Sampling Lyme Disease Reports to Calculate Disease Estimates: Further Evaluation of an Approach to Reduce Surveillance Burden, Minnesota, 2005-2010

BACKGROUND:  Many state disease surveillance programs are being challenged by marked increases in Lyme Disease (LD) reports. The Minnesota Department of Health (MDH), which conducts passive laboratory- and physician-based surveillance for LD, receives large numbers of LD reports; all require surveillance follow-up, but only a minority become confirmed cases (e.g., 1,293 [28%] of 4,551 reports in 2010). An increasing surveillance burden prompted an investigation of alternative approaches to Lyme disease surveillance in Minnesota in 2011, when a sampling methodology used by New York was adapted and used to retrospectively analyze Lyme disease reports from 2010 to determine whether key surveillance findings would be statistically indistinguishable. The current analysis expands to include LD data from 2005-2010.

METHODS:  Random samples of electronic, laboratory-only reports were abstracted at 1:2 and 1:5 ratios for each year from 2005-2010. To estimate confirmed case counts, we multiplied the number of sampled, confirmed reports by 2 and 5, respectively, and added this to the number of confirmed cases reported by medical providers. Estimated case counts for confirmed cases that had erythema migrans (EM), late manifestations (LM), were male, and were pediatric (≤12 years of age), and were residents of specific reporting regions were similarly calculated and compared to observed values using Chi-square goodness of fit tests.

RESULTS:  From 2005-2010, estimated confirmed case counts calculated from both 1:2 and 1:5 samples did not differ from observed confirmed case totals. For all years, the percentage of cases with EM (p=0.52-0.93), LM (p=0.62-0.99), were male (p=0.39-0.78), and were pediatric (p=0.36-0.85) in the 1:2 sample were not significantly different (p>0.05) from observed. In the 1:5 sample, only the percentage of cases that had EM or LM differed significantly from observed case counts in 2005 and 2009 (p<0.05), and the percentage of male cases differed from observed counts (p<0.01) in 2010. Estimated case totals by region were similar to observed values across all years (1:2, p=0.76-0.98; 1:5, p=0.08-0.94).

CONCLUSIONS:  A retrospective analysis of Minnesota’s confirmed Lyme disease cases from 2005-2010 found that random 1:2 and 1:5 samples of laboratory-reported LD resulted in estimated confirmed case counts very similar to those observed, as well as similar proportions of confirmed cases that had EM, LM, were male, or were pediatric. Geographic distribution of cases within Minnesota did not differ between estimated and observed values. Future application of sampling would conserve limited surveillance resources by reducing follow-up effort while providing accurate disease estimates.