Validation of Claims-Based Algorithm for Lyme Disease, Massachusetts, USA

Area of Science:

• Epidemiology
• Infectious Disease Surveillance
• Health Informatics

Background:

• Notifiable disease surveillance underestimates Lyme disease incidence.
• Claims-based algorithms offer a potential alternative for estimating disease burden.
• The accuracy of claims-based algorithms for Lyme disease surveillance is not well-established.

Purpose of the Study:

• To evaluate the accuracy of a previously developed claims-based algorithm for identifying Lyme disease cases.
• To determine the positive predictive value (PPV) of the algorithm in a high-incidence setting.
• To assess the utility of claims data in supplementing traditional Lyme disease surveillance.

Main Methods:

• A validated Lyme disease algorithm (diagnosis code + antimicrobial prescription) was applied to a Massachusetts administrative claims database (July 2000-June 2019).
• A subset of identified cases underwent medical chart review and adjudication by clinicians using national surveillance case definitions.
• Positive predictive values (PPVs) were calculated for different case classifications.

Main Results:

• The algorithm identified 12,229 Lyme disease episodes.
• Clinician adjudication of 128 medical charts yielded a PPV of 93.8% for confirmed, probable, or suspected cases.
• The PPV for confirmed and probable cases only was 66.4%.

Conclusions:

• A claims-based algorithm demonstrates a high positive predictive value for identifying Lyme disease, including suspected cases.
• This algorithm can effectively supplement traditional surveillance methods for assessing Lyme disease burden in high-incidence areas.
• Claims data provide a valuable resource for enhancing public health surveillance of vector-borne diseases.