Extended Cost-Effectiveness

Extended Cost-Effectiveness Analysis of Health Policies

Extended cost-effectiveness analysis, or ECEA, evaluates a health policy by asking what it does for household finances as well as for health. Next to deaths or DALYs averted, it estimates how much out-of-pocket spending the policy removes, how many people it keeps out of catastrophic spending or poverty, and how each result falls across income groups. It was developed for the third edition of Disease Control Priorities, for low and middle income countries where patients pay much of the cost of care directly. Because it reports several outcomes side by side instead of one incremental cost-effectiveness ratio, it informs benefit-package choices on the way to universal health coverage. It extends cost-effectiveness analysis and sits close to distributional cost-effectiveness analysis, which handles health inequality but not financial protection.

Why financial risk protection belongs in the evaluation

A standard cost-effectiveness analysis counts the health bought with a budget, such as cost per DALY averted. Where households pay for care at the point of use, a public policy also changes household finances, which an analysis of health alone leaves out. Verguet, Laxminarayan and Jamison, who first set out the method for universal public finance of tuberculosis treatment in India, described public finance as having consequences in several domains. It raises uptake and so health gains, it crowds out private expenditure, and it insures households against catastrophic expenditures that would otherwise push them into poverty.

The Disease Control Priorities chapter by Verguet and Jamison records that the World Health Reports of 1999 and 2000 made financial risk protection one criterion of good health-system performance, and that households facing out-of-pocket costs cope by borrowing or selling assets. Target 3.8 of the Sustainable Development Goals now names universal health coverage including financial risk protection.

The four quantities an ECEA estimates

The tutorial by Verguet, Kim and Jamison frames ECEA as health policy assessment, concerned with policy instruments. For each population subgroup $k$, usually an income quintile but possibly a region, ethnic group or sex, the analysis estimates health gains, private expenditure averted, financial risk protection and the net cost of the policy to the government.

Health gains in each subgroup

The health calculation is the one a standard analysis makes, repeated for each subgroup. With a simple static disease model the tutorial writes it as follows.

$$B_{H,k} = \text{Eff}_k \times \text{Cov}_k \times D_k$$

where $B_{H,k}$ is the health gain in subgroup $k$, such as deaths averted, $\text{Eff}_k$ is the effectiveness of the intervention, $\text{Cov}_k$ is the coverage brought about by the policy, and $D_k$ is the disease burden before the policy. The tutorial notes that this captures one equity impact only, the distribution of health outcomes.

Private expenditure crowded out by public finance

ECEA takes the perspective of the people who fall ill. Their private expenditure in subgroup $k$ before the policy is as follows.

$$PE_k = i_k \times u_k \times (c_{DM,k} + c_{DNM,k})$$

where $i_k$ is the incidence of illness, $u_k$ is the use of treatment given illness, $c_{DM,k}$ is the out-of-pocket direct medical cost and $c_{DNM,k}$ is the direct non-medical cost, mainly transport. A policy that pays for treatment removes part or all of this, and in the simplest form the expenditure averted is $PE_{av,k} = \text{Cov}_k \times PE_k$. The tutorial describes this as disaggregating the societal perspective, so that the gain to households is shown instead of netting out as a transfer.

Three ways to measure financial risk protection

Money saved is not yet protection, because the same payment is trivial for a well-off household and severe for a poor one. The tutorial therefore scales expenditure by income and offers three metrics, of which an analysis uses one.

1. Cases of catastrophic health expenditure averted. A case of catastrophic health expenditure is counted when $c_{DM,k} + c_{DNM,k} > Th \times y$, where $y$ is income and $Th$ is a threshold share of income, for example 10%, 20% or 40%.
2. Cases of poverty averted. A case of impoverishing health expenditure is counted when $y > Pl$ and $y - (c_{DM,k} + c_{DNM,k}) < Pl$, with $Pl$ the poverty line.
3. Money-metric value of insurance. This is the risk premium a risk-averse person would pay to remove the gamble of illness costs.

For the third metric, expected income and certainty-equivalent income are compared.

$$y^{*} = \left[ p_k (y - c_k)^{1-r} + (1 - p_k)\, y^{1-r} \right]^{\frac{1}{1-r}}$$

where $p_k = i_k u_k$ is the probability of falling ill and paying, $c_k$ is the out-of-pocket cost, $y^{*}$ is the certainty-equivalent income under constant relative risk aversion utility, and $r$ is the Arrow-Pratt coefficient of relative risk aversion. Expected income is $E(y) = p_k (y - c_k) + (1 - p_k) y$, and the value of insurance is $E(y) - y^{*}$, summed over the income distribution. The tutorial reports that $r = 3$ is often used, while opinions on its value differ.

Cost to government and the efficiency ratios

Net cost to the government is estimated by subgroup as in a standard analysis, with savings from disease averted subtracted for preventive interventions. The tutorial then defines a financial protection ratio in the form of an ICER.

$$ICER_{FRP} = \frac{C}{B_{FRP}}$$

where $C$ is the net cost of the policy and $B_{FRP}$ is the financial risk protection provided, measured with one of the three metrics.

Choices that move an ECEA result

Income enters the calculation directly, which adds analyst choices to those of a standard evaluation. The tutorial asks for individual incomes where data provide them and otherwise a constructed distribution, for example a gamma distribution built from gross national income per capita and the Gini coefficient.

The protection metric and its cut-off change the count. The UN metadata for SDG indicator 3.8.2, last updated in March 2026, counts people whose out-of-pocket health spending exceeds 40% of a household discretionary budget, meaning consumption or income minus a societal poverty line. An ECEA using a share of total income therefore measures something different from the current SDG indicator.

On costs, the tutorial records that most ECEAs have counted only out-of-pocket direct costs because of data limits, and that productivity losses and borrowing should be added when data allow. It assumed new budget money to be available, while noting that opportunity cost could be included as displacement of existing programmes where budgets are fixed. The assumed change in coverage in each subgroup drives both the health and the financial results.

Worked example of public finance for one treatment across five income quintiles

All numbers in this example are illustrative. A country has five income quintiles of 1,000,000 people. Annual incidence of a treatable illness is 500, 400, 300, 200 and 100 per 100,000, poorest first, giving 5,000, 4,000, 3,000, 2,000 and 1,000 cases, 15,000 in all. Untreated case fatality is 10%, treatment is 80% effective and costs USD 150 out of pocket, and 60% of cases are treated. Universal public finance makes treatment free and raises coverage to 80%. Annual income per person is USD 400, 900, 1,600, 2,500 and 4,000, and the catastrophic threshold is 10% of income.

1. Health gains. Deaths averted per case are 0.8 × 0.2 × 0.1 = 0.016. The poorest quintile gains 0.016 × 5,000 = 80, and the others 64, 48, 32 and 16, a total of 240. The poorest two quintiles receive 80 + 64 = 144, or 60%.

2. Private expenditure averted. Here $u_k = 0.6$, and because treatment becomes free for every patient the whole of $PE_k$ is averted, so the 60% already treated stop paying. The poorest quintile saves 0.6 × 5,000 × 150 = 450,000, and the others USD 360,000, 270,000, 180,000 and 90,000, USD 1,350,000 in all. The poorest two quintiles receive 810,000, again 60%.

3. Financial risk protection. The cut-offs are USD 40, 90, 160, 250 and 400, so a USD 150 payment is catastrophic only in the poorest two quintiles. Cases averted there are 0.6 × 5,000 = 3,000 and 0.6 × 4,000 = 2,400, or 5,400, all in the poorest 40%.

4. Cost to the government. Paying for 80% of cases costs 0.8 × 15,000 × 150 = 1,800,000. Of this, USD 1,350,000 replaces household spending and 1,800,000 − 1,350,000 = 450,000 buys the 3,000 extra treatments.

5. Efficiency ratios. The policy costs USD 1,800,000 ÷ 240 = 7,500 per death averted and USD 1,800,000 ÷ 5,400 = 333 per catastrophic case averted.

6. Interpretation. A government-perspective analysis would report only USD 7,500 per death averted, and a societal one treating the payments as a transfer would give 450,000 ÷ 240 = 1,875 per death averted. ECEA also shows the USD 1,350,000 as a gain to households. Health gains and savings follow the cases, 60% to the poorest two quintiles, but every catastrophic case averted is there, because the same payment is large only relative to a small income. The 3,000 newly treated paid nothing before because they went without care, so their benefit appears in deaths averted, not in the protection count.

Reading ECEA results in priority setting

Health and protection per unit spent can be plotted against each other for several interventions, and the rankings need not agree. In an ECEA of nine interventions Ethiopia aimed to make universally available, measles vaccination averted the most deaths per USD 100,000 spent (367), followed by pneumococcal conjugate vaccination (170) and caesarean section (141). The most poverty cases averted per USD 100,000 came from caesarean section (98), tuberculosis treatment (96) and hypertension treatment (84). Choosing a pathway to universal health coverage may therefore mean weighing competing objectives.

An ECEA of tobacco tax in China estimated that a 50% price increase through excise tax would bring 231 million years of life gained over 50 years, a third in the lowest income quintile, and financial risk protection worth USD 1.8 billion, 74% of it in the lowest quintile, so the authors called tobacco taxation a potentially pro-poor instrument.

The tutorial presents ECEA as an input to benefit package design. When two policies give similar health returns, it can point to the one with more poverty reduction or larger gains for the poorest 40%, and the final weighting is left to the decision maker's preferences. Uncertainty is handled as in a standard analysis, for example by probabilistic sensitivity analysis.

Where ECEA sits beside distributional analysis

ECEA keeps the general approach and guidelines of cost-effectiveness analysis. The Disease Control Priorities chapter describes it as building on cost-consequence analysis, which tabulates disaggregated results, and on frameworks that bring equity and financial protection into economic evaluation.

Distributional cost-effectiveness analysis, set out by Asaria, Griffin and Cookson, models the social distribution of health under each option and judges it against the twin goals of improving total health and reducing unfair health inequality, using dominance rules and social welfare indices such as the Atkinson index. Its authors note that, like standard analysis, it focuses on health benefits and opportunity costs falling on the health sector budget. ECEA adds financial protection as a non-health outcome but stops at disaggregated results without a welfare function to combine them.

Limits and frequent misreadings

The tutorial describes ECEA studies as highly context specific and data intensive, needing subgroup data on disease, treatment use and out-of-pocket costs. Protection matters less where universal insurance already prevents medical impoverishment, as post-war systems such as the NHS were designed to do, although ECEA can still show the protection gained by adding an intervention to an insurance scheme.

Four misreadings recur.

1. Treating private expenditure averted as protection. Money saved by richer households may not count as protection at all, as in the worked example.
2. Adding the efficiency ratios. The cost per death averted and the cost per catastrophic case averted share one budget, so they cannot be summed.
3. Comparing protection results across studies unchecked. Catastrophic counts depend on the threshold and on whether the denominator is total or discretionary budget.
4. Reading the protection count as the full welfare effect. People who go without care because of cost never pay, so the count never includes them, and their gain from the policy appears only in health.

Sources

• Asaria M, Griffin S, Cookson R. Distributional cost-effectiveness analysis: a tutorial. Medical Decision Making. 2016;36(1):8-19. https://doi.org/10.1177/0272989X15583266
• United Nations Statistics Division. SDG indicator metadata, indicator 3.8.2. Last updated 27 March 2026. https://unstats.un.org/sdgs/metadata/files/Metadata-03-08-02.pdf
• Verguet S, Gauvreau CL, Mishra S, et al. The consequences of tobacco tax on household health and finances in rich and poor smokers in China: an extended cost-effectiveness analysis. Lancet Global Health. 2015;3(4):e206-e216. https://doi.org/10.1016/S2214-109X(15)70095-1
• Verguet S, Jamison DT. Health policy analysis: applications of extended cost-effectiveness analysis methodology in Disease Control Priorities, third edition. In: Disease Control Priorities: Improving Health and Reducing Poverty. 3rd ed. Washington (DC): World Bank; 2017. https://www.ncbi.nlm.nih.gov/books/NBK525307/
• Verguet S, Kim JJ, Jamison DT. Extended cost-effectiveness analysis for health policy assessment: a tutorial. PharmacoEconomics. 2016;34(9):913-923. https://doi.org/10.1007/s40273-016-0414-z
• Verguet S, Laxminarayan R, Jamison DT. Universal public finance of tuberculosis treatment in India: an extended cost-effectiveness analysis. Health Economics. 2015;24(3):318-332. https://doi.org/10.1002/hec.3019
• Verguet S, Olson ZD, Babigumira JB, et al. Health gains and financial risk protection afforded by public financing of selected interventions in Ethiopia: an extended cost-effectiveness analysis. Lancet Global Health. 2015;3(5):e288-e296. https://doi.org/10.1016/S2214-109X(14)70346-8