| Popis: |
Background Lyme disease (LD) is the fifth most common notifiable disease in the US with 30,000-40,000 LD cases reported annually via public health surveillance. Recent healthcare claims-based studies utilizing case-finding algorithms estimate national LD cases are >10-fold higher than reported by surveillance. The reliability of claims-based data depends on the accuracy of the case-finding algorithms using the information available in the claims primarily generated for the administrative purposes. To assess the true burden of LD, it is imperative to use validated well-performing LD case-finding algorithms (“LD algorithms”). We conducted a systematic literature review to identify LD algorithms based upon healthcare claims data in the US and their respective performance. Methods We searched PubMed and Embase for articles published in English from January 1, 2000 through the most recent date as of February 20, 2021. We selected articles including all of the following search terms: (1) “Lyme disease”; (2) “claim*” or “administrative* data”; and (3) “United States” or “the US*”. We then reviewed the titles, abstracts, and full texts to identify articles describing LD algorithms developed for claims data. Figure 1 shows the flow diagram following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement. Results We found 15 articles meeting the inclusion criteria. Of these, 7 study algorithms used only LD diagnosis codes (ICD-9, 088.81; ICD-10, A69.2 or A69.2x), 4 studies additionally used antibiotic dispensing records, and 4 studies additionally used serologic test order codes (CPT 86617, 86618). Three studies used different algorithms for inpatient and outpatient settings. Only one study (in Tennessee, a low-incidence state for LD) provided validation results for their algorithm, which only used a LD diagnosis code (ICD-9, 088.81), with reported sensitivity=50% and positive predictive value=5%. Conclusion Validation data on the LD algorithms developed for healthcare claims data are limited, and suggest algorithms using only LD diagnosis codes may not perform well. Further validation of high-performance claims-based LD algorithms is critical to inform the true burden of LD overall and within subgroups. Disclosures Bradford D. Gessner, MD, MPH, Pfizer Inc. (Employee) James Stark, PhD, Pfizer Inc. (Employee) Sarah Pugh, PhD, Pfizer Inc. (Employee) |
