rb2 large gray Calculating the value of a year of human life in $US

I just returned from a week of amazing talks at the IEEE conference on computational intelligence in Paris (yay!). I won’t bore you with a full review but I thought I’d do a post on an interesting debate I had based on a paper presented at the conference. The debate was about how much a human life is worth, it’s a crucial variable if you are writing an algorithm for managing a rapid response to a state of emergency. It is still generally too much of a taboo to put a real number on the value, let alone seriously consider giving different lives different values. That said we know if we had one fire engine to send to a burning school or a burning office we would not make a random decision. At the moment it takes up to 16 minutes for emergency vehicles to be dispatched in a state of emergency, if the dispatch was fully automated decisions would be made in seconds (Mohammadi & Sadeghian, 2011).

It’s a relevant question for me because my interest is in brain computer interfacing (BCI), a technology that provides means of communication for the severely paralysed by linking the human brain directly to a computer. BCI isn’t widely used clinically yet largely because the medical grade EEG equipment required costs tens of thousands of dollars. What isn’t discussed is the question of whether health authorities are interested in extending lives using artificial means, it certainly doesn’t do health budgets any favours and for some people the prospect leaves a bitter taste in the mouth.

One reason I’m fully behind the (BCI) developments is that history shows clearly that the costs of these systems are destined to fall rapidly and if we invest in them their performance is likely to increase exponentially. If you were to plot a graph of the increase in computing power or the reduction in computing cost over time using a regular linear graph it would be a relatively straight line disappearing almost vertically in to the sky.

The graph below is a semi-logarithmic plot which means each level on the y-axis represents a number one hundred times the level below it.

exponential growth Calculating the value of a year of human life in $US

During some slightly off-topic research on this somewhat rhetorical question I stumbled across this rather morbid paper which presents some intriguing stats from a meta-analysis of peer reviewed journals. Below are just a few of the relative costs of treatments or state interventions that are in practice. They are presented in terms of the cost of the intervention for every “life-year” saved.The value of life1 Calculating the value of a year of human life in $US

I’ll be first to admit that some of the figures are more than a little bit odd (probably due to the infinite number of confounding factors in a comparative study like this) and I haven’t reviewed the sources of the paper but I recommend taking a glance at the original (PDF). It’s morbidly intriguing, if nothing else.

Tengs, T., Adams, M., Pliskin, J., Safran, D., Siegel, J., Weinstein, M., & Graham, J. (1995). Five-Hundred Life-Saving Interventions and Their Cost-Effectiveness Risk Analysis, 15 (3), 369-390 DOI: 10.1111/j.1539-6924.1995.tb00330.x

Mohammadi, & Sadeghian (2011). iFAST: An Intelligent Fire-Threat Assessment and Size-up Technology for First Responders Proceedings of IEEE Symposium Series in Computational Intelligence

Follow Simon on TwitterFacebookGoogle+RSS, or join the mailing list.

Tagged with:
  • http://www.snowflakehell.blogspot.com Ken Brody

    Shouldn’t the cost per life-year be calculated as the extended future cost of all dependent and influenced lives reduced to present value? For example: the ban on asbestos brake blocks seems cost-ineffective unless you also consider that at some critical mass point, manufacturers will switch to ceramics. Therefore you are not only saving the lives immediately exposed to asbestos, but also changing the industry to avoid that material.

    In fact, when you consider the discounted present effect of all probabilities and all possible causal chains, you come to the conclusion that you cannot do this calculation with any accuracy. So perhaps the best you can do is a ranking, based on the best judgements of experts, or models based on scenarios.

  • Peter

    I agree with Ken that for a proper benefit-cost analysis, you’re going to need to calculate a net present value. That would get you a step closer to the way that BCA is actually carried out.

    More importantly, as you alluded to (“leaves a bad taste in the mouth”) and as Ken mentioned, BCA is not a great decision-making engine. There are actually a lot of ways to subvert BCA in practice. At most, a BCA should only be used as one criterion in such a decision.

    Economist Frank Ackerman has written about BCA and its systemic biases. For a short intro, you could check here:

    So what, if not BCA, do we use? One (better) option is multicriteria evaluation. It tends to help illuminate good choices, and doesn’t overrepresent the validity of its conclusions the way BCA does. In other words, it doesn’t drop a “right” answer out the back. Still, very useful and much better than BCA:

  • http://abrahamjnunes.wordpress.com Abraham Nunes

    This is an intriguing question. I would suggest you look up the “Quality Adjusted Life Year” or “Disability Adjusted Life Year.” These are the primary health outcome measures used in the health economics literature. Being in medicine, I would suggest that measuring a human life in dollars and cents is impossible: but, for an emergency situation such as the one you posed originally, and for health care provision, decisions must be evidence based. Adoption of the QALY has been a great step forward in this respect. It’s weakness, of course, is that QALY valuation is relative.

    Keep up the great posts!

Looking for something?

Use the form below to search the site:

Still not finding what you're looking for? Drop a comment on a post or contact us so we can take care of it!

Visit our friends!

A few highly recommended friends...