Exploring Spatiotemporal Patterns in County-Level Incidence and Reporting of Lyme Disease in the Northeastern United States, 1990-2000

Content Provider	Semantic Scholar
Author	Waller, Lance A.
Copyright Year	2004
Abstract	We present an exploratory analysis of reported county-specific incidence of Lyme disease in the northeastern United States for the years 1990-2000. We briefly review the disease ecology of Lyme disease and the use of risk maps to describe local incidence as estimates of local risk of disease. We place the relevant elements of local environmental and ecological variables, local disease incidence, and (importantly) local disease reporting in a conceptual context to frame our analysis. We then apply hierarchical linear models of increasing complexity to summarize observed patterns in reported incidence, borrowing information across counties to improve local precision. We find areas of increasing incidence in the central northeastern Atlantic coast counties, increasing incidence branching to the north and west, and an area of fairly stable and slightly decreasing reported incidence in western New York. 1. Background Lyme disease is a zoonosis caused by the spirochete Borrelia burgdorferi, which cycles between tick vectors and dozens of species of vertebrate hosts. The first medical description of Lyme disease appears in Steere et al. (1977), although earlier reports describe incident cases of erythema migrans, the localized rash serving as a primary symptom of Lyme disease (Scrimenti 1970). The United States Centers for Disease Control and Prevention (CDC) began systematic surveillance for the disease in 1982 (Schmid et al 1985), requiring (but not necessarily enforcing) reporting of newly diagnosed cases by local health departments and health-care providers. Surveillance data remained imprecise until the Council of State and Territorial Epidemiologists approved a standardized case definition in 1990, with nation-wide implementation in 1991 (Orloski et al. 2000). Currently, Lyme disease is one of the most common vector-borne disease in the United States with almost 90,000 cases reported between 1992 and 1998 (Orloski et al. 2000). Although cases have been reported from 49 states and the District of Colombia (Orloski et al. 2000), the bulk of reported cases occur in the northeastern US (CT, DE, DC, ME, MD, MA, NH, NJ, NY, PA, RI, and VT) with a second concentration in the midwestern US (MI, MN, and WI). The ecology of Lyme disease depends on the life cycles and interactions between the causative agent, the vector, and the various hosts. Larval and nymphal tick vectors (the black-legged tick, Ixodes scapularis, in the eastern US) feed primarily on smaller vertebrates (Lane et al. 1991, Barbour and Fish 1993, Mather 1993). White-footed mice (Peromyscus leucopus) play a crucial role as both a preferred immature tick host and principal reservoir of Borrelia burgdorferi (Anderson and Magnarelli 1984, Donahue et al. 1997, Levine et al. 1985, Mather et al. 1989, Fish 1993, Mather and Ginsberg 1994, Keirans et al. 1996). Adult ticks feed primarily on white-tailed deer (Odocoileus virginianus) (Piesman and Spielman 1979, Wilson et al. 1990). Each generation of ticks must become infected via these vertebrate hosts since there is no transovarial transmission between egg-laying females and their offspring (Piesman et al. 1986, Patrican 1997). Once infected, ticks may infect a human host during a blood meal (Barbour and Fish 1993). There can be a great deal of spatial variability in Lyme disease incidence. The tick vectors are patchily distributed both regionally and locally, regardless of infection status (Kitron and Kazmierczak 1997, Wilson 1998). In particular, presence of I. scapularis is positively associated with sandy soils, woody or shrubby vegetation, and presence of deer (Ginsberg and Ewing 1989, Kitron et al. 1991, Kitron et al. 1992, Glass et al. 1994, Glass et al. 1995, Duffy et al. 1994). Countering this spatial variability in vector presence is the tendency of tick loads on mice to remain relatively constant even in the face of substantial variation in the densities of both mice and questing ticks (Goodwin et al. 2001). Host abundance and community composition also can vary dramatically in both space and time (van Buskirk and Ostfeld 1995, Giardina et al. 2000). Spatial variability of Lyme disease hosts and vectors suggests construction of a risk map, i.e., a map of the potential risk of (human) infection. Such a map indicates areas where tick control, public education, or other interventions may be most beneficial (Orloski et al. 2000, Kitron 2000). Maps of vector abundance (perhaps as functions of environmental variables), vector infection rates, human-vector interactions, and reported human cases all address different components of the spatial pattern of risk of Lyme disease. Each map requires different data and reveals different elements of the interacting processes determining Lyme transmission. The literature contains examples of risk maps for Lyme disease for particular communities (Glass et al. 1995, Dister et al. 1997), particular states (Frank et al. 2002, Kitron and Kazmierczak 1997) and for the entire nation (Dennis et al. 1998, Estrada-Pena 1998, US CDC 1999). Risk maps for Lyme disease have been constructed from incidence data (Frank et al. 2002, Kitron and Kazmierczak 1997), tick distributions (Dennis et al. 1998), or environmental risk factors (Glass et al. 1995, Dister et al. 1997, Estrada-Pena 1998, US CDC 1999). Risk maps tend to be constructed either from a snapshot of the disease at a particular point in time (Glass et al. 1995, Dister et al. 1997) or by pooling or averaging disease incidence across the entire period of surveillance (Kitron and Kazmierczak 1997, Estrada-Pena 1998, US CDC 1999, Frank et al. 2002). While risk maps are potentially useful tools for public health practioners, Lyme disease presents particular challenges in their creation, analysis, and interpretation. Surveillance data for Lyme disease can be problematic for a number of reasons: developing recognition of the signature symptoms by health-care providers over the study period, the associated potential lack of accurate diagnoses, imprecise serologic results, uneven case detection, reporting biases, and difficulty relating location of report to location of exposure (Kitron and Kazmierczak 1997). As a result, Lyme disease may be substantially under-reported, with the probability of reporting varying with stage of disease and age of the patient (Orloski et al. 1998, Naleway et al. 2002). Kitron (2000) provides a brief but thorough discussion of risk maps for vector-borne diseases which we expand on in discussions below. Despite these concerns, surveillance data show that both the incidence and geographic range of human cases has increased steadily, particularly in the northeastern US (White et al. 1991, Orloski et al. 2000), spreading outward from the initial diagnoses near Lyme, Connecticut. In this study, we explore spatio-temporal patterns in temporal changes in reported incidence for human cases for the years 1990-2000. Our aim is primarily exploratory, since any observed patterns in reported incidence (and changes in reported incidence) reflect a combination of patterns in true incidence and local variations in diagnosis rates and reporting rates. The descriptive goal is important as it provides insight into both the evolving geographic coverage of the disease, but also into evolving coverage of the surveillance system monitoring the disease. While we cannot entirely separate the two components, we can draw conclusions on particular aspects of each, as will be seen below. In particular, we report county-level trends in reported Lyme disease incidence (rates per 100,000 persons) in the northeastern US for the years 1990-2000, using data from the CDC data repository. The goal of this study is a descriptive analysis of countyspecific changes in annual rate of reported cases, in particular assessment of the spatial pattern of rate changes in attempt to identify areas experiencing fast growth (either in number of cases or in improved reporting practices). We also examine geographic patterns of missing county-level reports and discuss relationships between missing reports and observed patterns in the incidence of disease. Previous investigations by the CDC concentrate on the years 1994-2000 due to varying data availability prior to 1994. We include available data from 1990-1994 and explore its compatibility with the proposed models. 2. Lyme disease surveillance data Since 1992, the CDC has compiled reported cases of Lyme disease based on the standardized case definition approved by the Council of State and Territorial Epidemiologists (Orloski et al. 2000). Using this data base, we determined the number of cases of Lyme disease annual in each county in the US northeast (i.e., the states PA, MD, DE, DC, MA, NJ, NH, VT, NY, CT, RI, and ME) from 1990 to 2000. To standardize for differences in county population size, we convert raw case numbers into incidence proportions per 100,000 people at risk using the 1990 U.S. Census population sizes. For simplicity, we use the 1990 Census population sizes as denominators, noting that adjustments for population growth change results only slightly. Before examining the data, some review of the nature of public health surveillance data is in order (Teutsch and Churchill 1994, Brookmeyer and Stroup 2004) is in order to provide a context for the analyses below. In particular, Lyme disease surveillance data comprise a mix of passive, active and laboratory surveillance (Orloski et al. 2000). Passive surveillance relies on health-care providers to report any new diagnoses to either local or state public health departments. The health department in turn assesses the report with respect to the standardized definition (perhaps including confirmatory laboratory tests), then electronically submits those cases meeting the standard to the CDC via the National Electronic Telecommunication System for Surveillance (NETSS). Active surveillance involves proactive contact of health-care providers by local
File Format	PDF HTM / HTML
Alternate Webpage(s)	https://www.sph.emory.edu/departments/bios/documents/techdocs/2004/Tech_Report_04-11.pdf
Language	English
Access Restriction	Open
Content Type	Text
Resource Type	Article