Oral Health, U.S. 2002 Annual Report
Section F: TECHNICAL NOTES
Analytic Methods
Oral Health U.S., 2002 contains predominantly descriptive statistics
including frequencies, percentages, ratios, incidence, prevalence and mortality
rates, and calculations of cost. In general, only results with statistical
significance at the 95% confidence level are discussed unless otherwise stated.
SAS, versions 6.12 and 8.0 (SAS, 1999), and SUDAAN, versions 7.5.3A and 7.6
(Shah et al., 1997), were used for statistical programming with the exception
of data from SEER (NCI, 2000). SEER data are provided with a self-contained
statistical programming package that was used for analysis of these data (NCI,
2000).
Occasionally, data from published material were used. In these cases references
are provided and statistical analytic methods vary with the source of data.
Whenever possible, nationally representative data were used. The national
surveys collecting these data generally use multistage probability samples.
Resulting analyses employ sampling weights. In these instances SUDAAN was used
to obtain accurate variances and confidence intervals. Confidence intervals are
provided in tables whenever possible. In some cases, data were obtained from
published documents and confidence intervals were not available. The
significance of trends was examined by performing a regression analysis on the
data (Korn & Graubard, 1999). The most recent data available for each indicator
were used. In a few cases an indicator was considered important enough to
include even though the most recent data were more than 5 years old. Some data
are reported for which the numbers are too low for statistical evaluation. This
was done when these data were the only source available.
Differences between population subgroups were examined using demographic
variables. These variables include age, gender, race/ethnicity, education, and
federal poverty level. The age groups used for analysis vary depending on the
medical/health implications of the indicator. In some cases age groups vary
depending on how data in individual surveys were collected. Education is
presented with three categories: less than 12th grade or 12 years of education,
12th grade or 12 years of education, and more than 12th grade or 12 years of
education. For adults, the education of the survey participant is used. For
children and adolescents, the education of the person identified as head of
household is used. The categorization of race/ethnicity depends on the
individual survey. For instance, in NHANES III, Mexican Americans were
oversampled and race/ethnicity is categorized as non-Hispanic white,
non-Hispanic black, and Mexican American. In many analyses the numbers
available for other racial/ethnic groups do not allow statistical conclusions
to be drawn. In NHANES I, the groups of white, black, and other are most
frequently used. Data from the NHIS generally use race/ethnic groups of
non-Hispanic white, non-Hispanic black, and Hispanic. Other surveys may use
white, black, and other or in some cases group white and other ethnic groups
and consider black, or black and Hispanic as separate groups. Analyses are
generally limited to the race/ethnic groups provided on public use tapes or
CD-ROMs of survey data.
Federal poverty level is determined by income thresholds that vary by family
size and the number of related children in the household who are younger than
18 years. Poverty thresholds are derived from the measure of poverty used for
calculating the number of persons living in poverty in the United States or in
states or regions. The thresholds do not vary geographically, but they are
updated annually using the Consumer Price Index. See the Census Bureau's
website (http://www.census.gov/ hhes/poverty/threshld.html) for more
information. Different indicators of poverty status have been used by different
federal agencies in the surveys they conduct. In most national surveys a
poverty income ratio is presented. The poverty income ratio is the ratio of a
family's income to the poverty threshold defined by the U.S. Census Bureau
appropriate to the family's composition. For most analyses in this report, a
cut point of 1.00 was used to differentiate those living below the federal
poverty level (less than 1.00) from those living at or above the federal
poverty level (1.00 or more). This cut point is used in NHANES I and III and in
the NHIS. MEPS defines less than 100% of the federal poverty level as negative
or poor; 100% to less than 125% as near poor; 125% to less than 200% as low
income; 200% to 400% as middle income; and 400% and above as high income. BRFSS
reports only income categories with no associated poverty level categories.
The issue of separating the effects of race/ethnicity and income was approached
by stratifying the sample on one variable and looking at the second variable
within each stratum. For instance, comparisons between racial/ethnic groups
were done separately for those living below the poverty level and for those
living at or above it. These analyses are limited by the sample sizes in each
group.
If data suggested an age dependency or if various subgroups being compared
included individuals with large differences in age distribution, analyses were
age adjusted to evaluate the possible impact of different age distributions on
the observed results. Procedures based on those described by Anderson and
Rosenburg (1998) were used. Direct age standardization applies age-specific
rates from two or more sample populations to a standard age distribution in
order to eliminate differences in the observed rates that are due to
differences in age distributions between populations. This weighted average of
the category-specific rates provides a single rate that reflects the numbers of
events that would have been expected if the populations being compared had
identical age distributions (Hennekens & Buring, 1987). Age adjustments in this
report were made using the year 2000 U.S. population age distribution (Anderson
& Rosenburg, 1998), with the exception of data from SEER. In this case the
statistical analysis program provided calculates age adjustment to the 1970
U.S. population.
REFERENCES
Anderson RN, Rosenburg HM. Age standardization of death rates: implementation
of the year 2000 standard. National Vital Statistics Reports; vol. 47 no. 3.
Hyattsville, MD: National Center for Health Statistics, 1998.
Hennekens CH, Buring JE. Epidemiology in Medicine. Boston: Little, Brown and
Company, 1987. Korn EL, Graubard BI. Analysis of Health Surveys. New York:
Wiley, 1999.
National Cancer Institute. SEER Cancer Incidence Public-Use Database CD-ROM,
1973–1997. Bethesda, MD: U.S. Department of Health and Human Services, Public
Health Service, 2000.
SAS Institute Inc. SAS/STAT User's Guide, Version 8. Cary, NC: SAS Institute,
Inc., 1999.
Shah BV, Barnwell BG, Bieler GS. SUDAAN User's Manual, Release 7.5. Research
Triangle Park, NC: Research Triangle Institute, 1997.
|
