Picture a health commissioner with budget for one diabetes app and three to choose from. The one that costs least per managed case wins. That is cost-effectiveness analysis, or CEA, at its simplest, and a UK government guide on evaluating digital health products uses almost exactly that example.
CEA compares the costs and outcomes of different courses of action. It is not the same as cost-benefit analysis, which puts a money value on the effect. Health services lean on CEA because putting a price on a year of life or a premature birth averted can feel wrong.
The result is a ratio. The cost sits on one side and the health gain on the other, whether that gain is years of life, premature births averted or sight-years gained. The most common outcome measure is the quality-adjusted life year, or QALY.
Variations on the ratio
Cost-utility analysis is a special case. It measures effects in years of full health lived, using QALYs or disability-adjusted life years. Results are usually given as an incremental cost-effectiveness ratio, or ICER, which divides the change in costs by the change in effects. Analysts often plot the results on a plane with four quadrants, cost on one axis and effectiveness on the other.
Standard CEA aims to maximise the average level of an outcome. Distributional CEA also looks at how that outcome is spread across a population, which puts the trade-off between equity and efficiency in plain view. Researchers use it mainly on interventions meant to tackle health inequality.
In pharmacoeconomics the effect measure depends on the treatment. It might be the number of people cured, the reduction in diastolic blood pressure in mm Hg, or the number of symptom-free days a patient has. Clinical judgment decides which fits. Side effects, interactions with other prescriptions and even small matters like taste can get some weight too.
A worked example
The UK guide describes a 2018 trial of HeLP-Diabetes, a self-management website for people with type 2 diabetes. Researchers took the view of the NHS and personal social services. The main outcome was diabetes-related distress, scored on a 0 to 100 scale where lower is better.
At 12 months the programme cut the score by 2 units compared with usual care. It cost £111 more, which worked out to £58 per unit of improvement.
The guide notes the method’s limit. Because CEA measures effects in disease-specific units, it cannot guide choices across different diseases.
Beyond medicine
Other fields use the same logic. Military planners comparing tanks look at operating radius, top speed, rate of fire, armor and gun performance as well as price. A tank that performs about the same but costs far less and is easier to produce can come out ahead. Where the price gap is near zero but the dearer tank carries special ammunition, radar fire control and laser range finding, planners may pick it instead.
CEA has also been applied to energy efficiency in buildings, to calculate the value of energy saved in dollars per kWh.
Published studies carry their own bias risks. A 2006 systematic review found that industry-funded studies often reported ratios below $20,000 per QALY. A 1995 study of more than 500 life-saving interventions found a median cost of $42,000 per life-year saved.
