Introduction
What is new?
Key findings?The key findings in this study were that participants who responded to postal questionnaire differed in some of their demographic characteristics and baseline outcomes compared with those who responded by telephone interviews. These differences may be a potential source of bias, which occur because the methods of data collection differ in their implementation. However, if any degree of bias was present, this was consistent across the treatment arms. The implication of this was that an overall treatment difference for an outcome could be obtained irrespective of the way the data have been collected.
What this adds to what was known?In clinical trials, where the comparison of interventions is of focal interest, follow-up data are often collected using a single method, for example, postal questionnaires. In this article, we have illustrated that collecting data on respondents of postal questionnaires supplemented with collecting data on nonrespondents by telephone can increase response and increase generalizability.
What is the implication, what should change now?The implication of this article is: telephone interviews should be considered as an effective way in reducing loss to follow-up of participants in clinical trials; researchers should consider additional methods of collecting data from participants who are not responding to initial requests. However, if different methods are used to collect data in a clinical trial, then the data need to be carefully scrutinized for any biases and how these may affect the overall estimate of treatment effect.
Several methods for collecting follow-up data from participants who take part in research studies have been cited in the literature and these include the use of the postal questionnaires, face-to-face interviews, telephone interviews, and use of the Internet [1], [2], [3]. The postal questionnaire is the most frequently used and is considered to be the most cost effective but is often associated with the lowest response rate [4]. Poor response to questionnaires is known to reduce the statistical power of the study as the “effective” sample size is reduced. It can also introduce bias if the nonresponders are systematically different on outcomes of interest to those who respond to the questionnaires.
Much of the literature focuses on assessing ways of improving response to postal questionnaires at follow-up (e.g., giving incentives, use of shorter questionnaires, or use of reminders by telephone to return questionnaires) [5], [6]. The use of an additional method, such as a telephone interview, as a means of collecting follow-up data from mailed nonresponders has been reported in a limited number of social and health surveys [7], [8], [9], [10]. However, there are no studies that report the use of two or more methods of collecting follow-up data in a randomized clinical trial setting, where the comparison of interventions or treatments is of focal interest.
This observational study uses data obtained from a large clinical trial of back pain—the Back Skills Training Trial (BeST) [11]—where participants were randomized to one of two complex interventions. The use of the postal questionnaire was the primary method of data collection at follow-up, and if participants had not responded to mailed questionnaires, then attempts were made to capture data on a shorter version of the paper clinical forms through telephone interviews.
The primary aim of this article was to assess whether the estimated intervention effects differed for the two methods of data collection: postal questionnaires or telephone interviews, which were subsequently documented on paper questionnaires. A secondary aim was to explore how those who responded to the follow-up postal questionnaire differed to those who responded to a telephone interview. For both these aims, the following were evaluated: (1) demography and baseline outcomes; (2) response rates; (3) missing data; (4) internal consistency of each outcome, and (5) the 12 month outcome measures.