Reliability of life events and social networks questions.

A test-retest study design was completed to examine the reliability of life events and social networks questions. This study was implemented by conducting a primary interview (face-to-face) aimed at investigating the potential function of stress as a risk for cancer in females, followed by a second interview (telephone) developed to analyze the reliability of the questions employed in the primary assessment. The life events questions obtained from the Life Events Survey were slightly altered for the population involved in the stress and cancer study. The social networks questions from Berkman's Social Network Index were also incorporated into the stress and cancer study. Therefore, the reliability study was performed to assess if the questions were reliable in their modified format. Specifically, the data were analyzed to: (a) examine the reliability of the life events questions in terms of positive and negative impact responses; (b) measure the reliability of whether or not the life events were experienced; (c) determine the reliability of social networks questions; and (d) evaluate the reliability of the above questions for cases and control status, in addition to the demographic variables of age, income, education, and religious preference and activity. The study subjects consisted of women between the ages of 20 and 54 who resided in the Salt Lake City area. Cases included women diagnosed as having first primary breast, ovarian, or endometrial cancer identified through the Utah Cancer Registry. Controls were randomly selected from the same study population as the cases. A total of 27 women participated in the present reliability study. The data were analyzed using the kappa statistic, percent of agreement, and McNemar's statistic. Based on the methods used in this study, the results indicate that the life events perceived impact of the events, and the social networks questions exhibited moderate reliability. In addition, the case/control and demographic characteristics of the subjects displayed moderate reliability concerning these questions. Thus, the questions appear to be appropriate for use in future research and the study for which they were originally intended. There also appeared to be a response difference between the two interview administrations since a positive McNemar's value was observed in certain analyses. This may indicate that the subjects tended to respond differently on the two tests, which potentially may have affected the reliability of the questions in this study.