Background: Ratios of age-adjusted rates between a set of geographic units and the overall area are of interest to the general public and to policy stakeholders. These ratios are correlated due to two reasons-the first being that each region is a component of the overall area and hence there is an overlap between them; and the second is that there is spatial autocorrelation between the regions. Existing methods in calculating the confidence intervals of rate ratios take into account the first source of correlation. This paper incorporates spatial autocorrelation, along with the correlation due to area overlap, into the rate ratio variance and confidence interval calculations. Results: The proposed method divides the rate ratio variances into three components, representing no correlation, overlap correlation, and spatial autocorrelation, respectively. Results applied to simulated and real cancer mortality and incidence data show that with increasing strength and scales in spatial autocorrelation, the proposed method leads to substantial improvements over the existing method. If the data do not show spatial autocorrelation, the proposed method performs as well as the existing method. Conclusions: The calculations are relatively easy to implement, and we recommend using this new method to calculate rate ratio confidence intervals in all cases.
Zhu, L., Pickle, L. W., & Pearson, J. B. (2016). Confidence intervals for rate ratios between geographic units. International Journal of Health Geographics, 15(1). https://doi.org/10.1186/s12942-016-0073-5