In this large cohort of geriatric patients hospitalized with COVID-19 infection in Stockholm, Sweden, the presence of dementia was associated with significantly higher in-hospital, short-term and long-term mortality compared to patients without dementia. Moreover, the prescription of psychotropic medication—particularly antipsychotics—was correlated with significantly higher mortality.
Dementia comprises multiple layers of pathology, clinical presentations, and prognoses; therefore, the outcome of COVID-19 infection may significantly vary among individual patients in the dementia spectrum, and the information on specific dementia diagnoses is currently insufficient. Importantly, multiple studies concur with the increased risk of mortality among dementia patients. However, the magnitude varies substantially—while the early report from Zheng et al. did not recognize dementia as a major prognostic factor , 2020 reviews by Izcovich and colleagues and July and colleagues reported 1.54× and 1.80× increase in mortality associated with dementia, respectively [13, 27], and three recent meta-analyses by Hariyanto and colleagues, Saragih and colleagues, and Damayanthi and colleagues concluded 2.62×, 2.96×, and 3.69× higher risk of mortality, respectively [9, 28, 29]. Contextually, our finding of 1.59× times higher mortality may seem low; however, the inclusion of multiple important confounders, e.g., entry oxygen saturation and pharmacological management, likely diminished the association but provided a “clearer” picture of the dementia-COVID link. In addition, our data suggests that the mortality is substantially increased even 1 year after discharge from the hospital setting, suggesting residual cognitive, pulmonary, or other ailments.
Additionally, the approach to dementia care in each "country" is a major confounder and could not be summarized by a singular number. Thus, our relatively moderate mortality risk may reflect a high standard of dementia care in Sweden and improvement in treatment and care of COVID-19. The mortality is substantially modified by increasing vaccination rates which was also likely reflected in the non-significant mortality increase in the third COVID-19 wave (the COVID-19 vaccination began at the end of December 2020 in nursing homes). Although the vaccination rates among dementia patients could be higher, Sweden in general has a high overall immunization percentage among the elderly , so this problem is likely non-differential.
While the higher risk of death in COVID-19-infected dementia patients may be intuitive, the underlying connections interacting between the two disorders are complex. First, patients with dementia are more prone to being infected and hospitalized with COVID-19 [7, 8, 15, 31, 32] and suffer more frequently from severe outcomes of COVID-19 [7, 9, 10]. These associations are likely explained by overall higher comorbidity burden including respiratory complications [11, 12, 33], inability to follow social restrictions [8, 15], increased caregiver burden , and possibly even higher infectious load in apolipoprotein E epsilon-4 variant carriers [15, 34]. Second, in older patients, the symptoms of the infection often do not conform to the usual clinical presentation, and atypical forms with altered mental status, tachypnea, and delirium may be dominant [8, 35]. In combination with lower social interaction, decreased frequency of health check-ups, lower cognitive ability to recognize the worsened state , and generally higher propensity towards pneumonia  may explain the more prevalent oxygen desaturation below 90% at admission in the dementia group and may have particularly contributed to the increased in-hospital mortality.
Furthermore, the clinical course of dementia is complicated by a variety of behavioral and psychological symptoms , and the burden of neuropsychiatric symptoms has increased during the pandemic [14, 37]. The increase is caused either indirectly through social isolation and lockdown measures [38, 39] or directly by COVID-19, which has been associated with the development and deterioration of neuropsychiatric sequelae . Consequently, the presence of mental disorders may independently contribute to more severe outcomes in COVID-19 patients [41, 42], which would disproportionately impact patients with dementia.
Recent reports suggest the longitudinal functional and mental capacities in COVID-19 survivors may improve during the follow-up examinations [43, 44]; however, it is unclear whether such observations extend to dementia patients. Conversely, COVID-19 has been associated with exacerbations of cognitive and neurological symptoms [45, 46], which in the long term can be expected to result in an increase in future dementia cases, primarily through increasing the cardiovascular and cerebrovascular burden .
Therefore, the early identification of deviations from the normal clinical course and stricter observation of dementia patients are reasonable steps until the COVID-19 pandemic is completely under control.
In addition, a specific concern should be given to the finding of higher mortality among psychotropic drug users, particularly antipsychotics. Importantly, there are reports of increased antipsychotic drug prescription rates among dementia patients during the pandemic [16, 21] as well as associations between antipsychotics and higher probability of COVID-19 infection , though some reports dispute such claims . In our cohort, the use of antipsychotic medication at admission to hospital was associated with 70% higher overall mortality, with the highest increase observed among in-hospital patients (3.58× higher hazard, with a 5% decrease in risk with each survived day), and the risk decreasing in the post-discharge periods (1.39× aggregate increase). The mortality increase may reflect the necessity of psychotropic medication use in overall more severe and atypically presenting COVID-19 (e.g., delirium, altered consciousness, and psychotic states) [8, 35, 48, 49], or higher psychiatric burden among patients with severe COVID-19, or necessary pharmacological management of BPSD. Importantly, the association survived adjustment for dementia, comorbidities, and oxygen saturation. Other studies concur with the direction of the association; however, the risk estimates vary between 1.26× and 11.1× the risk of death depending on the population, with large heterogeneity between the studied mortality periods [21, 22, 32, 44]. Unfortunately, we had no information on the pre-admission or post-discharge usage of antipsychotic drugs, which would allow us to confirm the putative increase in infection rates or study time-dependent effects.
The mortality risk in antipsychotics depends on the agents prescribed and dosage [20, 50, 51], which could have further confounded the association. We hypothesize that patients with dementia may be better adjusted to psychotropic medication due to BPSD, which may be reflected in the significantly higher psychotropic drug prescription among dementia patients. The results suggest some difference in risk estimates between dementia and dementia-free (38% in absolute difference, favoring dementia); however, the interaction was not significant. Sweden is a country with high quality of dementia care, and the care standards and the Swedish Dementia Registry (largest dementia registry in the world)  discourage the use of antipsychotics in dementia patients. Therefore, a comprehensive evaluation of psychotropic medication including dose-response is necessary before a valid guidance can be provided for individual drugs.
Further findings in our study confirm male sex as an independent predictor for higher mortality among COVID-19 patients with dementia, and the sex-stratified adjusted analyses suggested that the dementia-associated mortality increase has been observed primarily in male patients. Male sex is an established risk factor for higher COVID-19 mortality, with several putative factors suggested for the elevated risk (protective inflammatory effect of estrogens, higher comorbidity burden among men, behavioral/lifestyle differences) . Importantly, the strength of the association in our study (55% increase) concurs with the findings of other authors [5, 13, 53] and suggests the male patients with dementia admitted for COVID-19 are more likely to need intensive treatment thus allowing for appropriate care allocation.
In conclusion, the study brings novel information on the short- and long-term mortality among the dementia population suffering from COVID-19. However, we would like to highlight the major role COVID-19 plays in the elderly population, and post hoc analyses of the pandemic’s impact on multimorbid and frail populations will be crucial .
Overall, our study provides robust evidence that dementia is a strong and independent risk factor for both short- and long-term mortality after COVID-19 infection, and the use of psychotropic medication, particularly antipsychotics, compounds the risk of dying. As the cognitive reserve is exhausted, even a mild form of COVID-19 will likely lead to a non-proportional decrease in cognitive functioning; therefore, dementia patients may benefit from heightened clinical vigilance and an individualized approach. Such efforts should be concurrent with the en-masse vaccination programs, particularly focusing on booster applications and repurposing the vaccines towards new variants of concern, where the immunization efficacy was not primarily studied [2, 55,56,57,58]. The improvement of social engagement, proper cardiovascular and cerebrovascular prevention, and cognitive training in the post-pandemic era will play a crucial role in the prevention of long-term damage in the already fragile population of dementia patients.
Strengths and limitations
The main strengths of the study comprise the inclusion of a large unrestricted sample of geriatric patients with COVID-19 and the long follow-up after the hospitalization. The COVID-19 diagnosis was based on RT-PCR analysis of a nasopharyngeal swab specimen, generally considered the standard approach in COVID-19 diagnosis . Moreover, the diagnosis of dementia was bi-specified by either the presence of ICD-10 codes or the use of anti-dementia medication, to increase the sample size. Another major strength is the division by periods of COVID-19 infection, where all three major infection waves were analyzed. Finally, the inclusion of comorbidity, medication, and clinical data allowed for more extensive adjustment and possible generalization to a larger population.
On the other hand, a careful interpretation of the medication findings is in order, as we had no information on pre-hospital medication usage, nor the indication data for the prescription. However, it is not clear how such misclassification would affect patients with dementia to a larger extent (specifically in the treatment with antipsychotics). In addition, the Swedish setting might not be representative of the antipsychotic medication usage elsewhere, as the two main clinical registries (Register for Behavioural and Psychological symptoms of dementia and Swedish Dementia Registry) advocate towards lower use of antipsychotics in dementia care. Conversely, this stresses the need for further reports from other European countries to enable comparisons.
Furthermore, we had no data on cognitive or functional performance measures in the patients; thus, no stratification or adjustment on dementia severity could have been made.
The inclusion of data only from the geriatric wards may limit the generalizability of the results to an older and more frail population; however, this should not alter the primary results that extend to the dementia population—a typical disorder of the advanced age. However, further post-discharge data are needed to comprehensively assess cognitive functioning in patients with dementia who overcame the COVID-19 infection.
Finally, as in all observational analyses, we acknowledge the possibility of residual and unknown confounding including the information on vaccination.