Sample selection

A sample of 156 patients attending outpatient health care at the Neurology and Cognitive Behavior Unit of Hospital das Clinicas (Sao Paulo, Brazil) was selected after individual assessment by a neurologist. Comprehensive interviews were performed between November 2011 and May 2015.

The inclusion criteria for patient participation in the study were having a diagnosis of dementia at mild, moderate or severe stages with formal or informal caregivers who cared for the patient. The exclusion criteria were patients diagnosed with Mild Cognitive Impairment (MCI).

The statistical precision for estimation of direct and indirect costs of dementia was based on hypothesis test to calculate sample size and infer expected values of the target variable: considering one single sample, the idea was to estimate the distance between true value (population) and the sample value by controlling the statistical significance and power. According to Chow et al. [19] and Chow and Liu [20], the sample size may be calculated from the following hypotheses:

Where is the meaningful difference between the population mean μ_o and the observed value (null hypothesis). The constant δ is the statistical precision measuring the distance between the real value and the value obtained from sample of size n_s.

The study was based on calculation of the sample size taking into account both the Type I error (α) and the Type II error (β). In this case, it is important to specify the statistical power (1 - β) 100% and the confidence level (1- α) 100%. [19] derived the following expression to the sample size n_s: [Eq 1]

Where . We denote by S the sample standard deviation and is the corrected variance. The correction factor is applied for finite population of size N, which is the present case. We assumed that the variance is unknown. The scores and are calculated from a t-distribution with (n_s-1) degrees of freedom.

The average value for treatment cost found from our sample was = $1,405.72; the standard deviation S = $1,376.28; corrected standard deviation S_N = $1,300.04; and the population N = 1,440.

Using Eq 1, we estimated n_s by controlling θ, the significance level α and statistical power 1-β simultaneously. Setting α = 5% and 1-β = 80% and θ = 0,2; the sample size n_s = 156. That means the precision of estimates is δ~$260.00; which leads to the conclusion that the population costs should not exceed by the amount of δ~$260.00; i.e., <260. Of course, if there is need of higher precision, the sample size must be increased according to Eq 1.

The population size was calculated using information from the Hospital das Clinicas, the largest hospital in Brazil, based on the assessment of 15 elderly patients per week during the period of two years (24 months).

The protocol of the study established that the nurse responsible for diagnosis of dementia among elderly patients in the hospital should invite all caregivers to participate in the study during the period of research. Caregivers who accepted participation in the study were included in the sample, except if patients were diagnosed with Mild Cognitive Impairment (MCI). Thus, every patient diagnosed with mild, moderate or severe dementia and his/hers relatives or caregivers were included in the study.

Patients and caregivers were verbally informed on the details of the project and interviewed only after signing a consent form agreeing to participate in the study. Caregivers responsible for patients’ care signed the consent form, in case of impaired capacity of the patient, being certified that caregivers understood the terms of the research and the rights of participants.

The study was submitted to the Ethics Committee of Hospital das Clinicas of the University of Sao Paulo (HC-FMUSP) and approved under process number 368.096/2013.

Data collection and processing

Data collection was performed using a research protocol based on semi-structured questionnaire designed to gather information on sociodemographic, economic and health characteristics of patients and caregivers, information on health resources utilization and results of clinical assessment of patients and their respective caregivers, applied by a nurse specialized in dementia.

Clinical data included variables related to disease progression and comorbidities of the patients and their respective caregivers. Detailed information on the research protocol is available within a study published with preliminary results of the research, referring to indirect costs of dementia in Brazil, including a fraction of the planned sample regarding data from caregivers [14].

Disease severity was categorized using the Functional Assessment Staging—FAST scale [21], which allows observation of cognitive impairment due to dementia and dependency stage of the patient, categorizing into seven stages (1 to 7f). In the study, the seven stages were aggregated into three levels (mild, moderate and severe), considering the following scores: “1–4” = mild; “5a-5e” = moderate; and “7a-7f” = severe.

Caregiver burden due to the patient condition was assessed by the ZARIT Burden Interview Scale and categorized into four levels: light to moderate burden (score ranging from 0 to 40), moderate to severe burden (score 41–60) and severe burden (score 61–88) [22].

Information on health resource utilization was based on the Resource Utilization in Dementia (RUD) scale, including variables related to direct costs associated with patient disease and indirect costs related to the relatives’ and caregivers’ job abandonment, early retirement and productivity loss due to burden associated with patient care [23].

Costs estimates were calculated using the information on health resource utilization for direct and indirect costs due to patient condition from a societal perspective. Values were expressed in Brazilian currency (Reais), corrected for inflation to February 2016, and converted into U.S. dollars using the official exchange rate released by the Brazilian Central Bank.

Direct costs included costs of patients’ medication, health care, and other resources (excluding medication and health care, e.g., diapers, transportation, and other). Costs on medication prescribed for patients’ treatment were calculated for daily recommended dosage based on patient medical records and caregiver information, using market prices. Costs of health service utilization by patients were estimated using information on health care provided by the Brazilian government within the national health system encompassing specialized inpatient and outpatient health care. In summary, the direct costs included direct private costs (covered by patients’ families) or direct public costs (covered by government funding of national health system).

Indirect costs included costs of caregivers’ health care due to health problems related to burden of patients’ care, costs due to job abandonment or early retirement related to patients’ care and productivity losses of relatives and/or caregivers due to the burden of patients’ condition.

Estimates of economic losses due to job abandonment and early retirement were based on hourly wages according to respective caregiver’s salary, and economic losses due to productivity losses of relatives and/or caregivers time for patients’ care were based on hourly wages per capita, according to reported household income and number of salaried workers reported in the family unit, in order to account for rotation of the task among diverse individuals in the household and to consider differences in wages of household members among formal and informal labor activities.

The calculation of monthly hours related to activities of daily living (ADL), instrumental activities of daily living (IADL) and supervision of the patient was made using previously tested methodology [14].

Statistical procedures

Data on patients’ and caregivers’ sociodemographic, economic and health characteristics, clinical assessments of patients, and direct and indirect costs due to dementia, according to FAST category, were analyzed by descriptive statistics and partial correlations in order to allow analysis of differences between groups of patients and caregivers, adopting a significance level of 0.05.

Additionally, robust multiple regression models were estimated using the Stata software version 11.2, in order to identify main determinants of dementia costs in the sample of patients in Brazil, adopting 0.05 significance level for inclusion of variables in the models.

Characteristics of patients and caregivers

Demographic data on patients and caregivers regarding age and gender proved similar to the findings of other studies conducted in LA [1,12,16] and in high-income countries (HICs) [1,24]. Patients and caregivers interviewed in the present study were younger in comparison to other studies conducted in HICs (72.9 years versus ≥75 years, and 54 years versus 65 years, respectively), and in Brazil (42.1 years) [3,25–28]; however, it was consistent with the previous study performed using subsample of the research [14]. Most patients and caregivers were female, and a higher percentage of retirees was observed among the patients in comparison to other studies [3,14].

The educational level of caregivers in the study was lower in comparison to previous studies [26–29]. Some sociodemographic characteristics were consistent with findings from other studies: married caregivers (66.7%), living with the patient (73.1%), spouses (35.9%) and sons/daughters (51.9%), reflecting cultural aspects of the duty to provide elderly with care seen in LA [14], as well as Asian and Eastern countries [27].

A high proportion of caregivers reported leaving formal jobs to take care of patients, and changes in type of formal/informal work were also observed among caregivers with patients at the mild stage (41% versus 11% in other studies [3], including 36% in the preliminary results of the study [14]). Caregivers supporting patients at the severe dementia stage in the sample worked 4.95 fewer hours per month (formal or informal), providing more than one hour of care above the average for the sample.

Comorbidities

A meta-analysis involving 13,978 elderly individuals in the Brazilian population showed a 68% prevalence of high blood pressure (HBP), a similar rate to the present study in which HBP was the most prevalent comorbidity [29]. Diabetes mellitus (DM) has 6.2% prevalence in Brazil’s adult population and ranges from 14.5% to 19.6%, depending on age group [30]; whilst in the sample analyzed the proportion of patients with DM was significantly higher (27.6%).

Considering evidence on average survival rates after diagnosis (48–96 months) [31,32], the evolution of dementia reported among patients in the sample (60.1 months) is consistent with other studies, including the preliminary results reported in previous study [14].

The long-term evolution of AD may lead to the occurrence of acute conditions coexisting with chronic diseases or, rarely, with dementia only, exposing patients to situations of advanced frailty [33,34]; resulting in additional hidden costs which cause collateral burden to society [35,36].

A recent study involving a sample of 68,844 elderly individuals indicated that patients with dementia had, on average, 3.69 comorbidities versus 2.44 comorbidities among individuals without dementia [25].

The present study found an average of 2.03 (±1.38) comorbidities and a lower prevalence of DM (27.6% vs. 36.1%) and HBP (60.3% vs. 83.5%) compared to the cited study. By contrast, there was a higher prevalence of cardiovascular disease (46.2% vs. 41.2%), and cerebrovascular disease (20.5% vs. 11.3%) [25].

The lower rates of HBP and DM may be due to actions in primary care [25,37]; although the significant prevalence of DM indicates the need for public policies to reduce and/or control complications (including cardiovascular and cerebrovascular diseases), thus avoiding overload of the health system at secondary and tertiary levels [25,27,29,31].

Caregiver overload

Psychological overload of caregivers in attending dementia patients, economic impacts of the disease, and insidious evolution of AD may be factors contributing to excessive burden among caregivers [14,38]. The time lag between the initiation of patient care and the onset of caregiver illness was 11.56 (±20.90) months, showing that one year of care may be sufficient to cause health problems in caregivers of dementia patients.

Caregivers reported mostly emotional overload and depression; conditions associated with low individual resilience capacity, related to the need to change work hours or to abandonment of profession, leisure activities and ongoing life projects [39–41]. Usual work capacity was preserved for most caregivers of productive age; however, a significant proportion of caregivers reported the need to abandon formal work due to the sick relative.

A pioneering study evaluating the presence of comorbidities among caregivers of dementia patients in Brazil compared to individuals from the general population, indicated higher prevalence of depression (23% versus 11%), major depression (29% versus 20%), decline in physical health (7.3% versus 5.5%), obesity (23% versus 18%), and HBP (23% versus 15%) in caregivers than in non-caregivers, respectively [28]. In the present study, there was higher prevalence of depression (66.7%) than that found by Laks et al. [28]; however, the results were based on self-reported information from the caregiver about diseases diagnosed and on assessment of caregiver burden using a psychometric instrument.

Informal caregivers were more likely to report health problems; information that should be analyzed prospectively by general practitioners, especially considering the need for disease prevention. The risk of onset of physical or mental illness among caregivers may be elevated for numerous reasons ranging from a lack of resources for appropriate caregiving to individual predisposition associated with previous history of life and feeling of loneliness due to the responsibility of caregiving [32,36,40].

Informal care

Estimation of time dedicated to care of dementia patients and its economic evaluation may be subject to controversies over the methods employed for assessing hours (including ADL, IADL, supervision, or all) and appraisal of hourly costs, including issues related to country economic classification [8,9,14].

This study was based on measurement of global number of hours per month dedicated to ADLs, IADLs and supervision of dementia patients, which totaled 373.04 hours per month as reported by caregivers, 23.5% higher than the average in North America (87.60 hours per month) [3]; however, lower in comparison to the preliminary results (438.37 hours per month) of the research previously reported [14].

The annual time spent by Brazilian caregivers in the sample (4,476.5 hours per year) represents 393% of the average time reported in studies performed in USA (1,139.0 hours per year) [1,3], and approximately double the daily hours (12.4 hours per day), which may explain the overload perceived by caregivers.

Direct and indirect costs of dementia

A recent 2015 study based on a top-down approach to estimate costs of dementia reported that a significant percentage (87.4%) of costs worldwide were attributable to high-income countries (HICs), comprising predominantly social direct costs (US$308.1 billion, 43.1%) and informal care costs (US$271.1billion, 37.9%), whilst direct medical costs (US$136.0 billion, 19.0%) accounted for a lower proportion [8]. Informal care of dementia patients is usual in LA, due to low family income and high costs of hiring formal caregivers [14,41,42].

In Brazil, the Ministry of Health provides without charge medication for treatment of dementia, hospitalization, specialist consultations and outpatient follow-up for patients and caregivers [43–45]; the present study is pioneer in estimating direct social costs of health care supported by government funding within the Brazilian health system.

Although estimation of costs using the top-down approach (macro-costing technique) may not be as good as estimates based on the bottom-up approach (micro-costing technique) used in the study, the results showed that the economic impact of health care provided by the Brazilian government (direct medical costs) is equivalent to approximately 24% of household income per capita at the mild stage, 21% at the moderate stage, and 43% at the severe stage; representing a significant reduction in health expenditure for dementia patients and their respective families, in contrast to conditions prevailing in other upper-middle income countries (UMICs) [46–48].

Direct costs of dementia reported in the study accounted for US$562.09 (40.0% of global dementia costs). A substantial proportion of dementia costs estimated in the context of the study were attributable to indirect costs; i.e., social costs associated with caregiver productivity loss and time spent by caregiver on patient support. It is important to highlight that total indirect costs per month reported in the study (US$843.63) were higher than family income per capita reported (US$286.41); representing losses of income of US$577.87 for patients at mild stage, US$1,123.68 at moderate stage, and US$628.75 for patients at severe stage.

In comparison to previous study based on preliminary data referring to indirect costs of caregivers [14], the present study showed higher costs due to time spent in patients’ support and productivity losses due to job abandonment (considering solely cases attributable to patients’ care); that is, the partial sample initially interviewed in the research probably present underestimation of indirect costs of dementia in the Brazilian health system.

Based on estimates provided for UMICs [8], global costs of dementia represent approximately US$5,284 per person, of which 57.1% represents informal care costs; 22.4% direct medical costs; and 20.5% social costs. The study indicated that, in Brazil, global costs of dementia are significantly lower (US$1,405.72 per person), with 56.6% attributable to informal care costs; 15.1% direct medical costs; and remaining 37.3% social costs.

Additionally, according to the regression model estimated, the progression of dementia (FAST score) had significant influence on economic impact of the disease, followed by use of medication and educational level of the caregiver. The last two variables represent increased costs associated with direct utilization of resources of the health system and the patient’s family, respectively; whilst the first variable is probably associated with the social burden of dementia, since FAST score showed a significant positive association with ZARIT score on the analysis performed.

The results obtained also highlight the need to improve the quality of basic health care, especially regarding dissemination of information to the population on protective factors to delay the onset of dementia [48,49]. Simple, low-cost interventions may be effective strategies for UMICs, resulting in policies with a positive impact on aging which is inversely associated with exposure to risk of dementia [47].

Global costs of dementia represent a substantial economic and social burden, and comparisons with the impact caused by other diseases may encourage discussion on societal onus, toward prioritizing long-term health strategies to reduce health care costs through disease prevention and health promotion, thereby minimizing the risks of deterioration in the quality of health care [10,33].

The study showed that the global costs of dementia in Brazil already outstrip available resources within Brazilian society, calling for the development of health care strategies based on education programs on dementia (especially in primary care) and training of health professionals to provide caregivers with information, constituting useful approaches to attenuate the current and future impact of dementia in UMICs over the coming decades [14,35,42,47,50].

Limitations

The main limitation of the study was the utilization of a cross-sectional study design with convenience sampling; given that longitudinal analysis would provide additional evidence on determinants of health care costs and progression of the disease, while random sampling could provide information to perform further extrapolation in the statistical analysis from a societal perspective.

However, considering that the data presented relates to the first stage of the CAAD project, the research group is expected to be able to carry out further economic analysis based on longitudinal data.

Thus, in view of the dearth of evidence on dementia costs in Brazil, it was considered important to disseminate these initial findings on the economic impact of dementia, including direct and indirect costs, in the context of a publicly financed health system with universal coverage.