We Need to Teach Our Health Science Students How to Handle Missing Data

• Activity Number: 435
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Teaching of Statistics in the Health Sciences
• Author(s): Charles Goldsmith*+
• Companies: Simon Fraser University
• Keywords: Missing data ; Imputation ; Sensitivity analysis ; Biased conclusions ; Data analysis ; Teaching

Abstract

The New England Journal of Medicine (NEJM) on October 4, 2012 contained a paper and an editorial on handling missing data. As a result we need to rethink how we introduce our health science students to handling missing data since NEJM made it clear that submitting studies with missing data were going to have extra scrutiny to make sure that sensible and supportable methods were used to handle missing data in the study. In many of the studies I have consulted on, many health scientists think it is acceptable to do what we might call an missing completely at random analysis, without doing a sensitivity analysis to measure the missingness impact on the conclusions drawn in the study. Indeed many claim they have not seen this in journals. We should revise our teaching to include missing data in all examples and use the reporting to discuss suitable strategies to support the conclusions in the study where there are missing data. Since NEJM is considered a leading journal for reporting on health science studies, other journals will likely follow their lead and make this a requirement for their journal as well. We need to be ahead of the curve to prepare our students for that eventuality.