Large-scale health disparities associated with Lyme disease and human monocytic ehrlichiosis in the United States, 2007-2013

Abstract

Promoting health equity is a fundamental public health objective, yet health disparities remain largely overlooked in studies of vectorborne diseases, especially those transmitted by ticks. We sought to identify health disparities associated with Lyme disease and human monocytic ehrlichiosis, two of the most pervasive tickborne diseases within the United States. We used general linear mixed models to measure associations between county-level disease incidence and six variables representing racial/ethnic and socioeconomic characteristics of counties (percent white non-Hispanic; percent with a bachelors degree or higher; percent living below the poverty line; percent unemployed; percent of housing units vacant; per capita number of property crimes). Two ecological variables important to tick demography (percent forest cover; density of white-tailed deer) were included in secondary analyses to contextualize findings. Analyses included data from 2,695 counties in 37 states and the District of Columbia during 2007-2013. Each of the six variables was significantly associated with the incidence of one or both diseases, but the direction and magnitude of associations varied by disease. Results suggested that the incidence of Lyme disease was highest in counties with relatively higher proportions of white and more educated persons and lower poverty and crime rates; the incidence of human monocytic ehrlichiosis was highest in counties with relatively higher proportions of white and less educated persons, higher unemployment rates and lower crime rates. The percentage of housing units vacant was a strong positive predictor for both diseases with a magnitude of association comparable to those between incidence and the ecological variables. Our findings indicate that racial/ethnic and socioeconomic disparities in disease incidence appear to be epidemiologically important features of Lyme disease and human monocytic ehrlichiosis in the United States. Steps to mitigate encroachment of wild flora and fauna into areas with vacant housing might be warranted to reduce disease risk.

Fig 1. Maps of county-level case counts and incidence of Lyme disease and human monocytic ehrlichiosis in the United States during 2007–2013.

Case counts were calculated as the sum of the annual numbers of reported cases of each disease across the seven-year period in each county (N = 2,695 in 37 states including the District of Columbia). Incidence (standardized to cases per 100,000 persons) was calculated as the ratio of case counts and county population size in 2010 (obtained from the 2010 decennial U.S. Census).

Fig 2. Results of county-level analyses using general linear mixed modeling to quantify associations between the incidence of Lyme disease and human monocytic ehrlichiosis with six racial/ethnic and socioeconomic variables (socioeconomic variables) and two ecological variables.

For each disease, results of univariable models (each socioeconomic variable individually) (left), the final (reduced) multivariable model including multiple socioeconomic variables together (middle), and the final (reduced) multivariable model including multiple socioeconomic variables and two ecological variables together (right) are shown. Incidence was modeled using case counts (annual numbers of reported cases of each disease summed during 2007–2013 in each of 2,695 counties in 37 states and the District of Columbia); county population size in 2010 was included in the models as an offset term. Values for socioeconomic and ecological variables were centered by subtracting the mean and scaled by dividing each value by its centered standard deviation. Sources and summary values of disease, socioeconomic and ecological data are provided in S1 Table.

Fig 3. Results of county-level post hoc analyses using general linear mixed modeling to quantify associations between the incidence of Lyme disease and human monocytic ehrlichiosis with four housing vacancy type variables.

Results of the non-reduced multivariable models (all four housing vacancy type variables together) are shown. Incidence was modeled using case counts (annual numbers of reported cases of each disease summed during 2007–2013 in each of 2,695 counties in 37 states and the District of Columbia) with county population size in 2010 included in the model as an offset term. Values for housing vacancy type variables were centered by subtracting the mean and scaled by dividing each value by its centered standard deviation. Source and summary values of housing vacancy type data are provided in S1 Table.