Recurring Survey Data Sets

Download Report

Transcript Recurring Survey Data Sets

Recurring Survey Data Sets
Christopher McCarty
PHC6716
June 29, 2011
Surveys
•
•
•
•
•
•
Behavioral Risk Factor Surveillance System (BRFSS)
National Health Interview Survey (NHIS)
National Immunization Survey (NIS)
National Health Care Surveys (NHCS)
Medicare Health Outcomes Survey (HOS)
National Health and Nutrition Examination Survey
(NHANES)
• Medical Expenditure Panel Survey (MEPS)
• Current Population Survey (CPS)
• American Community Survey (ACS)
Behavioral Risk Factor
Surveillance Survey (BRFSS)
Overview
• Sponsored by the Centers for Disease Control (CDC) and
each state or territory
• Link: http://www.cdc.gov/BRFSS/
• Established in 1984
– 1984-1993 monthly data on 15 states
– 1994-present monthly data on all states, Washington D.C.,
Puerto Rico, Guam and U.S. Virgin Islands
Purpose
• Core questions collect information about:
–
–
–
–
Health risk behaviors
Preventive health practices
Health care access
Core module averages 18 minutes
• States use this information to:
– Identify emerging health problems
• Example – Hurricane preparedness after 2004
– Establish and track health objectives
– Develop and evaluate public health programs
– State modules add additional 5-10 miunutes
Data Collection
•
CDC does sample generation (Marketing Systems Group) technical oversight and data analysis
•
State Departments of Health oversee actual data collection (five year contract)
– In 2010, 9 health departments collected the data in-house and 45 contracted
universities or commercial firms
– Interview is presented as though from the state’s health department with assistance
from CDC
•
Random digit dial of landline U.S. households with cell phone supplement for selected states
collected monthly
•
Interviewer elicits names of all adult males in order of age, then adult females in order of age
and randomly selects one
•
Interviews done in English and in Spanish
•
More than 350,000 interviews are done each year, with a minimum of 4,000 per state
•
46 states have disproportionate strata while 8 have one stratum
Sample size
BRFSS Response Rates
Funding
• Minimum funding of 4,000 completes per state
• For Florida maximum funding for last contract was $4
million over three years
– 30,946 (2010), 11,040 (2009), 9,953 (2008) – Total 51,939
– $77/complete (We charged $75/complete for cell complete)
– Approximate cost of 2010 BRFSS $32,726,001
• State Added Questions
– 1999 Arkansas charged $0.32 per question per respondent
– 2010 Virginia charged $3,000 per question
National Health Interview Survey
(NHIS)
Overview
•
Started in 1957 and overseen by CDC
•
Cross-sectional face-to-face and telephone survey conducted by the Census Bureau using
CAPI
– 400 interviewers in 12 regions
•
Covers non-institutionalized population
•
People excluded from the survey
– Patients in long-term care facilities
– Persons on active duty in the Armed forces
– Citizens living in foreign countries
•
Conducted in English and Spanish
•
Interview covers all household members rather than a single respondent
•
Interview averages one hour
– About 20 minutes are supplemental questions
– Most supplements are from federal agencies and NIH institutes
•
Cost – Who knows?
Sampling
•
Multistage area probability design
– First stage- 428 primary sampling units (drawn from 1,900 geographically
defined primary sampling units (PSUs)
– PSU is a county, cluster of counties, or a metropolitan area
•
Within PSU- two segments
– Area segments of 8 to 12 addresses
– Permit area segments (housing units built after the 1990 census) identified
using building permits
• Permit areas consist of 4 addresses
• Occupied households in the sample addresses, within each segment are
targeted
•
Sample consists of approximately 7,000 segments
•
35,000-40,000 households interviewed
•
Data about 75,000-100,000 people
Data collection
• Three modules
• Basic (core) module
– Family core- everyone in the family included, all
adult members answer for themselves
– Adult sample core- one adult from the household
– Child sample core-one child from the household
with adult as proxy
• Periodic module
• Supplemental module
• Data available at
http://www.cdc.gov/nchs/nhis/nhis_questionnaires.htm
Sample Size
NHIS Response Rates
How does NHIS differ from BRFSS?
• Face-to-face versus telephone
• Smaller sample
• Less coverage (only about 25% of counties)
• Higher response rate
• Longer interview
• NHIS currently used as benchmark to gauge effects of cell
phones
National Immunization Survey
(NIS)
Overview
• Sponsored by the CDC
• Conducted by National Opinion Research Center (NORC)
• Purpose: Monitor vaccinations among children in all 50
states, Washington, DC, and 27 urban areas
– Goal is 90% of all two-years-olds with all vacciniations
• Targets:
– Children 19-35 months
– Teens 13-17 years old
Data Collection
•
List assisted RDD landline survey of households with pre-notification letter when
possible
–
–
–
–
•
Screener for children
Respondent is adult most familiar with child immunization
Roster of all eligible children
Parents are asked to retrieve shot records
This is followed by mail survey to health care provider if parents consent
– Parent reports are checked against provider records to improve accuracy
– Average of 1.3 providers per child
•
In 2000 there were 10 million calls made to identify 35,000 children
•
Survey is administered quarterly but reported annually
•
Data available at http://www.cdc.gov/nis/datafiles.htm
Sample Size and Response Rates
• In 2009 there were 24,068 completes out of
28,913 eligible households (RR 83.2%)
• 24,809 completed child interviews
• 17,053 completed child and provider
interviews (68.7%)
NIS and SLAITS
• State and Local Area Integrated Telephone Survey
• Designed to collect data at state and local level
• RDD telephone survey
• NIS uses the same sampling frame as NIS
National Health Care Surveys
(NHCS)
Overview
• Set of provider surveys covering different health care
settings
• Sponsored by National Center for Health Statistics/CDC
• Data available at
http://www.cdc.gov/nchs/dhcs/dhcs_products.htm
NCHS Surveys
National Ambulatory Medical Care Survey (NAMCS)
• Purpose is to monitor types of and quality of medical visit
– Necessary with electronic records?
• Annual since 1989
• Records based survey
• Target are non-federally employed physicians involved in direct patient
care
• Census Bureau oversight of data collected by physicians and their staff
• Data available at :
http://www.cdc.gov/nchs/about/major/ahcd/ahcd1.htm#Micro-data
Sampling
• 112 Primary Sampling Units (PSUs)
• 3,000 Physicians divided into 15 specialties
• Physicians randomly assigned reporting week
• Each reports on 30 patient visits
• About 25,000 visits covered (units are visits)
• In 2004 response rate was 64.7 percent
• Attempt to use incentives in 2002 resulted in no difference between
control, gift and monetary incentive
NAMCS Response rate
(Source: Overview of the NAMCS
and NHAMCS, Linda McCaig and David Woodwell)
75
Percent
70
65
60
55
89 90 91 92 93 94 95 96 97 98 99 '00 '01 '02
Year
26
NCHS Surveys
National Hospital Ambulatory Medical Care Survey
(NHAMCS)
• Purpose is to monitor medical visits to emergency
departments and outpatient departments
• Records based survey
• Annual since 1992
• Target sample size:
– 400 Emergency Departments and 37,000 visits
– 250 Outpatient Departments and 35,000 visits
• Departments monitor records over four week period
NHAMCS Response rates
(Source: Overview of the NAMCS
and NHAMCS, Linda McCaig and David Woodwell)
100
ED
Percent
90
OPD
80
70
60
50
92
93
94
95
96
97
98
99
'00
'01
'02
Year
28
Other NCHS Surveys
• These surveys are not conducted annually
– National Survey of Ambulatory Surgery (NSAS)
National Hospital Discharge Survey (NHDS)
– National Nursing Home Survey (NNHS)
– National Home and Hospice Care Survey (NHHCS)
– National Survey of Residential Care Facilities
(NSRCF)
Medicare Health Outcomes
Survey (HOS)
Overview
•
Developed by Centers for Medicare and Medicaid Services (CMS) as a way to monitor
managed care, particularly for Medicare beneficiaries
•
In response to increasing enrollment of Medicare beneficiaries in managed care in the 1990s
•
CMS worked with National Committee for Quality Assurance (NCQA) to adapt Healthcare
Effectiveness Data and Information Set (HEDIS®) measures to Medicare
•
Goal is to monitor health issues with Medicare beneficiaries to make managed care plan
adjustments
•
Survey work is contracted by NCQA with independent vendor
•
Use NCQA protocol, a series of mail surveys followed by telephone conducted annually
•
Respondents pulled using NCQA protocol and oversight
•
Data available at http://www.hosonline.org/Content/DataFiles.aspx
Data Collection
Year
2011
2010
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
Baseline
Cohort 14
Cohort 13
Cohort 12
Cohort 11
Cohort 10
Cohort 9
Cohort 8
Cohort 7
Cohort 6
Cohort 5
Cohort 4
Cohort 3
Cohort 2
Cohort 1
Followup
Cohort 12
Cohort 11
Cohort 10
Cohort 9
Cohort 8
Cohort 7
Cohort 6
Cohort 5
Cohort 4
Cohort 3
Cohort 2
Cohort 1
• Baseline conducted annually with
two-year follow-up
• From 1998-2006 target was 1,000
respondents for each Medicare
Advantage Plan
• This was increased to 1,200 in
2007
• From 1998-2008 respondent had
to be a member for six months
(HEDIS and CAHPS)
• 2009-present the six month
requirement was waived
Sample Size
Response Rates
National Health and Nutrition
Examination Survey (NHANES)
Overview
• Purpose is to assess health and nutrition of adults and children
– This is the source for many health benchmarks
•
•
•
•
Growth charts
Cholesterol
Glucose levels
BMI charts
• Combines interview with physical exam
• Conducted annually since 1999, and sporadically since the 1960s
• Data available at
http://www.cdc.gov/nchs/nhanes/nhanes_questionnaires.htm
Data Collection
• Mobile health data collection centers
• Interview team consists of physician, nurses, dietician, interviewer
• Blood samples taken except for very young
• Target is 5,000 respondents a year over a two year data cycle, about
15 PSUs each year
• Pre-notification letter and local media ads (e.g. newspaper and
radio)
• Uses Current Population Survey total to calculate weights since
sampling is clustered and relatively small
Sample Size
Response Rates
Medical Expenditure Panel
Survey (MEPS)
Overview
•
Sponsored by the Agency for Healthcare and Research Quality (AHRQ)
•
Purpose is to assess the types of health services Americans pay for, how much they
cost and how they are paid
•
Unique because it includes both a household and an insurance component
– Household component includes data from respondent in household and medical provider
– Insurance component is a survey of insurance providers
•
Conducted annually since 1996
•
Other expenditure surveys conducted sporadically before then since 1977
•
Data available at
http://www.meps.ahrq.gov/mepsweb/data_stats/download_data_files.jsp
Data Collection for Household Survey
• To reduce costs MEPS uses NHIS respondents as the sample
frame
• Mode is Face-to-face and phone
• Overlapping panel design
• Each Two-year panel consists of five rounds over 2.5 years
• Each round refers to a specific reference period
• Survey covers all people in household
Sample Sizes
Response Rate
(Combines NHIS response rate with MEPS panel response rate)
Current Population Survey (CPS)
Overview
•
Joint effort between Census Bureau and Bureau of Labor Statistics (BLS)
•
Fieldwork conducted by Census Bureau
•
Purpose is to measure unemployment and labor participation U.S. population over age 15
•
Covers all 50 states and Washington DC
•
Also oversamples to create reliable estimates for some sub-state regions and urban areas
•
Conducted monthly since 1940
•
Given its size and frequency the CPS is often used in survey power calculations and weighting
between Census years
•
Data available at http://www.bls.gov/cps/data.htm
Data Collection
• Until 2001 target was 50,000 households each month
• In 2001 was increased to 60,000 households
• Mixed mode Face-to-face and telephone
• Sample size is based on achieving accuracy on unemployment rate
(tolerable margin of error)
• Divides country into PSUs with at least 7,500 in each
• PSUs are aggregated into strata based on demographic characteristics
• 754 PSUs are selected out of 2,025
Data Collection
(Continued)
• PSUs for all 150 MSAs are sampled, remainder are divided into units of 4560 housing units
• From PSUs a sample of Ultimate Sampling Units (USUs) of about four
households are selected
• Uses Census addresses supplemented with Building Permits
• Covers all people in the household over age 15
• 4-8-4 sampling strategy increases continuity and reduces respondent
burden:
– Households are in the sample frame for 4 months
– Out for 8
– In again for 4 months then out permanently
• Respondent is person who owns or rents unit
Response rate and sample size
• 7.5 percent typically not interviewed
– 92.5% response rate
American Community Survey
(ACS)
Overview
• Purpose is to provide much of the information
contained on the Census long form
• Data for small area analyses, particularly
between Census years
• The ACS is now used more than CPS to do
power analyses and weighting between
Census years
Data Collection
• Conducted by the U.S. Census
• Both telephone and face-to-face follow-up
• Beginning in 2010 the ACS aggregated 5 year
intervals to create small area estimates
Sampling
• Sampling occurs in two phases
– First phase assigns blocks to five strata
• Approximately 3 million addresses sampled each year
since 2005
– Prototype surveys conducted from 2000-2005
• Includes group quarters
• About 250,000 mail surveys are sent out each month
Sample Size
Response Rate