Randomized controlled trials of socially complex nursing interventions: creating bias and unreliability?
Article first published online: 9 DEC 2003
Journal of Advanced Nursing
Volume 45, Issue 1, pages 84–94, January 2004
How to Cite
Lindsay, B. (2004), Randomized controlled trials of socially complex nursing interventions: creating bias and unreliability?. Journal of Advanced Nursing, 45: 84–94. doi: 10.1046/j.1365-2648.2003.02864.x
- Issue published online: 9 DEC 2003
- Article first published online: 9 DEC 2003
- Submitted for publication 11 September 2002 Accepted for publication 13 August 2003
- randomized controlled trials;
- socially complex interventions;
- research methodology;
- nursing interventions;
- Consolidated Standards of Reporting Trials
Background. The randomized controlled trial is viewed by many researchers as the ‘gold standard’ research design. It is used increasingly to evaluate the effectiveness of socially complex activities such as nursing interventions. This use is seen by many commentators as problematic, while others are concerned about the quality of many published trial reports. One area of concern is that of intervention bias: the impact that a sentient intervention, such as a nursing one, may have consciously or unconsciously on study outcomes. This paper reports on an analysis of intervention definitions and possible intervention bias in 47 reports of randomized controlled trials of nursing interventions published in 2000 or 2001.
Aims. This study evaluates four characteristics of the included reports: intervention sample size, intervention definition, involvement of intervention nurses in other aspects of the trial, and the claimed generalizability of results.
Methods. Reports of randomized controlled trials published in 2000 or 2001 were identified. Full-text versions of 47 papers were obtained and information about the four characteristics was extracted and analysed.
Results. Problems relating to possible intervention bias were identified in each of the papers. Inadequate intervention definition was the commonest problem, leading to difficulties in calculating the ‘intervention dose’ and in replicating or generalizing from the studies.
Discussion. None of the included studies met the requirements of the Consolidated Standards of Reporting Trials. Four types of intervention bias were identified, and their possible implications for the reporting of trials of nursing interventions are discussed. This was a small-scale study, limited by time and resources. Its results are suggestive of a major problem of intervention bias but larger-scale investigations are necessary to quantify its extent.
Conclusions. Intervention bias is potentially a problem in randomized controlled trials. Lack of detail about interventions in published papers could be corrected by stricter adherence to guidelines such as the Consolidated Standards of Reporting Trials, but this will not correct the underlying problem of inadequate study design that appears to be widespread in randomized controlled trials of nursing interventions.