Validation of Random Sampling as an Estimation Procedure for Lyme Disease Surveillance in Massachusetts and Minnesota

Abstract

Many disease surveillance programs, including the Massachusetts Department of Public Health and the Minnesota Department of Health, are challenged by marked increases in Lyme disease (LD) reports. The purpose of this study was to retrospectively analyse LD reports from 2005 through 2012 to determine whether key epidemiologic characteristics were statistically indistinguishable when an estimation procedure based on sampling was utilized. Estimates of the number of LD cases were produced by taking random 20% and 50% samples of laboratory-only reports, multiplying by 5 or 2, respectively, and adding the number of provider-reported confirmed cases. Estimated LD case counts were compared to observed, confirmed cases each year. In addition, the proportions of cases that were male, were ≤12years of age, had erythema migrans (EM), had any late manifestation of LD, had a specific late manifestation of LD (arthritis, cranial neuritis or carditis) or lived in a specific region were compared to the proportions of cases identified using standard surveillance to determine whether estimated proportions were representative of observed proportions. Results indicate that the estimated counts of confirmed LD cases were consistently similar to observed, confirmed LD cases and accurately conveyed temporal trends. Most of the key demographic and disease manifestation characteristics were not significantly different (P<0.05), although estimates for the 20% random sample demonstrated greater deviation than the 50% random sample. Applying this estimation procedure in endemic states could conserve limited resources by reducing follow-up effort while maintaining the ability to track disease trends.