Physical frailty, impaired mobility, dysphagia and higher rates of comorbidity mean people with dementia have more healthcare needs.1 But dementia is not only a neurocognitive disorder but also a condition that severely impacts autonomy, communication, physical health, and social participation, all of which contribute to reduced ability to report needs and enhanced vulnerability.1,2 People living with Alzheimer's disease and related dementias have been found to have less access to healthcare interventions than the general population.3,4 These findings must be analysed using two different approaches. Firstly, their social isolation due to cognitive and functional impairment, as well as the association of dementia with the female gender and a low socioeconomic status, hinders communication. This means that their significant needs may not have been reflected in the amount of healthcare they received,5,6 and being an example of the disproportionate care law, as defined by Cookson et al.,5 which occurs when people receive less care than others as a proportion of their need for care, calling it the disproportionate care law.5,7 Secondly, overuse of healthcare services, which are more likely to cause harm than good, is recognised as an increasing problem, particularly among the elderly with multiple chronic conditions.8 Older adults with dementia often suffer from medical overuse, particularly in the form of inappropriate medication prescriptions and aggressive end-of-life interventions.8 Even though people with dementia use social benefits more frequently, resulting in higher formal social costs,9 it is unclear whether their hospital and primary care resource use are also higher.8

To assess inequities associated with dementia, we need to measure the use of resources, as well as variables indicating a need for healthcare in people with dementia,5,7 such as neuropsychiatric symptoms, age, socioeconomic status, and somatic comorbidity. At the same time, inequity is evident among dementia patients with a low socioeconomic status, in that they are more likely to be prescribed antipsychotics.9,10

This study hypothesises that people with dementia are a vulnerable group at risk of receiving inappropriate healthcare.1,2 Therefore, we aimed to analyse the use of healthcare resources among people with dementia in the Gipuzkoa population based on the presence of neuropsychiatric symptoms, gender, socioeconomic status, and comorbidity.

The description of the population cohort by dementia status (Table 1) shows that of the 215,859 over-60-year-olds in Gipuzkoa registered, 7553 (3.5%) had been diagnosed with dementia. The prevalence of dementia was 4.9% in the low socioeconomic status group, compared to 2.4% in the high socioeconomic status population. The same magnitude of inequality was evident when comparing by sex (4.4% in women and 2.4% in men). Furthermore, the prevalence of dementia was higher in individuals with greater comorbidity, advanced age and/or neuropsychiatric symptoms.

Table 2 shows the distribution of variables by the four categories based on dementia and neuropsychiatric symptoms status across the covariates. Of the total population over 60 years (215,859 individuals), 0.7% had dementia only, 31.9% had neuropsychiatric symptoms only, 2.8% had both conditions, and 64.6% had neither. Notably, 79% of people with dementia also had neuropsychiatric symptoms. The percentage of people in the dual diagnosis group rose from 1.8% in the high socioeconomic status group to 4.0% in the low socioeconomic status group, while the percentage in the “healthy” group (not diagnosed with either dementia or neuropsychiatric symptoms) dropped from 70.7% in the high socioeconomic status group to 54.6% in the low group.

The raw costs broken down by category of descriptive variable are listed in Table 3. The median cost was €743 (interquartile range [IQR]: 241-182) for the total population, which can be broken down into €311 (IQR: 0-1090) for hospital care and €321 (IQR: 160-601) for primary care. When comparing groups by dementia status, it was found that those with dementia had higher costs in three types of cost. Comparing groups by dementia status, those with dementia had higher costs in three types of cost. The characteristic that varied most between categories was CCI, the median cost ranging from €552 for the group with the lowest level of comorbidity to €3086 for the group with the highest. The total costs increased with age up to the 80–89 age group (median cost of €1,064) and then decreased in the ≥90-year-olds group (€802). When total hospital care costs were broken down by population group, the main differences emerged in hospital admissions (see Table S3 in Supplementary Material).

The Tweedie regression models (Table 4) indicate that the only-neuropsychiatric symptoms group used the most healthcare resources when adjusted for the other variables, with an estimator of €496 compared to €358 for the group with both conditions. The results for other variables show the same pattern as the raw costs listed in Table 3, with gradients according to CCI, socioeconomic status, and age. It is also apparent that the estimator is lower in the ≥90 age group than in the 80–89 age group. The complete Tweedie models of total, primary care and hospital costs without interactions and with interactions between population group and socioeconomic status are provided in the supplementary material (see Tables S4 to S9 in Supplementary Material) and showed significant interactions.

Figure 1 represents the mean of the estimated costs and upper and lower confidence limits for each population group for the reference categories of covariates obtained from total healthcare costs (Fig. 1 A), primary care (Fig. 1 B) and hospital costs (Fig. 1C).

Figure 1.

Adjusted mean health costs and confidence intervals by population group. Panels correspond to cost-specific models: total (A), primary care (B), and hospital costs (C). The reference subject is a male, 60-69-year-old with high socioeconomic status and a Charlson Comorbidity Index (CCI) of 0-1. NPS: neuropsychiatric symptoms; SES: socioeconomic status.

Figure 2 shows the same indicators for each of the cost-specific models, broken down by sex. These plots help visualise the higher costs in the only-neuropsychiatric symptoms group versus the both conditions group. Furthermore, women receive a lower amount of healthcare than men, despite this being adjusted according to need.

Figure 2.

Adjusted mean health costs and confidence intervals by population group and sex. Panels correspond to cost-specific models: total (A), primary care (B), and hospital costs (C). The reference subject is a 60-69-year-old with high socioeconomic status and a Charlson Comorbidity Index (CCI) of 0-1. NPS: neuropsychiatric symptoms; SES: socioeconomic status.

The main finding of this study is that, when resource use, or better expenditure adjusted for need, is adjusted for need, people with dementia receive less healthcare than people without dementia. However, it is difficult to determine whether this is harmful because it deprives people with dementia of necessary care,20 or beneficial because it spares them from receiving unnecessary treatment.21 In the former instance, the law on disproportionate care5 would be complied with, and overuse of health services would be avoided in the latter. Although the raw cost is higher in groups with dementia, in the populations characterised by the presence of neuropsychiatric symptoms, the total healthcare cost estimator adjusted for CCI, age and gender is higher in the group without dementia than in that with dementia (€496 vs €358). This difference between raw and adjusted costs can be explained by the fact that individuals in the dementia group are older and have more comorbidities. In accordance with the terminology of the disproportionate care law, this could mean that, their higher needs are not being met proportionally. Adequate access to healthcare resources is essential for enabling people with dementia to maintain their independence, as emphasised in the Global Action Plan on the Public Health Response to Dementia 2017–2025, adopted by World Health Organization member states at the 70th World Health Assembly.1 However, defining appropriate care is challenging because more is not always better in healthcare, especially for multimorbid elderly patients.21 The challenge lies in addressing all conditions that may reduce quality of life and avoiding interventions whose potential for harm exceeds possible benefit.20,21

Combating stigma is critical to ensuring equity of access to services for vulnerable groups, particularly those affected by social interactions associated with a stigmatised characteristic.22,23 The management of neuropsychiatric symptoms in dementia patients is an example of overuse of antipsychotics. Reducing the use of antipsychotics in low socioeconomic status patients by increasing the availability of non-pharmacological treatments could help reduce inequity in the management of neuropsychiatric symptoms 10. At the same time, in order to reduce their severity, expert recommendations emphasise that efforts must be made to control all conditions that may affect patients’ well-being such as pain from musculoskeletal problems and urinary tract infections.24,25

The higher healthcare costs in the low socioeconomic status group can be partly explained by the fact that 15% of the population in Gipuzkoa has dual healthcare coverage,26 and most of these people have a high socioeconomic status. In any case, only 15% of the population in Gipuzkoa had dual health coverage in 2020.26 On the other hand, the Basque Health Service is a tax-funded Beveridge-style system providing universal coverage.27 The benefits of this type of system in terms of greater equity have been widely demonstrated.28 Nonetheless, while universal health coverage may appear to guarantee equity, it is essential to assess resource use and ensure that vulnerable groups, such as people with dementia, receive adequate healthcare. As is well known, conditions not quantitatively measured are undervalued. Providing equitable care requires acknowledging people's vulnerability and specific care requirements. The interaction between the three analysed costs (total, hospital and primary care) and socioeconomic status was significant, showing that the medium socioeconomic status and neuropsychiatric symptoms group increased its costs. For individuals with neuropsychiatric symptoms, dementia status determines whether the level of healthcare they receive.

As expected, the differences according to CCI show that the greater the comorbidity, the higher the costs in both hospital and non-hospital settings.29 Although people with higher comorbidity tend to be men, older people and more likely to have dementia, the gradient in total and primary care costs as a function of CCI remains in the regression models. This can be interpreted as an absence of inequity.

Clear differences by gender indicate that women receive less care for equal needs.30,31 Inequities related to dementia in women are mediated by their lower income, older age, higher risk of dementia and lower level of comorbidity.6,10 Therefore, the relevant results are those adjusted for these variables, which reveal gender-related inequities at both hospital and primary care levels. Similarly, the conclusions of Cooper et al. suggest greater inequity among women with dementia, who receive less healthcare than men with dementia.4

The healthcare system also treats differently the ≥90 age group, as they use services less frequently than the 80–89 age group, both in raw terms and when adjusted for covariates indicative of need such as CCI, dementia, and/or neuropsychiatric symptoms. This finding is evident in both total costs and primary care costs. Within the context of overuse, this pattern could be attributed to the belief that it is better for this age group not to receive care in order to avoid potential harm.3,21

As Cookson et al.5 point out, measuring healthcare service delivery is easier than measuring need for care. One strength of this study is that population data were available from electronic health record databases. These sources provided data that can be used to assess the need for healthcare, such as the presence of non-communicable diseases, age, gender, comorbidity and socioeconomic status. Although they do not offer the depth of information provided by surveys,32,33 such records do contain data on entire populations, unlike survey-based studies.34

People with dementia, women and those aged 90 and over receive less care than others with comparable needs. This could be an example of the disproportionate care law, or it could be a way of avoiding unnecessary treatment and overuse of health services. In any case, reducing the stigma surrounding cognitive impairment would convey the message that it is not a contraindication to healthcare that can improve the quality of life of patients and their families.

All statistical analyses were performed using the open-source statistical software R (version 4.3.3, R Core Team, Vienna, Austria) with a significance level of 95%. The following packages were used for specific analyses: compareGroups package17 for descriptive statistical analyses, statmod package18 for the Tweedie distribution in the models, and ggplot2 package19 for graphical visualization. The code to reproduce these analyses is open and available at https://github.com/lzumeta/HealthCare_costs_Dementia_Gipuzkoa. The owner of the data is the Basque Health Service which maintains a policy of restricted data access for ethical reasons. Our agreement with the Basque Health Service establishes that individual level data cannot be shared publicly because of confidentiality. Applications for access to the dataset must be made with the approval of the Basque Health Service (https://www.osakidetza.euskadi.eus/servicios-para-profesionales/webosk00-sercon/es/).

Salvador Peiró.

The corresponding author, on behalf of the other authors guarantee the accuracy, transparency and honesty of the data and information contained in the study, that no relevant information has been omitted and that all discrepancies between authors have been adequately resolved and described.

J. Mar, M. Soto-Gordoa and M. Merino designed the study. J. Mar wrote the main manuscript text (Introduction and Discussion). L. Zumeta-Olaskoaga wrote the main manuscript text (Method and Results). L. Zumeta-Olaskoaga and O. Ibarrondo conducted the statistical analysis. M. Soto-Gordoa and O. Ibarrondo collected the data and reviewed the literature. All authors reviewed the manuscript.