Observed, calculated from FoodNet Reports 1996 to 1999 based on number of cases with known age; LCL, lower 95% confidence limit; UCL, upper 95% confidence limit.

Observed, calculated from FoodNet Reports 1996 to 1999 based on number of cases with known age; LCL, lower 95% confidence limit; UCL, upper 95% confidence limit.

7.7). Of the four variables considered in this study (age, gender, state, year), the influence of gender appears to be the least important.

Analytical results: The first analytical results are incidence rates for each surveillance year estimated from the age-stratified data (Table 7.4). The second column of Table 7.4 presents 'observed' crude rates, i.e. rates that were calculated based on the number of reported cases of known age and the population in catchment areas as reported in each FoodNet report for the years 1996 through 1999. (These figures differ from those reported in Table 7.1, where rates for only the five original sites are reported and cases with unknown age also are included.) 'Estimated' crude rates were determined using a model specifying only year as the explanatory variable, and can be directly compared to the observed rates. As one would expect, the point estimates are equal. However, the analytical approach has advantages in that it delivers confidence limits (columns 4 and 5 of Table 7.4), and allows one to formally compare the single point estimates. Statistical comparison of Salmonella incidence by year indicated no significant difference (p > 0.119).

By fitting a model with all three main effects (age, state, and year), 'age-state adjusted rates' were obtained (columns 6-8 of Table 7.4). While these rates are abstract (they can no longer be compared with the observed crude rates), the adjusted rates reflect the specific impact of time since age and state effects are controlled. The adjustment leads to minor rectification of the previous conclusions - the Salmonella rates for 1998 become lower than those of 1996 and 1999 (p = 0.012 and p = 0.019, respectively), but the other contrasts are statistically indistinguishable (p > 0.064).

Age: The incidence rate ratios for the univariate and multivariate analyses of the age-stratified data set are reported in Table 7.5. In these tables, the reference levels of each variable are an age between 20 and 29 years, site Oregon, and surveillance year 1999. (Note that the incidence rate ratio for the year effect from the univariate analyses reported in Table 7.5 is equivalent to the ratio of the rates listed in Table 7.4.) Where the confidence interval does not include the

Table 7.5 Incidence rate ratio for Salmonella (Poisson model with generalized estimating equations, independent correlation structure)

Parameter Univariate analysis Multivariate analysis

Parameter Univariate analysis Multivariate analysis

Table 7.5 Incidence rate ratio for Salmonella (Poisson model with generalized estimating equations, independent correlation structure)

Was this article helpful? |

## Post a comment