Technical notes

How data for King County were analyzed:

Each indicator was examined for the entire King County population and by the following demographics: age, gender, race, and (depending on data source) income or poverty level. Data are also presented by the following areas of residence: Region and Health Planning Area (see definitions below). Not all demographic and residence data were available for all indicators.

For the population as a whole and for each sub-group, the most recent 10-year and 5-year time trends are shown. The most recent year’s rate and number of people affected are also shown for each indicator. To create stable comparisons, differences between the demographic groups and for Region and Health Planning Area are presented using an average of the most recent 5 years of data.

How data for comparison areas outside of King County were analyzed:

Community Health Indicators presents King County health measures in the context of Washington State and 5 demographically similar counties in the United States. The demographically similar comparison counties are:

Demographically similar counties were chosen with the following criteria, based on data from the 2007 American Community Survey, U.S. Census Bureau:

Use of rates: Age-specific, age-adjusted and other rates:

Age-specific and Age-adjusted rates: A rate in this report is usually expressed as the number of events per 100,000 population per year. When the rate applies to a specific age group (e.g., ages 15-24), it is called the age-specific rate. Age-specific rates present the actual magnitude of an event within a population or age group. When comparing rates between populations, it is useful to calculate a rate which is not affected by differences in the age composition of the populations. For example, if one population has a higher death rate and more older people, it will not be easy to determine if the non-age-related risk is truly higher or just reflects the higher death rate among older people. The age-adjusted rate is a rate that mathematically removes the effect of the age composition. By convention, the rate is adjusted to the age distribution of the estimated 2000 U.S. population. In this report, age-adjusted rates per 100,000 are presented unless otherwise noted.

Other rates: Infant mortality, maternal smoking, and other maternal/child health measures are calculated with live births as the denominator and presented as a rate per 1,000 live births (infant mortality) or percent of births.

Prevalence rates from the Behavioral Risk Factor Surveillance Survey are expressed as percent of the population. These other rates are not age-adjusted.

Are areas and sub-populations different? Use of Confidence Intervals.

When comparing rates between different groups, the “95% confidence interval” or margin of error is given for each rate or percent to assess how much the rate is likely to vary due to chance. The margin of error is calculated for all rates based on more than five events, as recommended by the Washington State Department of Health (see www.doh.wa.gov/Data/Guidelines/SmallNumbers.htm#Guidelines and www.doh.wa.gov/Data/Guidelines/ConfIntguide.htm). When comparing two rates, if the confidence intervals do not overlap, the difference in the rates is considered “statistically significant,” that is, the difference between the rates is unlikely to be the result of chance or random variation. In most cases, when confidence intervals overlap, random variation cannot be ruled out as a cause of the difference.

For Community Health Indicators measures from the U.S. Census Bureau’s American Community Survey (ACS) tables, 90% confidence intervals are used since that is the level of error estimation provided by the Census Bureau. In some cases, the American Community Survey data presented here are grouped differently than those presented in Census Bureau tables (for example, age groups have been collapsed). In these situations, margins of error have been calculated using the ACS Statistical Calculator designed by the New York State Data Center's staff at Empire State Development, based on equations provided in the U.S. Census Bureau's 2006 ACS Accuracy statement (www.census.gov/acs/www/Downloads/ACS/accuracy2006.pdf). The calculator is located on the internet at: www.demography.state.mn.us/documents/StatisticalCalculationsMenu.xls

Has there been a trend over time? Use of statistical tests.

Only trends that are statistically significant are shown with up or down arrows in the indicator spreadsheets. The probability that a trend over time is simply due to random variation is calculated using objective statistical tests. If the probability that a trend is due to random variation is less than 5% (i.e., p<0.05), a trend is considered statistically significant. (For rates, the test for trend is found in the following citation: Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med 2000;19:335-51 (correction: 2001;20:655). For percents from surveys, trend over time is assessed with the svylogit command in Stata (StataCorp. 2003. Stata Statistical Software: Release 8.0. College Station , TX . Stata Corporation.))

Showing trends over time: Use of rolling averages.

Time trend graphs for all sub-populations in this report are shown as rolling averages to help the reader observe the time trend by smoothing the trend line. For relatively small populations, small changes in the number of events will cause the rate to fluctuate substantially and create jagged lines of statistical “noise”. To help stabilize the rate and observe the time trend of an event, rates are sometimes aggregated into “rolled” averages, such as in 3- or 5-year intervals, across the total observed period. For example, for events for years 1998 to 2007, the rates are reported as three-year rolling averages: 1998-2000, 1999-2001...2005-2007.

Showing health disparities: Poverty, income and race

Where individual measures of income or poverty level were not available, we looked at health outcomes in three types of neighborhoods:

High poverty neighborhoods, where more than 20% of the population lives below the Federal Poverty Level (FPL)
Medium poverty neighborhoods, where between 5% and 19% of the population lives below the FPL
Low poverty neighborhoods, where less than 5% of the population lives below the FPL.

The high-poverty area follows the definition of a Federal Poverty Area. The 5% limit for low-poverty areas was chosen to create a group markedly different from Federal Poverty Areas, and thus sensitive to differences in health outcomes that may be the result of socio-economic differences, while maintaining enough tracts in each group for robust comparisons.

For area-based measures of poverty, a census tract is considered a neighborhood. Data sources where census tract information are not available use zip codes to designate the neighborhood. To construct these poverty areas, we rank-ordered King County neighborhoods by percent living below the poverty level in 1999 and divided them into the three groups. We calculated rates, confidence intervals and a statistical test for trend over time in each of the three poverty areas.

Where an individual income measure was available, we looked at five levels of annual household income. The five levels are: less than $15,000, $15,000 to $24,999, $25,000 to $34,999, $35,000 to $49,999 and $50,000 and over. Individual data on poverty level were not available.

Presenting data by race/ethnicity:

Most researchers believe that race/ethnicity is a marker for complex social, economic and political factors that are important influences on community and individual health, and that differences in rates of most diseases and injuries are not due to biologic or genetic factors. Many people and communities of color in this country have experienced social and economic discrimination and other forms of racism, which can negatively affect the health of these individuals and their communities. We continue to examine and present data by race/ethnicity because we believe that it is important to understand which racial/ethnic groups are disproportionately affected by significant health issues. We hope this understanding will lead to strategies that address these issues, as well as the social and economic inequities which underlie them.

Analyses by race are presented by single-race categories to allow trend over time comparisons. When data are available, analyses using the newer multiple-race categories are also presented. A full discussion of how race was classified is beyond the scope of this document. Please contact data.request@kingcounty.gov if you have questions or want more information.

Geographic Areas: Health Planning Areas and Regions

In 2005, Public Health created revised Health Planning Area (HPA) boundaries to be as consistent as possible with current and anticipated suburban city boundaries. For Seattle, HPAs were created in consultation with the City of Seattle s Department of Neighborhoods.

HPAs were created from smaller foundation geographic units. For the most precise HPAs, block groups were aggregated to create the new HPAs. A ZIP code-based grouping is used where health outcomes by block group are not available.

Below is a list of HPAs created from aggregated census block groups. ZIP code HPAs vary slightly from this list. Not all data are available by HPA:

• Auburn
• Ballard
• Beacon/Georgetown/South Park
• Bellevue
• Bothell/Northshore
• Burien
• Capitol Hill
• Cascade-Fairwood
• Central Seattle
• Covington/Maple Valley
• Delridge
• Des Moines/Normandy Park
• Downtown/First Hill
• Federal Way
• Fremont/Greenlake
• Issaquah/Sammamish
• Kent
• Kirkland
• Mercer Island/Point Cities
• NE Seattle
• North Seattle
• NW Seattle
• Queen Anne/Magnolia
• Redmond/Union Hill
• Renton
• Riverview/Lower Valley
• SE County
• SE Seattle
• Shoreline
• Tukwila/SeaTac
• Upper Snoqualmie Valley
• Vashon Island
• West Seattle
• White Center/Boulevard Park

Regions

Data are presented also by Regions, which are aggregations of Health Planning Areas. There are four regions: Seattle, North, East and South. These Regions are equivalent to those used in the report Communities Count: Social and Health Indicators for King County. A list of block groups, HPAs, ZIP codes and cities that are included in the four Regions is available on request to data.request@kingcounty.gov.