The incidence of stroke seems stable, but the case fatality has decreased in recent decades [1–3]. This trend has been associated with better stroke care (e.g. stroke unit care) [4] and improved preventative treatment of cardiovascular risk factors [1, 3, 5]. Experiencing complications after stroke is, on the other hand, associated with increased mortality and length of hospital stay in acute stroke patients [6]. The most common complications include fever, pain, progressing stroke and infections, but complications such as myocardial infarctions, pulmonary embolisms and cardiac arrest may also occur [6–8]. Most complications have their onset within the first week after stroke [9]. Disability and mortality increase with an increasing number of complications experienced, especially progressing stroke, chest infections and other infections [10, 11].

One important consequence of stroke unit (SU) care has been the prevention and early treatment of complications after stroke, especially those related to immobility [12]. However, studies examining the frequency of complications early after stroke show great discrepancies, with frequency of falls ranging from 2 to 25 % and pneumonia from 9 to 22 % [6, 13]. Such inequalities could be caused by inconsistent definitions of complications, case mix, differences in the completeness of recording, or variation in prevention and treatment of complications among hospitals. Some of these variations might also reflect changes in stroke characteristics and treatment over time. Significant changes in recent decades include the extended use of computer tomography and magnetic resonance imaging, making stroke diagnostics more precise [14], and more frequent use of thrombolysis in acute treatment [15]. Improved physiological homeostasis and early mobilization have also received attention [16].

No systematic studies have thoroughly investigated whether the frequency of complications after stroke have changed in patients treated in dedicated stroke units. The stroke population has also changed in recent decades with a greater proportion of mild strokes probably due to better primary prevention [17]. Because severity of stroke is found to be strongly associated with the risk of complications [9], severity should be considered in comparisons of the frequency of complications in stroke patients over different periods.

The primary aim of this study was to assess the changes in frequency of patients with complications in the acute phase after stroke from 2003 to 2013, adjusted for the severity of stroke. The secondary aims were to assess the changes in complications within the different severity groups and to assess how the different types of complications changed from 2003 to 2013 after adjusting for stroke severity.

Our primary hypothesis was that the frequency of patients with one or more complications has decreased over the last decade in patients treated in a comprehensive SU.

In 2003, 664 patients with suspected stroke were admitted to the SU between January 1, 2002 and May 15, 2003. All stroke patients entering the ward during this 16.5 month period were screened for inclusion, and 489 patients were included.

In 2013, 402 patients with suspected stroke were admitted to the SU during the inclusion period in 2013, lasting from September 17, 2012 to December 13, 2013, except for a few weeks during the holidays; hence the inclusion period covered 13 months. A final group of 185 patients contributed to the data sample. Of these, 158 (85.4 %) arrived in hospital within 24 h after symptom onset. Ninety-four patients (50.8 %) were discharged during the first week, and were contacted by telephone to complete the registration. Two patients (1.1 %) were not reached by telephone and considered lost to follow-up, and we assumed that no new complications occurred in these patients after discharge. The flow of patients is shown in Fig. 1.

Data from the Norwegian Stroke Register enabled comparison of the 131 patients who were eligible but missed for inclusion with those included in the 2013 cohort. Those missed for inclusion were in mean (SD) aged 79.9 (8.8) years, which was significantly older, P = 0.004, with shorter length of stay (LOS) in hospital, mean (SD) 6.3 (6.7) days, P < 0.001. Their mean (SD) SSS-score was higher, 39.8 (16.5) points, which was borderline significant, P = 0.055. Categorization of SSS-score by severity groups showed significantly more patients in the group of very mild stroke among those missed for inclusion compared with those included, 29.8 % versus 15.1 %, P = 0.002.

To our knowledge, this is the first study to show 10-year time trends for complications in patients treated in a dedicated comprehensive stroke unit. The study shows that complications are still common in the acute phase after stroke, but are generally less common in 2013 than in 2003, also after adjusting for differences in stroke severity between the two cohorts. The reduction in complications was most pronounced in those with moderate stroke. Regarding specific complications, the reduction was significant for progressing stroke and myocardial infarction. There was also a trend towards fewer chest infections.

Despite an overall reduction in the risk of experiencing one or more complications from 2003 until 2013, it seems that prevention of complications was not improved for patients with severe strokes or mild strokes. However, patients with moderate strokes showed a 29 % risk reduction over this period. One possible explanation is the improved monitoring and stronger focus on physiological homeostasis that has been implemented in the ward over the past years, and an increased use of thrombolysis. In the present study, the effect of thrombolytic therapy on complication risk was assessed by adjusting for thrombolysis in the regression analysis.

The significant decrease in progressing stroke particularly in patients with moderate and severe stroke indicates that early detection of risk and treatment of progression might have improved considerably. It is a strength in our study, that progression was defined and assessed in the same way by well-trained nurses on duty both in 2003 and in 2013. The inter-observer agreement has shown to be quite good using this method and the reduction observed reflects probably a real difference [20]. Patients with moderate stroke seem to have the most pronounced reduction of frequency of progressing stroke. The reason for less reduction in the severe group might be that progression of symptoms in this group quite often is due to brain edema, which might not be reduced by monitoring and better control of the physiological homeostasis. Mild strokes on the other hand have quite a low risk of progression and due to low number of events it will be difficult to find differences in this group. Hence, the patients with moderate stroke might be the group where system factors like blood pressure, oxygen saturation and temperature are likely to be of most importance for progression. In a review from the Cochrane Library they also found a trend towards reduced rate of neurological complications like progressing stroke in stroke units with continuous monitoring which is quite similar to our result [27]. However in the Cochrane review no stratification by severity of the stroke was performed.

Previous studies have found frequencies of progressing stroke ranging from 11 to 43 % [28, 29]. The great variability in the reported incidence has different reasons. One important reason is that it has been difficult to construct precise definitions of exacerbations of a stroke and various terms such as “progressive stroke”, “early neurologic deterioration” or “stroke in evolution” have been used and have not made studies of these condition easier [30]. Furthermore the use of different scales, cut-off values and time perspectives in the definitions makes conclusions and comparisons even more difficult [31]. Hence, a consensus on the definition of progressing stroke is needed to enhance further research within this field.

A significant decrease was also seen for acute MI, from 22 cases (4.5 %) in 2003 to only one case (0.5 %) in 2013. This very low incidence makes it difficult to conclude on the association between MI and stroke severity and to adjust for thrombolysis, as the single patient receiving MI in 2013 received thrombolysis while none of the patients in 2003 did. The method for measuring Troponin T was changed between 2003 and 2013, but the criteria for Troponin levels according to MI were quite similar. Hence the reduction of MI is probably also a real difference.

The significant decrease in the occurrence of MI is supported by an Austrian study [32] carried out from 2006 to 2013, where MI was found in 1 % of stroke patients during hospital stay. Maintenance of physiologic homeostasis is one important factor contributing to the good outcome of SU care [16], and improvement in such regulation might influence the risk of post-stroke MI. Another important factor is continuous monitoring of patients (ECG and oxygen saturation) which makes it easier to detect signs of ischemia on the heart (atrial fibrillation and heart failure leading to hypoxia) [4], so that prophylactic treatment can be initiated early in risk patients. A report from the Norwegian Institute of Public Health shows that the incidence of first-time-MI is decreasing in the general elderly Norwegian population, which possibly also has influenced our findings [33].

In the 2003 cohort fever, chest infections, progression of stroke, MI and UTI occurred significantly more often in patients with severe than with milder strokes but in the 2013 cohort such an association to severity occurred only for fever, chest infection and UTI and a trend for MI while progressing stroke had no such association to severity. For MI with only one case, association to severity is not possible to assess. It is not possible to conclude whether the reduction in frequency of progression of symptoms have contributed to less association to severity of the stroke.

The use of thrombolysis in 30 patients (16 %) in 2013 versus no patients (0 %) in 2003 might have influenced the complication rate. However, after adjusting for thrombolytic therapy no significant changes were found. Hence, thrombolytic therapy has probably not contributed very much to the changes in complication rate between 2003 and 2013.

A major strength of this study was that patients were followed up daily through a prospective design, minimizing the probability of omitting incidents of complications. Having a comparable cohort admitted to hospital 10 years ago reinforces the comparison, as the definitions of complications were the same. The studies were performed in the same SU which offers evidence-based treatment according to the Norwegian guidelines.

The study had some limitations. Fewer patients were screened for inclusion in 2013 compared to 2003. This was due to a reduction in the entrance area and also to a reduction in the number of beds for the SU at St. Olavs Hospital between 2003 and 2013. Furthermore, a considerable amount of patients were missed for inclusion. LOS was significantly shorter in the 2013 cohort, but those who were missed for inclusion in 2013 had an even shorter LOS. Some patients with mild stroke were discharged so rapidly that it was difficult to get time to include them in the study. This probably explain why there were significantly fewer patients with very mild stroke in those included in 2013 than in those who were missed. Hence, we cannot say that the differences in stroke severity between our 2003 and 2013 cohort represent an actual difference in these populations. This is taken into consideration by adjusting for stroke severity in all analyses, as stroke severity strongly influences the risk of experiencing complications in the acute phase. Hence, we do not think the differences in stroke severity influence the results very much. One should consider if the shortened LOS and the follow-up by telephone in 2013 have limited the ability to capture some complications that might occur after discharge. However, most complications occur within the first days after stroke [9], and the follow-up via telephone was performed thoroughly by describing and asking for each complication specifically, hence we do not believe that these features have made a huge impact on our results.

Another limitation was that the time limit for arrival in hospital after onset of symptoms was expanded from 24 h in 2003 to 48 h in 2013, to include sufficient patients in the study and increase the power. This increases the risk of missing complications that occurs after the stroke but before admission to hospital, as these were not registered. This matter especially for progressing stroke and other temporary complications such as falls and seizures; as most of the other complications would still be present by admission and therefore be registered. However 85 % were admitted within 24 h and a post hoc analysis showed almost identical results when patients arriving after 24 h were excluded from the analyses, and this is considered not to have affected our results significantly.

Complications are still common in the acute phase after stroke in patients treated in a comprehensive stroke unit. However, after adjusting for age, thrombolysis and stroke severity, the odds of experiencing one or more complications have decreased significantly from 2003 to 2013. The reduction was greatest for progressing stroke and myocardial infarction. Overall reduction in complications was mainly seen for patients with moderate stroke. The reduction in complication rate cannot be explained by the introduction of thrombolytic therapy but might be explained by an increased focus on continuous monitoring and stabilizing physiological homeostasis over the past 10 years.

CT, computed tomography; ECG, electrocardiogram; LOS, length of stay; MI, myocardial infarction; RD, risk difference; SSS, scandinavian stroke scale; SU, stroke unit; UTI, urinary tract infection