Background: Accurate monitoring of health conditions and behaviours, and health service usage in the population, using an effective and economical method is important for planning and evaluation. This study examines the reliability of questions asked in a telephone survey by conducting a test/retest analysis of a range of questions covering demographic variables, health risk factors and self-reported chronic conditions among people aged 16 years and over.
Methods: A Computer Assisted Telephone Interviewing (CATI) survey on health issues of South Australians was re-administered to a random sub-sample of 154 respondents between 13-35 days (mean 17) after the original survey. Reliability between questions was assessed using Cohen's kappa and intraclass correlation coefficients.
Results: Demographic questions (age, gender, number of adults and children in the household, country of birth) showed extremely high reliability (0.97 to 1.00). Health service use (ICC = 0.90 95% CI 0.86-0.93) and overall health status (Kappa = 0.60 95% CI 0.46-0.75) displayed moderate agreement. Questions relating to self-reported risk factors such as smoking (Kappa = 0.81 95% CI 0.72-0.89) and alcohol drinking (ICC 0.75 = 95% CI 0.63-0.83) behaviour showed good to excellent agreement, while questions relating to self-reported risk factors such as time spent walking for physical activity (ICC 0.47 = 95% CI 0.27-0.61), fruit (Kappa(w) = 0.60 95% CI 0.45-0.76) and vegetable consumption (Kappa(w) = 0.50 95% CI 0.32-0.69) showed only moderate agreement. Self-reported chronic conditions displayed substantial to almost perfect agreement (0.72 to 1.00) with the exception of moderate agreement for heart disease (Kappa = 0.82 95% CI 0.57-0.99).
Conclusion: These results show the questions assessed to be reliable in South Australia for estimating health conditions and monitoring health related behaviours using a CATI survey.