Citation

Hsu P, Kraus JF, Afifi AA, Vaca FE, Schaffer K. Inj. Prev. 2010; 16(Suppl 1): A270.

Copyright

(Copyright © 2010, BMJ Publishing Group)

DOI

10.1136/ip.2010.029215.960

PMID

unavailable

Abstract

Introduction The number of subjects in a prospective cohort study often decreases over time (sometimes substantially). This "loss-to-followup" can lead to bias or distortion of results, and in studies of less serious conditions or injuries such as mild traumatic brain injuries (MTBI), may be especially problematic. This manuscript reports on the factors associated with the loss-to-followup that occurred in a cohort of emergency department-treated mild traumatic brain injuries, and on the use of attrition weights to adjust for the potential associated bias.

Methods The "exposed" group consisted of patients with MTBI (n=689 at initial assessment) and the "unexposed" were patients with non-head injuries (n=1318). All were drawn from five emergency departments in Southern California. The loss occurred over a 3-month period, when a second telephone survey was administered.

Results While the overall loss-to-followup rate was 37.8%, participant dropout occurred differentially (34.1% among MTBI patients vs 39.7% among the non-head-injured). A logistic regression model found several associated factors: age, education, race/ethnicity, employment and mode of injury, with assault having the highest odds of loss-to-followup (OR 2.23, 95% CI 1.46 to 3.39). Comparisons between unweighted versus attrition-weighted values resulted in minimal differences.

Conclusion Although it is preferable to minimize loss-to-followup through retention strategies and incentives (perhaps with additional focus on those factors identified in this study), employing attrition-weights may provide useful information.