Data Definitions and Analysis

Data Definitions

• Confidence intervals: Confidence intervals are statistical ranges that estimate where a true population measure, such as a rate, is likely to be. In the dashboards found on Whatcom Community Health Insights, a 90% confidence interval is used, meaning we can be 90% confident that the true rate falls within the provided interval. Comparing two confidence intervals helps determine if there is a statistically significant difference in rates between two groups. For most of the data analysis on the WCHI, if the confidence intervals DO NOT overlap, the difference is considered significant. Two data sources (Healthy Youth Survey (HYS) and Behavioral Risk Factor Surveillance System (BRFSS)) provide the ability to conduct additional statistical testing; please see the analysis section for more information. 

• Crude rates: Crude rates are basic measures used to describe the occurrence of a health event, such as a disease or injury, within a population over a specified period of time. They do not account for differences in the age distribution of populations, which can affect the interpretation of the rates, especially when comparing populations with different age structures.

• Age-adjusted rates: When comparisons are made between communities or populations that have different age structures, age-adjusted rates are used. Since the risk of many diseases and health outcomes varies significantly by age, age-adjustment allows for a more accurate comparison by removing the effect of age as a confounding variable. We report an age-adjusted rate in order to compare geographies and subpopulations with differing age structures fairly.

• Suppression of small numbers: The suppression of small numbers is used in public health and epidemiology to protect the privacy and confidentiality of individuals in small or sparsely populated groups when reporting health data. Reporting very small numbers can also lead to unstable or unreliable estimates of disease rates or other health indicators. In order to avoid suppression, a common practice is to aggregate data across multiple years or subpopulation groups. Across the WCHI dashboards, an effort is made to combine multiple years of data for several sub-group analyses (e.g., Race and Ethnicity) so that as much data as possible can be displayed. When counts within a sub-group are less than 10, the data for that category is not displayed to protect privacy and confidentiality and avoid displaying unreliable estimates.
  • More information on standards for reporting data with small numbers can be found here.

Data Analysis

• Trend Analysis: Joinpoint analysis is used for to analyze trends in disease incidence or mortality rates. It helps to detect significant changes in trends over time, allowing for the identification of where shifts occur and the assessment of whether they are statistically significant. It can help to identify points where a linear trend changes direction or “joins points” within a time series dataset.

• Analysis with the Use of 90% Confidence Intervals: It is often common practice within public health to use a 95% confidence interval for an estimate for comparison between groups. For Whatcom Community Health Insights, a 90% confidence interval is used. This is advantageous when there is limited data or when the focus is on providing a more precise estimate, although it can come with a slightly higher risk of detecting a false positive (identifying a difference between groups when there actually isn’t one). For data analysis purposes, we use 90% confidence intervals to:
  • Compare Whatcom County to Washington State
  • Compare Groups in Subpopulations

• Analysis for Healthy Youth Survey (HYS) and Behavioral Risk Factor Surveillance System (BRFSS): These data come from surveys that permit statistical analysis to examine trends and make sub-group comparisons. For indicators that use data from these surveys, STATA software is used to make trend analyses, compare estimates between Whatcom County populations and the Washington State population, and detect differences between sub-groups.  Analyses are made using 90% confidence (𝛂 = 0.10). However, with the precision of using a statistical test to detect differences between sub-groups, some examples may occur where evidence of a difference is found even though there is some overlap of confidence intervals. Summary statements for sub-group analyses for HYS indicators are provided for the most recent year of data.