Costs of Dementia in Low- And Middle-Income Countries: A Systematic Review

Search strategy

Bibliographic databases on health, health economics, and social sciences were searched: PubMed/Medline, Embase, Global health, PsycINFO, PsycARTICLES (via OVID), Web of Science, and Econlit. The main grey literature resources in the health field were also searched (Greylit, ProQuest, Georgetown University resource, and Google scholar). No language or date restrictions were applied. Search strategies consisted of combinations of different keywords related to the disease (dementia, Alzheimer’s, cognitive disorder/impairment), costs (cost, valuation, expenses, and burden), and settings (low-income, middle-income, resource limited, developing country, and the list of the 137 LMICs countries (according to the World Bank on January 1, 2020 [14])). Specific equations for each database as searched on March 2, 2020 are presented in Supplementary Material 1. An updated search using the same strategy was conducted on December 28, 2021 to identify potential new studies of interest.

Studies selection

Results obtained from the searches on the different bibliographic databases were imported and processed using the Rayyan QCRI tool [15]. In a first step, duplicates were removed. Then, following a blind selection method, two investigators (AKM and DA) assessed the eligibility of the selected articles based on titles and abstracts. The proportion of agreement between the two investigators was measured; conflicting articles at this stage were all retained for the next stage. Full texts of selected references were retrieved, and the corresponding authors were contacted to request the manuscript when those were not available. Full-texts were then evaluated by the same two investigators (AKM and DA) for the final inclusion decision. Conflicting articles were discussed, and the inclusion decision was made with a third investigator (MG or CT).

Protocols, reviews, and meta-analyses were not included, but were examined to identify any relevant study for consideration. In addition, references of selected articles were checked. For publications in languages other than English and French (those in Spanish, Portuguese, and Chinese), collaborators were asked to provide a translation and/or description of the studies in order to make a decision on eligibility and to extract the data of interest.

During the selection process, articles were excluded according to the Population, Intervention, Comparator and Outcome (PICO) framework. Laboratory experiments (i.e., fundamental research on Alzheimer’s disease/dementia, including in vitro studies and studies in non-human populations), studies not addressing dementia care, and studies conducted outside LMICs were considered as wrong population. Wrong publication types included narrative review, meta-analysis, overview article, letter to the editor, and commentaries. Studies without a direct measure of dementia cost were accounted as wrong outcomes. Studies for which the full-text was not retrieved despite multiple searches and contact with the authors were also excluded.

Quality assessment

Although various guidelines have been published for assessing the quality of economic evaluation in health care [16, 17], they mainly focus on cost-effectiveness analyses. In the absence of a standard instrument to evaluate cost-of-illness studies, we assessed the quality of articles using a checklist, derived from the guide proposed by Larg and Moss [7] and the checklist developed by Kline-Budde et al. [18] for their review on bipolar disorders. Full description of the quality assessment criteria used is provided in Supplementary Material 2.

Data processing and analysis

Full texts were reviewed and data extracted using a standardized extraction grid developed in Epi Info 7 and Microsoft Excel®. Data were collected on study design, population characteristics and costing methods. Particular attention was paid to the costing perspective (i.e., the point of view from which costs are considered). Studies may measure costs to a health care system, a third-party payers, commercial sectors, government, or participants and their families [19]. In this review, the perspective was collected as specified in the original article. Instruments used to collect and valuate health resource utilization were also explored, to check the use of standard validated tools such as the Resource Utilization in Dementia (RUD) instrument [20] or questionnaires specifically designed by researchers for the purpose of their study. Costing approaches used to value informal care time were also assessed. Various methods for the valuation of informal care are available in the literature. On the one hand, there are revealed preference methods that are based on real-life decision data: 1) the opportunity cost method, which aims to estimate the cost of the opportunity to earn income or partake in leisure activities that an individual forgoes when engaging in informal care; 2) the replacement cost, or proxy good method, which assigns costs based on the market value that the care services would cost if they were provided by a professional provider. On the other hand, stated preference studies obtain value by asking respondents directly what economic value they attach to informal care through contingent valuation or conjoint analysis [21, 22].

When information was not presented in the original article as required for extraction, available data were converted to the accurate value and unit. For example, all sex ratios and gender figures were converted to male percentage. When data was not available or presented in a non-transformable format, it was considered missing. When summarizing data, weighting was done using the representativeness of each category in the total sample. For example, to calculate the total medical costs of a given study presented by disease severity, the proportion of participants in each stage was taken into account. Monetary data were converted into local currency unit (LCU) using the exchange rate provided by the authors. Where these were not available, we used exchange rates provided by the World Bank for the country in the year of the study cost [23]. Costs were then updated to their 2020 value using country-specific deflators provided by the World Bank [24], and converted to 2020 dollar purchasing power parities (PPP) using official conversion factors [25].

All costs were collected under the original items as provided by studies. They were then grouped into a dozen of health resource services and finally into the three main cost-of-illness categories (i.e., direct medical care, social care, and informal care). Some studies presented costs results directly in the three main categories which does not allow for a desegregated presentation; these are presented only in total costs descriptions.

The protocol of this systematic review was registered in PROSPERO (registration number CRD42020169898) before data extraction began.

Characteristics of included studies

Characteristics of the 17 included studies are summarized in Table 1. Out of 137 low- and middle-income countries, only 11 were represented in this review, including six studies in China [31–37]; 2 in Brazil [30, 38]; and one in each of the following countries: Argentina [28], Colombia [39], India [40], Iran [41], Peru [42], Philippines [43], Romania [44], Thailand [45], and Turkey [46]. Our dataset covered all epidemiological regions of the Global Burden of Disease, with the exception of Africa. Regional coverage was as follow: 1 from Eastern Mediterranean [41]; 2 from South-East Asia [40, 45]; 2 from Europe [44, 46]; 5 from the Americas [28, 42]; and 7 from Western Pacific [31–37, 43]. Studies were conducted between 2003 and 2020. All studies performed a retrospective collection of data in the previous 1–12 months, except studies from Turkey [46] and Philippines [43], which added a 15-day and 5-year prospective data collection period on time spent in informal care, respectively.

The majority of studies focused on Alzheimer’s disease only (12/20), 5 studies included all types of dementia and the remaining four did not specify dementia subtypes in their study population. In almost all studies, case identification was based on a previous dementia diagnosis. A few updated or verified the diagnosis before inclusion in the studies. Disease definition followed standard criteria including ICD-10, DSM-IV-TR and/or NINCDS–ADRDA, plus various clinical or imaging investigations. Some studies excluded patients with mild cognitive impairment [29, 30], early onset dementia [41], or limited functional independence [42]. Only four studies reported the mean disease duration: 2.7 years, 51.0 months, 60.1 months, and 5.1 years in China [34], Romania [44], Brazil [30], and Thailand [45], respectively.

The majority of studies examined all three categories of costs (direct medical, social, and informal care costs). Three studies [36, 39] investigated only medical costs, while Wang et al. [35] only looked at informal care costs. Zencir et al. [46] and Mould-Quevedo et al. [33] did not collect data on social and informal care costs respectively. Items included in each cost category varied widely. For direct medical costs, most studies included outpatient, inpatient and medication costs. Some reported only medication [30, 39] or hospitalization [36] costs. Regarding medication costs for instance, Soares et al. [38] in Brazil reported the costs of dementia-specific drugs dispensed by a government dispensing center. Meanwhile in Colombia, Prada et al. [39] considered direct insurance reimbursement for all-condition drugs of dementia patients. In one study from China, Wang et al. [34] included the costs of official dementia drugs, but also those of traditional medicine. Social care costs included travel expenses and formal care services where appropriate, but also food expenses [31, 41]. With regard to informal care costs, most studies reported estimates of time spent on care, without much detail on the activities considered. Two studies considered the patient loss-of-productivity [30, 44] and one included the costs of health services for the treatment of the caregiver [30]. Although the majority of studies valuated caregiver time by the replacement cost methods (which assigns a monetary value to caregiver’s time based on what it would cost to replace informal care activities with formal care), the proxy used varied considerably, ranging from national minimum or average wage, to the salary of a nurse in a public institution, or to the caregivers’ salary. Table 2 provides more details on the design characteristics of the included studies.

Results from the quality assessment of the studies are presented in Table 3. Almost all studies provided basic information on the study design and population characteristics, such as objectives, sample size and demographics. Eight out of the 17 studies clearly indicated the cost estimate perspective. Only Allegri et al. [28] and Custodio et al. [42] included a group of healthy participants in their sample, although they did not use this information to calculate disease-specific net costs. Only the cluster-randomized study in China presented the results of a sensitivity analysis [31, 32]. Most studies (9/17) failed to present separate information on the number of services used before getting in their valuation. Very few studies performed sensitivity analyses to account for uncertainties in key indicators. Onetiu et al. [44] for instance estimated the cost of time spent on informal care by a replacement cost approach using the national minimum wage, the average national wage and the caregiver’s salary. Finally, all studies presented and discussed methodological limitations.

Costs estimates

Annual costs of dementia per patient according to disease severity are summarized in Table 4. Annual domestic government health expenditures in 2020 for each country are also shown in this table. Overall, in all costs categories, dementia costs tend to increase with disease severity. Prada et al. [39] calculated the ratio of direct insurer expense to mild stage at 1.62 for moderate and 4.87 for severe stages; i.e., compared to a mild patient, a moderate stage patient will use 62% more resources and a severe stage patient 387% more resources. Medical costs are generally highest in the mild stage, while social costs (including institution fees) and informal care costs become more important in the moderate and severe stages. Regardless of the disease stage, we found a wide range of estimates of annual per patient costs in the studies. Medical costs ranged from PPP$131.0 in India [40] to PPP$31,188.8 in China [45]; social costs from PPP$16.1 in Philippines [43] to PPP$10,581.7 in China [31, 32]; and informal care costs from PPP$140.0 in India [40] to PPP$25,798 in China [45]. Overall, dementia care can cost from PPP$ 479.0 in India [40] to PPP$66,143.6 in China [31, 32] per year for a single patient.

Findings

Studies included in this systematic review comprised 17 cost-of-illness studies on dementia in 11 LMICs across all GDB regions except Africa. These studies were conducted from 2003 to 2020 and focused mostly on Alzheimer’s disease. All estimates of dementia costs increase with disease severity and informal care costs tend to exceed medical costs.

Globally, very few studies have been conducted to assess dementia costs worldwide. About 10 years ago, Quentin et al. [47] reviewed cost-of-illness (COI) studies in Europe and North America, including 28 studies. A review published in 2012 by Knapp et al. [48] on the economic evaluation of dementia care identified only 29 individual studies. Similarly, the international review of COI studies on dementia, by Schaller et al. in 2014 [49] found 27 individual studies for the period 2003 to 2012. Our review focusing on LMICs included 17 studies. As reported by Hendriks et al. [50], beyond the complexity of COI studies, the overriding constraint in LMICs is data availability, including lack of accurate financial records and incomplete patient disease registers, as well as lack of expertise to conduct cost studies.

Costs of dementia are high and highly variable. They range from PPP$479.0 to PPP$66,143.6 per year for a single patient. These costs are colossal as families remain the main bearers of the burden. Indeed, according to the WHO, in 2016, half of the world’s countries spent less than US$350 per person annually on health-related expenses, including most LMICs [51]. In 2018, the mean per capita general government domestic health expenditure was US$134.7 in LMICs and US$5,562.3 in HICs [52]. In this study, the per-patient costs of dementia reached fifty times these public expenditures.

Dementia costs increase with disease severity; similar results have been reported in other studies and can be explained by the rise in care needs as dependence and impairment increase [47, 54]. Evidence from the United States suggests that informal care costs decrease with disease severity as patients need more attention and are therefore institutionalized [55]. This was not observed in this review, where informal care costs remained the highest in most studies. Actually, in LMICs, about 95% of patients with dementia live with their families, compared to 69% in HICs [5]. Thus, in these countries where universal health coverage is not achieved, families bear the burden [56].

Limitations

The majority of included studies reported convenience sampling and/or small sample size as a study limitation in their publications. Unfortunately, these two limitations restrict further use of the results as they are difficult to generalize. Half of the included studies (9/17) focused on Alzheimer’s disease only, as this is the main pathology of dementia. It should be noted that, the disease considered for cost evaluation and its definition criteria strongly influence costs estimates. In fact, Custodio et al. reported that fronto-temporal dementia costs were higher than Alzheimer’s disease or vascular dementia ones [42].

Considering there is no standard methodology for cost-of-illness studies and no reference tool for quality assessment, available checklists [7, 18] were adapted to assess the studies included in this review, which appeared to be of satisfactory quality. However, very few studies included a detailed description of service utilization and a clear definition of items considered in each cost category. The way in which services are defined and valuated is of crucial value in understanding and comparing the resulting estimates, particularly in the context of LMICs where health care and health financing systems vary. The lack of methodological standard in COI studies is a common and challenging issue to address, as described in other reviews on dementia [6, 53]. It has also been described in other diseases, including colorectal cancer [8] and diabetes [57].

Collection and valuation of time spent by caregivers on care is rather difficult. First, many activities and services (activities of daily living and instrumental activities of daily living) are done for the older people regardless of their health status and are not considered by caregivers as an additional workload, even when the person can no longer perform them themselves. This is reinforced by the fact that many people still consider dementia to be a normal part of aging [58]. Secondly, a joint effect should be considered, that of achieving many activities simultaneously, such as shopping for the PWD and caregiver themselves or supervising the PWD while cooking for the entire household. According to monetary valuation, a good number of methods exist, with the difficulties of disaggregating time spent (paid work time, domestic work, and leisure time) and defining the unit cost to be applied to each category [22, 59]. In the included studies, both opportunity cost and proxy good approaches were used, but were not sufficiently robust and/or their methods not well described. The quality of disease-specific costs estimates strongly rely on the accuracy of data collected and therefore of the collection tools. Therefore, harmonizing the costs items definition and collection is strongly recommended to ensure accuracy and comparability with other studies. Standard collection tools exist, including the Resource Utilization in Dementia (RUD) Instrument [20]. This is a dementia-specific tool that has been widely used and validated in global settings [60]. Otherwise, a minimum set of items should be collected, such as outpatient, inpatient and medication for medical costs, transport, and accommodation for social costs. Attention should be paid to informal care time too, specifying which ADL, IADL, or supervision activities were carried out, and how caregiving time was valued in money.

In addition, it was not always clearly stated in all studies which perspective was considered for the costs or whether these were dementia-specific or total patient costs. Key distinctions that may affect the interpretation and use of costs. The majority of studies presented estimates directly in US$ or local currency. It was therefore necessary to know the precise cost reference date, exchange rate, country-specific inflators and purchasing power parities to convert all estimates into international dollars. This process allowed relevant comparisons of dementia costs between the studies included in this review. However, it limits comparisons with studies from other settings or global estimates which usually present the results in US dollars. Due to the wide heterogeneity of study designs, methodological and costing-related limitations cited above, disease costs summaries provided in this review should be taken with caution.

In this review, modelling studies were excluded. Modelling is a good way to appraise the reality without investing huge resources in complex field surveys. However, the likelihood of a model relies on the proxies used, for example, utilization of health services, market price of health services or reference wage for workers. These data are often neither standardized in resource-limited settings nor available [50]. As a result, many modelling studies may not provide an accurate picture of reality. Thus, in order to identify accurate methods for calculating dementia costs in LMICs, and to summarize primary costs estimates, we considered only studies with field collection of health services use and costs. Similarly, studies with a top-down cost approach were excluded as they did not include a direct measurement of dementia costs. Although top-down studies are highly valuable, they mainly capture disease-specific medical costs from the state or third party payer perspective, with very little information on social and informal care costs. They also require a number of indicators like the number of dementia cases or heath resource utilization, which may not be available in the countries of interest, leading to the use of additional proxies and imputations.

Our search strategy included the most plausible sources of publications for the topic. The high rate of duplicates bears out the exhaustiveness of this review. In addition, no language or date restrictions were applied to the searches or results. Moreover, all studies that included cost estimation of any component of dementia care were included and all the costs estimates were converted from US dollars or local currency to purchasing power parity dollars to allow for cost comparisons between studies.

Conclusion

To the best of our knowledge, this is among the first reviews to focus on the cost of dementia in LMICs. It is valuable regarding appraisal of methodological constraints when evaluating the costs of dementia in resource-limited settings. Findings highlighted once more the scarcity of valuable cost-of-illness studies in these countries, particularly in Africa where original research for primary data collection is essential. The costs of dementia were found to be high, placing an enormous burden on families. These results should serve to raise the alarm on the urgency of dementia and to advocate for dementia-oriented programs and policies.