Cost-of-Illness Study

This page explains what a cost-of-illness study measures, how its boundaries and methods affect the result, and how the estimate should be interpreted. It also follows the analysis through resource measurement, valuation, attribution, calculation, uncertainty and transparent reporting.

Cost-of-illness study is commonly shortened to COI study. It describes economic burden associated with a disease, injury or risk factor, but it does not compare the costs and outcomes of competing interventions. A large burden estimate therefore does not establish that a particular intervention is cost effective or should receive funding.

What a cost-of-illness study tells us

A cost-of-illness study estimates the economic burden attributable to a disease, injury or risk factor within a defined population, perspective and period.

The study may help planners:

• Describe the scale of economic burden.
• Show how burden is distributed across cost categories.
• Identify populations or services experiencing substantial burden.
• Identify major resource-use or productivity drivers.
• Reveal gaps in available data.
• Provide inputs for later economic models.
• Support service planning and research prioritisation.

A COI study does not compare the incremental costs and outcomes of competing interventions. It cannot determine whether a treatment is cost effective, affordable or the best use of limited resources.

Which boundaries define the result

A cost-of-illness estimate does not represent one fixed property of a disease. Its meaning depends on the study boundaries.

The study should define:

• The condition, injury or risk factor.
• The case definition.
• The population.
• The jurisdiction and setting.
• The analytical perspective.
• The costing period.
• The prevalence- or incidence-based approach.
• The included resource and cost categories.
• The attribution method.
• The valuation methods.
• The currency and price year.

A broader boundary may produce a larger estimate because it includes consequences falling outside the healthcare system. Including more categories does not automatically make a study more accurate or useful if those categories are poorly measured, overlap or do not match the study’s purpose.

Whose costs are counted

The analytical perspective identifies whose costs and economic consequences enter the study. Two estimates for the same condition can differ substantially because they use different perspectives without either calculation necessarily being incorrect.

Healthcare-system perspective

A healthcare-system perspective may include:

• Hospital care.
• Outpatient care.
• Primary and community care.
• Medicines.
• Diagnostic testing.
• Rehabilitation.
• Medical devices.
• Healthcare staff time.
• Condition-related healthcare services.

Payer perspective

A payer perspective may include:

• Reimbursed healthcare services.
• Covered medicines.
• Programme expenditure.
• Contracted services.
• Other payments borne by the specified payer.

Patient or household perspective

A patient or household perspective may include:

• Out-of-pocket healthcare payments.
• Travel.
• Accommodation.
• Unpaid patient time.
• Informal caregiving.
• Household adaptations.
• Other condition-related household expenses.

Employer perspective

An employer perspective may include:

• Work absence.
• Reduced productivity while working.
• Employee replacement.
• Workplace accommodations.
• Other workplace disruption.

Societal perspective

A societal perspective may include:

• Healthcare resource use.
• Patient and household costs.
• Informal caregiving.
• Productivity consequences.
• Other relevant effects falling across sectors.

A study claiming a societal perspective should explain which societal consequences were included and which were excluded. The label alone does not demonstrate that every relevant consequence was measured.

Prevalence-based and incidence-based studies

The epidemiological approach determines which people and costs are followed. Prevalence- and incidence-based studies answer different questions and should not be treated as interchangeable.

Prevalence-based approach

A prevalence-based study measures costs occurring during a specified period among people living with the condition, regardless of when the illness began.

This design is useful for describing:

• Current annual burden.
• Expenditure faced by a service or payer during a budget period.
• The distribution of present resource use.
• Short-term planning needs.

Incidence-based approach

An incidence-based study begins with new cases arising during a specified period and may follow their costs over the future course of illness.

This design is useful for describing:

• The lifetime burden associated with new cases.
• The potential consequences of preventing a new case.
• The timing of costs over a disease pathway.
• Long-term consequences of current incidence.

An incidence-based study may require extrapolation, survival assumptions and discounting. A prevalence-based annual estimate should not be added to an incidence-based lifetime estimate.

Bottom-up and top-down estimation

Cost-of-illness studies can construct burden from individual resource use or allocate burden from aggregate totals. A study may combine the approaches, but their boundaries must remain clear.

Bottom-up estimation

A bottom-up study constructs the estimate from person-level resource use, unit costs and other individual losses.

A simplified calculation is:

Population Burden = Mean Attributable Cost per Affected Person × Number of Affected People

Bottom-up estimation can show variation across people or subgroups, but it requires detailed and representative data.

Top-down estimation

A top-down study begins with aggregate expenditure or another population total and attributes a share to the condition.

A simplified calculation is:

Attributable Burden = Aggregate Expenditure × Disease-Attributable Fraction

Top-down estimation may be practical when individual records are unavailable, but the result depends heavily on the attribution assumptions and completeness of the aggregate data.

Combined approaches

A study may use bottom-up estimation for some categories and top-down estimation for others. When methods are combined, the study should demonstrate that the categories do not overlap and that the population, period and valuation basis are compatible.

Separating attributable costs from all observed costs

People with a condition may use healthcare or experience productivity losses for unrelated reasons. A COI study must explain how it separates condition-attributable burden from all observed resource use and loss among affected people.

Possible approaches include:

• Diagnosis-related attribution: Identify services or costs explicitly recorded as related to the condition.
• Excess-cost analysis: Compare affected people with an appropriate population without the condition.
• Matched analysis: Match affected and unaffected individuals on relevant characteristics.
• Adjusted analysis: Control statistically for observable differences between groups.
• Attributable fractions: Apply epidemiological estimates of the proportion associated with the condition.
• Expert attribution: Use transparent expert judgement when stronger evidence is unavailable.

The appropriate method depends on the available evidence and causal question. Even careful attribution may remain uncertain when comorbidity, shared risk factors, incomplete coding or unobserved differences affect the comparison.

Measuring resource use

Resource use and unit costs should remain separate wherever possible. This allows readers to understand whether a large estimate arises from frequent resource use, high unit costs or both.

For each resource:

Resource Cost = Resource Quantity × Unit Cost

Resource-use data may come from:

• Administrative claims.
• Electronic health records.
• Registries.
• Clinical studies.
• Patient or caregiver surveys.
• Time-and-motion studies.
• National statistics.
• Published literature.
• Expert elicitation.

For each important resource, report its quantity, unit, population, setting, period, source, missing-data treatment and attribution method.

Assigning monetary values

The valuation method should match the resource or economic consequence being measured. Charges, prices, reimbursements and economic opportunity costs are not automatically equivalent.

Report:

• The unit cost or monetary value.
• Its source.
• The jurisdiction.
• The currency.
• The price year.
• Any inflation adjustment.
• Any currency conversion.
• Whether taxes, discounts or overheads are included.
• Whether the value represents a price, charge, reimbursement or estimated opportunity cost.
• The uncertainty assigned to the value.

Monetary values from different years should be converted to a common price year when appropriate. Currency conversion alone does not make values transferable between healthcare systems or jurisdictions.

Valuing productivity consequences

Productivity consequences may include work absence, reduced productivity while working, early retirement or premature mortality. The method used to value these consequences can materially affect the total.

Common approaches include:

• The human-capital approach.
• The friction-cost approach.
• Direct employer costs.
• Self-reported income or productivity loss.
• National wage statistics.

The study should explain whose productivity is included, which valuation method is used, the time period applied and whether unpaid work is represented.

Productivity loss should not automatically be added to income loss, benefit payments and employer costs without checking for overlap.

Valuing informal care

Informal care may include unpaid help from family members, friends or other caregivers. The study should identify what activities count as care and how the time was measured.

Possible valuation methods include:

• Replacement cost using the price of equivalent paid care.
• Opportunity cost based on the caregiver’s forgone activity.
• A stated-preference valuation.
• Another justified local method.

Caregiver time, caregiver health effects and paid care costs are different consequences. They should not be combined without checking for double counting.

How the main calculations work

A cost-of-illness total combines the mutually exclusive cost categories permitted by the study’s perspective and boundary.

A broad societal total might be expressed as:

Total Economic Burden = Direct Medical Costs + Direct Non-Medical Costs + Productivity Costs + Informal-Care Costs + Other Included Consequences

Not every study should include every category. The components must match the stated perspective, purpose, population and costing boundary.

The analysis should show how each component was calculated rather than presenting one unexplained final total.

How a cost-of-illness study is carried out

A cost-of-illness study moves from a clearly defined question to data collection, valuation, attribution and interpretation. Decisions made early in the process determine which data are appropriate and what the resulting estimate can legitimately mean.

1. Define the purpose. State the planning, research or policy question the estimate is intended to inform.
2. Define the condition and population. Specify the disease, injury or risk factor, case definition, geography and affected population.
3. Select the perspective. Determine whose costs and consequences belong in the analysis.
4. Choose the epidemiological approach. Select a prevalence- or incidence-based design that matches the question.
5. Set the period and costing boundary. Identify the observation period, settings, services and included cost categories.
6. Choose the estimation route. Select bottom-up, top-down or a justified combination of methods.
7. Measure resource use and losses. Use appropriate clinical, administrative, survey, registry or economic data.
8. Assign monetary values. Apply valid unit costs, prices, wage values and other valuation methods.
9. Estimate attributable burden. Separate condition-related or excess costs from unrelated resource use where necessary.
10. Aggregate or extrapolate results. Produce population and subgroup estimates using appropriate weights and epidemiological data.
11. Examine uncertainty. Test plausible changes in data sources, attribution methods, assumptions and parameter values.
12. Validate the analysis. Check categories, formulas, denominators, units and potential overlaps.
13. Report the methods and limitations. Provide enough information for readers to understand, compare and reproduce the estimate.

Worked example of an annual burden estimate

Consider an illustrative prevalence-based study using a societal perspective. The study includes healthcare expenditure, patient costs, unpaid care and productivity losses occurring during one year.

The figures are synthetic and demonstrate the calculation rather than describing a real disease or jurisdiction.

Hospital and outpatient care

• Annual cost: £48 million
• Included category: Direct medical cost

Medicines

• Annual cost: £12 million
• Included category: Direct medical cost

Patient travel and out-of-pocket costs

• Annual cost: £4 million
• Included category: Direct non-medical cost

Informal caregiving

• Annual cost: £16 million
• Included category: Informal-care cost

Productivity losses

• Annual cost: £30 million
• Included category: Productivity cost

Total economic burden

£48 million + £12 million + £4 million + £16 million + £30 million = £110 million

The estimated annual burden is £110 million under the study’s definitions and assumptions. This does not mean that eliminating the condition would release £110 million in cash. Some resources may be fixed, shared, unavoidable during the period or redeployed to other healthcare needs.

Economic burden is not recoverable savings

The estimated burden represents valued resources and losses associated with the condition. It does not automatically equal the savings that an intervention could generate.

The amount recoverable through prevention or treatment may be smaller because:

• No intervention eliminates every case or consequence.
• Some costs are fixed in the short term.
• Freed capacity may be used by other patients.
• Treatment itself has costs.
• Health services may not reduce staffing or infrastructure expenditure.
• Productivity gains may not become cash savings.
• Costs may occur in a different sector from the savings.
• Preventing one condition may increase later healthcare use through longer survival.

A claim about avoidable or recoverable burden requires explicit evidence about the intervention, its effectiveness, coverage, implementation and resulting resource changes.

How the results can be used

Cost-of-illness results can support planning by showing where economic consequences occur and which populations or services contribute most to the estimate.

A COI study can:

• Describe the scale and distribution of attributable economic burden.
• Identify major cost categories.
• Identify affected groups and services.
• Highlight important data gaps.
• Provide resource-use or cost inputs for later analyses.
• Support workforce, service or research planning.
• Show how burden is distributed across sectors.

A COI study cannot:

• Establish cost-effectiveness.
• Determine which intervention should be funded.
• Show that the entire burden is preventable.
• Show that the entire burden can be converted into cash savings.
• Rank diseases as funding priorities using cost totals alone.
• Replace evidence about health outcomes, opportunity costs, equity or feasibility.
• Make the final resource-allocation decision.

Distribution and subgroup burden

A total population estimate may conceal substantial variation among groups. Report subgroup estimates when they are supported by appropriate evidence and relevant to the study purpose.

Possible subgroup dimensions include:

• Disease severity.
• Age.
• Sex or gender where relevant.
• Socioeconomic position.
• Geography.
• Comorbidity.
• Employment status.
• Type of payer.
• Healthcare setting.
• Patient and caregiver group.

Subgroup comparisons should use consistent case definitions, costing boundaries and valuation methods. Differences should not be interpreted causally unless the study design supports that conclusion.

Examining uncertainty

Uncertainty can arise from disease frequency, case definition, resource use, unit costs, attribution, productivity valuation, informal-care valuation, missing data and population extrapolation.

Relevant approaches may include:

• Confidence or credible intervals.
• One-way sensitivity analysis.
• Scenario analysis.
• Probabilistic analysis.
• Alternative case definitions.
• Alternative attribution methods.
• Alternative productivity methods.
• Alternative assumptions about missing data.
• Alternative population weights.

Report which inputs and assumptions drive the result. A single total without an uncertainty assessment may imply more precision than the evidence supports.

Why estimates from different studies may disagree

Cost-of-illness estimates are sensitive to differences in perspective, population, case definition, epidemiological design, attribution method, data source, valuation approach, currency and price year.

A larger estimate may reflect:

• A broader analytical perspective.
• A larger population.
• A more inclusive case definition.
• An incidence-based lifetime horizon.
• Inclusion of productivity or informal care.
• A different attribution method.
• Higher local prices.
• A more recent price year.
• Different epidemiological assumptions.

Comparisons should begin with methods and scope rather than headline totals. Converting results to a common currency and price year does not remove differences in population, healthcare system, valuation method or study design.

Preventing double counting

The total should use mutually exclusive categories. Potential double counting should be assessed whenever several data sources or valuation methods are combined.

Common overlaps include:

• Hospital charges and separately counted hospital resource costs.
• Paid care and informal caregiving.
• Work absence and lost income.
• Caregiver time and the monetary value of caregiver productivity.
• Healthcare expenditure and condition-specific programme spending.
• Mortality costs and lifetime productivity losses.
• Patient expenditure already included in payer totals.
• Total disease costs and separately calculated comorbidity costs.

Maintain a clear cost inventory showing each category, source, attribution method and overlap check.

Common problems that can distort the result

Methodological problems can make a COI estimate appear more precise or comprehensive than its evidence supports.

• An unclear case definition can place the wrong people or events inside the study population.
• An unstated perspective makes it impossible to determine whose costs were counted.
• Mixing prevalence- and incidence-based estimates can combine incompatible time frames.
• Counting every observed cost as disease-attributable can overstate burden.
• Overlapping categories can count the same resource or loss more than once.
• Treating charges or reimbursement as economic costs without explanation can misstate value.
• Claiming a societal perspective while omitting relevant patient, caregiver or productivity consequences can mislabel the study.
• Combining monetary values from different years without adjustment can create misleading totals.
• Extrapolating a non-representative sample can distort the population estimate.
• Presenting burden as fully avoidable expenditure can imply savings an intervention could not deliver.
• Comparing headline totals without examining methods can produce invalid rankings.
• Treating missing cost categories as zero can understate burden.
• Presenting productivity losses without identifying the valuation method can prevent meaningful interpretation.

What a transparent report should include

A transparent report should separate observed data, modelling assumptions and interpretive judgement. It should make uncertainty and exclusions visible so readers can determine whether the estimate is relevant to a particular planning question.

Report:

• The study purpose.
• The condition and case definition.
• The population, setting and jurisdiction.
• The perspective.
• The prevalence- or incidence-based design.
• The observation period and time horizon.
• The costing boundary.
• Included and excluded cost categories.
• Resource-use sources.
• Unit-cost and valuation sources.
• Attribution methods.
• Productivity and informal-care methods.
• Currency and price year.
• Inflation or currency adjustments.
• Population extrapolation and weights.
• Subgroup analyses.
• Uncertainty analyses.
• Overlap and double-counting checks.
• Material data gaps.
• Limitations.
• The distinction between total burden, avoidable burden and recoverable savings.