Direct costs

Direct costs are incurred when providers diagnose and treat illnesses. These include the costs of hospitalization, outpatient visits, diagnostic tests, pharmaceuticals, psychosocial treatments, preventive care, ancillary services, and other healthcare interventions or services ( ). Direct costs also include those associated with commonly-used nonhealthcare resources such as transportation, informal cares, cost of relocation due to illness, childcare expenses, and related expenditures ( ; ). Direct costs are usually the easiest of the three types of costs to estimate, given the availability of structured billing information or standardized cost data for most healthcare services, tests, and treatments. Therefore, direct costs are the most common types of costs ascertained in COI studies.

Indirect costs

Indirect costs are those related primarily to work productivity losses due to absenteeism (missed days of work due to illness or injury, disability, or discretionary time), presenteeism (reduced productivity when working), and early mortality. There is far less structured or standardized information available on indirect costs for most illnesses. Therefore, researchers infer or impute indirect costs. For example, counted time off due to disability as a missed full day of work in their depression COI study. An outpatient office visit on a working day was assigned the equivalent of a half-day missed (even if the cohort member’s office visit occurred outside of working hours or on breaks), and hospital stays and emergency room visits were counted as a full day of missed work. They multiplied each whole- or half-day missed by the individual’s daily wage to estimate indirect costs due to absenteeism. In that same study ( ), researchers assigned a value of 6.1 times the cost of injury/illness-related absenteeism to calculate the presenteeism costs ( ). As highlighted by this example, indirect cost estimation is often more of an approximation of the likely “true” value than is the estimation of direct costs.

Intangible costs

Intangible costs involve disease-related pain and suffering or loss of functioning, typically assessed by measuring their adverse impacts on quality of life or related metric. Researchers and other stakeholders typically measure the adverse impacts of illnesses on the quality of life or related metrics to calculate the intangible dimensions illness. Estimating the monetary value of intangible costs constitutes a considerable methodological challenge, given the absence of a well-defined market. Measuring effects on quality of life or conducting extensive surveys that assess respondents’ willingness or desire to avoid certain morbidities and what they are willing to pay to do so are often required to derive intangible cost estimates. As such, methods for estimating intangible costs in monetary terms are often more difficult to employ in COI studies than those used for determining direct and indirect costs.

Data sources

Numerous types of data exist for conducting COI studies. These include patient or caregiver surveys, prospective studies (e.g., prospective cohort studies, clinical trials, etc.), chart review studies, and retrospective cohort studies that utilize large administrative claims databases. Each type of data source has certain advantages and disadvantages. For instance, surveys and prospective studies have the advantage of yielding data that were collected specifically for research purposes, thus providing some assurance of high data quality and completeness. However, statistical power may be limited for some questions due to insufficient enrollment, poor participation, or high attrition; and survey respondents or prospective study participants may not be sufficiently representative of the underlying population of interest.

Retrospective chart review and administrative database studies have the advantages of convenience (due to the lack of need to recruit study cohorts and follow them over time when longitudinal data is required) and high statistical power due to the large sizes of study cohorts that can be created using data that are often already available in a structured form. However, the information in health records or administrative databases was collected for purposes other than for conducting research and are thus subject to misclassification and to limitations related to incomplete or missing information on key study variables ( ).

Exposure groups: People with the disease of interest and those without

To contextualize COI data and increase its interpretability, disease-associated costs are often compared between at least two groups of individuals—those with the disease of interest, and those without the disease of interest—over a defined time period. In the case of treatment-resistant depression, COI comparisons are often made between people with depression that meet a prespecified definition of treatment resistance and those with depression who do not meet the prespecified definition of treatment resistance. Some studies include a second referent group composed of persons with no evidence of depression during the observation period. Researchers can use such an approach to compare raw (unadjusted) costs. However, persons with treatment-resistant depression are much different than persons with treatment-responsive depression, analogous to how people with a disease of interest and those without the disease may differ on a number of baseline characteristics. This difference in baseline characteristics may systematically influence cost differences between groups. As such, investigators will often attempt to balance exposure groups (or adjust cost estimates) on these other factors using matched designs, multivariable modeling approaches, other types of modeling approaches, or a combination of these.

Endpoints

As mentioned earlier, the “inputs” for COI studies are the types of costs associated with specific exposure groups (e.g., persons with treatment-resistant depression vs persons with treatment-responsive depression, etc.) and the “outputs” (endpoints) for COI studies are the comparative costs associated with each, in monetary terms. The comparative costs of illness are calculated for a defined time period, often expressed as a monetary value per patient per year (PPPY) or per patient per month (PPPM). For each exposure group, costs are typically presented in aggregate (e.g., total direct healthcare costs for patients with treatment-resistant depression vs those same costs for people with treatment-responsive depression). However, cost outputs may be stratified according to the type of cost (direct costs, indirect costs, etc.) or costs in specific subgroups (e.g., behavioral health costs, depression-related costs, or intervention-specific costs such as those related to hospitalization, etc.).

Heterogeneity across cost-of-illness studies

As briefly reviewed above and elsewhere, there are several approaches to estimating the economic burden from MDD and other depressive syndromes ( ). COI studies are often descriptive in nature, unlike other types of studies that estimate illness burden using more standardized measures such as years of life lost, years lived with disability, disability-adjusted life-years, and health-related quality of life ( ; ). As of this writing, there is no current consensus on which specific elements or measures should be used to derive COI estimates for depression. As a result, individual COI studies are subject to wide variation (heterogeneity) with regard to the types of data used, the perspective of the study (e.g., patients, families/caregivers, employers, insurance providers, etc.), the types of costs that are calculated (e.g., direct costs, indirect costs, etc.), how calculated costs are stratified (e.g., behavioral health-specific costs, depression-specific costs, costs related to suicide, etc.), the duration of follow-up, the types of depression included (e.g., major depression only, multiple unipolar depressive syndromes, unipolar and bipolar depressive syndromes, etc.), the type(s) of payor(s) from which data on costs of treatments and cares are derived (commercial insurance databases, public insurance databases, prospective datasets, etc.), and other factors ( ).

Not surprisingly, cost estimates can vary widely for a given condition, given the methodological heterogeneity in COI studies. The heterogeneity in cost estimates for even a single condition raises questions about the reliability and utility of COI studies for guiding practice and policy decisions ( ; ). This may be particularly true of COI studies for depression—a condition that in-and-of-itself is subject to high levels of interindividual variation in clinical presentation and treatment response ( ; ). In our opinion, this is perhaps the biggest limitation of COI studies of depression. Despite this limitation, it remains the case that COI studies provide important estimates of the burden of depression and treatment-resistant depression from a societal perspective—both in terms of the magnitude of costs in economic terms and how those costs are distributed.