Vasileios-Arsenios Lioutas, MD


de Man-van Ginkel JM, Hafsteinsdóttir TB, Lindeman E, Ettema RGA, Grobbee DE, and Schuurmans MJ. In-Hospital Risk Prediction for Post -stroke Depression: Developmentand Validation of the Post-stroke Depression Prediction Scale. Stroke. 2013

Depression is a well known post-stroke complication. What predicts development of depression in an individual patient is less well understood. In this interesting study, Dr. van Ginkel et al. attempt to develop a predictive model that will allow in-hospital identification of patients that will later develop depression. 

To briefly describe the study methodology, a number of clinical (mainly stroke-related) and socio-demographic factors were recorded within the first post-stroke week and before patient’s discharge. Sociodemogrpahic factors included marital status and perceived level of social support. Subsequently, diagnosis of depression was made in the 6th-8th post stroke week. The prediction model was then internally validated.
As one would perhaps intuitively expect, past history of depression was strongly associated with post-stroke depression (OR 7.22). Angina pectoris showed a positive and hypertension an inverse association, although clearly no causative implications can be made on the basis of this study only. One interesting fact that emerged was the strong association of the “Dressing” element of Barthel Index with subsequent development of depression: Patients completely dependent on others’ help were more likely to be depressed in follow up (OR 1.57 but with CI 0.80-3.09) and more importantly, those needing only partial help were much less likely to develop depression (OR 0.26, CI 0.08-0.82). Although it is difficult to know whether there is a neurobiologic underpinning, it seems plausible from a purely psychologic standpoint that the subjective feeling of “helplessness” in performing such a rudimentary daily task plays a significant role in developing post-stroke depression.
The study has several limitations: Only communicative patients were included, therefore the sample is not accurately representative of the whole spectrum of a stroke patient population. Although internally validated, rigorous external validation is necessary to further assess its utility and explore its generelizability. It would be interesting to include the stroke location in the model, as it is very likely that certain brain areas are more strongly associated with post-stroke than others. Lastly, it would be interesting to alter the time-point at which diagnosis was made – perhaps choosing a later time (eg 10-12 weeks or later) would yield different results.