Methodological considerations in cost of illness studies on Alzheimer disease

To conduct a COI study, it is necessary to define disease, epidemiological approach, type of costs, and study perspective. Subsequently, resource consumption data and unit costs can be gathered, and the results presented and methodically discussed, in conjunction with sensitivity analysis to test their robustness. A checklist (Table  1), containing its items, was developped on the model described by Drummond et al. [11] and adapted to COI by Molinier et al. [12]. An equal weight was given to each item. The final score was the sum of the eleven individual items.

Illness costs widely depend on how the disease is defined. AD diagnosis is based on Alzheimer’s Association criteria (NINCDS-ADRDA) [30] and Diagnostic and Statistical Manual of Mental Disorders criteria (DSM IV) [31]. As costs increase with disease severity [32‐35], disease stage must be specified and measured using validated tools, as Clinical Dementia Rate [36] (CDR) or Mini Mental State Examination (MMSE) [37]. Cost components may vary depending on the living conditions (e.g. home, institution), and therefore must be specified [33, 38].

Prevalence-based COI studies estimate the economic burden to society during a period of time as a result of the prevalence of the disease (e.g. in a given year). Incidence-based studies estimate lifetime costs, and measure the costs of an illness from diagnosis until recovery or another endpoint (e.g. death).

A COI study may be conducted from several perspectives that must be specified to check that relevant costs are included. From the healthcare payer perspective, only direct costs incurred by a payer (e.g. national health insurance) are considered. Indirect costs and the patient “out-of-pocket” must also be included in a study which uses societal perspective.

Resource consumption estimates vary depending on the data available, but validated tools exist to collect them [39]. In prospective COI studies, events have not occurred yet, so data collection is done by the patients’ follow-up, medical records, data from clinical trials and patients or caregiver questionnaires. Conversely, in retrospective COI study, events have already occurred, so data collection must refer to recorded data, either using “Top-down” method (aggregate figures from hospital admissions, mortality, etc.) or “bottom-up” method (by referring to the patients sample record).

The COI is estimated by identifying the cost-generating components and by attributing them a monetary value. This is the opportunity cost, the value of the forgone opportunity to use in a different way those resources that are lost due to illness [11]. Direct and indirect costs should be valued to assess the total economic COI. Direct costs measure the resources used to treat an illness and can be estimated by per capita expenditures, national tariffs, market prices, and published studies. Patient charges and tariffs do not give an accurate estimate of the underlying costs. Market prices can be used to value some cost categories like drugs, rehabilitation items (e.g. eyeglasses, etc.). Direct costs can also be valued with care facilities estimates, through the analytical account system which specified distribution properties. Indirect costs measure the loss of productivity, the effect of the illness on the patient or caregiver abilities to work. Three methods are used to value indirect costs: the Human Capital Approach (HCA) [40], the Friction Cost Method [41] and the Willingness to pay approach [42]. Informal care is an unpaid care often provided by relatives and plays a substantial role in the AD patient’s total care. Two methods are mainly used to value the shadow price of informal care time. The opportunity cost approach values the opportunity forgone as a result of caregiving and the replacement cost approach values the caregiving time spent at a price of a close substitute [9, 43, 44].

Discounting captures individual preferences for income today rather than income in the future and is frequently applied when COI studies are considered over several years. In the USA, they estimate the discount rate at 3% [45]. The following equation is applied to estimate costs:

Where: C _a  = present value of cost strategy, C _t  = value of cost strategy in year t, r = discount rate, t = time period.

Sensitivity analysis is recommended because COI studies contain uncertainties. It allows testing the robustness of the results by varying in range key variables (e.g., prevalence, unit costs, etc.) [46]. It seems more credible to present a range of possible illness costs.

The presentation of COI results should be consistent with collected data and should disaggregate results into as many components as possible with full explanations given for clarity (Table  1).

The adopted perspective was the society in 13 studies [13‐16, 18, 20‐24, 27‐29] and respectively the family and the healthcare payer in two studies [17, 19]. Fourteen studies quantified direct medical and non-medical costs and informal costs [14, 15, 17, 18, 20‐29]. Two studies did not include inpatient costs [17, 25]. Informal costs were defined as direct non-medical costs in seven studies [14, 15, 17, 19, 22, 23, 25] and as indirect costs in seven studies [18, 21, 24, 26‐29]. One study quantified all the costs (i.e. direct, informal and indirect) [13], indirect costs were defined as patient early retirement and informal costs as direct non-medical costs. Danish study quantified only direct medical and non-medical costs [16]. German study quantified only direct medical costs [19].

Three studies estimated resource consumption retrospectively [15, 21, 29]. Two of them used bottom-up approach to gather activity data through questionnaires [15, 29]. The other used a top-down approach, using published national indicators, national surveys and published studies [21]. Eleven studies estimated resource consumption prospectively [13, 14, 16‐18, 20, 22‐24, 26, 28]. Two studies used mainly the Resource Utilisation in Dementia (RUD) [48] to gather activity data [14, 18], completed with report forms, medical records and questionnaires on 560 patients in the Spanish study [14]. Nine studies gathered activity data mainly with questionnaires [13, 16, 17, 20, 22, 23, 25, 26, 28]. In the Danish study, 164 AD couple (i.e. patients/caregivers) was interviewed at home about Activities of Daily Living (ADL), use of health care and community services [16]. Two US studies measured direct costs using caregiver’s interviews and Medicare Current Beneficiary Survey (MCBS) [13, 23]. Rice et al. gathered activity data through monthly caregiver’s telephone interviews, billing records and with calendar given to caregiver at the baseline visit [23]. Israeli study used a baseline questionnaire to record time spent on caring and use of goods and services, and secondly recorded the same items with 5 monthly telephone interviews [28]. Two studies recorded data via mailed questionnaires filled out by caregivers [13, 17]. Turkish study used a questionnaire and daily time sheets for caregiving time [25]. Wang et al. interviewed 66 AD couples for filling out the resource use’s questionnaire [26]. Two studies used decision modelling [19, 27] and estimated resources with published sources and national surveys [27] or with data extractions of a German retrospective analysis [19]. One study did not precise the approach used to gather activity data [24]. Eight studies specified the number of caregiver included, several for 4 studies [13, 15, 17, 22] and only one for others [18, 23, 25, 28]. Eight studies recorded AD net costs [15, 16, 21, 22, 24, 26, 28, 29] either by subtracting healthy patient costs or by asking AD couple about resource’s used exclusively for AD.

The follow-up period was the lifetime in the study which adopted the incidence-based approach [21] and was one year in six studies [13, 14, 16, 22, 26, 27], but was frequently reduced to six, three or one month [14, 18, 20, 25, 28, 29]. One study used a sixty months follow-up period [19].

Direct costs were estimated from published data, national estimates and Medicare-over charges in one American study [21], and from unit costs using the MCBS and national estimates for adult day care in another one [20]. Rice et al. used charges and bills provided by the caregivers [22]. Two Spanish studies used the Spanish database on medical costs (SOIKOS), patient’s reports and the Spanish Vademecum to estimate unit costs [13, 14]. Direct unit costs were based on reimbursement tariffs used by the French social health insurance and on the French Disease related group (DRG) for inpatient care [15]. The Danish study used the reimbursement tariffs of social insurance, hospital accounts, gross wage rates of professionals and amortization procedure to value direct costs [16]. Institutional costs were valued with the average cost per day without food and beverage. Zencir et al. valued unit costs with the average price of public and private physician visits and with the average price per day for medication [25]. In the Chinese study, unit costs were valued with drugs prices, transportation reimbursements and local fees for home help [26]. Tariffs of social health insurance and full drugstore prices were used to unit costs valuation in the Argentinian study [28]. Meserton et al. valued unit costs with ward-specific per diem costs from regional price, the lowest available price for medication and with per visit costs [21]. For residential care, the number of days in institution was multiplied by the corresponding unit costs. Two studies used national estimates to value unit costs [23, 27]. No information relating to direct costs valuation was reported in four studies [17‐19, 24]. Twelve studies used the replacement cost approach to value informal care [13, 15, 17, 20‐26, 28, 29]. Seven studies used national estimates of a close substitute [17, 20‐23, 25, 29]. Among these, 5 studies used different national wages to value different caregiver’s activities [17, 20, 22, 23, 28]. Israeli study obtained different hourly wage rates for each activity from the Central Bureau of Statistics [28] and Leon et al. used hourly wage of home health aids to value ADL time and homemaker’s hourly wage to value IADL time [23]. Two studies used hospital nurse hourly wage [21, 25], another one, local wage rate without specifying the type of professional caregiver [26] and the Canadian study used published data [24]. Three studies did not specify the sources used to value informal care time [13, 15, 29]. Nevertheless, one study used professional caregiver monthly salary [29], another the cost per hour of a domestic cleaner gross wage [13] and the last one the average between housekeeping and paid assistance wage [15]. One study chose opportunity cost approach to value informal care, using the mean informal caregiver own salary [14]. Both approaches were used in one study [27]. The time spent by working caregivers was valued with national estimates and the time spent by working caregivers was valued with the own caregivers salaries. Opportunity cost and revealed preference approaches were used to value informal costs in one study [18]. Caregiving time was valued either with the HCA (working caregivers) or with a monetary estimate value of one hour of leisure time (not working caregivers). One study valued indirect costs with national estimates on employment and wages [13].

One study has discounted costs without specifying the discounting rate [21]. German study performed costs discounting in the sensitivity analysis [19]. All the other studies used a short follow-up period and had no need to discount costs.

Only three studies performed a sensitivity analysis [14, 27, 29]. One study analyzed the impact of informal costs variation [14], another one the variation of AD sufferers’ proportion according to the need of full time care level [27], and the last one the discounting of the incurred costs [19].

Most studies presented their results clearly. They were mainly well explained and consistently set out in relation to the methods adopted. Five studies did not sufficiently disaggregate costs, so the information strength was reduced (Table  3) [15, 20, 23, 24, 27]. All studies presented results in terms of cost per patient. Four studies proposed also AD total costs for the country [13, 23, 24, 27]. According to the key methodological points, we have drafted a checklist of questions related to the eight items analyzed (Table  1). For 9 studies, the answers of at least seven to eleven questions were “yes” [13, 14, 16, 18, 20, 21, 26, 27, 29].