The effect of sampling bias on generalizability in intervention trials after brain injury

ABSTRACT

Objective and design: The first objective of this descriptive study was to explore the effect of strict inclusion and exclusion criteria on patient recruitment in cognitive rehabilitation; the second was to analyse the representativeness of study patients for all eligible stroke and traumatic brain injury patients referred to the department from September 2011 to November 2014.

Methods: The setting was multi-professional in- and outpatient brain injury rehabilitation in a university hospital. Participants were enrolled consecutively in acute (< 4 months) or sub-acute (4–12 months) phases of rehabilitation after injury.

Results: Nine-hundred-and-forty-four patients were screened. Many failed to meet the inclusion and exclusion criteria (acute phase = 90.5%; sub-acute phase = 85.3% excluded). The most frequent exclusion criteria were age and multiple co-morbidity. The most striking differences between eligible participants and non-participants were the latter’s lower educational and occupational levels.

Conclusion: Results indicate the need for a closer connection between selection criteria and study aims. Strict inclusion and exclusion criteria render the study population less representative and, thus, reduces generalizability. Data collection is prolonged. This study, instead, advocates the use of broad inclusion criteria and common data elements and the need to report data from the total patient population. Increased awareness of the participants’ socioeconomic backgrounds is also needed when recruiting to research studies.

KEYWORDS:

• Sampling bias
• brain injuries
• treatment outcome
• cognitive training

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.