Open Access1338. Development of a Novel Application for Differential Diagnosis of Tick-borne Diseases
Author(s) -
Corey Meyer,
Jaleal Sanjak,
Audrey Cerles,
Christian H. Garnier,
Laurel MacMillan
Publication year - 2019
Publication title -
open forum infectious diseases
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.546
H-Index - 35
ISSN - 2328-8957
DOI - 10.1093/ofid/ofz360.1202
Subject(s) - medicine , incidence (geometry) , disease , tick borne disease , risk factor , tick , virology , physics , optics
Background Early diagnosis and treatment of tick-borne diseases (TBDs) is critical for mitigating their adverse health outcomes, but the differential diagnosis of TBDs is challenging because many symptoms are nonspecific and commonly used diagnostic assays have significant shortcomings. Furthermore, although the local incidence of TBDs is recognized as an important factor in diagnosis, tools to help clinicians formally consider surveillance data in their decision-making are not available. To address these gaps, Gryphon Scientific developed a differential diagnosis application (app) for TBDs that calculates a patient’s likelihood of infection with specific TBDs based on their symptoms, risk factors, and state of suspected tick exposure. Methods A differential diagnosis model for TBDs was developed using data on: (1) TBD symptom and risk factor prevalence in TBD patient populations, collected from clinical studies; and (2) human TBD incidence data from notifiable disease surveillance systems and tick infection prevalence data from reports and public databases, which were combined to develop an environmental risk measure. These data were used to build a Bayesian Belief Network (BBN) model that predicts TBD infection probabilities based on a patient’s symptoms, risk factors, and state of suspected tick exposure. Performance of the model was validated using case studies from the biomedical literature. The model was incorporated into an app developed using R-shiny, called TBD-DDx (Figures 1 and 3). Results A pilot application was developed that includes 10 states (AR, CT, MA, ME, MN, MO, NH, RI, VT, and WI) and the 11 TBDs endemic to those states. The differential diagnosis model identified the patient’s true disease as the top-predicted disease in 56% of cases and within the top three predicted TBD in 84% of cases. The inclusion of incidence factors in the model improved performance (Figure 4). Conclusion These results demonstrate that the TBD-DDx app is a promising tool for informing clinical diagnoses of TBDs to guide selection of diagnostic testing and treatment. This study represents the first use of a BBN modeling approach that incorporates an environmental risk measure and could be adapted for differential diagnosis of other diseases with environmental or other exposure risks. Disclosures All authors: No reported disclosures.