Death and ADL Dependency After Scoring Zero on the NIHSS
A Swedish Retrospective Registry-Based Study

The National Institutes of Health Stroke Scale (NIHSS) has proven to be a good predictor of outcomes after stroke, where lower points are associated with more favorable outcomes.¹ However, a caveat of the scale is that it is weighted toward language and gross motor function, while being less sensitive for brainstem and cerebellar strokes as well as cognitive functions.^2,3 This bias is also supported by studies showing that the NIHSS underestimates infarct size in posterior circulation and right hemisphere strokes.^4,5

Previous studies have shown that up to one-third of all patients with “mild strokes” (no universal definition exists in this article, defined as 5 points or less on the NIHSS) at presentation still exhibit significant deficits 3 months after the stroke.^6-8 Of all strokes, mild strokes are in the majority with proportions up to two-thirds of all strokes.^9,10 The mildest among mild strokes are those with an immeasurable neurologic deficit, defined as zero points on the NIHSS (0-NIHSS) at hospital admission. Studies focusing on 0-NIHSS strokes are scarce, and one reason for this could be that these strokes occasionally are considered as transient ischemic attacks or simply have been omitted from previous studies.^11,12

Previous studies on patients with 0-NIHSS strokes have had study samples ranging from 12 to 108 patients.^13-16 The largest among these studies found that 47% had a modified Rankin Scale (mRS) score of 1 or more and 29% had a mRS score of 2 or more 3 months after a 0-NIHSS stroke.¹³ The other studies found varying degrees of cognitive and motor impairment, suggesting that these strokes are not entirely harmless.^14-16 Owing to the previous studies relatively small sample sizes on these presumably benign strokes, we aimed to use registry data to further characterize this group of patients, study outcomes, and investigate factors associated with death and ADL dependency at 3 months.

We found that among previously ADL-independent patients who had a 0-NIHSS stroke, 6% had either died or were ADL-dependent 3 months after the stroke. We also found that old age and prestroke physical inactivity significantly increased the odds of an adverse outcome among these patients.

In total, 24% of the patients in our study had a 0-NIHSS stroke at admission, which contrasts previous studies where incidence rates have ranged from 1% to 4%.^13,16 The discrepancy in incidence compared with previous studies is not entirely clear but could be attributable to a few different factors. First, the previous studies have included patients between 2003 and 2013.^13,16 During the same time, stroke severity has decreased, increasing the incidence rates for mild strokes.^22,23 Second, a 27-month long national stroke awareness campaign was conducted in Sweden between 2011 and 2013, increasing the proportion of patients seeking medical care for their stroke regardless of severity.^23,24 Third, Swedish health care is tax funded, apart from a minor copayment made by the patient, making it available and affordable for all citizens. This is not the case in the previous studies which were from Switzerland and the United States, where health care instead relies on a health insurance system where patients, although they are covered, often have to pay a certain percent of the total cost by themselves. This makes health care relatively expensive for the individual, which probably make patients with perceived mild symptoms less prone to seek medical attention compared with patients with severe symptoms.

However, when we compare our results with national data in Sweden, 0 point on the NIHSS has been the most common score among all stroke patients since NIHSS was introduced in the annual reports 2012 (before that stroke severity was only measured by the level of consciousness in Riksstroke), with percentages rising from ∼13% of all strokes in 2014 to ∼17% in 2019.^10,25 At the same time, the missing data levels of admission NIHSS have decreased from 52% to 39%, indicating a bias toward higher NIHSS with more missing data, which probably partly explain the relatively high proportion of 0-NIHSS strokes in our study where the levels of missing data for NIHSS are considerably lower.^10,25 Comparable national data from Norway show similar proportions as Sweden, with 0-NIHSS strokes accounting for 15% of all strokes in 2020, while data from England and Wales are lower at 7% between 2013 and 2017.^26,27 Finally, because this is a retrospective study with no standard protocol for all patients, differences in NIHSS status assessment probably differ more than in prospective studies with more rigid assessments. However, all physicians caring for stroke patients in the emergency setting in Gothenburg receive training in how to perform the NIHSS since 2008, which vouches for better quality of the data compared with whether the physicians had not received any training.

Looking at patient characteristics, one of the most prominent differences is that atrial fibrillation was only half as prevalent among patients with a 0-NIHSS stroke compared with those with 1 point or more. This is not surprising because atrial fibrillation is a risk factor for cardioembolic strokes, which in turn are associated with more severe strokes.²⁸ Another interesting difference is the gap in prestroke physical activity, where 6 of 10 patients with a 0-NIHSS stroke were physically active, compared with 4 in 10 among those with 1 point or more on the NIHSS. This is also in line with previous studies from the same study population showing that prestroke physical activity reduces stroke severity.^29,30 The difference in ADL status before the stroke between 0-NIHSS strokes and those with 1 point or more on the NIHSS at admission is also noteworthy. We believe that this is probably because ADL status, just like physical activity, to some extent is a measure of a person's overall health, where those with better health seem to do better in stroke severity.

The finding that the median length of hospital stay is shorter among 0-NIHSS patients compared with those with 1 point or more on the NIHSS was expected because the need for rehabilitation would be less extensive and the risk of in-hospital complications was lower.³¹ The finding that fewer receive reperfusion therapy among patients with 0-NIHSS stroke was also expected because the risk vs reward of thrombolysis among mild strokes is less clear compared with more severe strokes. However, the use of thrombolysis has increased in recent years even among mild strokes, largely after the insight that these strokes also can give rise to disabling symptoms.³² Recent findings also show that among mild strokes, patients with both disabling and nondisabling symptoms benefit equally from reperfusion therapy.³³

The difference in the rates of myocardial infarctions between those with 0-NIHSS strokes and those with 1 point or more on the NIHSS was interesting but not surprising given the variations in patient characteristics between the groups, where the latter were older and more ADL-dependent before the stroke, indicating a less favorable prestroke health status. It was reassuring, however, that only 1% suffered a myocardial infarction during acute care, which is consistent with a study from Austria.³⁴ The finding that 1 in 3 reports verbal problems 3 months after the stroke indicates that the NIHSS, although focused on language, is not perfect in detecting all problems. Considering the levels of dysphasia and dysarthria diagnoses at only 5% and 7%, the high levels of patient-reported problems hint that more extensive symptom screening is warranted during the acute care episode.

The only previous study on death and functional outcomes after a 0-NIHSS stroke showed a mortality rate of 4% 1 year after the stroke.¹³ Because we studied mortality at 3 months, our finding of a mortality rate of 2% is not entirely comparable. Regarding functional outcomes, ADL dependency as defined in our study would be equal to at least 4 points on the mRS, and hence, comparisons to the previous study are not possible. However, the patient-reported outcome regarding whether patients still experience problems from the stroke is similar to 1 point or more on the mRS. Our finding that 59% of the patients with a 0-NIHSS stroke experienced problems is slightly more than the previous study which reported that 47% had a mRS score of 1 or more at 3 months.¹³

A major strength of our study is that all stroke units in Gothenburg report to both Riksstroke and the Väststroke register, where the former maintained a coverage (the proportion of stroke patients admitted to hospitals that is also registered in Riksstroke) of 85%–87% during the study period for the included sites. Validation studies have also shown a slight overdiagnosis of stroke in routine care; hence, the true coverage is estimated to be in excess of 90%.³⁴ The excellent coverage in combination with the structure of Swedish health care, where access to care is available to every citizen, minimizes selection bias. Other major strengths are that the data are prospectively collected with low levels of missing data for most variables, which together with extensive internal and external validation processes vouches for high-quality data.³⁵

As in all observational studies, the choice of variables is limited. To facilitate comparison with previous studies, it would be desirable to compare prestroke mRS with mRS at 3 months. This, however, was not possible because prestroke mRS was unavailable. Other drawbacks are that prestroke physical activity and the variables from the three-month follow-up were patient-reported and not objectively measured. Another limitation is that there was a small discrepancy in patient registrations between Riksstroke and the Väststroke register which gave rise to missing data in some of the background variables for a small proportion of the patients. However, the total level of missing data was low, and the similar results between the imputed data set and the complete case analysis support the accuracy of the findings.

Death is impeccably reported to Statistics Sweden. However, ADL dependency at 3 months had 14% missing data among patients with a 0-NIHSS stroke. We compared the patients who answered the three-month follow-up with those who had missing data on the variable and found that the latter were 6 years younger on average. In addition, they were twice as often born abroad, more often had diabetes, smoked, and were physically inactive. The lower age would decrease the risk of a negative outcome, whereas the sedentary lifestyle would increase the risk; hence, it is hard to predict the direction of possible bias. However, 14% missing data are still relatively acceptable, which in combination with no clear direction of the bias makes us believe that the results are robust. By including all patients in a large area, we believe that our results are generalizable to patients in Sweden and in similar health care organizations abroad.

We found that 6% of the patients were either dead or ADL-dependent 3 months after a 0-NIHSS stroke, highlighting that these strokes are not always benign, and more extensive symptom screening might be needed. The older and physically inactive are at greater risk of an adverse outcome.

We thank the patients, the Väststroke collaboration, Riksstroke, and the included stroke units.