSIDER
 SEER-Medicare Linked Data
 Health & Retirement Survey – Medicare Linked
Dataset
 NCHS Surveys linked with Medicare Data
 CAHPS and HOS linked with Medicare Data
CONTACTING RESDAC ASSISTANCE DESK
 Phone
- Toll free: 888-9ResDAC (888-973-7322 )
 email
- [email protected]
 WEB
- www.resdac.org (information)
- resdac.oit.umn.edu (request assistance)