**5. Limitations**

This study has several flaws that should be overcome in future studies. First, cases and controls were identified in accordance with the results of antibody detection. However, IgG-positive samples may have been derived from patients who were infected with DF several years ago. Thus, the results of the completed questionnaires of these patients may be incompatible with their situations during their initial infection. Second, recall biases stemming from the inaccuracy of memory could also reduce the validity of the questionnaire. Third, newly infected individuals may have been misdiagnosed as controls because IgG and IgM antibody titers are present at undetectable levels during the initial stages of infection with dengue [38]. Thus, misclassification bias may have been introduced. Fourth, volunteer bias might have been caused by participants who volunteered to become part of the samples. Additionally, the information provided by the volunteered participants might have been different from the general population. Fifth, some definitions like the definitions of "good indoor daylight quality" and "good ventilation" were given in the questionnaire, but respondents might have had a subjective understanding of these definitions, which could have led to inaccuracy of the relevant information. Finally, the model's general applicability was difficult to evaluate at present given that the data used

in this study were obtained on the basis of serum samples collected from residents in Guangdong Province over the period of 2013 to 2015 without other extrapolated studies. Given that other studies have shown that the community wherein residents live remains an independent factor associated with DF infection [23], the results of this study could be used as the reference for the development of personalized protective measures against DF infection in tropical and subtropical countries with high densities of *Aedes* mosquitoes.
