Handling of Missing Outcome Data in Traumatic Brain Injury Research: A Systematic Review

Abstract

Traumatic brain injury (TBI) research commonly measures long-term functional outcome, but studies often suffer from missing data as patients are lost to follow-up. This review assesses the extent and handling of missing outcome data in the TBI literature and provides a practical guide for future research. Relevant electronic databases were searched from January 1, 2012 to October 27, 2017 for TBI studies that used the Glasgow Outcome Scale or Glasgow Outcome Scale-Extended (GOS/GOSE) as an outcome measure. Studies were screened and data extracted in line with Cochrane guidance.A total of 195 studies, 21 interventional, 174 observational, with 104,688 patients were included. Using the reported follow-up rates in a mixed model, on average 91% of patients were predicted to return to follow-up at 6 months post-injury, 84% at 1 year, and 69% at 2 years. However, 36% of studies provided insufficient information to determine the number of subjects at each time-point. Of 139 studies that did report missing outcome data, only 50% attempted to identify why data were missing, with just 4 reporting their assumption on the “missingness mechanism.” The handling of missing data was heterogeneous, with the most common method being its exclusion from analysis. These results confirm substantial variability in the standard of reporting and handling of missing outcome data in TBI research. We conclude that practical guidance is needed to facilitate meaningful and accurate study interpretation, and therefore propose a framework for the handling of missing outcome data in future TBI research.

Emergency department observation of mild traumatic brain injury with minor radiographic findings: shorter stays, less expensive, and no increased risk compared to hospital admission

“…Indeed, this is a very frequent problem in the TBI population. 35 Our study demonstrates the importance of a thorough follow-up plan, as 2 patients developed a chronic subdural hematoma requiring delayed surgery. Although our protocol specified a 1-month follow-up in the neurosurgery clinic, it did not provide a formal mechanism for scheduling or confirming the visit, a notable weakness in its design.…”

Imputation of Ordinal Outcomes: A Comparison of Approaches in Traumatic Brain Injury

“…1,2 However, missing outcome data is a common problem in TBI research, and for longitudinal studies completion rates at six months can be lower than 70%. 3 This is important, since completecase analyses may introduce bias and at least reduce power. 4 Last observation carried forward (LOCF) is a recommended single-imputation method for dealing with missing data in TBI research clinical trials because it is conservative with respect to evaluation of the intervention.…”

“…+/-2 weeks; https://tracktbi.ucsf.edu/researchers), but in some contexts this can lead to a substantial amount of missing data). 3 Consequently other studies have defined more pragmatic protocol windows (e.g. -1 month to +2 months, see also 14 ).…”

“…It is thus of the utmost importance to implement measures for avoiding missing data in the first place. Nevertheless, in practice, missing values due to loss-to-follow-up will always occur and should be addressed effectively 3. There is a wide consensus that statistically sound imputation of missing values is beneficial for both the reduction of bias and for increasing statistical power.…”

Focus on neuro-critical care: combined interventions to improve relevant outcomes

“…In that perspective, relatively short follow-up intervals to assess outcomes are not always appropriate. Long-term follow-up, however, is hard to achieve, and at least 15% loss to follow-up at 1 year is to be expected [19].…”