Recruitment bias resulted in poorer overall health status in a community-based control group

Abstract

Objective: During a prospective community-based incidence study of parkinsonism, a control group was recruited for comparison with the incident patients. This study compared the demographic and health status of recruited vs. nonrecruited controls.

Study design and setting: For each incident patient, attempts were made to recruit an age-gender matched control from the same general practice or, failing that, from a previously identified community cohort of people aged over 64 years who had expressed an interest in taking part in future research. Recruited controls were compared with those who were approached but not recruited in terms of age, socioeconomic status, gender, several measures of health status, and survival.

Results: A total of 74 controls (40%) were recruited out of 186 potential controls who were approached. Recruited controls scored slightly worse than nonrecruited controls on every measure of health status, which reached statistical significance for numbers of acute prescriptions and major surgical procedures. There were no significant differences in age, gender, socioeconomic status, or survival.

Conclusion: The control cohort was affected by recruitment bias, which suggested that recruited controls had slightly poorer health compared to nonrecruited controls. This bias may reduce differences in health when comparisons are made between the controls and the parkinsonian patients.