APPENDIX C: TECHNICAL NOTES
Rate Calculation. All West Virginia rates were generated from
mortality data gathered by the Health Statistics Center, West Virginia Bureau for Public
Health. Actual counts of firearm-related deaths from 1985-94 are provided in Table A-1.
Other counts are available from the Health Statistics Center. All United States rates from
1970-90 were calculated from mortality data found in Vital Statistics of the United
States, Volume II: Mortality, Part A for the years under consideration (14).
Provisional rates for 1994 were obtained from the monthly Vital Statistics Report (15).
For comparison purposes, all West Virginia and United States rates were age and sex
adjusted, where appropriate, using the 1990 U.S. census population distribution (16).
Caution is urged in the interpretation of some county-level and age-, race-, and
sex-specific rates due to the small numbers of events.
Population Denominators. Denominators for the calculation of
single-year firearm-related mortality rates among the West Virginia population were taken
from the census year population figures for 1970, 1980, and 1990, with interpolated
population figures used for non-census years (1975, 1985, and 1994). Census population
figures for 1990 were used as denominators for the aggregated 1985-94 rates. Table A-2
contains 1990 population distributions for the state and the nation. Additional population
figures are available from the Health Statistics Center.
Determination of Statistical Inference. Determination of
statistical inference for data in this report was based on overlapping 95% confidence
intervals. Significance testing was performed only on those rates and estimates having
numerators of sufficient size to provide valid results. (Rates based on small numerators
can include wide confidence intervals, which can render such testing too imprecise to be
useful.) The terms "more likely" and "less likely" indicate that the
difference in rates or estimates was statistically significant. A lack of comment
regarding the difference between any two rates or estimates, however, does not mean that
the difference was tested and found to be not significant.
Data Limitations. Analyses of data from the Behavioral Risk Factor
Surveillance System are subject to at least two limitations. First, telephone-based
surveys undersample populations that do not participate or cannot be reached by telephone
(17). According to the 1990 Census of Population and Housing, approximately 90%
of the housing units in West Virginia have telephones (18). Second, surveys based
on self-reported data can yield inaccurate information due to societal stigmas or support
attached to certain behaviors.
The results of the Youth Risk Behavior Survey are also based upon self-reported data,
allowing for the same data limitations as the BRFS. Research has shown, however, that
school-based surveys are more likely than non-school-based surveys to elicit candid
responses for various reasons, including the fact that students feel they are responding
as part of a group, they consider school-based surveys to be a common school practice,
and/or they felt there is more confidentiality connected with such surveys (19).
Health Statistics Center
Office of Epidemiology and Health Promotion
West Virginia Bureau for Public Health
Comments
Last updated 02/07/06