Non-response Bias in Patient Reported Outcomes Research for Breast Reconstruction: Lessons from a Multi-institutional Prospective Cohort Study
Nicholas L. Berlin, MD, MPH, Jennifer B. Hamill, MPH, Ji Qi, MS, Hyungjin M. Kim, ScD, Andrea L. Pusic, MD, MHS, Edwin G. Wilkins, MD, MS.
University of Michigan, Ann Arbor, MI, USA.
PURPOSE: Survey-based research has emerged as a primary method of evaluating the impact and outcomes of breast reconstruction in an era of patient centered care. Despite increased utilization of patient reported outcomes measures, such as the BREAST-Q surveys, there are few studies assessing non-response among study participants. Differential non-response to surveys threatens the generalizability and validity of studies evaluating patient reported outcomes. The objectives of this study were to identify independent variables associated with non-response to surveys following breast reconstruction to aid in the design and improvement of future patient-reported outcomes research. METHODS: The Mastectomy Reconstruction Outcomes Consortium is a multi-institutional prospective cohort study involving 11 leading medical centers from the United States and Canada. Data were collected through a combination of self-reported patient surveys and medical record review. Patient reported outcomes including satisfaction with breast and with care (BREAST-Q), depression (PHQ-9), and anxiety (GAD-7) were collected preoperatively and at several intervals postoperatively. Non-response rates to BREAST-Q, PHQ-9, and GAD surveys were measured at one week, three months, one year, and two years following breast reconstruction. Complication rates between questionnaire responders and non-responders were compared at one year. Multivariable logistic regression models were designed with non-response as the dependent variable. RESULTS: Among 3,083 women included in the analytic cohort, 2,060 (66.8%) underwent implant-based, 857 (27.8%) underwent autologous, and 166 (5.4%) underwent mixed breast reconstruction. Non-response rates to BREAST-Q surveys at one week, three months, one year, and two years postoperatively were 14.5%, 21.8%, 27.8%, and 34.4% respectively. At one year following reconstruction, the rate of non-response ranged from 8.0% to 41.4% among participating centers. Patient race and ethnicity, annual household income, and education were all associated with non-response to the postoperative surveys. Patients undergoing implant-based breast reconstruction were approximately half as likely to respond to BREAST-Q surveys at one year when compared with women who underwent autologous reconstruction. Neither satisfaction with care nor incidence of clinical complications was predictive of non-response to the surveys. CONCLUSION: Although patient reported outcomes are increasingly utilized to assess and compare the impact of breast reconstruction procedures, there are few studies evaluating differential non-response rates to survey instruments among study participants. In a multi-institutional, prospective cohort study, racial and ethnic minorities, and women from lower income and education groups were less likely to respond to follow-up surveys. Additionally, women who underwent implant-based breast reconstruction were approximately half as likely to respond to BREAST-Q surveys at one year when compared with women who underwent autologous reconstruction. Researchers studying patient-reported outcomes in breast reconstruction should assess and adjust for non-response bias. Future studies are needed to design and implement effective approaches to improving response rates for patients at high risk of loss to follow-up.
Back to 2017 Program