What is the return on a life saved?

Q: You both study the problem of HIV/AIDS through policy modeling. How does that approach help you to come to grips with a problem that’s as big and complicated as HIV/AIDS?

David Paltiel: Let me give you the first reaction I had when I saw the title “Can good health be good business?” The first thing that I would want to do is frame the question in the terms that students at SOM learn. When one thinks about whether things are good business, typically we think about factors like return on investment. Curiously, in the realm of health, people seem to want health interventions and health programs to pay for themselves. When it comes time to think about health, people just say “Is it budget neutral? Does every dollar we invest in intervention x actually return more than one dollar?” And it’s a curious standard to which we hold health. It’s a standard to which we don’t hold other sorts of investments.

Edward Kaplan: Typically, if you say we’re going to invest a certain amount of a resource, you ask, what do we get in return? So the idea here is you’re actually purchasing a good, which is improved health. I would argue that’s a better way of thinking about it.

DP: That’s what we do. We try to think about what is the return on investment, and how does it compare to the other ways that we might be allocating our health resources? Compared to the other opportunities for health investments that we’re forgoing when we decide to invest in intervention x as opposed to anything else, is the return on investment reasonable? So much of the work that we do is trying to figure out credible ways of formalizing the return on investment in terms of the things that people really care about: infections averted, years of life saved, even quality-adjusted years of life saved.

EK: A healthy year is worth more than a sick year.

DP: I would want to ask, can good business practice be applied to health? At present, it isn’t. It’s being unreasonably applied. As a result, we’re forgoing all sorts of lifesaving preventive and therapeutic investments and, in their place, doing all sorts of things that don’t pay a reasonable return on investment, but seem to appeal to passion and our intuitive sense of duty to lives in peril. The little girl at the bottom of the well, for example.

EK: Let me just change the focus slightly. I want to contrast the kinds of things we’ve done to what goes on in the more orthodox world of public health and medicine, which are disciplines that are largely phenomenological. They are science-based, they’re observational. They’re interested in learning what the relationships are between various factors, biological and human, to tell you how diseases spread, how diseases progress, and so on. They’re not focused around making decisions. The analyses themselves are sort of decision-neutral. They’re just trying to describe what is.

It’s one thing to talk about what an HIV epidemic, if left unchecked, does to you, but it’s another thing to talk about what the right kinds of decisions are that one could make. How do you know if an intervention is sensible or not? And it’s in support of those decision-making questions that you really would like to have criteria by which you can evaluate alternative choices.

The choices get made anyway, right? So are people supporting programs just because they believe a particular constituency needs help? Or are they supporting programs because they really think they’re going to get the best overall health outcome for the collection of resources they have? Usually it’s not the latter.

That creates all sorts of interesting paradoxes, not only within the world of HIV, but across diseases. I’ll just toss this out. There are large funds that sponsor public health interventions to slow the spread of HIV, slow the spread of malaria, slow the spread of tuberculosis — all these kind of things. It actually turns out that relative to, say, stopping the spread of malaria, HIV interventions are pretty expensive. Malaria, you do things like use bed nets. You can spray an area.

There are things you could do that actually would stop a lot more infections per dollar spent — there’s your efficiency measure — and then when you start actually equating what the consequences of the diseases are, you may actually start to think that, gosh, focusing only on this one disease could put you out of whack. So it’s very, very difficult to take all relevant margins into account. The problem with what I’m saying is that the argument never ends. You can talk about disease versus disease, but what about health versus education versus highways? So it’s not like you can really write down some kind of simple budget optimization for saying what the right amount is to spend on everything under the sun. But at least we can try to promote some kind of more reasoned thinking about whether different activities make sense.

DP: Let me pick up on two threads that I want to pull a little harder on. As Ed was saying, much of what people in the realm of health and medicine do is observational, is phenomenological. They run clinical trials. They study what diseases do to people. And Ed has often said, what we’re trying to add to the discussion is what people can do to diseases.

We don’t try to tell doctors what to do. Typically, what we try to do is add new information and ask questions that they’re not accustomed to asking — things that businesspeople who read this will probably be able to relate to: What are the returns on investment? What is the value of this information? What would you do with the results of this clinical trial if you had them?

EK: David did a very interesting piece of work in exactly that vein, where before the trial was even launched, he asked the question, “Is there any possible outcome of this trial that would change your decisions about treatment?” People wanted to do this trial because of a scientific question.

DP: Because they didn’t know the answer.

EK: They want to know the answers. Are we going to get a better outcome in a statistical sense, with p less than .05? David was able to show that there’s no outcome to this trial which would actually change behavior in the world. From an SOM standpoint, the value of the information would be zero.

DP: In the realm of health and medicine, that idea of abstracting and saying “I don’t know what the right answer is, but it would have to be bigger than x for me to care” makes them nuts. They want to know what the truth is. Sometimes the search for better decisions is not necessarily the search for scientific truth.

There are certain questions, which, either because of the long time-horizons or the costs or the ethical dilemmas that they pose, you would never be able to mount a clinical trial to investigate. But you can build a model-based analysis that gives you a sense of whether this is data you even want to go out and get if you have the opportunity. And there have been several instances in our own work where that kind of analysis really has changed the way people practice and deliver healthcare.

EK: The important point to make is that the operational or the logistical detail matters just as much in the outcome as the epidemiological detail.

When I studied the needle-exchange program here in New Haven starting in the 1980s, it turned out that understanding how the program physically interrupts the circulation patterns of the needles used by intravenous drug users was very important. It mattered just as much as understanding how the disease gets transmitted from an infectious person to a syringe and back to a person again.

DP: I’m glad you mentioned that example, because Ed came up with a very simple model that basically says the less time a needle spends in circulation, the better we are in terms of the transmission of the disease. Now, ex post, you stand back and say “duh!” But ex ante, nobody noticed that circulation time was critical. And there are certain levers you can pull to affect it. Throw more needles into the system. Or figure out how to get needles back and replaced with clean ones faster.

EK: Or make them more accessible. Improve the hours of operation.

DP: Once you’ve got your performance measure, you’re off to the races in terms of managing it, right? But it took that little insight.

I’ll give you a second example. This is going to seem obvious, too. I thought a lot about the mechanics of HIV testing and its cost structure. And one of the things that we figured out was that the cost of an HIV test itself is not the critical driver of the economics of HIV testing. It’s the fact that every now and then you identify an HIV-infected person, and when you do, you’re triggering a pathway of costs of antiretroviral therapy and patient care. In fact, whether an HIV test costs a dollar or eight dollars or even a hundred dollars a pop doesn’t matter. The important question is, how many people are you actually detecting?

EK: By the way, there’s another thing that also comes out of some of David’s work — it’s one thing just to find people, but now what are you going to do about it? If you get a bunch of people who are newly tested positive and you can’t link them to care, then what’s the point? You haven’t really blocked anything.

What you’re going to do with your next step becomes very, very important. If you were going to use your resources to screen fewer people but end up putting more people through, you could actually end up with a much better balance. You can say that the benefits come completely from doing something with those infected individuals you find.

DP: That sounds awfully obvious once you hear it, but people were investing as much money as they could in testing, trying to find as many infected people as possible, and not investing in the follow-up and linkage to care. For example, a surprising number, 25%, of people who test HIV-positive never return to get their test results. Even among those who return, a frightening proportion never actually go to care. It was like investing a ton to find needles in a haystack and then tossing them away.

EK: The same thing happens with a lot of the other sorts of interventions. With the needle exchange, for example, people who are advocates for drug injectors are of the view that these programs should be widely available everywhere. Well, I could agree with that, to a point, but not absolutely. Obviously, if the rate of infection among drug users was zero, everybody would agree that we shouldn’t do anything. But the switch point isn’t zero or not zero. It could be a little higher than that. So, this question of when it is worth it to do something is an important question.

And this is natural stuff in economics. How many franchises do you want? What’s the right market size for something?

DP: I have a collection in my files of instances where important people — usually politicians — have said, more or less, “Even if it saves one life, it’s worth it.” Now, that sounds beautiful, and there’s even a biblical reference — “He who saves just one life saves the entire world.” But I think most of us coming out of SOM recognize that one wants, when thinking about whether something is “worth it,” to ask the question “compared to what?” One wants to ask about opportunity costs, and what else might I have been able to purchase with that investment instead? That’s the kind of question one does not hear in the discourse around health and medicine.

EK: If it saves one life, it’s worth it. So what’s it worth to save two?

David gave the example of the little girl at the bottom of the well. So I think he meant the rule of rescue.

DP: It’s a human proclivity — a nice one, by the way — to feel a sense of duty to the identifiable life in peril. But the curious thing is that even when you tell people that you can be building fences around wells and save seven or eight little girls...

EK: Right, that’s exactly the point.

DP: Please do not think we’re saying you shouldn’t be yanking the little girl out of the well, but...

EK: Given that you’re willing to spend this much to get the little girl out of the bottom of the well, why aren’t you willing to spend comparable amounts to build fences around wells and save many more little girls?

DP: Now we know something about why we don’t: We know that little girls tend to elicit public sympathy. We know that identifiable lives in peril tend to elicit more sympathy than statistical lives in peril.

Q: Do you get resistance to your analysis from the medical community?

EK: We have. It’s not so bad now, because we’ve been at this game for awhile. But getting started was very difficult. With the needle-exchange stuff, as an example, all of the competing analyses at the time we started working on this were behaviorally based, and our approach was just so radically different from anything that they had looked at before that it was hard to accept.

DP: I don’t want to claim that we’ve had a huge influence, although we’ve had a few home runs. And in each case I think it was very important, number one, that we collaborated with clinicians, and number two, that we devoted inordinate — at least inordinate by the standards of reckoning of many of our colleagues in our own home field — amounts of energy to learning the language, translating into either doctor or public health talk, and then back-translating.

And the other thing I want to say is that the push-back that we get from the field is healthy. As much as I like to talk about cost-effectiveness and return on investment, it’s an efficiency-driven approach to health and medicine, and it ignores many important ethical and humane considerations. It is completely agnostic on the question of distributional equity. It doesn’t matter whether I can save 10 years for one person or one year for 10 people. And yet most of us would clearly have a preference for one of those two things over the other.

To the extent that one looks at quality-adjusted life-years, that’s a criterion that, inevitably, discriminates against both the elderly and the disabled. And so as upset as I am when people fail to bring questions of return on investment to the table, I would be even more horrified at a health system that relied entirely upon return on investment, to the exclusion of the things that make us human.

EK: Right. A good friend and former colleague, Al Novick, who passed away a few years ago, used to tease both of us on this very point. He would say that we would design interventions that would only prevent the easy infections, because if you’re trying to prevent as many as you can, in some sense it biases you towards the ones that are easiest to prevent.

Another instance of that sort of problem came up in some work I did a few years back with a panel at the Institute of Medicine, where the issue was how the CDC should allocate its federal prevention funds across different risk groups in different locations. And the issue here was that if you did take seriously the cost-effectiveness idea, it was clear that you could prevent more infections with the same money. But the resulting allocations would break almost every rule of fairness that people had previously imagined. You would have some risk groups in certain parts of the country that would not get a penny.

DP: This is a global issue, too. We had exactly this on an Institute of Medicine panel I was on. The question was HIV-prevention resources in Africa. The fact is if you try to prevent the greatest number of infections, you absolutely focus your efforts on the big cities. If you believe that an infection prevented is as valuable whether it’s in one person or another, then, again, you would focus on the big cities and let the rural areas go. It’s just too inefficient, by comparison. And because of the demographics and how HIV is transmitted, the net result of that is that you would be discriminating not only against the rural areas but against women.

So it’s a mess if one uses only the efficiency-driven criteria.

EK: Equity is an important consideration. How do you actually measure equity? That’s a subject in itself.

DP: Let me toot Ed’s horn for a minute. Here’s what Ed did, which I loved: He said, I recognize that if we were only fixating on the idea of preventing the greatest number of infections, we might, for example, end up allocating a disproportionate amount of money to the prevention of transmission among injection drug users. He then said, let’s suppose that we introduced some distributional equity criterion, which said we need to allocate at least some fraction of the budget to these other risk groups as well. Now we would have a diminution in the overall number of infections averted, but I can tell you how many infections you failed to avert as a result of that, thus actually putting a quantitative, analytic framework on top of what otherwise remains in the squishy world of qualitative hand-waving. This is how many infections you’re failing to avert — the opportunity cost.

EK: We put a shadow price on equity. There was an existing procedure which everyone involved seemed to think was equitable. And basically what it says is we’re going to dole out dollars in proportion to AIDS cases. And at first blush it doesn’t seem like a crazy idea. If I’ve got twice as many AIDS cases, maybe I’ve got twice as many infections going on, so I should get twice as much money.

But when you really think about it, you say, well, if you’re trying to prevent infections, programs have different efficiencies. But, having said that, people don’t like the idea of there being losers. If you have some kind of budget allocation process, not everybody can get everything, obviously. Some people can get more, and some people can get less. But these proportional formulas give everybody something, and people had the view that this was fair. So what we were able to do was to basically say, well, suppose we reserve or earmark a certain fraction of the budget using that old proportional formula, and do everything else in what we called a discretionary way, meaning by cost-effectiveness. You get most of your cost effectiveness for the first couple of dollars because you’re basically picking out the glaring errors that are made in the proportional approach. So you could end up doing not too badly by maybe going up to as much as 60% of the funding in the old-fashioned, “fair” way.

The techniques that we use can price out other things. Here’s another example: You have regulations that say you can’t do certain things. You know, the only kind of sex education which can be federally sponsored is abstinence-only education, and so on. When you take a look at these larger allocation processes, you can say, well, what are the best results you could get in terms of life-years expanded, infections averted, whatever your measure is, with the regulation in place? What’s the best you could get without the regulation in place? What’s the difference? I’ve just told you what the health cost of the regulation is. Now you have to argue that there is some other benefit which is worth more than that cost.

DP: I was just going to mention one other one, because I know Ed did this. Recall that Al Novick said, “You’re cherry-picking. You’re picking easy infections.” And I think Al went on to say, “These are all our brothers and sisters. You should be valuing them equally.” And Ed’s point was, “I’m trying to treat every single infection as being equally desirable to avert when I’m allocating my resources. Now, why don’t you tell me what these lives are worth?” And of course, people typically shut down and say, “I can’t do that.”

EK: “Who do you think I am, God?”

DP: So then Ed said, “Well, here’s how you’re allocating your resources, currently. Implicit in that allocation is a valuation. By allocating your resources, say, across the board equally, both to those groups that are easy to reach and those groups that are hard, you are implicitly valuing the lives in the hard-to-reach group at some multiplicative factor that’s 10 or 20 times as big as the other group’s.”

EK: Let’s take two drastically different situations. And, by the way, in what follows I’m not going to claim that the values should be the same. But the question is, whatever differences there are, should they be as great as the ones that we actually see?

On the one hand, preventing infections among drug users probably can be done for anywhere between $50,000 and $100,000 per infection, something like that — it’s relatively cheap. The size of the expense, of course, comes from having interventions that are targeted at a lot of people to prevent a few infections. On the other hand, we have screening of the blood supply. And the screening tests keep getting ever more sophisticated and costly. So now, you can ask yourself, what’s the incremental number of infections you’ve prevented in the blood supply by going from the cheaper test to the more expensive test, and what’s the total incremental dollars that you spent? Instead of preventing infections for $50,000 to $100,000, you’re preventing infections for, say, $7.5 million.

Now you can come back and say that you have drug injectors, who are knowledgably taking on risks, versus people who are only getting blood transfusions because they absolutely have to, who are not knowledgably taking on risks. Fine, there’s a difference, but should it be the difference between $7.5 million per infection and $50,000 to $100,000? It’s not obvious that it should be. But that’s how we behave.

DP: You know, one of the things that I meant to mention earlier is that in health and medicine, many of the decisions that we make and whether they seem attractive or unattractive hinges critically on the perspective that you’re taking. Let me give you an example: a policy to send mom home from the hospital after delivering a baby one day earlier than before. Viewed from the perspective of the insurance company, that’s a pure savings. Viewed from the perspective of the hospital, it’s probably a wash. Viewed from the perspective of mom and her family, and her caregivers, it’s absolutely a cost. And so, the perspective changes your view of whether this is an unbelievably attractive idea or an unbelievably unattractive idea.

Q: What are some of the big questions now in HIV?

DP: There are two worlds right now in HIV. There’s what’s happening in Western countries and there’s what’s happening in the resource-constrained settings. The questions in resource-constrained settings are how to allocate prevention dollars, how to scale up antiretroviral therapy now that we have a price structure that permits us to roll out drugs to a broader swath of the population, how to get them there logistically, and then once you’ve got them there, how to monitor and treat optimally in settings where the laboratory monitoring and the availability of professional clinical care becomes the constrained resource.

In this country, there are also important questions that we can be looking at. If I had to list them, operational question number one right now would still be when to start therapy, when to stop therapy, and when to switch therapy — optimal timing of care.

EK: One thing I will say, even though I haven’t been as involved in it lately, is that there’s been, in my view, a real lack of creativity on the prevention program side in the last several years. I haven’t seen anyone really come up with a brand new idea.

In the developing world, there’s also a big trade-off between treatment and prevention. You have that issue even in the places where people are trying to provide treatments. I think logistics come into play. David can tell you these stories better than I can — about pills getting shipped over and then spoiling for lack of refrigeration.

DP: Or diverted to the black market.

EK: Right.

DP: You go into a clinic anywhere in Africa, and the Fritos are there and they’re not crushed. The packages are intact.

EK: And there’s a soda machine, and yet there’s a worry about the pills spoiling. So why can’t you put the pills in the soda containers?

There are also unintended consequences which are just really wrenching, really hard to deal with in this area. One of the more cost-effective treatments that has been developed over the years has to do with preventing the infection of newborns from infected mothers. It turns out to be quite cheap in the sense that you’re not treating someone over a long course of time. You’re basically treating a woman over the course of her pregnancy, really right around her delivery time.

The problem though is that it becomes possible to prevent a lot of these mother-to-child transmission events, but there isn’t enough money to continue to treat the mothers. So the unintended consequence is that you save a lot of babies who are then destined to become orphans. I did some back-of-the-envelope calculations a few years ago which suggested that if people were going to really go gung ho on preventing these mother-to-child interventions, they easily could be creating 10 million new orphans a year.

Here’s a case where it becomes very, very difficult. We’re not just talking about money anymore or talking about resource allocation. Should you try to save as many kids as possible, or should you save fewer kids, but those who are saved have their families preserved? How do you make decisions like this? This is hard. This is not something that I can offer an answer to.

DP: I’ll give you one more thought: the distribution channels.

We spend a lot of time at SOM talking with students in operations and production classes about the fact that distribution channels can have different attributes, as do the things that flow through those distribution channels. Information can go instantly. It doesn’t have a shelf life, and it doesn’t have any weight. There are perishables, there are non-perishables. There are things that need to be refrigerated.

There are all sorts of things, currently, that need distributing in the realm of health and medicine. So one of the things we teach students here in production classes is that when systems are working at capacity, stuff can blow up, and that you want to maintain, as much as you can, a certain amount of excess capacity to deal with the ups and downs in the demand for the service. Might it be possible to piggy-back on to some of that excess capacity in situations where stuff really needs to be transported for humanitarian purposes? And might it be possible, actually, to create a taxonomy that links particular items that need shipping to the distribution channels that might be particularly suited to them?

So, for example, Ed has thought a lot about the distribution of vaccines and treatments for different diseases in the context of bioterrorism. Anthrax and smallpox have completely different attributes that commend one to delivery to the home, while the other might be delivered to central community points that people can come in to.

EK: With anthrax, the issue is contamination of the whole area. In Washington, D.C., now, as a result of that work, the plan would be that police-escorted postal workers would deliver antibiotics.

DP: Right, so the post office is perfect, because it goes to every home.

EK: They do it now. That’s what they do.

DP: Right. But you can’t send bulky stuff that way. However, you can send bulky stuff via tanker to any big transshipment point, just by sticking it on top of a British Petroleum tanker that’s going there, or something like that. And so, if we could think about which attributes commend a particular product to a particular distribution channel that would be a really nice SOM-type exercise.

Interviewed by Jonathan T.F. Weisberg