Assessing health burden risk and control effect on dengue fever infection in the southern region of Taiwan

Cheng, Yi-Hsien; Lin, Yi-Jun; Chen, Szu-Chieh; You, Shu-Han; Chen, Wei-Yu; Hsieh, Nan-Hung; Yang, Ying-Fei; Liao, Chung-Min

where b mh and b hm represent transmission probabilities from mosquito to human and from human to mosquito per mosquito bite, n m and n h stand for extrinsic and intrinsic incubation rates, ie, the reciprocals of mean periods in dengue viral shedding for mosquito and human, respectively, B is daily mosquito biting rate, μ m and μ h are death rates for female mosquitoes in adult stage and human, S m (0) represents susceptible female mosquitoes that can be estimated by the product of female mosquitoes surrounding a human (m), S L , as well as total people in Kaohsiung (N h ), and h h is the dengue fever recovery rate for human. Note that R 0 represents transmission potency of dengue virus in the vector-host dynamic model constituted with transmissibility of dengue from mosquito to human (R mh ) and from human to mosquito (R hm ) populations (ie, R 0 2 =R mh ×R hm ). 16, 27 R 0 >1 indicates that dengue fever is epidemically spreading within the human populations with increasing incidences, whereas R 0 <1 means the disease is dying out. An average R 0 of 1 implies that the disease is in endemic equilibrium. In addition, to investigate the contribution of each parameter to the variance of temperature-dependent R 0 , uncertainty and sensitivity analyses were conducted using Monte Carlo (MC) simulation by a commercial software Crystal Ball (Version 2000.2, Decisionerring, Inc., Denver, CO, USA). Specifically, each parameter was defined with an optimally fitted distribution (ie, lognormal distribution) by assigning with either its mean and SD or mean and 95% CI values (ie, define assumption). Temperature-dependent R 0 was then assigned as the integrated forecast of each defined parameter distribution that can be estimated using Eq. (1). A total of 10,000 iterations were performed in the MC simulation through Crystal Ball to ensure the stability of probability distributions. A built-in function, sensitivity analysis, in Crystal Ball was then adopted to examine the contribution of each parameter to the variance of temperature-dependent R 0 .