Missing Data in Patient-Reported Outcomes Research: Utilizing Multiple Imputation to Address an Unavoidable Problem

Patient-reported outcomes (PROs) have become a focus in postoperative surgical care. Unfortunately, studies using PROs can be subject to missing data, which may lead to biases or inaccurate conclusions. Multiple imputation (MI) is a statistical method for addressing missing data in clinical research. The aim of this study was to explore MI as a way to address missing data in PRO research.

A working example of MI using real-world data was performed using the BREAST-Q PRO measure in postmastectomy reconstruction. A retrospective review of immediate tissue expander breast reconstruction patients in 2019 was conducted to compare BREAST-Q physical well-being of the chest scores between prepectoral and subpectoral cohorts at 2 weeks postoperatively. The observed dataset and three hypothetical missingness situations were created to assess how increasing missingness affects MI results.

Overall, 916 patients were included in the analysis. When excluding patients with missing information and solely performing analysis on the completed cases, prepectoral patients had significantly higher physical well-being of the chest scores at 2 weeks postoperatively; however, this trend was reversed with increasing missingness scenarios, where subpectoral patients had higher scores. In comparison, all MI results showed that prepectoral patients had higher scores on average compared with subpectoral patients regardless of missingness scenario.

MI demonstrated consistent results with increasing missingness scenarios, whereas performing analysis in higher missingness scenarios without MI led to varying results. This working example emphasizes the need for missing data methodology to be considered in PRO research.