OBJECTIVES: Responses to health-related items on the Community Health Survey (CHS) provide evidence that is used to develop community-based health policy. This study aimed to assess the test-retest reliability of selected health behavioral items on the CHS according to item category, response period, and response scale. METHODS: A sample of 159 men and women 20 to 69 years of age participated in a test-retest with an interval of 14 to 21 days. A total of 28 items relating to smoking, alcohol consumption, diet and weight control, and mental health were selected. We evaluated the test-retest reliability of the items using kappa statistics. RESULTS: Kappa values ranged from 0.44 to 0.93. Items concerning habits had higher kappa values (mean, 0.7; standard error, 0.05) than items concerning awareness or attitudes (p=0.012). The kappa value of items with two- to four-point scales was 0.63, which was higher than the value of 0.59 for items with scales involving five or more points, although this difference was not statistically significant. Different kappa values were observed for each reference period, but no statistically significant differences were noted. CONCLUSIONS: The test-retest reliability of the CHS items that we studied was associated with item category. Further study of the relationship between item category and reliability in domains other than health behaviors is required.