[Delirium in stroke patients : Critical analysis of statistical procedures for the identification of risk factors].

backgroundDelirium is a relevant complication following an acute stroke. It is a multifactor occurrence with numerous interacting risk factors that alternately influence each other. PROBLEM: The risk factors of delirium in stroke patients are often based on limited clinical studies. The statistical procedures and clinical relevance of delirium related risk factors in adult stroke patients should therefore be questioned.

methodThis secondary analysis includes clinically relevant studies that give evidence for the clinical relevance and statistical significance of delirium-associated risk factors in stroke patients. The quality of the reporting of regression analyses was assessed using Ottenbacher's quality criteria. The delirium-associated risk factors identified were examined with regard to statistical significance using the Bonferroni method of multiple testing for forming incorrect positive hypotheses. This was followed by a literature-based discussion on clinical relevance.

resultsNine clinical studies were included. None of the studies fulfilled all the prerequisites and assumptions given for the reporting of regression analyses according to Ottenbacher. Of the 108 delirium-associated risk factors, a total of 48 (44.4%) were significant, whereby a total of 28 (58.3%) were false positive after Bonferroni correction. Following a literature-based discussion on clinical relevance, the assumption of statistical significance and clinical relevance could be found for only four risk factors (dementia or cognitive impairment, total anterior infarct, severe infarct and infections).

conclusionsThe statistical procedures used in the existing literature are questionable, as are their results. A post-hoc analysis and critical appraisal reduced the number of possible delirium-associated risk factors to just a few clinically relevant factors.