Dying Too Soon: How Cost-Effectiveness Analysis Can Save Lives
Table of Contents
The Harvard Life-Saving Study
The lifesaving interventions referred to above represent just a fraction of those surveyed by the Harvard Life-Saving Team. Funded by that National Science Foundation, my colleagues and I amassed cost-effectiveness information for hundreds of different interventions.
Although the full data set contains cost-effectiveness estimates for 587 interventions, national annual cost and effectiveness estimates were available for only 185 of these interventions. For each, we supplemented cost-effectiveness data with information on the degree to which that intervention was implemented. [See the sidebar on Methodology of the Harvard Life-Saving Study.]
To learn more about the economic efficiency of societal investments, we contrasted the current pattern of investment in these 185 interventions with the hypothetical "optimal" pattern of investment that cost-effectiveness would dictate. Like the interventions in Table II, some of the 185 were implemented fully, some partially and some not at all. Further, as Figure II illustrates, there was no relationship between implementation and cost-effectiveness. We estimated the resources currently consumed by these interventions at $21.4 billion and the life-years currently saved at 592,000.
Our research revealed that if the entire $21.4 billion were spent on the most cost-effective interventions, and none on the cost-ineffective interventions, we could save 1,230,000 years of life annually. That is 636,000 more than the 592,000 we are currently saving. Roughly, we could double the survival benefits of our investments at no additional cost!
"Basing decisions on cost-effectiveness will always save the most years of life."
The Efficient Frontier. This phenomenon can be understood by referring to the diagram in Figure III. The curve represents the maximum number of life-years that could be saved for a given level of resources consumed. This curve is called a "cost curve" or "efficient frontier" because it represents an efficient use of resources. It would be impossible to be above the curve, because the maximum survival benefits for each level of resource consumption is plotted. However, it is possible to be inside the curve -- by failing to maximize survival benefits for a given level of expenditures. Notice that the curve increases at a decreasing rate. This reflects "decreasing marginal returns." That is, the first few interventions cost very little relative to the survival benefits they achieve, but as we spend more money, although we realize more survival benefits, the amount we gain with each added dollar declines.
"We currently invest in many interventions that are not cost-effective and ignore many that are."
It is clear that the efficient frontier is a good place to be. But it is not clear where we should be on the frontier. That choice depends upon the maximum we are willing to spend to save one year of life. If that value is approximately $600,000, then we would want to be at point B because the slope of the curve at that point is 1/600,000. That is, B represents the point where the last and least cost-effective lifesaving intervention funded costs $600,000 per year of life saved. B would be the right choice if our willingness to pay was $600,000. If we were not willing to spend as much as $600,000/life-year, then we might choose something like point C, where the cost per year of life saved is, say, $100,000. If our willingness to pay were even lower than that, we might choose point D, where the cost per year of life saved is $10,000. Economists have estimated that people tend to make trade-offs between survival and money at the rate of $3 million to $7 million per life saved.5 A figure of $5 million per life saved would translate into a few hundred thousand per year of life saved, assuming 10-20 discounted years of life saved when a premature death is averted. Thus some point between B and C might be a reasonable choice.
"Failure to invest wisely in lifesaving interventions costs about 60,000 lives per year."
Unfortunately, our current pattern of investment puts us at point A, reflecting our current expenditures of $21.4 billion and saving of 582,000 years of life. If we divide $21.4 billion by 582,000 years of life we obtain an average of $37,000/life-year. This result may appear attractive, but in fact we can do much better. If our current lifesaving portfolio did not ignore many cost-effective investment opportunities and contain many cost-ineffective interventions, we could be at point B, saving 1.2 million years of life. The portfolio at point B could be achieved by holding expenditures constant at $21.4 billion, investing in all interventions with marginal cost-effectiveness ratios less than $600,000/life-year and none of the interventions with higher marginal cost-effectiveness ratios. The vertical distance between points A and B represents the 636,000 years of life lost annually due to our failure to invest wisely in lifesaving interventions.
If we preferred to spend less than $21.4 billion on promoting survival because, for example, we valued life-years at only $100,000 each, we could be at point C. If this were our choice, we would spend less, yet gain survival benefits relative to the status quo.