Abstract:Human cases of Lyme disease, caused by Borrelia burgdorferi, are well-documented in California, with increased risk in the north coastal areas and northwestern slopes of the Sierra Nevada range. Borrelia miyamotoi, a more recently identified zoonotic spirochete causing a relapsing-fever type illness, has been documented as causing human disease in the eastern United States and Europe, but human cases have not been identified in California to date. The Ixodes pacificus, is the primary tick vector of these two zoonotic spirochetes in California. Lyme disease became a state reportable condition in 1989 and nationally notifiable in the United States in 1991. Lyme disease is the most common tick-borne disease in the US with over 30,000 cases reported annually. However, studies have shown that Lyme disease is subject to underreporting, with estimates of up to 500,000 cases annually in the US by multiple data sources. The incidence of Lyme disease in California is low, with approximately 100 confirmed cases reported annually (0.2 confirmed cases per 100,000 population). However, California’s unique ecological diversity contributes to focal highly endemic areas. The goal of this dissertation is to investigate the human epidemiology of these two zoonotic borreliae in California first with a focus on Lyme disease to reduce the burden of Lyme disease surveillance through predictive modeling, then investigating physician practices and finally by assessing human exposure to both these agents through serosurveillance.
In Chapter 1, we obtained data from the California Reportable Disease Information Exchange (CalREDIE), a secure system implemented by the California Department of Public Health (CDPH) for electronic disease reporting and surveillance. Currently, the investigation of Lyme disease is time intensive. Due to a variety of reasons, including the amount of follow-up information required due to the complexity of the Lyme disease case definition, many cases are not followed-up completely to obtain all relevant information to correctly classify a case. In high incidence states in the US for Lyme disease, an estimation sampling approach was performed where 20% of all positive laboratory results were fully investigated which yielded accurate estimates of Lyme disease case numbers. We proposed that automatically reported information, such as lab results, and demographic risk factor information augmented with tick surveillance data would provide estimates of Lyme disease incidence similar to what would be obtained through full investigations requiring intensive follow-up. We created four predictive models using logistic regression starting with a simple model with positive and specific lab data, then successively added automatically reported and easily obtainable contextual information to each model. Our predictive models estimate was validated using k-fold cross validation with constructed ROC curves. Each of the four predictive models had very low sensitivities, which demonstrated that models based on subsets of surveillance data would underestimate the incidence of Lyme disease in California.
In Chapter 2, we surveyed physicians in California to understand knowledge and practices for testing and treating Lyme disease based on the expectation that physician awareness of recommended practices could be limited in low-incidence states like California. We compared knowledge and practice scores of physicians practicing in higher-endemic counties compared to lower-endemic counties. The risk of Lyme disease varies in California and this variation in risk can impact choices about diagnostic testing and interpretation. We found that our physicians in this study deviated from IDSA national guidelines in diagnostic testing for LD when patients sought care for both symptomatic disease and asymptomatic tick bites. Our survey results demonstrated that physicians in California could benefit from targeted education to better understand disease risk in California and to improve recognition of symptoms and appropriate use and interpretation of serologic testing.
In Chapter 3, we evaluated and compared human exposure to B. burgdorferi and B. miyamotoi over a broad geographical range in California. We assessed human exposure to B. burgdorferi and B. miyamotoi by testing 1,700 blood bank serum samples from both higher and lower Lyme disease endemic counties in California with the hypothesis that counties with higher endemicity of Lyme disease would also have a higher endemicity of B. miyamotoi disease because this disease shares the same tick vector. Two of the 1,700 samples had detectable antibodies against B. miyamotoi (0.12%, Exact 95% CI: 0.01%, 0.42%). Both samples tested positive by C6 ELISA, GlpQ ELISA and B. miyamotoi whole cell western blot. Eight of 1,700 samples had detectable antibodies against B. burgdorferi (0.47%, Exact 95% CI: 0.20, 0.93). Samples tested positive by C6 ELISA and IgG western blot for Borrelia burgdorferi. Given the few seropositive samples, we could not characterize the geographic concordance between B. burgdorferi and B. miyamotoi, although we confirmed that exposure to these disease agents is low in California.
Taken together, the results of this dissertation provide insights that model-based methods using limited follow-up on cases underestimated the incidence of Lyme disease in California, which is a low-incidence state. Therefore, complete individual case follow-up as required by the current case definition is necessary to gather adequate information for accurate surveillance. Accurate surveillance information is crucial for physicians in their assessment of suspected Lyme disease patients in this low-incidence state and for monitoring Lyme disease incidence geographically. Overall, this research validated that the risk of human infection by Borrelia burgdorferi and Borrelia miyamotoi in California is low and focal.
