Incidence, characteristics and clinical relevance of acute stroke in old patients hospitalized with COVID-19

This monocentric cross-sectional retrospective study analyzed patients hospitalized in geriatric wards of Geneva University Hospitals. Hospitalized patients with COVID-19 had one or more of these clinical features: a) pneumonia with a severity assessed by the CURB-65 score ≥ 2, b) new dependence on oxygen or increase of oxygen needs, c) a respiratory rate ≥ 20 breaths/minute, d) a decompensated chronic disease, e) severely altered general state of health, f) deteriorating clinical course.

We analyzed data from a total of 265 patients hospitalized with SARS-CoV-2 infection between March 13th, 2020 and May 17th, 2020. The geriatric hospital is part of Geneva University Hospitals, which were responsible during the first wave of the COVID-19 pandemic for hospitalizing all patients of its served population of approximately 500,000. This study was carried out in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement. This study was authorized by the institutional board of the University Hospitals of Geneva and accepted by the Geneva’s committee for research ethics (project ID: 2020–00819).

Data regarding demographics, comorbidities, clinical symptoms, and laboratory analyses were retrospectively collected by one research nurse. In general, all information was obtained within the first 24 h of hospitalization in an acute ward. The geriatric hospital’s wards are managed by medical staff trained in internal medicine and geriatrics and can provide general medical care, such as intravenous pharmacotherapy and non-invasive oxygen support (by nasal cannula and face mask, allowing an increase of the fraction of inspired oxygen from 24 to 65%). Discharge criteria were an improvement of acute illness with subsequent transfer to a rehabilitation ward, or discharge to home or to a nursing home. For patients presenting worsening symptoms and unfavorable evolution, end-of-life care was implemented on the ward with consultation of palliative mobile teams. The diagnosis of COVID-19 was defined as a positive reverse transcription polymerase chain reaction (RT-PCR) test for SARS-CoV-2 on a nasopharyngeal swab. Patients with negative virus detection by RT-PCR (18/265; 6.8%) but a high clinical suspicion of disease were also diagnosed with COVID-19 [11].

The Functional Independence Measure (FIM) was performed by a nurse, based on observation during the first 24 h after hospital admission. It reflects functional status and physical function, with a point score ranging from 18 to 126, higher scores corresponding to better functionality [12]. The Clinical Frailty Score (CFS) was calculated by the physician in charge, reflecting the condition of older patients before hospitalization [13]. Higher CFS scores (from 5 to 8) correspond to more severe degrees of frailty, with a terminally ill patient assigned the highest score of 8. The Cumulative Illness Rating Scale-Geriatric (CIRS-G) was performed at admission, measuring a patient’s comorbidity burden by taking into account chronic diseases as well as the severity of acute illnesses, with higher scores representing a higher overall disease burden [14]. The severity of respiratory symptoms was assessed using the Pneumonia Severity Index (PSI) [15] and the abovementioned CURB-65 [16], which varies from 0 to 5, with higher scores being associated with higher mortality. For patients diagnosed with stroke, we computed the Modified Ranking Score (MRS) [17], which assesses functional recovery after a stroke event, as well as the CHA₂DS₂-VASc [18] and HAS-BLED [19] scores for thromboembolic and bleeding risks, respectively.

Cerebrovascular complications were defined as a diagnosis of stroke with neuroimaging confirmation by CT or MRI, including both ischemic stroke and intracranial hemorrhage. We included all cases of stroke having occurred during hospitalization, between the moment of COVID-19 diagnosis and hospital discharge or death. Hence, stroke diagnoses documented during acute and rehabilitation hospital stays were included in the analysis. For all stroke cases identified, an extensive review of the electronic medical record and imaging results was performed by a geriatrician and, for the purpose of this study, all images were reviewed by the same neuroradiologist.. Stroke cases were classified with respect to their localization, extent of ischemia or hemorrhage, and vascular territory [20]. In general, all patients that were suspected to present a stroke underwent standardized neuroimaging by CT or MRI, routinely performed according to in-house protocols.

Categorical variables are expressed as numbers and proportions, while continuous variables appear as means and standard deviations. A two-group comparison (patients with and without stroke) was performed using Fisher’s exact test and independent t-test for categorical and continuous variables, respectively. For the CHA₂DS₂-VASc and HAS-BLED scores from the stroke group, results are expressed as medians with minimum and maximum values. Because of the low number of stroke occurrences, only univariate logistic regression models were calculated to identify stroke predictors. The variables selection was based on statistically significant differences observed in the two-group comparison (patients with and without stroke). Furthermore, for the significant variables, we performed a trend analysis to study the relationship with stroke occurrence [21].

Statistical analysis was performed using Stata software version 16.1 (Stata Corp., College Station, TX, USA).

The study population consisted of 265 patients with a mean age of 85.9 ± 6.5 years, 43% were male, and the in-hospital mortality rate was 32.1%. Acute stroke was confirmed in 11 (4.15%) patients (Fig. 1). Mortality was similar between patients with (32.1%) and without stroke (27.3%, p > 0.999), and there were no differences regarding age, sex, or length of stay in an acute ward.

Stroke patients had a higher prevalence of active smoking (27.3% vs. 4.8%; p = 0.019), as well as a history of previous stroke (45.5% vs. 13.8%; p = 0.014). On the other hand, tiredness was less frequently reported in the group of stroke patients (9.1% vs. 50.2%; p = 0.010). Interestingly, stroke patients had a lower BMI than those without stroke (20.7 ± 3.5 vs. 24.9 ± 6.4; p = 0.002). While their mean BMI value was within the “normal” range, 45.5% of calculated BMIs were in the underweight range (< 20 kg × m^− 2), while no patient with acute stroke was classified as obese (BMI ≥30 kg × m^− 2) (Fig. 2). The trend towards a lower BMI in the stroke group was statistically significant (p = 0.038).

There were no differences regarding other cerebrovascular risk factors, except for dyslipidemia, which was more frequent in stroke patients, but with a borderline statistically significant p-value (63.6% vs. 33.2%; p = 0.051). Similarly, we did not observe differences in comorbidity burden, functional status, frailty, the severity of COVID-19, or the occurrence of other thromboembolic events between the two groups (Table 1).

In an univariate logistic regression model of stroke prediction, active smoking and previous stroke remained significant predictors, increasing by more than seven times and by more than five times, respectively, the risk of stroke. Similarly, each additional point in BMI was associated with a reduced risk of stroke by approximately 14% (Table 2).

Of the 11 patients with acute stroke, 81.8% were ischemic (9/11) and 18.2% hemorrhagic (2/11). The stroke events occurred after an average of 15.2 days from COVID-19 diagnosis, and the majority of patients presented with stroke during the acute care stay (8/11; 72.7%). Three patients had a simultaneous diagnosis of COVID-19 and acute stroke, whereas another three patients suffered a stroke on a geriatric rehabilitation ward (Supplementary Table S2). An altered state of consciousness and/or delirium were the most frequent clinical manifestations of stroke, reported in 81.8% of cases (9/11). In five patients (45.5%), a focal neurological deficit was present at the time of imaging by CT or MRI. Thromboembolic risk assessed by the CHA2DS2-VASc score showed a median score of 5 (range, 3–7), and a HAS-BLED score of 3 (range, 1–5).

Large vessel occlusion was reported in 22.2% of ischemic stroke cases (2/9). Furthermore, strokes were mainly limited to one side (5/9 right, 3/9 left) and the middle cerebral artery territory was affected in more than half of all cases (5/9; 55.5%), followed by the posterior cerebral artery (3/9), and vertebrobasilar territories (2/9). Ischemic lesions in multiple territories were found in two cases (Supplementary Table S1).

A cardioembolic cause was identified in 3 cases (44.5%), followed by suspected arterio-arterial embolism in 2 cases, and small-vessel occlusion in 1 case. Furthermore, one patient presented a border zone infarct as a consequence of systemic hypoperfusion. Ischemic stroke was classified as cryptogenic in two cases, with no etiology determined in acute workup.

By the end of the hospital stay, three patients (27.3%) died between 3 and 6 days after the stroke occurrence. In-hospital mortality rates were similar between patients with and without acute stroke, as well as institutionalization rates at hospital discharge (21.1% vs 36.4%; p = 0.258). All survivors presented moderate to severe disability at discharge (Supplementary Fig. S1).

There was a high burden of cerebral small vessel disease in patients with stroke, with more than half of patients presenting concomitant lacunes (54.5%). Significant white matter lesions were described in 81.8% of cases and 45.5% of patients had at least one cerebral microbleed, in a classical deep or lobar topography. There was no evidence of features such as meningitis, encephalitis, or vasculitis.