Post on 23-Dec-2015
The Medical Expenditure Panel Survey: Data Resources to
Inform Research & Policy
Jeffrey Rhoades, Ph.D.
MEPS History
1977 National Medical Care Expenditure Survey
1987 National Medical Expenditure Survey
1996 Medical Expenditure Panel Survey (annual)
MEPS Survey Components
MEPS-HC -- Household Component
MEPS-MPC -- Medical Provider Component
MEPS-IC -- Insurance Component
Medical Expenditure Panel Survey – Household Component
Annual Survey of 14,000 households: provides national estimates of health care use, expenditures, insurance coverage, sources of payment, access to care and health care quality
Permits studies of: Distribution of expenditures and sources of
payment Role of demographics, family structure,
insurance Expenditures for specific conditions Trends over time
MEPS-Household ComponentSurvey Design
Sub-sample of respondents from the previous year’s National Health Interview Survey (NHIS), sponsored by NCHS
Representative of the civilian non-institutionalized population of the US
Collects data for 2 years of healthcare usage from each panel
5 in-person interviews over 2 ½ year period using CAPI technology
Person and family level data collected Interviews average 90 minutes with a range of
one to four hours
MEPS Panel Design: Data Reference Periods
2008 2009 2010
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
Panel 12 Round 3 Round 4 Round 5
Panel 13 Round 1 Round 2 Round 3 Round 4 Round 5
Panel 14 Round 1 Round 2 Round 3 Round 4 Round 5
Panel 15 Round 1 Round 2 Round 3
Sample Size
N = 31,262 N = 34,920 N = 31,228
N is equal to the number of people with a positive person weight on the file.
MEPS-HC Core Interview Content
Demographics Charges and Payments Health Status Conditions Utilization Employment Health Insurance
MEPS- HC Supplemental CAPI Sections
Sections asked in rounds 2 and 4: Access to care Child preventive health Satisfaction with health plans &
providersSections asked in rounds 3 and 5: Assets (round 5 only) Income Preventive Care Priority conditions
MEPS-HC Supplemental Paper Questionnaires
Diabetes Care Survey (DCS)• Given once a year to each adult identified as
having diabetes
Adult SAQ• Given once a year to each adult 18 years old
and older
Cancer SAQ– Given only in Panel 15 round 5 and Panel 16
round 3 to each person identified as having cancer
Levels of MEPS-HC Public Use Files
Person Level - detailed person information
Event Level - detailed event level information
Condition Level - detailed condition information
Job Level - detailed job information
MEPS-HC Caveats and Limitations
Sample size limitations preclude some analyses
Typically, one respondent provides data for the entire household
Household respondents may not be able to report accurately certain types of information– type of health plan– detailed event information– diagnoses
MEPS Medical Provider Component (MPC) - Purpose
Compensate for household item non-response
Accuracy and detail
Imputation source
Methodological studies
MEPS-MPC (Medical Provider Component)
Survey of medical providers linked to respondents of the HC
Collects data that household respondents cannot accurately provide, such as dates of visit, diagnosis and procedure codes, charges and payments
The Pharmacy Component (PC), a subcomponent of the MPC, collects drug detail information, including National Drug Code (NDC) and medicine name, date filled and sources and amounts of payment
The MPC is not designed to yield national estimates. It is primarily used as an imputation source to supplement household reported expenditure information.
Medical Provider Component (MEPS) Data
Collected for:– Physician Office Visits– Outpatient Department Visits– Hospital Inpatient Stays– Emergency Room Visits– Prescribed Medicines (Pharmacy Component)– Home Health Agency Care
Not collected for:– Non-physician Office Visits– Dental Visits– Home Health - Independent Providers– Other Medical Expenses
MEPS-IC Survey
Nationwide, annual survey of both private and public sector establishments
Funded by the Agency for Healthcare Research and Quality (AHRQ)
Conducted by the U.S. Census Bureau Survey data available for 1996 through
2010 (except 2007)
MEPS-IC (Insurance Component)
An independent survey of employers and unions not linked to the household survey
The sample contains information from about 44,000 establishments and supports national and state-level estimates for all 50 states.
Employer-sponsored health insurance measures:– Availability– Enrollment– Benefit and payment provisions– Cost
MEPS-IC Sample Design
Sample of private establishments drawn from the Census Bureau’s Business Register– Approximately 42,000 establishments sampled
Sample of state and local governments drawn from the Census Bureau’s Census of Governments– Approximately 2,000 governmental units
sampled (No longer collected) Sample of employers
directly linked to the MEPS Household Survey – Last year available is 2001
MEPS-IC Sample Design
Designed to make National, State, and some Metro area estimates
Designed to make year-to-year estimates
Data is Census Bureau Confidential - Public Use Files are not available
Methodology Reports available on website
Types of Information Collected
Establishment-level (location) characteristics
Health insurance plan characteristics
Firm-level (company) characteristics
Medical Expenditure Panel Survey
Medical Conditions
MEPS Condition Roster
One roster per person Cumulates medical conditions reported across MEPS interviews Source of conditions
–Following 4 sections: Condition Enumeration Priority Conditions Medical Events Disability
General File Structure
Each record represents a unique condition or procedure reported by a household respondent
Depending on the number of conditions reported, persons may be represented on the file– once– several times – not at all
Condition Enumeration
We're interested in learning about health problems that may have bothered (PERSON) {since (START DATE)/between (START DATE) and (END DATE)}. – Health problems include physical conditions,
accidents, or injuries that affect any part of the body as well as mental or emotional health conditions, such as feeling sad, blue, or anxious about something.
Condition Enumeration
Asked in every round
Has a time frame (since last interview until today)– Responses recorded verbatim and coded into 5
digit ICD9 codes– Responses directly linked to conditions roster– Chronic/priority conditions appear only once on
the roster– Acute conditions can have many records– Responses aggregated across rounds for the
annual responses
Priority Conditions
New section since Panel 12 Separate section of questionnaire Series of questions asking if “ever” had
condition– “yes/no” responses, no ICD9 coding– responses of “yes” with a current utilization
record will appear on person’s condition roster Factors used in determining priority
conditions– Prevalence– Expenditures– Policy relevance
Priority Conditions List
Heart disease Heart attack Angina High cholesterol Cancer Stroke High blood pressure Diabetes Asthma Arthritis/Joint pain Emphysema Chronic bronchitis Attention deficit disorder
Medical Events
What conditions were discovered or led (PERSON) to make this visit? PROBE: Any other condition? IF CONDITION IS ALREADY LISTED, ASK: Is this the same (NAME OF CONDITION) that we have already talked about before?– Types of visits
Inpatient Outpatient Emergency Room Office Based Home Health
Conditions associated with Prescribed Medicine purchases:– What health problem is (MEDICINE) prescribed for?
PROBE: Any other health problems?
Disability Days
Missed school or work– What are the health problems that caused
(PERSON) to miss work/school on those days? PROBE: Any other health problems?
Bed days– What are the health problems that caused
(PERSON) to spend half day or more in bed on those days? PROBE: Any other health problems?
Reporting and Recording Conditions
Respondents may report having the same condition more than once– Interviewer verifies that these are
different occurrences of the condition– each unique episode of a condition is
recorded only once person may have more than one cold in a
year each cold has a separate record
Condition Coding
Coding and Editing
– Text strings coded into fully specified ICD-9 CM codes (up to 5 digits)
– Collapsed into 3 digits to maintain confidentiality
– Approximately 10% of condition codes are collapsed further by combining 2 or more 3-digit codes
Clinical Classification System (CCS codes) - ICD-9 codes aggregated into clinically meaningful categories
Limitations: clustering of ICD-9 codes in NEC (not elsewhere classified) and one respondent provides information for the entire household
Medical Expenditure Panel Survey
Health Care Utilization and Expenditures
Health Care Utilization
MEPS household respondents asked to report all health care use for family members during reference period
Utilization is called an “event” in MEPS Event type categories:
– Office-Based Medical Provider Visits (OB)– Hospital Inpatient Stays (IP)– Outpatient Department Visits (OP)– Emergency Room Visits (ER)– Dental Visits (DN)– Prescription Medicine Purchases (RX)– Home Health Care (HH)– Other Medical Expenses (OM)
Health Care Expenditures Collected at the event level Represent payments to providers of
the health care Payments are reported by source
(e.g., out-of-pocket, private insurance, public program)
Total expenditure is the sum of payments across all sources of payment
Source of Payment Categories
Self or family Medicare Medicaid Private
insurance VA
TRICARE Other federal
gov’t State or local
gov’t Worker’s comp Other insurance
Sources of Expenditure Data
Expenditures derived from two survey components:– Household Component (HC)– Medical Provider Component (MPC)
MPC data used to replace and/or supplement household-reported expenditures
Incomplete data on expenditures for an event is imputed - no missing values
Sources of Expenditure Data by Event Type
Event type HC MPCOB: Physician yes yesOB: Non-Physician yes noIP yes yesOP yes yesER yes yesDN yes noRX no yesHH: Agency no yesHH: Paid independent yes noOM yes no
Annual Expenditure Data
Annual data cumulated across approx 2 ½ rounds of data collection
Event level files– Separate by type of service– Unique record for each reported event
Some persons have no events Some persons have multiple events
Person-level file (full year consolidated) Variables derived from event level
Event Level File Record Units
EVENT TYPE RECORD UNIT
OB visit
OP visit
ER visit
IP stay
DN visit
HH month
RX original script or refill
Flat Fees
What is a Flat Fee?– A fixed dollar amount paid for a group of
health care services– Common examples: orthodontic,
prenatal care Flat Fee ID: FFEEIDX Flat fee structure (FFevTYPE)
– Stem - Initial medical visit - expenditures– Leaf - subsequent medical visits - zero
expenditures
Zero Dollar Events
Reasons for $0 total expenditures– Leaf event in flat fee bundle from prior
year– Follow-up visit without extra charge– Free care – Bad debt
Medical Expenditure Panel Survey
DISSEMINATION OF INFORMATION AND DATA
PRODUCTS
MEPS Website www.meps.ahrq.gov
Overview of MEPS and Frequently Asked Questions (FAQs)
Staff Reports using MEPS Findings/Statistical Briefs/Chart books
Data Tables of Estimates Public Use Files (microdata) MEPSnet Interactive Query Tool Survey Methodology Reports Survey Questionnaires and Other Collection Materials Data product availability and ordering information MEPS data workshop information and schedule Mailing list, List server and e-mail for technical
assistance Data Center Information
Data User Workshops
Information will be posted on Workshops and Events section of web site
For inquiries please e-mail: Workshopinfo@ahrq.hhs.gov
MEPS Publications Statistical Briefs: Easy-to-read, concise graphical
summaries of MEPS data Research Findings and Highlights: Tables and
summaries of descriptive statistics Methodology Reports: Detailed information on MEPS
sample design and survey methods Chartbooks: Policy-sensitive topics in an accessible
question-and-answer format Working Papers: Preliminary analyses of methodological
and technical issues by AHRQ staff Research in Action: Analyses using research results from
AHRQ-sponsored studies, including MEPS data
Micro Data Files
Public Use Files (Microdata) – Available for downloading from web site (Household survey only)
Restricted Access Files (Microdata) – MEPS-HC – Available for Use at AHRQ
Data Center– MEPS-IC – Available for Use at Census
Research Data CentersFor more information go to www.ces.census.gov
Contact Information
MEPS e-mail address - MEPSPD@ahrq.gov
MEPS Information Coordinator:– (301) 427-1406
AHRQ Data Center
Provides researchers access to non-public use MEPS data (except directly identifiable information)
Located in Rockville, MD Applications/procedures on MEPS web site User fee of $300.00 includes up to 2 hours
of programming (fee waived for full-time students)
AHRQ Data Center Facilities
Secure room Terminal connected to secure LAN SAS, STATA, GAUSS, Stat Transfer,
SUDAAN, Limdep, EQS software available, and others upon request
Limited staff support by people who know:– the data– the confidentiality issues– the software
ADC Guidelines
Researcher may bring data in, but not out Researcher has access only to data
needed for approved project All tabular data will be reviewed for
confidentiality before release from Center Only approved tables can leave the
Center Center will store data files, foreign merge
files, and all outputs needed for replication
ADC Limited Remote Access
Once you have an established data center project, and have worked on site to develop and debug programs, jobs may be submitted to our Data Center Supervisor to run. Output will be reviewed for confidentiality and mailed to you.
ADC Application Procedures
Application procedures are on the MEPS web site
Submit proposal to data center coordinator
Review within 1 week for feasibility, and data availability
Institutional Review Board (IRB) review required
Data Center Questions
Contact Data Center Administrator by e-mail at: CFACTDC@AHRQ.HHS.GOV
Census Bureau Research Data Centers (RDC)
Access to MEPS-IC data files and other files (including MEPS-HC)
All work takes place at a RDC operated by the Census Bureau’s Center for Economic Studies – University of Washington
(coming Summer 2012)– Texas (coming Fall 2012)
– Atlanta (Federal Reserve Bank)
– Boston (NBER)– UC-Berkeley – UCLA – Stanford– Washington DC (Census
HQ)– Chicago (Federal Reserve
Bank)– University of Michigan– University of Minnesota– Baruch School of Public
Affairs (New York City)– Cornell University– Duke University – Research Triangle (RTI)
Medical Expenditure Panel Survey
MEPS Web Analytical Tools
MEPS Web Analytical Tools
The MEPS web site contains two web analytical tools:
MEPSnet Query Tools
Customizable Summary Data Tables
MEPSnet Query Tools
MEPSnet is a collection of analytical tools offering online capability to generate MEPS estimates. The tools are divided into two sections:
MEPSnet/Household Component MEPSnet/HC
MEPSnet/Insurance Component MEPSnet/IC
Customizable Summary Data Tables
The following MEPS Summary Data Tables are customizable:
All of the Expenditures by Health Care Service Tables
All of the Expenditures by Medical Condition Tables
All of the Quality of Care Tables
Only Table 1, Usual Source of Health Care and Selected Population Characteristics, United States from the Access to Care Tables
Medical Expenditure Panel Survey
Considerations for Developing an Analytical File
Major Areas of Health Research Topics Using MEPS Data
Access Use Expenditures Health insurance Health status and conditions QualityFirst order of business is to define goals of analysis
using MEPS data as clearly as possible!
Sources of Useful Information
MEPS-HC Questionnaires Public Use File Documentation
– General information about MEPS– File-specific general information– File specific variable information
Person-level or Family-level Condition-level or Event-level MEPS supplement questions
Public Use File Codebooks– Good source of overview information– Formatted frequencies for all variables on file
Both weighted and unweighted
Variable Naming Conventions
Edited Variables end in an “X”– For example: RACEX
Names of year specific variables use last two digits of year– For example: TOTEXP09, PERWT09F, AGE09X
For round specific variables, round designation is indicated at the end of the variable or immediately before the “X” in the case of edited variables– For example: AGE31X, AGE42X, AGE53X– Certain questions or instrument sections are only asked
in certain rounds, e.g. the Self-Administered Questionnaire in rounds 2 and 4
General Tips
Clearly define research objectives Read the documentation Subset to only the variables you need for
your analysis Do not subset to specific populations prior
to running statistical analyses Compare program output with codebooks Use the correct weight, stratum and psu
variables (including supplement weights) Read the documentation!