Impact of Dementia on Mortality Due to Coronavirus Disease 2019: Propensity-Score-Matching Study.

BACKGROUND AND
PURPOSE: Patients with dementia are particularly vulnerable to coronavirus disease 2019 (COVID-19) because they tend to be older and often have concomitant diseases. Previous studies have investigated the impact of dementia on COVID-19 outcomes, but the evidence is not robust for Asian populations. We aimed to determine the relationship between dementia and COVID-19 outcomes using data from a large-scale nationwide public database.
METHODS: Data on patients with COVID-19 who were released from quarantine between January 1, 2020 and April 30, 2020, published by the Korea Disease Control and Prevention Agency, were divided into two groups based on the dementia status. Propensity-score matching was used to adjust for multiple confounders between the dementia and no-dementia groups. Binary, ordinal logistic regression and multivariate Cox proportional-hazards models were used to compare mortality, quarantine duration, and clinical deterioration according to the dementia status in the two groups.
RESULTS: Males and older individuals (age ≥60 years) constituted 41.5% and 32.9%, respectively, of the 5,299 patients. The prevalence of dementia was 4.2%, and 4.5% of the participants died during hospitalization. In multivariate analysis, dementia was significantly associated with increased mortality (odds ratio [OR]=2.80, 95% confidence interval [CI]=1.60-4.60), longer duration of quarantine (hazard ratio=1.69, 95% CI=1.16-2.45), and larger shift to a worse clinical severity (common OR=1.74, 95% CI=1.18-2.61).
CONCLUSIONS: After adjusting for important clinical predictors, dementia was associated with increased in-hospital mortality, duration of quarantine, and clinical deterioration during hospitalization in COVID-19 patients.

INTRODUCTION

Coronavirus disease (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has massively affected the world, with there being 197,767,859 cases and 4,215,888 deaths as of August 1, 2021 (https://www.worldometers.info/coronavirus/). The first COVID-19 patient was reported in January 2020 in Korea,1 and the cumulative total numbers were 198,345 patients and 2,095 deaths as of July 31, 2021 (http://ncov.mohw.go.kr/). Advanced age is a well-known risk factor for mortality in COVID-19 patients.2 Soneji et al.2 reported that the age-standardized death rate per 100,000 person-years increased from 0.1 deaths in patients aged 30–39 years, to 9.5 deaths in those older than 80 years. In addition to older age, pre-existing comorbidities such as hypertension and cardiovascular and pulmonary diseases have been reported to be associated with higher rates of mortality.3

Patients with dementia are particularly vulnerable to COVID-19 because they tend to be older, often have concomitant diseases, and in many cases live in chronic-care facilities.4 Bauer et al.5 reported that dementia patients were more likely to have comorbidities such as stroke, diabetes, atherosclerosis, electrolyte imbalance, and pneumonia. It has consistently been shown that dementia increases mortality in Western countries.67 Docherty et al.6 reported a dementia rate of 13.5% in 17,459 COVID-19 patients in the United Kingdom, and found that the presence of COVID-19 increased the mortality rate significantly (hazard ratio [HR]=1.40, 95% confidence interval [CI]=1.28–1.52). However, information on the association between dementia and coronavirus disease outcomes in Korea has been scarce and contradictory. Hwang et al.8 found that dementia increased the risk of mortality, in contrast to the findings of another study.9 We think that the small number of enrolled patients (11 out of 103 in one study and 35 out of 352 in the other), the inclusion of small regional cohort studies, racial differences, and analyses performed without adjusting for confounding factors contributed to these conflicting results. In addition, a community-based study showed that mortality risk due to COVID-19 differs with race,10 with an Asian population exhibiting a higher mortality rate than non-Hispanic white people (relative risk [RR]=1.38, 95% CI=1.25–1.53). However, in patients without COVID-19, the mortality rate due to dementia in the Asian population has been found to be lower than in non-Hispanic white patients with dementia, especially in those younger than 70 years.11 In-hospital mortality in patients with dementia has been associated with advanced age, concomitant cardiovascular diseases, and malnutrition, which are risk factors that also strongly overlap with poor outcomes from COVID-19. It is therefore necessary to adjust for confounding factors when attempting to accurately assess the association between dementia and mortality.

This study performed a detailed analysis of how dementia influences outcomes in the Korean population using a large-scale nationwide public data set. We accounted for factors that could contribute to outcomes, and performed a propensity-score matching (PSM) analysis to adjust for confounding variables.

METHODS

Data sources

The Institutional Review Board of Hallym University Chuncheon Sacred Heart Hospital (No. 2020-07-016) and the Korea Disease Control and Prevention Agency (KDCA) approved this study and waived the need to obtain informed consents. The Korea National Committee for Clinical Management of COVID-19 had previously constructed a registry to collect the clinical data of patients hospitalized for COVID-19. That registry utilizes a standardized clinical record form (CRF) that is a modified version of the World Health Organization global CRF for COVID-2019 (https://www.who.int/teams/health-care-readiness-clinical-unit/covid-19).12 We extracted data for our analysis from patients confirmed as being released from quarantine between January 1, 2020 and April 30, 2020, according to clinical epidemiological information of COVID-19-confirmed patients published by the KDCA (http://www.kdca.go.kr/).13 Dementia was defined based on the judgement of the clinician who had made the past diagnosis, as well as the medication history. Clinical severity is recorded on a 10-point scale score in the WHO CRF.14 However, we revised this score and divided it into three classes: 1) no need for oxygen therapy, 2) need for oxygen therapy, and 3) ventilator support or death.

The primary outcome was a comparison of in-hospital mortality rates in COVID-19 patients with and without dementia. Secondary outcomes were how the duration of quarantine and the need for admission to an intensive-care unit (ICU) due to clinical deterioration differed according to the dementia status. In particular, the duration of quarantine was defined as the period of release from isolation from the general treatment center where patients with mild symptoms stayed, or as the period from hospital admission to discharge.

Statistical analysis

We compared baseline clinical and demographic characteristics between COVID-19 patients with and without dementia using the χ2 test, t-test, or Mann-Whitney U test, as appropriate. Missing values were sparsely distributed in our COVID-19 cohort, and so we applied multiple imputation (5 repetitions with 10 cycles) in such cases. The baseline mortality, quarantine duration, and maximum clinical severity during hospitalization differed significantly between the dementia and no-dementia groups. We compared these clinical outcomes after applying PSM (1:1 ratio).

Binary logistic regression analysis was performed to compare predictors of in-hospital mortality between the two groups. A multivariate logistic regression model was applied to independent variables for which the p value was <0.05 in the univariate model or that had clinical relevance. We used a Cox proportional-hazards (CPH) survival model of in-hospital mortality and quarantine at the end of the follow-up period, according to the dementia status. All independent variables were entered into the CPH model for mortality, and significance was assessed using the HR and 95% CI. Cumulative survival plots for the quarantine duration were created using Kaplan-Meier statistics to compare the quarantine duration between the dementia and no-dementia groups. The maximum clinical severity scores during admission were divided into tertiles, after which ordinal logistic regression analysis was performed to identify predictors of a larger shift in the distribution of the maximum clinical severity. All statistical analyses were performed with R software (version 3.6.3 R Core Team, R Foundation for Statistical Computing, Vienna, Austria), with statistical significance set at p<0.05.

RESULTS

Among the 5,628 patients who were confirmed as having COVID-19, 329 for whom information on dementia was not available were excluded. Supplementary Table 1 (in the online-only Data Supplement) compares the clinical characteristics of confirmed COVID-19 patients with dementia (n=224) and those without dementia (n=5,075). The proportion of patients with dementia was 4.2% (224 out of 5,299), 2,197 subjects were male (41.5%), 32.9% of the subjects were aged ≥60 years, and 192 subjects were hospitalized in ICUs. The most common COVID-19 symptom was sore throat (84.6%), followed by fever and cough.

In-hospital mortality

The in-hospital mortality rate was 4.5% (241 out of 5,299 patients). Supplementary Fig. 1 (in the online-only Data Supplement) shows that missing variables were randomly distributed. After multiple imputation, we compared the clinical characteristics and laboratory values between the dementia and no-dementia groups (Table 1).

Table 1

Comparison of clinicodemographic differences between the dementia and no-dementia groups before and after PSM

		Before PSM			After PSM
		Dementia (n=224)	No dementia (n=5,075)	p	Dementia (n=224)	No dementia (n=224)	p
Age, years				<0.001			0.986
	0–9	0 (0.0)	66 (1.3)		0 (0.0)	0 (0.0)
	10–19	0 (0.0)	195 (3.8)		0 (0.0)	0 (0.0)
	20–29	0 (0.0)	1,009 (19.9)		0 (0.0)	0 (0.0)
	30–39	0 (0.0)	523 (10.3)		0 (0.0)	0 (0.0)
	40–49	1 (0.5)	684 (13.5)		1 (0.5)	1 (0.4)
	50–59	9 (4.0)	1,068 (21.0)		9 (4.0)	10 (4.5)
	60–69	22 (9.8)	857 (16.9)		22 (9.8)	21 (9.4)
	70–79	52 (23.2)	489 (9.6)		52 (23.2)	56 (25.0)
	≥80	140 (62.5)	184 (3.6)		140 (62.5)	136 (60.7)
Sex, male		71 (31.7)	2,126 (41.9)	0.003	71 (31.7)	90 (40.2)	0.076
BMI, kg/m2							0.501
	<18.5	33 (14.7)	296 (5.8)		33 (14.7)	24 (59.8)
	18.5–22.9	121 (54.0)	2,118 (41.7)		121 (54.0)	116 (51.8)
	23.0–24.9	39 (17.4)	1,204 (23.8)		39 (17.4)	48 (21.4)
	25.0–29.9	27 (12.1)	1,209 (23.8)		27 (12.1)	29 (12.9)
	≥30	4 (0.4)	248 (4.9)		4 (0.4)	7 (3.1)
Outcome		149 (66.5)				174 (77.7)
	Death	75 (33.5)	166 (3.3)	<0.001	75 (33.5)	50 (22.3)	0.008
	Duration of quarantine (days)	24.0 (13.0–33.2)	24.0 (18.0–32.0)	0.065	24.0 (13.0–33.3)	23.0 (17.0–33.0)	0.372
	ICU admission	16 (7.1)	176 (3.5)	<0.001	16 (7.1)	26 (11.6)	0.145
Clinical severity score				<0.001			0.056
	No need for oxygen therapy	107 (47.8)	4,377 (86.2)		107 (47.8)	123 (54.9)
	Need for oxygen therapy	41 (18.3)	500 (9.9)		41 (18.3)	48 (21.4)
	Ventilator support or death	76 (33.9)	198 (3.9)		76 (33.9)	53 (23.7)
BPsys ≥130 mm Hg		133 (59.4)	2,744 (54.1)	0.136	133 (59.4)	154 (68.8)	0.049
BPdia ≥80 mm Hg		118 (52.7)	3,152 (62.1)	0.004	118 (52.7)	132 (58.9)	0.216
Body temperature (°C)		36.8±0.6	36.9±0.6	<0.001	36.8±0.6	36.9±0.6	0.157
Comorbidities
	Diabetes	63 (28.1)	618 (12.2)	<0.001	63 (28.1)	57 (25.4)	0.594
	Hypertension	126 (56.2)	1,045 (20.6)	<0.001	126 (56.2)	125 (55.8)	1.000
	Heart failure	12 (5.4)	47 (0.9)	<0.001	12 (5.4)	11 (4.9)	1.000
	Coronary artery disease	15 (6.7)	162 (3.2)	0.007	15 (6.7)	16 (7.1)	1.000
	Asthma	8 (3.6)	116 (2.3)	0.308	8 (3.6)	7 (3.1)	1.000
	COPD	8 (3.6)	32 (0.6)	<0.001	8 (3.6)	4 (1.8)	0.380
	CKD	10 (4.5)	45 (0.9)	<0.001	10 (4.5)	8 (3.6)	0.810
	Malignancy	5 (2.2)	140 (2.8)	0.792	5 (2.2)	3 (1.3)	0.721
	Chronic liver disease	6 (2.7)	76 (1.5)	0.261	6 (2.7)	8 (3.6)	0.786
	Rheumatic/autoimmune disease	0 (0.0)	38 (0.7)	0.370	0 (0.0)	0 (0.0)	-
Laboratory parameters
	Hemoglobin (g/dL)	12.0±1.9	13.3±1.7	<0.001	12.0±1.9	12.2±1.9	0.230
	Lymphocytes (%)	35.7±5.7	39.4±4.9	<0.001	24.6±13.1	23.3±11.9	0.291
	Platelet count (×103/µL)	204±81	238±82	<0.001	204±81	224±91	0.016
	White blood cell count (×103/µL)	6.1±3.0	6.1±2.8	0.922	6.1±3.1	6.7±3.4	0.040

Categorical variables are number (percentage), and continuous variables are mean±standard-deviation or median (interquartile range) values.

BMI, body mass index; BPdia, diastolic blood pressure; BPsys, systolic blood pressure; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; ICU, intensive-care unit; PSM, propensity-score matching.

There were significant differences in clinical characteristics between the two groups, and so we applied PSM (1:1 ratio) to the independent variables. Supplementary Table 2 (in the online-only Data Supplement) indicates that the independent variables were balanced for matched data after PSM, with the variables used having similar distributions in the dementia and no-dementia groups. Table 2 lists the results of the multivariate logistic regression analysis of predictors of in-hospital mortality in the matched population. Dementia was a significant predictor of mortality after adjusting for significant variables (odds ratio [OR]=2.80, 95% CI=1.60–4.90) and in the CPH model (HR=1.85, 95% CI=1.27–2.70) (Fig. 1). Body mass index (BMI), lymphocyte count, and platelet count were also significant predictors of mortality in the CPH model.

Table 2

Results of the binary logistic regression analysis of predictors of mortality in the participants

Model	OR (95% CI)	p
Dementia (crude model)	1.90 (1.24–2.90)	0.003
Model 1	2.12 (1.35–3.32)	0.001
Model 2	2.18 (1.36–3.48)	0.001
Model 3	2.80 (1.60–4.90)	<0.001

Model 1 was adjusted for variables included in the crude model and additionally for age and sex. Model 2: Model 1 + BMI, diabetes, hypertension, heart failure, coronary artery disease, bronchial asthma, COPD, CKD, malignancy, and chronic liver disease. Model 3: Model 2 + hemoglobin and lymphocytes.

CI, confidence interval; OR, odds ratio.

Fig. 1

Results from the multivariate Cox proportional-hazards model for in-hospital mortality in the participants. Each black rectangle represents the hazard ratio (HR) for in-hospital mortality. Horizontal bars represent the 95% confidence intervals. An HR of 1 is equivalent to no difference in the hazard rate of each risk factor versus control. HRs higher than 1 indicate increased hazard rates of in-hospital mortality resulting from the risk factor, while HRs lower than 1 indicate decreased hazard rates of in-hospital mortality. BMI, body mass index; BPdia, diastolic blood pressure; BPsys, systolic blood pressure; WBC, white blood cell count.

Duration of quarantine

Participants were treated in isolation for a median of 24.0 days. The Kaplan-Meier plot analysis revealed that patients with dementia had a higher probability of being quarantined than did no-dementia patients (p in log-rank test=0.004) (Fig. 2). In the CPH model, dementia was a significant predictor of the duration of quarantine in matched data (HR=1.69, 95% CI=1.16–2.45) (Supplementary Fig. 2 in the online-only Data Supplement).

Fig. 2

Kaplan-Meier curves displaying the estimated quarantine duration probability for the dementia and no-dementia groups. Each vertical step in the curve indicates the event (end of quarantine period). A p value of <0.05 indicates a significant difference in the log-rank test between the survival curves.

Distribution of maximum clinical severity

Fig. 3 compares the distributions of maximum clinical severity scores between the dementia and no-dementia groups during hospitalization. The proportion of patients who received ventilator support or who died was higher for those with dementia than for those without dementia. Table 3 presents the association between dementia and a higher maximum clinical severity score. Dementia was a significant predictor in the ordinal logistic regression model (common OR=1.74, 95% CI=1.18–2.61).

Fig. 3

Distribution of the maximum clinical severity in the participants during hospitalization. Ventilator support indicates the use of a high-oxygen-flow-rate nasal cannula or mechanical ventilation. The number (percentage) of the patients with the maximum clinical severity during hospitalization is given in each cell. The data show that a larger proportion of patients in the dementia group received oxygen or ventilatory support, or died.

Table 3

Results of the ordinal logistic regression analysis of a larger shift in the maximum clinical severity during hospitalization of the participants

Model	OR (95% CI)	p
Dementia (crude model)	1.36 (0.96–1.93)	0.088
Model 1	1.46 (1.02–2.10)	0.041
Model 2	1.44 (1.00–2.09)	0.001
Model 3	1.74 (1.18–2.61)	0.006

Model 1 was adjusted for variables included in the crude model and additionally for age and sex. Model 2: Model 1 + BMI, diabetes, hypertension, heart failure, coronary artery disease, bronchial asthma, COPD, CKD, malignancy, and chronic liver disease. Model 3: Model 2 + hemoglobin and lymphocytes.

CI, confidence interval; OR, odds ratio.

DISCUSSION

Dementia is believed to be closely associated with mortality in COVID-19 patients. A recent meta-analysis revealed that dementia significantly increased mortality due to COVID-19 (RR=2.62, 95% CI=2.04–3.36).3 However, the results of that study may be unreliable since it did not adjust for confounding factors such as patient age and comorbidities that can affect the prognosis. In addition, high heterogeneity (I2=96%) was observed across the studies included in that meta-analysis, and it also did not accurately reflect outcomes in the Asian population because it included only 3 studies involving a total of 17 dementia patients from Asian countries.3 We therefore considered it necessary to investigate how dementia affects outcomes after SARS-CoV-2 infection in a study design that adjusted for confounding variables and drew from a Asian population.

Our study found that the baseline characteristics of COVID-19 patients differed significantly between those with and without dementia. Patients with dementia were older, less obese, more likely to be female, and had higher maximum clinical severity scores and a higher rate of comorbidities (Table 1). In particular, higher mortality rates in COVID-19 patients have been found to be associated with medical comorbidities, advanced age, and female sex.6 Thus, confounding variables should be adjusted for when assessing the impact of dementia on mortality. In our analysis, after adjusting for possible confounding variables in the matched population using PSM, we observed that dementia was still a significant risk factor for in-hospital mortality (OR=2.80, 95% CI=1.60–4.90). This was true in various statistical models (Table 2), which provides robust evidence for this relationship being present in Korean subjects.

Most previous studies have focused on simple associations between mortality and dementia in COVID-19 patients.8 In addition to the current evidence, we further analyzed associations of dementia with the quarantine duration and the maximum clinical severity during hospitalization. Our study clearly demonstrated that COVID-19 patients with dementia require a longer quarantine duration than those without dementia. Moreover, the need for ventilator support and the occurrence of death were both more frequent among those with dementia. These findings indicate that hospital admissions of COVID-19 patients with concomitant dementia require significant additional medical resources and longer inpatient stays. This can contribute significantly to the likelihood of healthcare management systems being overwhelmed when the number of COVID-19 patients continues to increase, especially if dementia patients are not being managed efficiently. The cognitive decline of dementia patients makes it difficult for them to follow hygiene rules, and they are more susceptible to infection because they often live in chronic-care facilities. It is therefore important for dementia caregivers and medical professionals to cooperate in order to develop protocols for preventing COVID-19 in these patients, and actively comply with them.

The impact of dementia on long-term outcomes has not received sufficient attention in COVID-19 patients. COVID-19 patients have accompanying neurological symptoms and complications. Reportedly 36% of COVID-19 patients experience decreased consciousness, seizures, and acute cerebrovascular diseases, and such symptoms are more pronounced in the presence of severe infection.15 Due to its structural similarity with SARS-CoV, SARS-CoV-2 is believed to bind to angiotensin-converting enzyme 2 (ACE2), which is expressed in various organs including the lungs, blood vessels, and brain.41617 Brain injury can result from excessive immune and inflammatory responses and an altered coagulation system, which can in turn contribute to vascular injury and subsequent damage to the blood-brain barrier.415 The SARS-CoV-2 protein also interacts with age-related proteins.18 Neuroinflammation involving reactive gliosis, oxidative damage, and mitochondrial dysfunction is a common feature of dementia,1920 and so brain damage and inflammatory reactions caused by COVID-19 could exacerbate existing dementia. Long-term follow-up of changes in cognitive function and brain structures using radiological tests—even after COVID-19 patients are released from quarantine—is becoming necessary.

We found a U-shaped relationship between BMI and mortality in COVID-19 patients (Fig. 1). Since it was reported that high BMI was associated with an increase in influenza mortality during the H1N1 pandemic in 2009,21 a relationship between obesity and COVID-19 mortality is not unexpected. Soeroto et al.22 found that obesity was associated with increases in ICU admissions, acute respiratory distress syndrome, COVID-19 severity, hospital admissions, and mortality in a systematic review of 16 studies. In addition, obesity can reportedly affect the immune system and impair the host defence mechanism.23 Therefore, our result is consistent with previous reports of obesity being associated with poor outcome in COVID-19 patients. Additionally, the platelet count was negatively associated with COVID-19 mortality (Fig. 1, Table 1). COVID-19 patients often have mild thrombocytopenia and appear to exhibit increased platelet consumption,24 and low platelet count has been associated with severe SARS-CoV-2 infection.25 Our finding of an association between platelet count and COVID-19 mortality can be interpreted similarly to those previous observations.

This study was subject to some limitations. First, it did not analyze data from a large number of laboratory tests. Compared with non-COVID pneumonia, COVID-19 pneumonia significantly decreases the leukocyte, neutrophil, and platelet counts.26 Although multiple imputation was performed to handle missing data, individual risk assessments performed using laboratory tests would have yielded more accurate information. Second, there was unreliable information about the place of residence of the patients before they were infected by SARS-CoV-2. Matias-Guiu et al.27 reported that the place of residence is more closely associated with COVID-19 than dementia, suggesting that it can affect mortality. However, residence information was missing from the KDCA database, making it impossible for us to evaluate its association with mortality. Third, we did not distinguish between different dementia subtypes. Compared with Alzheimer’s disease, frontotemporal dementia can occur at a younger age and have a lower mortality rate after COVID-19.27 Accordingly, future studies are needed of mortality rates in COVID-19 patients while taking dementia subtypes into account.

In conclusion, this study found dementia to be associated with increased in-hospital mortality in COVID-19 patients. Longer duration of quarantine as well as clinical deterioration during hospitalization were also closely associated with dementia status in COVID-19.