Exploratory spatial analysis of Lyme disease in Texas –what can we learn from the reported cases?

This was the first study that examined the geographic distribution of reported human LD cases in Texas using exploratory spatial data analysis. We detected significant variation in the spatial distribution of reported cases in the state, and some illuminating patterns emerged. First, after adjusting for the population at risk, we delineated a high-risk cluster in Central Texas corresponding to the Cross-Timbers eco-region, which is a region that is considered to be beyond the geographical range of the main vector, Ixodes scapularis [19, 20]. Second, the eastern part of Texas, which is considered to provide the most suitable habitat for I. scapularis, did not appear to be a high-risk area. Third, LD cases were reported from several counties in western Texas, a region considered unsuitable for the survival of Ixodes ticks. There are a number of alternative explanations for the observed pattern of reported cases, namely 1) clustered misdiagnosis of cases, 2) clustered infections that were acquired elsewhere but were reported in the patients’ county of residence and 3) clustered local populations of infected, vector-competent and human-biting ticks in the high incidence counties. The significance and implications of these findings for public health, and the interpretation of the ongoing surveillance and reporting activities in the state for these possibilities, are outlined below.

The study was based on LD cases as they were reported to the public health system. It is possible, therefore, that surveillance case definitions were interpreted differently in different jurisdictions due to unequal knowledge and practices of physicians, resulting in regional over- or under-reporting bias. For instance, an erythema migrans rash without rigorous laboratory evidence might be reported as a LD case, even though it may not actually be LD. This can happen because STARI meets the surveillance definition of LD, and doctors are required by law to report patients with erythema migrans rash as having LD, which could contribute to over-reporting of LD [6]. This is particularly problematic in the southern United States, including Texas, where STARI is thought to be more common than LD, due to the abundance and aggressive biting behavior of Lone Star ticks in this state [6]. At the same time, physicians in non-endemic states may be told that there is no LD disease in their area, and that any LD diagnosis (whether true LD or misdiagnosed STARI) is in error. This might confuse physicians and they may question the importance of reporting a disease that is not officially recognized to exist in their state [6]. This in turn, could lead to underreporting of both STARI and LD. The findings of this study highlight the need to revise the surveillance definition for LD, and to clarify whether LD cases reported in Texas are true cases or STARI cases. If indeed most reported LD cases have been misdiagnosed (and are in fact false positives), then the surveillance and case definition criteria of LD should be carefully revised in order to assist future epidemiologic investigations, and to improve targeting of interventions to truly high-risk areas. Also, more research is needed to better understand the distribution and etiology of STARI.

The current study identified several low-high and high-low outlier counties where incidence of reported LD sharply differed from the surrounding counties. Also, the eastern part of Texas did not appear to be a high-risk area for LD. It would be of high interest to compare the diagnostic and reporting criteria for LD between the eastern part of Texas and the Cross-Timbers cluster identified in the current study, and also between the outlier and the surrounding counties, to assure that diagnostic and reporting criteria are uniformly applied across the whole state of Texas and so to maximize the benefits of the ongoing surveillance activities.

Regional epidemiologic patterns of human LD are influenced both by the heterogeneity of ecological risk factors and also by human risk behavior, particularly in terms of traveling to, and outdoors activities (such as hunting) in areas with endemic LD. There was no information about traveling history for the cases analyzed in this study to further explore this risk factor. Presumably, not all cases that were diagnosed and reported in Texas had been exposed to infected ticks in Texas. At the same time, it seems reasonable to assume that cases living in metropolitan areas were more likely to have been exposed far away from their home area, compared to cases reported in the remote areas where out-of-state travel to endemic areas may not be as likely or easily accessible. Interestingly, however, the high-risk cluster in Central Texas was located away from any metropolitan areas. Also, there are no known endemic LD areas in the neighboring states. Although this finding is intriguing, without the knowledge of travel history, no conclusions can be drawn about the place of exposure. This highlights the importance of systematic collection of information about the location of exposure of LD cases whenever known (e.g., outside vs. inside of the state) to enhance LD surveillance in Texas. In the event that the travel history and misdiagnosis (and the resulting under- and/or over-reporting bias) could not explain the observed patterns in human infection, then the pattern reflects either an advance of the tick-reservoir cycle from eastern deciduous forests, or long-distance dispersal of infected I. scapularis by birds [21]. Given that the eastern part of Texas did not appear to be a high-risk area for LD, local advance from the east is less likely but still possible. It might also be possible that long distance dispersal via birds could have established a new disease focus in the Cross-Timbers region, initiating a spatial cluster of infection. In the absence of I. scapularis, there may be alternative tick vectors (for example, Ixodes affinis, a competent vector of B. burgdorferi sensu stricto in some parts of the southern United States [22]) and host species involved in the maintenance of enzootic cycles of B. burgdorferi that underlies the observed patterns of human infection [5]. Currently, there is no tick surveillance in Texas which could provide insights into the distribution and possible expansion of I. scapularis, or into the presence of alternative tick vectors for B. burgdorferi in the state. Also, little is known about the vector specificity of B. burgdorferi to certain species of ticks. For example, in Europe, Ixodes ricinus in a competent vector, but the other common ticks are not. Sampling of potential vector ticks and reservoir hosts at targeted sites would be needed to determine if alternative species of vectors and reservoir hosts exist and maintain B. burgdorferi in Texas.

The Cross-Timbers is a North American eco-region that stretches from southern Oklahoma into Central Texas, forming a transitional area between the eastern deciduous forests and the grasslands of the southern Great Plains. The Cross-Timbers ecosystem is a vast mosaic of forest, woodland, savannah and prairie. The dominant trees are post oak and blackjack oak with an understory of shrubs and grasses. The variety of ecosystems in the Cross-Timbers provide diverse habitats for wildlife, and the abundant acorns are a staple food source for wild turkeys, prairie chickens, raccoons, squirrels, and deer [12]. As such, this ecoregion may well provide suitable habitat for a variety of potential host and vector species for B. burgdorferi and this area thus merits further investigation.

While the Cross-Timbers may provide a suitable habitat for a range of potential tick vector and mammalian reservoir host species for B. burgdorferi, the arid conditions and sparse vegetation in West Texas limits the survival of several species that inhabit the eastern half of the state. It has been suggested that, in parts of the southern United States, host composition may be dominated by non-mammalian species such as lizards [23]. However, the role of lizards in maintaining B. burgdorferi is unknown. It is also unclear which tick species might serve as vectors in these arid ecosystems. More studies are needed to shed light on the presence of an enzootic cycle of this pathogen in the southern United States.

In recent years, the annual incidence of LD in Texas has typically varied between 0.4-0.6 cases per 100,000 population. To put these numbers into perspective, a national study of 15-year duration (1992–2006) revealed mean annual cumulative incidence of LD for all states to have ranged from <0.01 cases per 100,000 in Montana and Colorado to 73.6 per 100,000 cases in the highly endemic state of Connecticut, with a median of 0.5 cases per 100,000 population [24]. Thus, the reported incidence of LD in Texas typically has been around the national median rate.

Some limitations to this study exist. While exploratory spatial data analysis is particularly suitable for visualizing and exploring spatial data and for detecting interesting spatial patterns, an inherent limitation of this method is that it does not explain the patterns it reveals [15]. Thus, as discussed above, further studies are needed to explain the spatial patterns we observed and to determine the specific anthropogenic, environmental and ecological factors responsible for distribution of LD in Texas.