Results of a Prevalence Survey
of Macomb County, 1999

The Behavioral Risk Factor and Health Needs Survey of Macomb County was conducted as a series of Computer Assisted Telephone Interviews (CAT) by the Survey Research Division of Michigan State University's Institute for Public Policy and Social Research (IPPSR). The study was designed as a cross-sectional survey of adults and households in the county, with interviews completed for 1,003 adult residents of the county. The completion rate was 47% and the refusal rate was 19%. Interviewing began February 25, 1999 and continued through June 18, 1999.
MCHD Explanation: In an effort to obtain information that can be generalized to the entire county's population, the survey was administered to a group of adults in households representing a cross-section of the county population. The response rate was similar to other studies of this type.

IPPSR weighted the final data set to correct for unequal probabilities of selection and to maximize the representativeness of the sample findings for the Macomb County population of adults. Initially, the data set was weighted by the inverse of the selection probabilities across strata from which the cases originated and the number of adults living in the household since these would have been sources of unequal probabilities of selection into the sample. This initial weight was then adjusted with a post-stratification factor to match as closely as possible the demographic profile of households and the adult population of the county with respect to gender, age, and education based on the 1990 Census of Macomb County.
MCHD Explanation: Since it is difficult to exactly duplicate the demographics of a large community with a study of this size, the results of this study were weighted to more accurately represent the county as a whole. For example, if female respondents were under-represented in the study verses the population of the entire county, the weight placed on the responses of women can be increased while the weight of the over-represented sample of male responses can be decreased.The final working sample size was 1,003. In general, the overall margin of sampling error for a sample of this size is + 3.1% or less at the 95% confidence level. For samples of 605 which is the number of interviews with women, the margin of sampling error is + 3.9% or less, and, for the sample of 398 men, the margin of sampling error is + 4.9% or less. Nearly a third (33.1%) of the final sample was individuals older than age 55. A sample of 331 individuals has a margin of sampling error of + 5.4%. MCHD Explanation: Sampling error is a term used to describe the range of possible results when survey findings are generalized to the entire population of the county. In this case, the sampling error estimates the most accurate percentage for the result and the range within which we would expect the true value to fall 95 times out of 100.

IPPSR used SPSS to manage, transform, and analyze the data. All results presented in this report are based on the weighted sample. Throughout the analysis and the reporting, IPPSR has endeavored to code, categorize, and report results consistently with the procedures adopted by the Michigan Department of Community Health in its reporting of the Michigan Behavioral Risk Factor Survey. This is intended to enable direct comparison of the prevalence rates and findings of the Macomb BRFS to those for the state as a whole. Specifically, this means that in calculating the percentages of individuals "at risk" because of a particular health behavior or condition, individuals who refused to answer a question or who responded that they did not know the answer were excluded from the computation of the percentage. This was also done for the responses to the demographic questions. That is, the percentages reported are based on the valid responses to the questions where "don't know" and "refused" are treated as invalid responses. Exceptions to this will be noted in the text and the tables.
MCHD Explanation: MCHD Explanation: Traditionally, in survey research, non-responses and/or refusals are excluded from analysis because the rate of non-response can be biased.

Following the format of the Michigan Behavioral Risk Factor Survey reports, the percentages of respondents giving each of the answers or who fall into various response or risk categories are reported by categories of respondents based on sex, age, education, and income.
MCHD Explanation: The categories were duplicated from the State of Michigan study in order to allow for comparisons with other studies. The income categories represent that of the respondents' entire households, whereas each of the other variables refer to the characteristics of the respondents. We have not bothered to represent the breakdowns of results by race since there are so few non-white residents in the county that a random sample of all residents produces too few non-white respondents to generate stable, reliable statistics for non-white portions of the population. Consequently, all racial groups are included together. To facilitate comparisons to the statewide rates, we have grouped individuals within categories of these demographic variables and report them in the same way as is done in the Michigan Behavioral Risk Factor Survey reports with two notable exceptions. Whereas that series of reports aggregates all individuals with household incomes of $35,000 or more into a single category, we have elected to separate these individuals into two categories, those with incomes of $35,000 to $49,999 and those with incomes of $50,000 or more.

Decisions as to differences among categories of respondents are based on the test statistic 2 (Chi-square) or F (for continuous variables). For the most part, the categories will be judged to be different from each other in terms of their respective prevalence rates or frequencies if the chance of observing the amount of difference found by sampling error alone is less than one in twenty (i.e., if the probability of observing a 2 or F value as large as that found would be expected to occur in less than five percent of all samples as a result of sampling error alone when the categories did not really differ in fact). This will be noted as being a "statistically significant difference."

In the tables throughout this report, results that differ significantly across categories of demographic variables, such as gender or race, will be noted by an asterisk. When the columns in a table represent categories of a single variable such as perceived health status (i.e., excellent, good, fair, poor), the asterisk will be located behind the name of any demographic variables where differences are statistically significant (see, for example, Age in Table 2). When the columns in a table represent results for two or more variables, asterisks indicating significant differences among categories of a demographic variable will be located by the upper most percentage in the column on which the groups differ (see, for example, Perceived Poor Health by Age in Table 3). We will differentiate three different levels of statistical significance -- * p < .05 (i.e., probability is less than 5 chance in 100 of occurring by sampling error alone), ** p < .01 (i.e., probability is less than 1 chance in 100 of occurring by sampling error alone), and *** p <.001 (i.e., probability is less than 1 chance in 1,000 of occurring by sampling error alone.
MCHD Explanation: Statistical tests performed measured the differences in the mean responses of the groups defined in the preceeding paragraph. If any of the subgroups varied significantly, that difference is indicated by asterisks on the table. The presence of asterisks identifies questions where there was a significant difference observed between one or more of the response categories.

Statistical significance is partly a reflection of sample size. That is, with very large samples, almost all differences in the rates of various categories of respondents will be statistically significant even if the differences in the actual percentages are rather small. The same magnitude of difference in the rates would probably not be statistically significant if the sample sizes are relatively small. With relatively small samples, substantial differences in the rates among various categories of respondents will typically be required before they can be judged to be anything more than just sampling error. Thus, when the sample sizes are smaller, there is a greater chance of failing to note a substantively important difference among categories because the magnitude of the difference is not great enough to be statistically significant. Furthermore, in this report a large number of comparisons will be made, thus increasing the possibility that some of the statistically significant differences noted may in fact be the result of simply sampling error.

In some portions of the report, we project the actual numbers of individuals or households in the population of the county who do some particular behavior or have a particular condition. These projections are extrapolations from the sample findings to the estimated Macomb county population according to U.S. Census Bureau numbers taken from the 1990 look-up facility on the Census Bureau's Internet web page (http://venus.census.gov/cdrom/lookup). The STF1A series counts were used for demographic totals. The population projections to 1999 were extrapolated from the growth rates calculated based on the Census counts for 1990 and the Census Bureau's projected counts for 1997. The 1990 Census indicated that there were 545,747 adults (i.e., individuals 18 years of age or older) in Macomb County. The total number of individuals in the county in 1990 was 717,400 and the Census Bureau projects the number to be 784,095 in 1997 This represents an average annual growth rate of 1.278%. We used this growth rate to project the 1999 population to be 804,262 of which 615,214 were estimated to be adults. Similarly, the projected number of households in the county for 1999 is 297,076 (264,991 in 1990 x 1.278% x 9 years).

Back To Top