1 INTRODUCTION

Caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the COVID-19 pandemic has swept the world for more than 1 year, with more than 79 million confirmed cases and more than 1.7 million deaths worldwide as of December 29, 2020.1 Male sex, African ethnicity, and age, among other sociodemographic and clinical factors, have been established as risk factors for poor COVID-19 outcomes.2-4 Older individuals have been found to be particularly vulnerable to COVID-19: for example, in the UK 90% of COVID-related deaths have been among people more than 60 years old, and individuals aged 80 or over have been found to be at a 20-fold higher risk than 50- to 59-year-olds.3

Dementia has been identified as a risk factor for COVID-19, in addition to other chronic comorbidities, including cardiovascular and respiratory diseases, hypertension, diabetes, obesity, and cancer.3, 5 However, the causes for this association remain unclear. It has been suggested that increased prevalence of COVID-19 in individuals with dementia may be, at least in part, accounted for by living in care or nursing homes, by depending on external caregivers, and/or by lack of independence and ability to maintain hygiene/preventative health measures in isolation,6, 7 all of which increase the risk of infection. Moreover, age is the major risk factor for development of dementia (notably Alzheimer's disease [AD]), raising the possibility that the association between COVID-19 and all-cause dementia or AD might be explained by dementia patients being older on average than patients without dementia. Finally, it is still unclear whether AD and Parkinson's disease (PD), the two most prevalent neurodegenerative disorders in the elderly, are specifically associated with COVID-19 infection or its outcomes, or whether the association between COVID-19 and all-cause dementia results from contributions from other forms of dementia (non-AD, non-PD).

To address these questions, we investigated the prevalence of COVID-19 in a community-living cohort from the UK Biobank (UKB). An advantage of using data from the UKB is that detailed clinical data are available for infection (COVID-19 positivity), hospitalization (used as a proxy of disease severity), and deaths, thus allowing independent assessment of risk factors associated with infection and disease development. Between March 16 and August 24, 2020, a total of 12,863 individuals above 65 years old (1814 aged 80 years old and older) were tested for COVID-19 in our studied cohort, with 1167 positive. This allowed us to examine the associations of several risk factors including all-cause dementia, and AD and PD in particular, with COVID-19 positivity, severity (hospitalization), and deaths. To account for age as a risk factor, we stratified our cohort into three groups comprising individuals 66 to 74, 75 to 79, and those aged 80 years old and older. Results indicate that all-cause dementia, and AD in particular, are associated with higher risk of COVID-19–positive diagnosis and hospitalization independent of age, and are strong risk factors for severity and death among hospitalized patients even in individuals 80 years old and older.

RESEARCH IN CONTEXT

Systematic review: We searched PubMed for papers (original articles, reviews, editorials, commentaries) published between January 1 and November 30, 2020, using search terms “COVID-19″ AND [“dementia” OR “Alzheimer” OR “Parkinson”]. Available evidence indicates that patients with dementia or Alzheimer's disease (AD) are more susceptible to COVID-19. However, whether this reflects higher infection rates, older age, or increased disease severity in such patients is not clear.

Interpretation: Among several other demographic characteristics and comorbidities in a community-dwelling cohort, only all-cause dementia and AD remained as risk factors for COVID-19 diagnosis in individuals ≥ 80 years old. All-cause dementia was further associated with age-independent higher risk of COVID-19–related hospitalization and death, and was the only risk factor for COVID-19–related death in the oldest individuals (≥ 80 years old).

Future direction: These findings point to all-cause dementia or AD as targets for stringent preventive measures, higher surveillance, and early intervention in COVID-19.

HIGHLIGHTS All-cause dementia and Alzheimer's disease (AD) are risk factors for COVID-19 diagnosis in individuals ≥  80 years old. All-cause dementia and AD are age-independent risk factors for COVID-19–related hospitalization. All-cause dementia is a risk factor for COVID-19–related death in inpatients ≥  80 years old. 2 METHODS 2.1 UK Biobank cohort

We accessed the UKB database under application ID 64777. The UKB is a large prospective cohort comprising extensive phenotypic and genotypic data from approximately 500,000 individuals in the UK.8 International Classification of Diseases (ICD) versions 9 and 10 (ICD 9/10) codes for AD, PD and other forms of dementia (e.g., vascular dementia, Lewy body disease, Pick's disease) were retrieved from the electronic health records of patients, based on prior examination or on evaluation by a research nurse. In some cases, but not all, this may reflect prior imaging or biomarker results. Definitions used here are largely clinical and do not incorporate the recently proposed National Institute on Aging–Alzheimer's Association research framework for biological classification of AD on the basis of objective biomarkers.9 Participants were recruited from 2006 to 2010, and their current ages range from 49 to 86 years. Detailed information is available on sociodemographic characteristics, lifestyle, and health care of participants. Data from individuals > 65 years of age (n = 297,854) who were tested for COVID-19 between March 16 and August 24, 2020 (n = 12,863 tested individuals) were retrieved for the current analysis. COVID-19–positive individuals in the analysis (n = 1167) corresponded to patients with at least one positive test under diagnosis reference code “U07.1″ according to ICD-10,10 representing COVID-19 virus detection confirmed by reverse transcription quantitative polymerase chain reaction. Samples were collected at various times in relation to the course of SARS-CoV-2 infection. Out of the 12,863 samples analyzed, 12,274 (95%) were collected from the upper respiratory tract (e.g., nasal or nasal/oropharyngeal swabs). Additional samples included serum (2%) and other types (3%). COVID-19 tests in the UK from March 16, 2020 onward were initially largely restricted to those with symptoms in hospital, that is, swab testing in Public Health England (PHE) labs and National Health Service (NHS) hospitals for those with a clinical need. As of March 29, 2020, testing capacity was increased to include swab testing for the wider population, which included asymptomatic people. Test samples were recorded in UKB as being from a hospitalized inpatient if marked as originating from an acute (emergency) care provider, an Accident & Emergency department, an inpatient location, or from health care–associated infection; some individuals may not have been hospitalized after being seen at emergency facilities. Tests marked as being from “Healthcare Worker Testing” were not recorded as inpatient samples.

We used data on hospitalization (n = 932 COVID-19–positive hospitalized patients) as a proxy of disease severity, by consolidating hospitalization information from two UKB registries, namely “covid19_result.txt” and hospital episodes statistics (HES) “hesin.txt.” For hospital inpatient data records, the admission date was March 16, 2020 or later. Deaths registered in UKB with diagnosis “U07.1″ as primary or secondary cause were considered positive cases of death caused by COVID-19 (n = 397). To control for a possible bias associated with younger individuals experiencing milder (or no) symptoms after infection and never getting tested, we excluded younger individuals (aged 49–65) from the analyses, and only included individuals 66 years old and older tested for COVID-19. When age stratification was performed, we considered three groups, one aged between 66 and 74 years (n = 6182), another aged between 75 and 79 years (n = 4867), and a third aged between 80 and 86 years (n = 1814).

2.2 Statistical analyses

Statistical analyses were performed using R version 4.0.2.11 A binary logistic regression was performed in a univariable model using sociodemographic and clinical variables, including sex, age, ethnic group (self-declared), blood type, body mass index (BMI), high blood pressure, diabetes, asthma, heart attack, cardiovascular disease, chronic obstructive pulmonary disease (COPD), reported cancer event, all-cause dementia, AD, and PD. In this model, age was initially considered a numerical continuous variable, and subsequently stratified into three groups as described above (and see Results).

Binary logistic regression was also performed using a multivariable model stepwise regression strategy with the stepAIC function of the MASS package from R.12 In the multivariable stepwise regression, results are only shown for variables that composed the full model with the lowest Akaike information criterion (AIC) score. The AIC allows testing how well a model fits the data set without overfitting it: the model with the lowest AIC score is expected to strike an optimal balance between its ability to fit the data set while at the same time avoiding overfitting. When dementia was analyzed together with AD and PD in a multivariable stepwise regression model, we excluded AD and PD individuals from the “all-cause dementia” dataset and created an “Other dementia” group with the remaining individuals to have only independent variables in the model.

3 RESULTS 3.1 All-cause dementia is associated with higher risk of COVID-19 diagnosis

To investigate the impact of sociodemographic characteristics and clinical comorbidities on the diagnosis of COVID-19, we analyzed the entire cohort of tested individuals > 65 years old (n = 12,863), comprising 1167 COVID-19–positive and 11,696 negative individuals. This initial analysis showed that age, male sex, African ethnicity, BMI, high blood pressure, diabetes, angina, stroke, all-cause dementia, AD, and PD were associated with a higher risk of COVID-19 diagnosis, while asthma and cancer were inversely associated with risk of COVID-19 diagnosis (Table 1; Figure 1A). Several of these risk factors remained significant in a multivariable stepwise regression model, notably age (odds ratio [OR] = 1.019, 95% confidence interval [CI] = 1.004–1.034, P = .011), AD (OR = 5.700, CI = 3.709–8.762, P < .001), PD (OR = 2.242, CI = 1.511–3.328, P < .001) and other forms of dementia (OR = 3.412, CI = 2.234–5.213, P < .001; Figure S1, Table S1 in supporting information), indicating that older individuals and those with dementia, notably AD and PD, are at a higher risk of positive diagnosis of COVID-19.

TABLE 1. Univariable analysis of risk factors for COVID-19 positivity Variables COVID-19– positivea Negativea OR [95% CI] Pb Age 74.86 (4.46) 74.45 (4.33) 1.022 [1.008–1.037] .002 Sex Female 484 (41.47) 5787 (49.48) Male 683 (58.53) 5909 (50.52) 1.382 [1.224–1.562] <.001 Ethnicity White 1059 (90.75) 11176 (95.55) African 42 (3.6) 142 (1.21) 3.121 [2.175–4.388] <.001 Asian 37 (3.17) 218 (1.86) 1.791 [1.238–2.52] .001 Others 29 (2.49) 160 (1.37) 1.913 [1.257–2.809] .002 Blood type A 497 (44.57) 4986 (44.03) AB 39 (3.5) 399 (3.52) 0.981 [0.687–1.362] .910 B 113 (10.13) 1078 (9.52) 1.052 [0.845–1.298] .646 O 466 (41.79) 4861 (42.93) 0.962 [0.842–1.098] .564 BMI 29.02 (5.24) 28.28 (5.02) 1.028 [1.016–1.04] <.001 Diabetes No 1010 (86.55) 10518 (89.93) Yes 157 (13.45) 1178 (10.07) 1.388 [1.157–1.655] <.001 Heart attack No 1096 (93.92) 11106 (94.96) Yes 71 (6.08) 590 (5.04) 1.219 [0.939–1.561] .126 Angina No 1059 (90.75) 10819 (92.5) Yes 108 (9.25) 877 (7.5) 1.258 [1.015–1.545] .032 Stroke No 1117 (95.72) 11321 (96.79) Yes 50 (4.28) 375 (3.21) 1.351 [0.988–1.809] .050 High blood pressure No 633 (54.24) 7003 (59.88) Yes 534 (45.76) 4693 (40.12) 1.259 [1.115–1.42] <.001 Cancer event No 908 (77.81) 8672 (74.15) Yes 259 (22.19) 3024 (25.85) 0.818 [0.707–0.943] .006 Pulmonary embolism No 1149 (98.46) 11487 (98.21) Yes 18 (1.54) 209 (1.79) 0.861 [0.512–1.358] .546 Asthma No 1152 (98.71) 11352 (97.06) Yes 15 (1.29) 344 (2.94) 0.43 [0.244–0.697] .001 COPD No 1160 (99.4) 11602 (99.2) Yes 7 (0.6) 94 (0.8) 0.745 [0.313–1.495] .453 All-cause dementia No 1099 (94.17) 11532 (98.6) Yes 68 (5.83) 164 (1.4) 4.351 [3.239–5.783] <.001 Alzheimer's disease No 1133 (97.09) 11632 (99.45) Yes 34 (2.91) 64 (0.55) 5.454 [3.546–8.245] <.001 Parkinson's disease No 1132 (97) 11549 (98.74) Yes 35 (3) 147 (1.26) 2.429 [1.647–3.486] <.001 aResults of univariable analysis of risk factors for positive diagnosis of COVID-19 in tested individuals (n = 12,863 individuals, of whom 1167 were positive for COVID-19 and 11,696 negative). bBold P-values are significant. Table shows absolute numbers of individuals (% in brackets). Abbreviations: BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; OR, odds ratio.

Forest plots showing associations between demographic/clinical variables and COVID-19–related outcomes. Points represent odds ratio (OR) and horizontal lines represent 95% confidence intervals (CI). A, Risk factors for positive diagnosis of COVID-19; solid lines are for the analysis using the entire dataset (n = 12,863 individuals), dashed lines represent the groups of individuals 66 to 74 years old (n = 6182), 75 to 79 years old (n = 4867), and ≥ 80 years old (n = 1814), as indicated in the figure. B, Risk factors for COVID-19–related hospitalization; black symbols/lines correspond to overall hospitalization (n = 12,863) and red corresponds to COVID-19–related hospitalization (n = 6232). C, Risk factors for COVID-19–related death; black symbols/lines indicate overall deaths (n = 12,863) and red lines represent COVID-19–related deaths (n = 976). In all panels, variables with large 95% CIs (> 30) are not included for clarity. BMI, body mass index; COPD, chronic obstructive pulmonary disease

3.2 Age as a confounding factor

Age is the main risk factor for AD, and is also a risk factor for PD. Indeed, we confirmed that all groups of individuals with dementia (all-cause dementia, AD, and PD) in our cohort were on average older than non-demented individuals (Figure S2 in supporting information), with AD individuals exhibiting the highest median age (78 years [interquartile range (IQR) = 75–80] for AD vs. 75 years [IQR = 71–78] for controls, P < .001), followed by all-cause dementia (77 years [IQR = 74–80] vs. 75 years for controls, P < .001) and PD (76 years [IQR = 73–79] for PD vs. 75 years for controls, P < .001). This indicates that, as expected, all-cause dementia and age are not independent variables in our analysis.

To evaluate possible interactions between the risk factors of our interest and age, we built three different interaction multivariable models using all-cause dementia, AD, or PD each as independent categorical variables and age as continuous numerical independent variables, and COVID-19–positive diagnosis as dependent variable (Table S2). A significant interaction was observed between all-cause dementia and age for COVID-19–positive diagnosis (OR = 1.16, CI = 1.03–1.22, P = .010). No significant interactions with age were observed for AD (P = .13) and PD (P = .80; Table S2).

To reduce the impact of age as a confounder, we stratified the study cohort by age into three groups with age intervals of 66 to 74, 75 to 79, and 80 to 86 years old. In the univariable model, all-cause dementia and AD remained associated with a higher risk of positive diagnosis of COVID-19 in all age groups, while PD was a risk factor only in the two younger age groups (Table 2). It is noteworthy that, among all comorbidities examined, only all-cause dementia (OR = 5.837, CI = 3.386–9.880, P < .001) and AD (OR = 6.551, CI = 2.892–14.398, P < .001) were associated with higher risk of COVID-19–positive diagnosis in the elderly group (80 years old and older; Table 2).

TABLE 2. Univariable analysis of risk factors for COVID-19 positivity stratified by age group 66–74 y (n = 6182) 75–79 y (n = 4867) 80–86 y (n = 1814) Variables COVID-19 postive Negative OR [95% CI] P-valuea COVID-19 postive Negative OR [95% CI] P-valuea COVID-19 postive Negative OR [95% CI] P-valuea Sex Female 217 (43.23) 2925 (51.5) 189 (39.87) 2100 (47.8) 78 (40.84) 762 (46.95) Male 285 (56.77) 2755 (48.5) 1.394 [1.161–1.678] <.001 285 (60.13) 2293 (52.2) 1.381 [1.139–1.677] .001 113 (59.16) 861 (53.05) 1.282 [0.947–1.743] .110 Ethnicity White 427 (85.06) 5382 (94.75) 446 (94.09) 4223 (96.13) 186 (97.38) 1571 (96.8) African 26 (5.18) 68 (1.2) 4.819 [2.985–7.558] <.001 13 (2.74) 55 (1.25) 2.238 [1.163–3.999] .010 3 (1.57) 19 (1.17) 1.334 [0.311–3.96] .646 Asian 29 (5.78) 134 (2.36) 2.728 [1.772–4.063] <.001 8 (1.69) 63 (1.43) 1.202 [0.529–2.379] .626 0 (0) 21 (1.29) 0 [0–5.05e+06] .978 Others 20 (3.98) 96 (1.69) 2.626 [1.563–4.201] <.001 7 (1.48) 52 (1.18) 1.275 [0.525–2.64] .550 2 (1.05) 12 (0.74) 1.408 [0.218–5.214] .656 Blood Type A 226 (47.88) 2436 (44.2) 186 (40.61) 1879 (44.1) 85 (45.95) 671 (43.23) AB 14 (2.97) 182 (3.3) 0.829 [0.453–1.401] .512 20 (4.37) 162 (3.8) 1.247 [0.744–1.987] .375 5 (2.7) 55 (3.54) 0.718 [0.245–1.681] .490 B 44 (9.32) 531 (9.64) 0.893 [0.631–1.238] .510 43 (9.39) 399 (9.36) 1.089 [0.76–1.529] .633 26 (14.05) 148 (9.54) 1.387 [0.85–2.2] .176 O 188 (39.83) 2362 (42.86) 0.858 [0.701–1.049] .136 209 (45.63) 1821 (42.74) 1.159 [0.942–1.428] .163 69 (37.3) 678 (43.69) 0.803 [0.573–1.122] .200 BMI 29.25 (5.51) 28.32 (5.22) 1.032 [1.015–1.049] <.001 29.12 (5.19) 28.28 (4.92) 1.033 [1.014–1.052] <.001 28.2 (4.55) 28.11 (4.57) 1.004 [0.971–1.037] .793 Diabetes No 441 (87.85) 5171 (91.04) 406 (85.65) 3924 (89.32) 163 (85.34) 1423 (87.68) Yes 61 (12.15) 509 (8.96) 1.405 [1.05–1.851] .018 68 (14.35) 469 (10.68) 1.401 [1.058–1.832] .016 28 (14.66) 200 (12.32) 1.222 [0.783–1.846] .357 Heart attack No 477 (95.02) 5469 (96.29) 444 (93.67) 4143 (94.31) 175 (91.6