Poster presented at International Symposium on Tick-borne Pathogens and Disease (ITPD) 2023
Introduction: Lyme disease (LD) is common in the US but the total burden is unknown. We previously validated a claims-based algorithm (LD diagnosis code and indicated antibiotic) via medical record review (PPV 93.8%). The current work explores additional claims-based algorithms to identify LD cases without LD-specific diagnoses.
Methods: We developed algorithms using non-LD diagnosis codes, LD-indicated antibiotics, and procedure codes for LD diagnostic tests. Diagnosis codes reflected musculoskeletal, nervous system, and cardiovascular manifestations and were not specific to LD. For each manifestation, we assessed 3 algorithms; each required a diagnosis code and: (1) antibiotic and diagnostic test; (2) antibiotic, with no diagnostic test; (3) diagnostic test, regardless of antibiotic. We applied all algorithms to claims from Massachusetts residents, where LD is endemic.
Results: We identified 144,058 individuals who met ≥1 algorithm, among whom only 10% had a LD diagnosis code within +/-90 days. The algorithms identified the greatest patient count for musculoskeletal manifestations, followed by neurologic manifestations, then cardiovascular. The most common diagnoses identifying cases were: musculoskeletal: knee pain; neurologic: radiculopathy, Bell’s palsy, polyneuropathy; cardiovascular: atrioventricular block, other specified conduction disorders.
Among patients identified by algorithms requiring an antibiotic (algorithms 1 and 2), patients with a diagnostic test (algorithm 1) were more likely than those without a diagnostic test (algorithm 2) to receive doxycycline (43% vs 18%); most other patients received amoxicillin. Algorithm 3 had a low prevalence of treatment (16%). Algorithms 1 and 3, both requiring a diagnostic test, reflected LD seasonality, with higher burden in summer; algorithm 2, which did not require a diagnostic test, had an inverse seasonal pattern.
Conclusion: The claims-based algorithm requiring a diagnosis code for a LD manifestation, antibiotic treatment, and LD diagnostic test identified patients whose characteristics were consistent with those of LD cases. We will validate this algorithm via medical record review.
