Improving Data Sharing in Research with Contect-free Encoded Missing Data

Abstract

Lack of attention to missing data in research may result in biased results, loss of power and reduced generalizability. Registering reasons for missing values at the time of data collection, or — in the case of sharing existing data — before making data available to other teams, can save time and efforts, improve scientific value and help to prevent erroneous assumptions and biased results. To ensure that encoding of missing data is sufficient to understand the reason why data are missing, it should ideally be context-free. Therefore, 11 context-free codes of missing data were carefully designed based on three completed randomized controlled clinical trials and tested in a new randomized controlled clinical trial by an international team consisting of clinical researchers and epidemiologists with extended experience in designing and conducting trials and an Information System expert. These codes can be divided into missing due to participant and/or participation characteristics (n=6), missing by design (n=4), and due a procedural error (n=1). Broad implementation of context-free missing data encoding may enhance the possibilities of data sharing and pooling, thus allowing more powerful analyses using existing data. Keywords: missing data; data pooling; data sharing; context free encoding ; HATICE (www.hatice.eu) is a collaborative project co-funded by the European Union's Seventh Framework Program (FP7, 2007-2013), under grant agreement No 305374. The research leading to these results has also been funded by the "Multimodal preventive trials for Alzheimer´s Disease: towards multinational strategies-programme: MIND-AD", Academy of Finland (291803) and VTR, Kuopio University Hospital (5772815).