Spatiotemporal Bayesian analysis of Lyme disease in New York state, 1990-2000.

Mapping ordinarily increases our understanding of nontrivial spatial and temporal heterogeneities in disease rates. However, the large number of parameters required by the corresponding statistical models often complicates detailed analysis. This study investigates the feasibility of a fully Bayesian hierarchical regression approach to the problem and identifies how it outperforms two more popular methods: crude rate estimates (CRE) and empirical Bayes standardization (EBS). In particular, we apply a fully Bayesian approach to the spatiotemporal analysis of Lyme disease incidence in New York state for the period 1990-2000. These results are compared with those obtained by CRE and EBS in Chen et al. (2005). We show that the fully Bayesian regression model not only gives more reliable estimates of disease rates than the other two approaches but also allows for tractable models that can accommodate more numerous sources of variation and unknown parameters.

Epidemic and spatial dynamics of Lyme disease in New york State, 1990-2000.

Lyme disease (LD) occurrence in New York State (NYS) has not only increased over time but also spread throughout the state from the original disease focus in southeastern NYS. Few studies have investigated this epidemic and spatial dynamic in great detail. Using data from the NYS Department of Health Lyme Registry Surveillance System, we summarized epidemic and spatial characteristics of LD in NYS for the 11-yr time period from 1990 through 2000. New epidemiological trends associated with age, sex, and residential influences on LD over time were found. An empirical Bayes approach was used to produce maps of smoothed incidence at different time points to give a foundation for future state and local health funding plans and education programs.