Our investigation has some limitations. First, interviews were conducted several weeks after illness onset, which enabled accurate classification of patients by hospitalization status and data collection on all symptoms and their duration (
45). However, this timing might result in incomplete recall and recall bias, which could affect the accuracy of reported symptoms and their timing, particularly among hospitalized patients, who might be more likely to remember more severe symptoms (
46). Future prospective studies using methods such as symptom diaries or serial interviews could reduce recall bias. Second, a higher proportion of proxies were interviewed on behalf of hospitalized case-patients. However, when proxies were removed from the reduced multivariable model, the ORs were relatively stable, indicating the proxies did not affect the association of symptoms with hospitalization. In addition, although clinical manifestation of viral respiratory diseases can differ by age, we were unable to compare symptoms across different age groups because of the high percentage of proxy interviews for patients >65 years of age, which resulted in fewer symptoms being reported in that age group. Our findings might not apply to all populations because of differences in age distribution, disease severity, testing practices, and socioeconomic status. Finally, because symptoms such as seizure and hemoptysis were experienced by a small number of participants, we were limited in our ability to draw conclusions about their duration and associations with hospitalization status.