Borrelia or Bell’s? Development of BoB: a clinical decision support tool to differentiate Lyme neuroborreliosis facial palsy from Bell’s palsy

BACKGROUND: Differentiating Lyme neuroborreliosis (LNB) from Bell's palsy (BP) in adults with peripheral facial palsy remains challenging. We aimed to develop and internally validate clinical decision support tools for European adults to reduce unnecessary lumbar punctures and support timely treatment. METHODS: We conducted a nationwide retrospective cohort study including adults (≥18 years) with facial palsy and confirmed LNB or BP from two Danish cohorts (2015-2024). All patients underwent lumbar puncture. Data were extracted from medical records, and 29 candidate predictors were selected for analysis. The dataset was randomly split into training (80%) and test (20%) sets. Key predictors were identified using random forest and used to develop a decision tree. Analyses were repeated excluding serum Borrelia burgdorferi (Bb) IgG as results may be unavailable at presentation. RESULTS: We included 683 patients (422 LNB, 261 BP); 23% of LNB patients were initially misdiagnosed as BP. Key predictors of LNB were radicular pain, serum Bb IgG positivity, symptom duration ≥7 days, and forest/grass exposure. The resulting decision tree showed high performance (accuracy 0.93; sensitivity 0.93; specificity 0.94). It could reduce lumbar punctures by 40% and initially missed LNB to 8%, while 9% of BP would still undergo lumbar puncture. Excluding serum Bb IgG, radicular pain and symptom duration ≥10 days remained key predictors, with similar performance (accuracy 0.93; sensitivity 0.92; specificity 0.93). CONCLUSIONS: We developed two accurate tools using a small set of variables, with radicular pain emerging as an important predictor. External validation is required before clinical implementation and is underway through a European collaboration.