Delirium in older patients with COVID‐19: Prevalence, risk factors and clinical outcomes across the first three waves of the pandemic

Delirium is a serious condition, which poses treatment challenges during hospitalisation for COVID‐19. Improvements in testing, vaccination and treatment might have changed patient characteristics and outcomes through the pandemic. We evaluated whether the prevalence and risk factors for delirium, and the association of delirium with in‐hospital mortality changed through the pandemic.


| INTRODUCTION
A delirium is a serious condition frequently observed in patients during hospitalisation. 1 It mainly affects frail older adults and severely ill patients and is associated with increased risk on adverse clinical outcomes. 2Delirium also plays a significant role in the context of older patients hospitalised for Coronavirus disease 2019 (COVID- 19) and its estimated prevalence is between 14.2% and 54.9%. 3,4vere Acute Respiratory Syndrome Coronovirus-2 (SARS-CoV-2) can induce the onset of metabolic encephalopathy in patients with COVID-19 by a cytokine storm, hypoxemia and the direct effect on the central nervous system. 5,6Delirium during COVID-19 hospitalisation poses several specific challenges. 7Medication that is often prescribed for COVID-19 such as morphine for dyspnoea and/or pain and dexamethasone are known to increase the risk for delirium. 8en patients are isolated, contacts are minimised and healthcare professionals look all the same in isolation clothes potentially provoking disorientation.Furthermore, agitation can complicate treatment with for example, invasive devices such as airflow or the use of catheters and intravenous infusions. 9ditionally, over time, the severity of disease in COVID-19 patients decreased. 10Over the course of the pandemic there have been many developments in areas such as vaccination and treatment.
After the introduction of COVID-19 vaccinations, the most frail older adults were vaccinated first, which could have decreased disease severity in this specific patient population. 11Besides, treatment strategies became more effective on limiting severity of disease over time, for example, with the introduction of corticosteroid treatment. 12We know that compared to the first wave, in-hospital mortality decreased in the second wave, whereas no differences in prevalence of comorbidities or frailty were observed. 13It is likely that because of the aforementioned developments, the prevalence of delirium changed in the second and third wave.
The aim of this study was to compare the prevalence of delirium among older hospitalised COVID-19 patients through the first, second and third wave of the COVID-19 pandemic.Furthermore, we aimed to compare risk factors for delirium, and the possible association of delirium with in-hospital mortality through different waves of the pandemic.

| Study design
The COVID-OLD study is a retro-and prospective, multicentre observational cohort study among patients ≥70 years who were hospitalised for a COVID-19 infection. 14In this study, we included patients admitted between February 27th 2020-May 14th 2020 (first wave), September 1st 2020-December 31st 2020 (second wave) and September 1st 2021-December 31st 2021 (third wave) in the Netherlands.During the first wave, the wild-type virus variant was dominant, in the second wave the Alpha variant and in the third wave the Delta variant. 15Data for the first wave was collected in 19 hospitals, for the second wave in 10 hospitals and for the third wave in 5 hospitals (Supplemental Table S1).An opt-out procedure was applied for data collection of the first and second wave.The medical ethics committees of all hospitals waived the necessity for formal approval of the study, as data collection followed routine practice and took place until hospital discharge.When applicable, informed consent has been asked for patients included in the third wave.All data were treated according to the European privacy regulations and the study was performed in accordance with the declaration of Helsinki.

| Data collection
Demographics, clinical parameters and laboratory measurements at hospital admission and hospital outcomes were collected from electronic medical records.The prevalence of delirium during hospitalisation was assessed using the Delirium Observation Screening Scale (DOSS) which was ideally scored three times daily and supplemented with medical record review.A DOSS score >3 suggested the presence of delirium, which was confirmed by a clinical evaluation.7][18] ADL were assessed using the Katz-ADL-scale. 19Comorbidities were evaluated using the Charlson Comorbidity Index (CCI). 20Frailty was measured with the Clinical Frailty Scale (CFS) score. 21The CFS was prospectively assigned during the first day of hospital admission and noted in the medical record.If not prospectively assigned, the CFS was determined retrospectively based on available data about functional status 2 weeks before admission, and was scored by a geriatrician or internist-geriatrician, or researcher trained by a geriatrician or internist-geriatrician. 22 Data on CFS were considered missing if information from the health record was not sufficient to determine the CFS score retrospectively.According to the Dutch guidelines, the CFS was categorized in three groups: fit (CFS 1-3), pre-frail (CFS 4-5) and frail (CFS 6-9). 23Disease severity indicators were the registered vital signs and laboratory results collected within the first 24 h of admission.Medication use was extracted from the medical records throughout the hospital admission.Data were collected using Castor Electronic Data Capture (Amsterdam, The Netherlands) (2022). 24

| Statistical analyses
Demographics, clinical parameters and laboratory measurements were summarised using descriptive statistics stratified for pandemic wave.Baseline characteristics in patients with and without delirium during hospitalisation were compared using an independent t-test for continuous parameters with a normal distribution, a Mann-Whitney U-test for continuous parameters with a non-normal distribution, and a chi-square test for categorical parameters.The association between potential risk factors and delirium was assessed using binary logistic regression analyses.First, a univariable model was built to assess associations between potential risk factors and delirium for each wave separately.Second, in multivariable analyses, associations were adjusted for identified covariates (p < 0.05) in univariable analyses (age, frailty, previous episode of delirium and history of memory problems).To determine if the associations between potential risk factors and delirium changed through waves, we used multiplicative interaction terms to provide statistical evidence whether the effect sizes between the waves were different.
The association between delirium and in-hospital mortality was assessed using binary logistic regression analyses.First, a univariable model was built.Second, a multivariable model was built in which was adjusted for sex, age and frailty.To determine if the association between delirium and in-hospital mortality changed through waves, we included a multiplicative interaction term between wave and delirium in the multivariable-adjusted analysis.We used multiplicative interaction terms to provide statistical evidence whether the effect sizes between the waves were different.Inhospital outcomes for patients with and without delirium were compared, also stratified for wave, using an independent t-test for continuous parameters with a normal distribution, a Mann-Whitney U-test for continuous parameters with a non-normal distribution and a chi-square test for categorical parameters.A p-value <0.05 was considered statistically significant.Statistical analyses were performed using SPSS, version 28.0.1.0.

| RESULTS
3461 patients were included: 1874 in the first, 1121 in the second and 466 in the third wave.We excluded 334, 237 and 96 patients per wave respectively, because of a PCR diagnosis >24h after admission and/or missing delirium status and/or because of discharge to another hospital.2794 patients were available for baseline-and outcome analysis; 1540 in the first, 884 in the second and 370 in the third wave (Figure 1).Baseline characteristics were stratified by wave and compared between patients with and without delirium in Table 1.
Furthermore, comorbidities from the CCI that showed significant differences between patients with and without delirium are presented in Table 1.
The prevalence of delirium in the first, second and third wave was 22.5%, 23.5%, and 12.7%, respectively.There was a significant difference in delirium prevalence between wave 1 and wave 3 (p < 0.001), and between wave 2 and wave 3 (p < 0.001).
In the first wave and second wave, patients with delirium had lower C-reactive protein (CRP) levels than patients without delirium (first wave 63 vs. 80 mg/L, p = 0.003, second wave 59 vs. 73 p = 0.057), whereas no numerical difference was seen in the third wave.Furthermore, in the first and second wave, patients with MINNEMA ET AL. delirium had a shorter duration of symptoms before hospital admission (first wave 5 vs. 7 days, p < 0.001, second wave 4 vs. 6 days, p < 0.001) than those without.Finally, only relevant in wave 3, patients with delirium were more frequently unvaccinated to COVID-19 compared to those without (64.3% vs. 78.2%,p = 0.047).).The significant multiplicative interaction between previous episode of delirium and wave, showed that the OR of the second wave was significantly lower compared to the OR of the first wave (p = 0.008).Frailty was not independently associated with risk of delirium.

The association between risk factors and delirium is shown in
The association between delirium and the risk of inhospital mortality is shown in Table 3. Patients with delirium had an increased risk of in-hospital mortality in the first wave (OR 1.47, 95% CI 1.15-1.87).After adjustment for age, sex and frailty, delirium was no longer associated with the risk of in-hospital mortality.Frailty was independently associated with in-hospital mortality in all waves.The association between risk factors and in-hospital mortality did not change across waves (all multiplicative interaction terms p > 0.05).
The association between delirium and the risk of other adverse in-hospital outcomes is shown in Table 4. Patients with delirium had a longer length of hospital stay than patients without (8 vs. 6 days, p < 0.001) in the first and second wave, but not in the third wave (7 vs. 7 days, p = 0.094).Patients with delirium were more often admitted to the Intensive Care Unit (13.8% vs. 10.0%,p = 0.049) compared to patients without in the first wave, but not in the second and third wave.

| DISCUSSION
This study has three main findings.First, the prevalence of delirium in the third wave was lower compared to the first and second wave.
Second, risk factors for delirium remained consistent across waves, except for a previous episode of delirium, which had the highest delirium risk in the first wave and a significantly lower risk in the second wave.Third, delirium did not independently increase the risk for in-hospital mortality.
We found a delirium prevalence of respectively 22.5%, 23.5% and 12.7% in the first three waves of the COVID-19 pandemic.The prevalence in the first wave was similar to a systemic review including 48 studies that estimates delirium prevalence of 28.2% in older COVID-19 patients. 25The lower delirium prevalence could be explained by improved treatment strategies resulting in decreased disease severity (e.g.anti-IL-6 receptor tocilizumab and corticosteroids).Additionally, the introduction of vaccinations could have decreased disease severity and thus decreased delirium prevalence in the third wave.Another explanation could be that patients in the third wave were less prone to delirium, since the traditional risk factors (previous episode of delirium and earlier memory problems) were less prevalent in the total patient group in the third wave.
Patients with delirium had similar values of vital signs at admission and thus no more severe disease than patients without delirium.We observed slightly lower CRP levels in patients with delirium than without in the first wave.Possible explanations could be that patients with delirium were admitted to the hospital earlier in the disease course compared to patients without, or patients with delirium had a different immune response.
In wave 1 and 2 but not wave 3, patients with delirium were more frequently frail than patients without delirium.Frail patients in the third wave got vaccinated first, which may have led to a decreased disease severity in this group without need for hospitalisation.Additionally, frailty itself was not found to be an independent -5 of 10 risk factor for delirium.This is contrary to findings in literature.7][28] In our study, the CFS was used to score frailty which also takes the degree of dementia into account.Hence, the CFS is a combination of both physical and cognitive frailty.Since cognitive frailty yields a larger risk for a new episode of delirium, this might be an explanation for our finding that a previous episode of delirium and history of memory problems representing cognitive frailty were independently associated with delirium risk.
Delirium was associated with increased in-hospital mortality risk in univariable analyses, but not after adjustment for covariates such as frailty.The independent relationship between delirium and inhospital mortality in COVID-19 patients is still debatable.3][34] Our study implies that delirium is a less accurate predictor for in-hospital mortality compared to frailty, possibly due to the fact that frailty as syndrome is more comprehensive than delirium in itself.
Our study has several limitations.The sample size in each subsequent wave was almost twice as small, which might have affected the power to find significant differences between patients with and without delirium in the second and third wave.Therefore, we used multiplicative interaction terms to test whether associations significantly changed through waves, or whether a non-significant association in the third wave was the result of a smaller sample size.T A B L E 3 Univariable and multivariable-adjusted binary logistic regression analyses for the possible association of delirium and in-hospital mortality.

Patients
≥70 years who were hospitalised due to a confirmed COVID-19 infection were eligible to participate.Patients required a positive reverse-transcriptase polymerase chain reaction (RT-PCR) test from an oropharyngeal and/or nasal swab or if the diagnosis of COVID-19 infection was based on typical findings on computerized tomography scan and/or chest X-ray.Patients were excluded if diagnosed with COVID-19 during hospital admission for another illness, defined as a positive RT-PCR test ≥24 h.Additionally, patients were excluded in case they were transferred between hospitals due to missing baseline and outcome data.

Table 2 .
In multivariable-adjusted regression analyses, history of memory problems was associated with delirium in all waves (wave 1

4 of 10 -
Baseline characteristics compared between patients with and without delirium in the first, second and third wave.
pitals in the first and second wave, this most likely did not affect study results.Besides, in the third wave, the ethical procedure changed in one of the participating centres.In this hospital, (Erasmus MC University hospital) only a small proportion of the third wave patients (7 out of 62 (11.3%)), did not give informed consent retrospectively or did not return the informed consent letter.Delirium T A B L E 1 (Continued) Note: Missing values: Wave 1: Missing numbers in patients with a delirium (N = 347): 1 Model 1: univariate regression analyses for in-hospital mortality.Model 2: multivariate regression analyses for delirium and in-hospital mortality in which is corrected for sex, age and frailty.Both models are presented with Odds Ratio's and 95% Confidence Intervals.Interaction terms in univariate analyses (variable*wave-categorical, variable, wave-categorical): Delirium*wave, p = 0.518; Sex*wave, p = 0.365; Age*wave, p = 0.364; BMI*wave, p = 0.926; Frailty*wave, p = 0.760; CCI*wave, p = 0.919.Interaction terms in multivariate analyses (variable*wave-categorical, variable, wave-categorical, and identified covariates): Delirium*wave, p = 0.451; Sex*wave, p = 0.250; Age*wave, p = 0.517; Frailty*wave, p = 0.736.Missing values: wave 1: 2 Sex, 341 BMI, 427 Frailty, 563 CCI; wave 2: 148 BMI, 151 Frailty, 300 CCI; wave 3: 77 BMI, 61 frailty, 136 CCI.Univariable and multivariable analyses for the possible association of risk factors and delirium.Note: Model 1: univariate regression analyses for delirium.Model 2: multivariate regression analyses for delirium in which is corrected for age, frailty, previous episodes of delirium and history of memory problems.Both models are presented with Odds Ratio's and 95% Confidence Intervals.Interaction terms in univariate analyses (variable*wave-categorical, variable, wave-categorical): Age*wave, p = 0.742; Sex*wave, p = 0.833; CFS*wave, p = 0.050; CCI*wave, p = 0.823; Previous episodes of delirium*wave, p = 0.032; History of memory problems*wave, p = 0.310.Interaction terms in multivariateTo conclude, we observed a decreased prevalence of delirium through COVID-19 waves.This decrease was probably due to differences in virus variant, but also to the start of vaccinations and developments in treatment.These findings suggest that early developments of therapeutic programs can change disease severity and specifically delirium as manifestation.For future potential pandemics and other severe infectious diseases, it remains critical to recognize delirium risk factors in an early stage and diagnose delirium.In hospital outcomes compared between patients with and without delirium in the first, second and third wave.Note: Missing values: Wave 1: Missing numbers in patients with a delirium (347): 13 ICU admission, 1 Ventilator assisted breathing, 8 Readmission.Missing numbers in patients without a delirium (1193): 30 ICU admission, 44 Readmission.Wave 2: Missing numbers in patients with a delirium (208): n. a. Missing numbers in patients without a delirium (676): 1 ICU admission, 1 Readmission.Wave 3: Missing numbers in patients with a delirium (47): 13 ICU admission.Missing numbers in patients without a delirium (323): 30 ICU admission.
analyses (variable*wave-categorical, variable, wave-categorical, and identified covariates): Age*wave, p = 0.909; CFS*wave, p = 0.437; Previous episodes of delirium*wave, p = 0.028; History of memory problems*wave, p = 0.862.Missing values: wave 1: 2 Sex, 427 Frailty, 563 CCI, 312 previous episodes of a delirium, 284 history of memory problems; wave 2: 151 Frailty, 300 CCI, 132 previous episodes of a delirium, 118 history of memory problems; wave 3: 61 frailty, 136 CCI, 71 previous episodes of a delirium, 63 history of memory problems.prevalence was not different in patients who provided informed consent versus patients who did not.Therefore, it is unlikely that this three pandemic waves and its association with mortality.Nevertheless, certain factors remain unclear and further research is needed to investigate long-term outcomes of delirium in COVID-19.Besides, important parameters such as quality of life and patient experiences are not considered, which were suggested by a Seniors Advisory Board.Additionally, more research is needed tailored towards older patients hospitalised for COVID-19.Recent studies predominantly consist of cohorts with mixed age groups, whereas risk stratification and treatment decisions might be even more relevant in older and severely ill patient groups.T A B L E 4 Significance: * = <0.05,** = <0.01*** = <0.001.