AI-Based Diagnostic Platform Capabilities With Lyme Disease as a Use Case: Integrative Exploration

Lyme disease (LD) is the most common vector-borne disease in the United States. It is difficult to diagnose because it can mimic numerous other conditions, and testing protocols may not be sufficient. Although the Centers for Disease Control and Prevention (CDC) recommend a 2-tiered serologic testing approach for LD diagnosis, many patients are diagnosed clinically based on criteria such as the erythema migrans rash and, particularly when the rash is not present, various symptom patterns, exposure history, other information, and clinician observations. Against this backdrop, online symptom checkers, using artificial intelligence (AI) processing techniques, are increasingly used to obtain diagnostic information and resources. With LD as a use case, this research applied a modified capabilities approach to explore the relative effectiveness and utility of AI-based tools in application and comparison to serologically (CDC+) and clinically based diagnoses.

The overarching goal of this research was to provide a baseline exploration of AI-assisted diagnostic tools relative to more traditional medical assessment approaches in detecting complex infectious diseases. To assess the potential diagnostic utility (DU) of online symptom checker platforms with LD as a use case, this study aimed to (1) evaluate platform performance in identifying LD across different diagnostic cohorts; (2) compare symptom patterns and severity distributions among online LD diagnoses; and (3) identify the most frequently co-occurring conditions potentially misclassified as LD or vice versa.

Data were drawn from a limited structured survey of patients with confirmed or probable LD, including diagnostic pathways (CDC+ and/or clinical), symptom profiles and severity, treatment history, and time to diagnosis. These patient cases were then entered into 3 leading AI-based symptom checker platforms-MediFind, Isabel, and WebMD-to examine diagnostic performance. Descriptive analytics, logistic regressions, and postestimation analyses were used to identify patterns of DU and interaction effects among cohort type, symptom severity, and AI-based platforms. This survey-based research was not intended to serve as an experiment or clinical trial.

DU varied significantly across platforms and symptom severity thresholds. DU improved at higher symptom severity thresholds (≥3; P<.001) and was slightly higher among clinically diagnosed cohorts compared to CDC+ cohorts (P=.21). Marginal analyses revealed that clinically diagnosed respondents were more sensitive to changes in severity, but with different levels of platform consistency across conditions. Note that DU was analyzed principally for exploratory purposes.

This research serves as a preliminary and directive step for expanded data collection and a larger, more comprehensive study. The findings suggest that AI-based symptom checkers may supplement early diagnostic reasoning in complex conditions such as LD, particularly when symptom severity is high. Inconsistency across platforms and diagnostic categories highlights the need for algorithmic refinement and standardized validation frameworks to enhance diagnostic reliability in AI-based tools.