. 2019 Oct 24;7(10):e14769. doi: 10.2196/14769

Table 3.

Generalized linear mixed model for the number of users per county versus the county population size and the Lyme disease status of the county based on Lyme disease incidence and percent change of cases from 2013 to 2017, adjusting for the regions’ random effects. We used a negative binomial model, and the effect of each independent variable is expressed as incidence rate ratios.

Variables^a		Incidence rate ratio	95% CI	P value
County population size (per 100,000)		1.3	1.2-1.4	<.001^b
County Lyme disease status (2013-2017)
	No Lyme disease cases reported	1	—^c	—
	Low incidence—no change	0.8	0.4-1.7	.60
	Low incidence—greater than 1-fold higher increase	1.8	1.1-3.2	.03^d
	High incidence—no change	4.2	2.1-8.1	<.001^b
	High incidence—greater than 1-fold higher increase	3.5	1.8-7.2	<.001^b

^aRegion (random effect) coefficient=0.6 (95% CI 0.1-4.9); Log-likelihood ratio test, P<.001.

^bP<.001.

^cNot applicable (reference category).

^d.001≤P≤.05.