In-hospital stroke care pathways: possibilities for improvement

The aim of this thesis was to identify specific areas for improvement along the in-hospital care pathways for stroke.

In the Emergency Department (ED), we investigated whether we could develop an accurate prediction model for patient presentations to the ED for stroke using public meteorological and calendar data. With ongoing climate change, rising temperatures and reduced air quality have the potential to increase the incidence of stroke. In a national cohort study of more than 250,000 ED patient presentations for acute stroke, we found clear temporal patterns, with more strokes on Mondays and pronounced seasonal effects. The variation in the absolute number of ED patient presentations was too small to contribute to ED crowding and to be relevant for ED staff planning decisions.

In a hospitalised stroke population we found that acute illness of stroke patients, so called vital instability defined using an internationally renowned Early Warning Score (EWS), is associated with death or dependency at 3 months after stroke. This association is mainly driven by the dichotomous consciousness component of the EWS used in our study, resulting in high alert rates in patients with stroke. Consequently, we proposed four possible adaptations to current EWS scoring systems to capture (the change of) vital instability in stroke patients.

In a large Dutch Intensive Care (ICU) cohort, we examined long-term mortality of ICU-admitted stroke patients. 30-day mortality of patients with an ischemic stroke was 31% and in patients with an intracerebral haemorrhage 42%. Mortality risk decreased with increasing duration of survival time after ICU admission. This means that short-term mortality rates cannot be extrapolated to the longer term in ICU-admitted stroke patients. A developed and validated prediction model shows the following important predictors of 30-day mortality: age, the lowest consciousness score within 24 hours after ICU admission, acute physiological disturbance, the application of mechanical ventilation and the occurrence of acute renal failure.

Finally, we analyse how and whether mortality prediction can be useful in the clinical practice of ICU-admitted stroke patients. We identify the challenges of successful clinical implementation of prediction models and discuss the role of the so-called 'Withdrawal of Care' bias, which can confound the modelling of mortality predictions.