|
|
|
2008 Atlas of Stroke Hospitalizations Among Medicare BeneficiariesAppendix B: Technical Notes and Methods
Study PopulationThe Atlas of Stroke Hospitalizations Among Medicare Beneficiaries presents stroke hospitalization data for fee-for-service Medicare beneficiaries ages 65 years and older who resided in the United States, Puerto Rico, or the U.S. Virgin Islands during 1995–2002. On average, there were 27,759,446 Medicare beneficiaries in our study population per year during 1995–2002. Nearly 60% of the beneficiaries were women and approximately 40% were men. The racial/ethnic composition of the Medicare beneficiaries during this period was 7.8% black, 1.5% Hispanic, 87.7% white, and 3% other. Hispanics may be underrepresented in this Atlas in part because of the methods used for reporting race and ethnicity (see Definitions, Race/Ethnicity, for more details). The age distribution of beneficiaries was 52.6% in the youngest age group (65–74 years), 34.8% in the middle age group (75–84 years), and 12.6% in the oldest age group (>85 years). For more information about the study population, see Section 1: National Patterns of Stroke Hospitalizations. Data SourcesMedicare Hospital Claims Data We included all hospitalizations among Medicare beneficiaries meeting our inclusion criteria for which the principal (i.e., first-listed) diagnosis was stroke, as indicated by International Classification of Diseases, 9th Revision, Clinical Modification1 (ICD-9-CM) codes 430–434 and 436–438, but excluded those for which the principal diagnosis was transient ischemic attack (ICD-9-CM code 435). The data in the MEDPAR Part A files do not distinguish between repeat hospitalizations and first-time hospitalizations. Therefore, the data presented in this Atlas refer to the number and rate of hospitalizations for stroke rather than to the number of people who were hospitalized for stroke, since some people were hospitalized more than once during the period covered in this Atlas (1995–2002). Medicare Denominator Files Area Resource File We obtained the 2002 data on the types of health care facilities (i.e., all short-term general hospitals and those with neurological services, emergency departments, and rehabilitation care services) from the 2004 Area Resource File. We obtained the 2002 data on the number of stroke specialists (i.e., neurologists and neurosurgeons) from the 2005 Area Resource File. We used the contiguity matrix from the 2002 Area Resource File to spatially smooth the stroke hospitalization data. DefinitionsSelected Comorbidities Atrial Fibrillation (427.3)
Diabetes (250)
Hypertension (401–405)
Hospital Charges Hospital Discharge Status
Race/Ethnicity This reporting practice can result in misclassification of race and ethnicity. According to 1996 data, the probability that the racial/ethnic designation on Medicare claims data is correct is 96.6% for whites and 95.5% for blacks, but only 19.4% for Hispanics.4 At the same time, the probability that a person identified as Hispanic in the Medicare data set is actually Hispanic is 98%. Together, these data suggest that Hispanics are underreported in the Medicare data sets and that this underreporting could introduce bias into the results presented here. We recognize that racial categories reflect socially distinct groups of people but not biological differences.5,6 Moreover, we recognize that each racial/ethnic group includes people of considerable diversity with regard to culture, socioeconomic status, heritage, and area of residence. Stroke Subtypes
Statistical AnalysisRationale for Spatial Smoothing Several techniques exist for spatially smoothing data; each technique has its own set of strengths and limitations.7 We chose to spatially smooth the data using a spatial moving average. Numerators (e.g., the number of stroke hospitalizations) and denominators (e.g., number of Medicare beneficiaries) for each county were combined with the numerators and denominators of immediate neighboring counties (i.e., contiguous counties), and then divided by the number of years in the study period to produce an average annual estimate. Thus, a single county’s smoothed estimate represents an average of the rates of that county and all its contiguous neighbors for the study period. As with all small area estimates of health outcomes, the county-level estimates of stroke hospitalization data presented in this Atlas are understood to be approximations of the actual geographic disparities that exist.8 Calculation of Spatially Smoothed and Age-Adjusted Stroke
Hospitalization Rates To calculate spatially smoothed and age-adjusted rates, we first summed the number of stroke hospitalizations (the numerator) in each county during 1995–2002 by age group (65–74, 75–84, and >85 years) to create summary numerators for the 8-year study period. Likewise, we summed the number of Medicare beneficiaries (the denominator) in each county during 1995–2002 by age group to create summary denominators for the 8-year study period. Next, for each age group, we combined the summary numerator of each index county with the summary numerators of its neighboring counties, and then divided this number by 8. We also combined the summary denominator of each index county with the summary denominators of its neighboring counties, and then divided this number by 8. Next, we divided each newly calculated numerator by each newly calculated denominator to produce average annual spatially smoothed stroke hospitalization rates for each age group. Finally, we used the 2000 U.S. standard population weights for ages 65 years and older to calculate age-adjusted rates. The result was average annual spatially smoothed and age-adjusted estimates of stroke hospitalization rates at the county level. We repeated these calculations to produce rates by stroke subtype for each of the three racial/ethnic groups. Two constraints were applied to the calculation of county stroke hospitalization rates. For each racial and ethnic group, a stroke hospitalization rate was not calculated for any county for which the total number of race-specific stroke hospitalizations in that county and its neighboring counties was fewer than 20 during the study period (1995–2002).9 To avoid calculating rates for counties that had no members of a particular racial/ethnic group but whose neighbors did have significant population sizes, race/ethnicity-specific rates were not calculated for a county with fewer than 10 Medicare beneficiaries of a particular racial/ethnic group during 1995–2002. Standard Population Weights First, we summed the published standard population weights for ages 65–74, 75–84, and >85.10 Second, for each age group, we divided the original weight (based on the total age distribution) by the sum of the weights for ages 65 and older. The resulting quotients are the standard population weights used in this Atlas. 2000 U.S. Standard Population Weights
Source: National Center for Health Statistics, CDC, 200110
|
Age Group (years) | Population | Weight |
---|---|---|
65–74 | 18,136,000 | 0.522501 |
75–84 | 12,315,000 | 0.354797 |
≥85 | 4,259,000 | 0.122702 |
Total ≥65 | 34,710,000 | 1.000000 |
Percentages of Stroke Hospitalizations with Selected Discharge
Outcomes
We calculated spatially smoothed percentages of stroke hospitalizations with
selected discharge outcomes at the county level by race/ethnicity for each
of the following discharge outcomes: home, skilled nursing facility, other
care facility, any care facility, died before discharge, and died within 30
days of stroke hospitalization.
To calculate spatially smoothed percentages, we first summed the number of stroke hospitalizations for each discharge outcome (the numerator) in each county during 1995–2002 to create summary numerators for the 8-year study period. Likewise, we summed the number of stroke hospitalizations (the denominator) in each county during 1995–2002 to create summary denominators for the 8-year study period.
Next, for each discharge outcome, we combined the summary numerator of each index county with the summary numerators of its neighboring counties, and then divided this number by 8. We also combined the summary denominator of each index county with the summary denominators of its neighboring counties, and then divided this number by 8. Finally, we divided each newly calculated numerator by each newly calculated denominator to produce average annual spatially smoothed estimates of stroke hospitalization percentages at the county level for each discharge outcome by race/ethnicity.
Three constraints were applied to the calculation of county stroke percentages for each discharge outcome. For each racial and ethnic group, a stroke hospitalization percentage was not calculated for any county for which the total number of race-specific stroke hospitalizations in that county and its neighboring counties was fewer than 20 during the study period (1995–2002).9 To avoid calculating percentages for counties that had no members of a particular racial/ethnic group but whose neighbors did have significant population sizes, race/ethnicity-specific percentages were not calculated for a county with fewer than 10 Medicare beneficiaries of a particular racial/ethnic group during 1995–2002. Furthermore, county percentages with a relative standard error >30% were considered unstable11 and were categorized as “insufficient data.”
Percentages of Stroke Hospitalizations with Selected Comorbidities
We calculated spatially smoothed percentages of stroke hospitalizations with
selected comorbidities at the county level by race/ethnicity for each of the
following comorbidities: hypertension, diabetes, and atrial fibrillation.
To calculate spatially smoothed percentages, we first summed the number of stroke hospitalizations for each comorbidity (the numerator) in each county during 1995–2002 to create summary numerators for the 8-year study period. Likewise, we summed the number of stroke hospitalizations (the denominator) in each county during 1995–2002 to create summary denominators for the 8-year study period. Next, for each comorbidity, we combined the summary numerator of each index county with the summary numerators of its neighboring counties, and then divided this number by 8. We also combined the summary denominator of each index county with the summary denominators of its neighboring counties, and then divided this number by 8. Finally, we divided each newly calculated numerator by each newly calculated denominator to produce average annual spatially smoothed estimates of stroke hospitalization percentages at the county level for each comorbidity by race/ethnicity.
Three constraints were applied to the calculation of county stroke percentages with selected comorbidities. For each racial and ethnic group, a stroke hospitalization percentage was not calculated for any county for which the total number of race-specific stroke hospitalizations in that county and its neighboring counties was fewer than 20 during the study period (1995–2002).9 To avoid calculating percentages for counties that had no members of a particular racial/ethnic group but whose neighbors did have significant population sizes, race/ethnicity-specific percentages were not calculated for a county with fewer than 10 Medicare beneficiaries of a particular racial/ethnic group during 1995–2002. Furthermore, county percentages with a relative standard error >30% were considered unstable11 and were categorized as “insufficient data.”
Contiguity Matrix
We used the contiguity matrix for all U.S. counties from the Area Resource
File2 database to spatially smooth stroke hospitalization data.
Counties are considered contiguous by water rights to other counties when
they both border the same body of water. Islands and peninsulas are
considered contiguous to neighboring counties on the basis of shared water
rights or accessibility.
Spatial Geometry
The geographic data used for the county-level maps in this publication came
from the Environmental Systems Research Institute’s (ESRI) ArcUSA database,
which includes the spatial geometry and characteristics of all U.S.
counties. ESRI produced this database by updating 1973 digital line-graph
data produced by the U.S. Geological Survey to reflect changes in county
boundaries through 1988. The maps were drawn on a scale of 1:2,000,000,
which is sufficiently detailed to identify major county features.
Hospitalization data were linked to county-level jurisdiction on the basis
of the modified Federal Information Processing Standard (FIPS) codes.
Map Projections
We used several different map projections to produce the county-level maps
in this publication. For the contiguous United States, we used an Albers
conic equal area projection; for Alaska, Puerto Rico, and the U.S. Virgin
Islands, we used a Miller’s Cylindrical projection; and for Hawaii, we used
geographic coordinates (latitude and longitude). Alaska, Hawaii, Puerto
Rico, and the U.S. Virgin Islands are not in proper geographic scale
relative to the continental United States on the national maps. By using a
combination of different projections and scales, we were able to present a
relatively familiar picture of these regions.
The coordinate information for the contiguous 48 states was derived from an Albers conic equal area projection with the following parameters:
Spheroid: Clarke 1866
Central Meridian: -96.0
1st Standard Parallel: 29.5
2nd Standard Parallel: 45.0
False Easting: 0.0
False Northing: 0.0
Reference Latitude: 37.5
The coordinate information for Alaska, Puerto Rico, and the U.S. Virgin Islands was projected using the Miller’s cylindrical projection with the following parameters:
Spheroid: Sphere
Central Meridian: -121.8497772217
County FIPS Code Modifications
We used Federal Information Processing Standard (FIPS)12 codes to
link county-level data from multiple data sets. However, to ensure that data
from these multiple data sets were linked accurately, we had to modify the
FIPS codes by “incorporating” cities and some entire counties into adjacent
counties. The total number of counties included in this Atlas is 3,187.
Alaska |
|||
---|---|---|---|
Original County |
Original County FIPS Code |
Incorporated into Adjacent County |
Modified FIPS Code |
Aleutians East | 2013 | Aleutian Islands | 2010 |
Aleutians West | 2016 | Aleutian Islands | 2010 |
Skagway-Hoonah-Angoon | 2232 | Skagway-Yakutat-Angoon | 2231 |
Yakutat | 2282 | Skagway-Yakutat-Angoon | 2231 |
Florida |
|||
---|---|---|---|
Original County |
Original County FIPS Code |
Incorporated into Adjacent County |
Modified FIPS Code |
Dade | 12025 | Miami-Dade | 12086 |
Hawaii |
|||
---|---|---|---|
Original County |
Original County FIPS Code |
Incorporated into Adjacent County |
Modified FIPS Code |
Kalawao | 15005 | Maui | 15009 |
Montana |
|||
---|---|---|---|
Original County |
Original County FIPS Code |
Incorporated into Adjacent County |
Modified FIPS Code |
Yellowstone National Park (Part), Montana | 30113 | Park | 30067 |
Virginia |
|||
---|---|---|---|
Independent City |
Independent City FIPS Code |
Incorporated into Adjacent County |
Modified FIPS Code |
Bedford | 51515 | Bedford | 51019 |
Bristol | 51520 | Washington | 51119 |
Buena Vista | 51530 | Rockbridge | 51163 |
Charlottesville | 51540 | Albemarle | 51003 |
Clifton Forge | 51560 | Alleghany | 51005 |
Colonial Heights | 51570 | Chesterfield | 51041 |
Covington | 51580 | Alleghany | 51005 |
Danville | 51590 | Pittsylvania | 51143 |
Emporia | 51595 | Greensville | 51081 |
Fairfax | 51600 | Fairfax | 51059 |
Falls Church | 51610 | Fairfax | 51059 |
Franklin | 51620 | Southampton | 51175 |
Fredericksburg | 51630 | Spotsylvania | 51177 |
Galax | 51640 | Grayson | 51077 |
Harrisonburg | 51660 | Rockingham | 51165 |
Hopewell | 51670 | Prince George | 51149 |
Lexington | 51678 | Rockbridge | 51163 |
Lynchburg | 51680 | Campbell | 51031 |
Manassas | 51683 | Prince William | 51153 |
Manassas Park | 51685 | Prince William | 51153 |
Martinsville | 51690 | Henry | 51089 |
Norton | 51720 | Wise | 51195 |
Petersburg | 51730 | Dinwiddie | 51053 |
Poquoson | 51735 | York | 51199 |
Radford | 51750 | Montgomery | 51121 |
Richmond | 51760 | Henrico | 51087 |
Roanoke | 51770 | Roanoke | 51161 |
Salem | 51775 | Roanoke | 51161 |
South Boston | 51780 | Halifax | 51083 |
Staunton | 51790 | Augusta | 51015 |
Suffolk | 51800 | Suffolk | 51123 |
Waynesboro | 51820 | Augusta | 51015 |
Williamsburg | 51830 | James City | 51095 |
Winchester | 51840 | Frederick | 51069 |
*Links to non–Federal organizations are provided solely as a service to our users. Links do not constitute an endorsement of any organization by CDC or the Federal Government, and none should be inferred. The CDC is not responsible for the content of the individual organization Web pages found at this link.
Back to Top |
Page last reviewed: April 9, 2008
Page last modified: April 9, 2008
Content source: Division for Heart Disease and Stroke
Prevention,
National Center for Chronic Disease Prevention and
Health Promotion
|
||||||||||||
|