Document Type

Capstone Experience

Graduation Date


Degree Name

Master of Public Health



First Committee Member

Jane Meza, PhD

Second Committee Member

Lorena Baccaglini, PhD

Third Committee Member

Kaleb Michaud, PhD


Introduction: Participant attrition is a problem common to many longitudinal studies, and when it occurs nonrandomly, it can impact the study’s validity and generalizability. Identifying factors associated with attrition can help to detect bias and aid in developing targeted interventions to reduce attrition. Methods: Logistic regression and Cox proportional hazards regression models were constructed separately for patients with rheumatoid arthritis, systemic lupus erythematosus, and other rheumatic diseases who participated in the FORWARD study between 1998 and 2018. Results: Patient characteristics associated with attrition included male sex, younger age, non-White race, and less education, each of which was identified in multiple models. Score indicating poorer function or greater disease activity on the Health Assessment Questionnaire Disability Index (HAQ), Short Form Health Survey Mental Component Scale (MCS), and Rheumatic Disease Comorbidity Index (RDCI) were associated with dropout in certain groups. Method of recruitment to the study was significant, though its specific impact varied by diagnostic group and analytical technique. Discussion: Male sex and less education as predictors of dropout concurred with previous studies, as did the relatively greater significance of socioeconomic factors than health-related factors. The associations with poorer scores on health indices were consistent with researchers’ logic that patients in poorer health would be more likely to drop out. Conclusion: Patients of male sex, non-White race, younger age, and less education and patients with poor score on health indices were more susceptible to early dropout. FORWARD is advised to develop interventions targeted at retaining at-risk participants, such as achievement tracking to engage younger audiences, accommodations for patients with less education, outreach to patients who report infections, and automated communications triggered by poor scores on health indices.