4.1 Cost Estimates for MCI

The results of this study suggest that resource utilization in AD increases early in the disease stage, with 1.3 times higher costs in MCI than in the corresponding control group. Lin et al. analyzed direct medical costs in MCI in the US setting and showed that the costs for MCI are higher than for controls even 12 months before the diagnosis [43]. Another study by Frahm-Falkenberg et al. in a Danish setting shows that people with dementia incurred higher costs than controls even 1 year before the diagnosis [29]. This pre-diagnosis period could be representative of the MCI state. In our analysis, we used a 5-year cut-off to retrospectively identify AD patients who had a medical visit with an MCI diagnosis. This 5-year threshold was chosen based on literature that shows the duration of MCI varies widely depending on factors like age, sex, biomarkers, and APOE 4 status. One study combining data from six cohorts suggests the average time spent in MCI is 4 years (95% CI 3–5) for individuals aged 70 years entering preclinical AD [44]. Another study showed that the predicted time spent in MCI is 5.51 years (95% CI 4.78–6.25) for women and 4.15 years (95% CI 3.71–4.59) for men aged 79 years when entering the MCI stage [45].

We also conducted a sensitivity analysis on MCI costs using different cut-off periods. The results showed that MCI costs were highest when only considering clinic visits within 1 year before AD dementia diagnosis and lowest when including MCI visits up to 5 years before AD diagnosis. This could be because individuals diagnosed with MCI within 1 year of a mild AD diagnosis may be awaiting further clinical evaluation for AD and are likely at the more severe end of the MCI spectrum. These findings highlight the need for longitudinal studies that account for all costs and time spent in MCI to calculate an average annual cost.

4.2 Cost Estimates in Amyloid-Positive Populations

We identified only two studies that estimate resource utilization in the amyloid-positive AD population [46, 47]. Both studies are based on the GERAS-US cohort and provided cost estimates for MCI and mild AD but are different in study design and data sources used. Robinson et al. conducted a prospective study in 679 MCI and 650 mild AD patients and provided monthly societal cost estimates in 2017 USD [46]. To make it comparable to our study, we extracted direct medical and social costs and converted them to 2021 annual costs, which gave rise to a cost of USD $16,519 for MCI and USD $21,057 for mild AD. Looking at the amyloid-positive population, the study provides a cost estimate of USD $16,237 for amyloid-positive MCI and USD $15,176 for amyloid-positive mild AD. Chandler and Kubisiak used Medicare claims data to investigate the healthcare costs (n = 174). They showed a statistically significant lower cost in amyloid-positive than in amyloid-negative mild AD, which might be explained by the higher comorbidity profile in amyloid-negative AD in their study [47]. We did not directly compare our cost estimates with these studies due to differences in care structure and cost components.

Our study shows that cost estimates for amyloid-positive AD are lower than those for clinically diagnosed AD populations. However, we did not make a direct comparison between these two estimates, as the amyloid-positive group in our study is a subset of the total study population and not from a separate cohort. Additionally, the primary goal of the study was to provide cost estimates for economic modeling rather than to make direct comparisons.

It is worth noting that only 19% (N = 6059) of our study population had recorded CSF p-tau/Aβ42 results, and the majority (N = 5610) of them were amyloid-positive. Individuals who underwent CSF examinations tended to be younger, had higher baseline MMSE scores, and had lower comorbidity profiles (as shown in Table 9 in the ESM). This aligns with Sweden’s diagnostic guidelines, where biomarker confirmation is not mandatory, and only patients without a definitive AD diagnosis or younger patients with suspected dementia are referred for extended testing, including CSF examinations, at specialist clinics [34]. Therefore, the cost estimates for amyloid-positive AD in this study may not be fully representative of individuals who did not undergo biomarker confirmation.

4.3 AD Versus Control

The study showed that people with AD consistently incur higher costs compared with their corresponding controls. Specifically, the cost incurred by AD patients is 1.3 times, 2.6 times, 3.9 times, and 7.3 times higher in MCI, mild, moderate, and severe AD than in the control population. This cost difference is apparent across all cost components, with the greatest differences observed in costs for home care and institutionalization. These results align with findings from other studies [48, 49]. A systematic review by Sontheimer et al. showed that dementia patients incur 1.2 times higher costs than controls, particularly in home care (4 times higher) and institutionalization (3 times higher) [49]. In Sweden, Sm-Rahman et al. found that about 75% of dementia patients stay in nursing homes compared with 23% of those without dementia during the 1 month preceding death [50]. For those living in the community, Wimo et al.’s analysis indicated that people with cognitive impairment use formal and informal care 1.8–5.2 times more than those with no cognitive impairment [51]. They also showed that the extent of informal care is 1.9–4.4 times higher than formal care use [51]. However, our analysis did not include informal care, so a comparison of informal care costs was not made.

4.4 Cost by AD Severity

The estimated annual cost of care in AD increased with disease severity, ranging from SEK 99,906 in MCI to SEK 795,617 in severe AD. The cost increment with disease severity is mainly driven by direct social costs, consistent with existing literature [12, 14, 15, 52]. A Swedish population-based study of individuals aged > 75 years residing in nursing homes showed that the cost increment is not only due to the dementia disease itself but also due to ADL dependency [53, 54]. A systematic review by Cantarero-Prieto et al. estimated the cost in 2015 EUR as 16,968 EUR, 24,024 EUR, and 28,716 EUR for mild, moderate, and severe dementia [14]. Another systematic review by Tay et al. provided cost estimates in 2021 USD as $18,344, $35,486, and $46,353 for mild, moderate, and severe AD [15]. However, a direct cost comparison with the present study is difficult due to the differences in care structure and definitions of cost components. For instance, Tay et al.’s review included countries in Asia, Europe, and the US [15] and Maldonado et al. showed very low costs in institutionalization but included informal care costs in the direct non-healthcare costs estimation [12]. This highlights the importance of country-specific estimates and the need for a transparent definition of cost components.

We compared our cost estimates with the meta-analysis of the cost of dementia in the Nordic region, which estimated the costs in 2021 EUR [13]. The meta-analysis provided an estimated societal cost of 20,876 EUR, 37,540 EUR, and 58,198 EUR for mild, moderate, and severe dementia. The costs from our analysis were higher than that study in all disease severities even when we did not include the cost of informal care. The main differences observed were higher inpatient costs and direct social costs in our study. This could be due to the study design. Our study is a retrospective study capturing all costs incurred from diagnosis until death or censoring, while the studies involved in the meta-analysis were either cross-sectional studies, which estimate the cost within 1 year, or prospective studies, which asked patients or caregivers about the resource utilization within a short period, which might contribute to recall bias. The costs captured by our study might be more accurate due to the use of real-world resource utilization data.

4.5 Cost Difference by Setting

The costs of dementia care are consistently higher for those institutionalized across all stages of AD severity compared with those residing in the community. The main cost driver is institutionalization, as confirmed by a review conducted by Schaller et al. [10]. A comparative study of costs of care in eight European countries showed that Sweden bears the highest institutionalization costs among the countries [11]. Another study investigating the factors influencing long-term care (LTC) utilization across 16 European countries found that variations in LTC use are mainly attributable to a country's public spending on institutional care [55]. Sweden is a well-developed welfare state with an established LTC system that provides health and social care services for individuals aged 65 years and older. The 290 Swedish municipalities are legally obliged to provide LTC services, which are publicly financed by local taxes and services and are accessible to all socioeconomic groups. The access to LTC is needs-tested and includes both home care and institutionalized care [56]. Therefore, the use of institutionalization in Sweden is higher than in other European countries, which is also evidenced in the meta-analysis by Jönsson et al. [13].

4.6 Cost Difference by Sex

Our study showed that female AD patients incurred higher costs for institutionalization and home care while males incurred higher direct medical costs and costs for short-term care. These findings are consistent with existing literature [48]. Epidemiology studies have shown that AD affects men and women differently. Women are at greater risk and have a higher prevalence of AD [57]. They tend to live longer and spend more time in institutions than male AD patients [45, 50]. Moreover, women are often placed in caregiving roles for those with dementia, contributing to 60–70% of all primary caregivers in dementia [58]. As a result, women who develop dementia are more likely to end up in institutional care often because they are caring for a spouse with dementia or because their spouses have already passed away, leaving them living alone [59, 60]. Studies on LTC usage in the final years of life in Sweden support this hypothesis [59, 61]. This pattern is also evident in Wimo et al.’s investigation in a population-based study, which found that women are more likely to receive formal care, whereas the extent of informal care use is over twice as high in men (69 h) than women (32 h) [51]. The study did not identify significant sex differences in the extent of formal home care, likely because institutionalized individuals were excluded from the analysis.

4.7 Cost Difference by Age Group

We did not identify studies that estimate costs by age group. Schwarzkopf et al. analyzed the impact of age on healthcare expenditures and found that total care costs and long-term care costs increase while medical care costs decrease with age [48]. Our study showed that the costs for inpatient care, institutionalization, and home care increased with age, whereas those for outpatient care decreased with both age and advanced disease. This might be due to functional limitations in older age and more severe diseases, which hinder healthcare access in outpatient settings. Overall, our findings indicate that total costs increase with age across all stages of AD severity. Health economic models that involve cost parameters should consider the effect of patient age on costs when modeling interventions for the AD population.

4.8 Strengths and Limitations

This is the first study using real-world registry data to provide cost estimates across the full AD continuum, from MCI to severe AD, and for the amyloid-positive AD population. In the economic evaluation of health care interventions, it is imperative that costs are estimated in a population that corresponds to the product label. Previous studies usually did not include estimates for MCI and amyloid-positive individuals, who are now the target populations for new DMTs. Our cohort includes those with both clinical and biomarker-confirmed AD, offering valuable data for cost-effectiveness analyses of DMTs. The cost estimates in this study are based on routinely collected data and, therefore, are more accurate and reflective of real-world resource utilization. We report costs by AD severity, care settings, cost components, sex, and age, allowing economic modelers to tailor costs to their specific populations. Additionally, our MCI population is retrospectively identified before the diagnosis of mild AD, representing those who could benefit from DMTs. By including all MCI visits within 5 years of AD diagnosis, we comprehensively captured resource use, providing balanced cost estimates that avoid the overestimation seen with shorter windows, such as the 1-year time window in the sensitivity analysis.

This study has several methodological limitations. First, the costs of informal care were not included. Previous studies have provided information on informal care costs [16,17,18,19,20], where the authors prospectively collected the data using the RUD instrument [62]. Our study is a retrospective analysis of registry data where routine data collection of informal care costs is not feasible. However, the information can be retrieved from a recently published report in Sweden, where the costs of informal care were estimated to account for 16% of the total costs [9]. The report also estimated that institutionalized people required 0.3 hours of informal care, while people who are not institutionalized required 1.9 hours of informal care per day [9]. We have provided the cost estimates by cost components, which allows easy subtraction and addition of any cost components to the total cost based on different care settings. Therefore, adding an informal care component from another study to our cost estimates is feasible. Second, the costs of primary health care (PHC) visits were not included in the analysis. This is because the NPR in Sweden does not record PHC visits. Since MCI is the early stage of AD, people in this stage might present their minor symptoms to a primary care physician. Excluding this might lead to underestimating the direct medical cost in this particular population. Lastly, we treated the MMSE measurement to be valid for 2 years. Thus, any costs incurred within 2 years from the last MMSE measurement were treated as costs for that MMSE-defined AD state. This might lead to an overestimation of costs in milder AD states since patients might have progressed to more severe disease if they don’t come to the clinic for the next MMSE assessment. However, considering the slow progressive nature of the disease, assuming the patient maintains the same MMSE for 2 years seems to be reasonable.

4.9 Future Research

Our study is the first study estimating cost by dementia severity in the amyloid-positive AD population in Europe. Since there is continuing interest in estimating the cost effectiveness of interventions in early AD, these cost estimates are essential for an accurate evaluation of these interventions. More cost estimates for MCI and amyloid-positive populations from different countries and care settings are needed since our estimates can only reflect Sweden and Nordic countries at best. Moreover, future studies should integrate the costs of informal care into the cost estimates.