Welcome to our section on advanced models and system dynamics, and in this section we are going to discuss the biggest picture of a health system that has to provide both cure and prevention. I'd like to start by just noting that there is often in the world a common budget for both cure and prevention. Typical ministry of health has to do it all, has to take care of the sick as well as preventing the spread of diseases through the use of sanitation and health education, making the water and foods and medicine and healthcare safe. They have to do both with the exact same budget. But cure and prevention are really separate social goals. And when they're placed in that common budget of the health budget, there could be a zero sum political game played between the two sides of the health system. Where curing people don't see themselves as in the same interest group as the prevention people. Whenever there's a dispute between the curative side of the health care system and the preventative side of the health care system, we just see this as something we see again and again that prevention will lose the zero-sum politics. And this is an application of system dynamics modeling to try to figure out why a common budget for cure and prevention can lead to unpleasant outcomes. So we're going to be using system dynamics as a cautionary tale. Something that we're just illustrating what could go wrong when we're not paying attention to aspects of our health care system. So I wanted to set up this model to show what could happen when the government is a good government, the government is trying to do the rational thing. Allocating it's funds to prevention and cure according to strict cost benefit principles. Not being biased by emotion or anything crazy like that, they simply want to do the right thing. We'll also have foreign NGOs in the system that are also donating their money proportionally, according to rational cost-benefit strategies. But we're going to put in this wrinkle in the system that there can be lobbying by the doctors, and the doctors might take up some of their money and go to the government and try to lobby for their particular side of the health care system. That lobbying has the potential to lead to unintended consequences, where the ministry of health might tilt the balance of spending away from prevention and towards cure. You'll see in this model that if the doctors can accumulate more political power than the prevention workers, they can distort the system so much that it can be quite harmful to the people. How can the doctors accumulate extra power? Well, there's asymmetry that I'll put in to the model where the doctors are able to charge fees because sick people are willing to pay fees. But the prevention workers are not able to charge fees, and that's because healthy people don't really feel like paying anybody to be healthy. It's very hard to get money from healthy people to keep being healthy and it's very easy to get money from sick people to offer them health. And inside that basic asymmetry, these charitable organizations and NGOs, once they inject money into the system it simply amplifies the lobbying power of the medical lobby. And the prevention workers don't have a base in order to keep up with the prevention workers, and we see a system that can actually end up hurting the people. So let's go through a very complicated system dynamics model, and show that one can break it apart and understand it, and even use it in our work. So in this projection, the model is simply being shown to you in all of its gritty detail, and don't try to read each word here yet. I simply want to show you that there are four main areas, and I'll go through each area one by one. I've circled them and given them separate colors. In the blue, there is going to be a population area where we keep track of the people, how many people there are, and whether they are sick or healthy. We're going to have two diseases to keep track of, curable disease called disease A, and a preventable disease called disease B and I'll come to that in just a second. There is a yellow part of the model, which is the resource allocation, where the money could either go to the left and go to the doctors, or go the right and go to the preventive worker. And then if we look at the left we see the doctors side of the module and the doctors get money and spend money on taking care of patients. If we look at the right, we see this pink area where the hygienists, or prevention workers, also get money, and either use their resources to take care of preventing diseases. But they also can use the money to lobby the system. And that's what we'll go through in detail. But I wanted to show you that this breaks down into simply four domains that are going to be running in the system. So let's first look at subsystem one. In subsystem one, we have a susceptible population on the left hand side of the slide and they are running out. They can die of only one thing. They can die of this sudden lethal illness called disease B. At the top right of the model, they can get this curable disease, called Disease A, and you can see them flowing to the right into Disease A, and flowing to the left out of Disease A as they are recovered. The rate of recovery, you can see some blue arrows flowing into that recovery, that's definitely going to come from doctors who are taking care of them. The rate of dying from disease B is also influenced by factors, and you'll see in the full model that the death rate from this preventable disease B is being driven by the preventive workers trying to do something to prevent that sudden death. Disease B should be thought of like a bus crash, the bus runs into you and you die instantly, the preventive worker can try to make the bus's safer so they don't hit you. But you’re definitely going to die if you get condition B. Disease A is something that never kills anybody, it just makes them sick for a long time and if they don't get any medical care, they stick sick for a long time. If they get medical care, they can recover much faster, they don't spend as much time staying sick from disease A. Both disease A and condition B go to the right, and we'll keep track of how many disability adjusted life years we're losing in the system, both due to the temporary disability from disease A and the lost life years from diseased B. So we can keep track of how the population's doing at all times. There's some system for the doctors. The doctors get money on the right from NGOs and from the Ministry of Health gives them money that they'll use, and we'll call that doctor power. They get money from their private patients. At the bottom, you can see private pay to the doctors flowing into doctor power. They spend their money on lobbying. And you can see an arrow on their lobbying time. And then on the left, the doctors use their power and their money taking care of patients who are sick with the Disease B. Through sub-system two B, which is the hygiene workers, the preventive workers. And again, the preventive workers get money from NGOs. They get money from the Ministry of Health. They spend some of their money on lobbying to protect their interests, and they spend some of their money and their power taking care of preventing disease B. And the more they spend preventing disease B, the less people die from disease B. So we've gone through three of the four subsystems, the last of the four subsystems is the political allocation. So we have a public fund in the middle with arrows going left and arrows going right. The public fund going to the right pays the hygienist and that leads to hygienist power. Also the NGO can give money to the hygienist and that will flow to the right. The public fund, if it goes to the left, can pay the doctors, and the NGOs can also pay the doctors, and that would flow into doctor power. Now there's this critical asymmetry that the gray arrow's pointing to where consultation fees that come from the number of patients with disease A will give money to the doctors. They'll get private pay, at the bottom left of the slide. The hygiene workers have no way to get any private pay. That's the system. And if you turn it on in Vensim, you can start running it and playing with it. There are some policy models that we've done with this model that we published this year, in the 2014 publication by Dr. Pinot, Dr. Peters, and myself. The model of the policy we chose to look at was to think about an NGO that was trying to donate it's money in proportion to the burden of disease in this country. And this is a very rational NGO. They look at how much burden is from disease A, and they donate $D sub A in proportion to the DALYs and they donate $D sub B in proportion to the DALYs from disease B. And so for this very rational NGO, we want their policy maker to set up a policy of donating to each disease to try to minimize the total disability adjusted life years at minimum cost. So when they're trying to do this, what we'll show is that an unanticipated problem emerges. Now this is not a crazy type of policy to do. We've just seen, after the Ebola epidemic of 2014, that the US government decided to donate money to the effected nations in proportion to the size of the epidemic. And the Gates Foundation also donated money in proportion to the size of Ebola epidemic. NGOs and governments do this all the time. They look at how much disease is in a population and they allocate in proportion to that disease. So setting up the amount to donate in proportion to the DALY is something that we see, and we are just simply modeling that phenomenon. So what's inside the model is a set of epidemic spikes, and I've shown you pictures here of those epidemics. We set up the system so that if nothing happens, nothing happens and there will be no disease A or disease B, and everything runs flat. And that's really boring. In order to get some dynamics, we program in a set of four epidemic spikes of disease A, in red, and four epidemic spikes of disease B, in blue. And the last one, we make them come together. And we modeled the system over a period of 150 years, simply to give enough time for everything to reach an equilibrium after the epidemics. So these are the step functions that will trigger the response by the donors and the governments, and we wanted to see what happens. We can track the cumulative burden. We can track the total number of deaths of Disease B, in blue. And every year that time goes on, especially after epidemics, the death count goes up. And in red, we're tracking the total cumulative number of cases of A, and each case of A, depending on how long it lasts, generates disability adjusted life years, and every time years go by, the cumulative burden of both diseases will rise. So what I want to show you in this slide is a graph of spending on the vertical axis, against the total cumulative disease burden. And in blue, we're showing spending on doctors going up at a quite similar slope to the slope of spending on health workers, as a proportion of the disease burden. So both burdens are triggering a similar response in the long run between giving money to doctors and giving money to healthcare workers. This goes back to what I said. We program them at the start to be very rational, where they don't privilege either disease A or B. As the burden goes up, both healthcare workers and hygiene workers are getting the same amount of dollars per unit of disease over the long term. So we go back to the question for the NGO. If we see an epidemic of A, how many dollars per case of A should be send? That is going to symbolized by D sub A, the dollars per case of A that we should send in response to an epidemic. Similarly, if we see an epidemic of B, how many dollars per case of B should we send? And the answer will be symbolized by D sub B. And the example would be like the Ebola outbreak. You've seen Ebola outbreak? How many dollars in proportion to the caseload? Suppose we see a homicide outbreak. Again, we donate in proportion to the caseload. Homicide outbreak kind of means an insurgency or a terror episode occurring. And again, donor governments do respond in proportion to the size of burden. So the policy consideration that one might think about, one might take an ethical view of things and say a DALY is a DALY is a DALY, and that the D sub A should always equal D sub B. That all disability adjusted life years are the same, whether it comes from Ebola or homicide. However, we know from our experience in public health systems that curing sick people gives the warmer glow than preventing disease. And foundations need to go to their board and talk about how the board should really have a big warm glow from taking care of a sick baby rather then preventing something that no one ever sees. So just to be realistic about things, it might be okay for people to be ethically unfair to DALYs of prevention and give more money to DALYs of cure. We can get over that because we just see it as part of human nature. But the systems question is does that type of disease exceptionalism, of treating these curable diseases as exceptional and worth the extra money, could it possibly offer unintended consequences? And could it potentially actually create more death and suffering than it's worth? And we're going to answer that question with a stress test of the model. We're going to try out different pairs of D sub A and D sub B, and see what happens. So fortunately for you, our team has run this model hundreds and hundreds of times. And each dot on this chart represents the outcome of a full run of the model over its 150 years, and we've summarized the outcomes of each of these models. So if we pick a point on the far right, you'll see it numbered 10:0. That 10:0 point represents a policy that has had D sub A set = 10 and D sub B set = 0. So we run the model for 150 years with DA = 10, DB = 0, and what came out that time were DALYS on the horizontal axis and it looks like we got about 34,900 DALYS that time. And in order to do that we had to spend a total of about $30,100 on the vertical axis, and that's where the point is placed on the graph, at about $30,100 of spending in order to achieve 34,900 DALYS. And so we can try out many other points. Now, if you're thinking about a graph like this, DALYs are bad, and spending money is bad, and the best place to be on the graph, Is as close to the bottom left as you can possibly be to have very low DALYS and very low spending. So 10:0 is a very low cost setting. But 10:5 might be a little bit better as long as we're only willing to spend $10 for disease A. Now if we're willing to spend more per case if disease A. If we're willing to spend $20 per case of disease A. We can shift over to the yellow curve, and 20:5 is where we spend DA equals 20 and DB equals 5, and that looks pretty nice. And if we move even further to the left, if they're willing to spend $30 per case of disease A and $5 per case of disease B. We can get to 30:5 which is a pretty attractive number, very close to the bottom left, and that would be a good thing. You can see how it's possible to overspend on either prevention or on cure. You could see very high cost settings where DA equals 30 and DB equals 70 at the very top of the model. And you can see it can be very wasteful to spend too much on, prevention. So this is the type of policy outcome that we can help guide these DA versus DB decisions. So holding DA fixed and increasing DB from 0 to 70. One learns that you can never make people worse by spending more on prevention, you can just spend inefficiently. As we move up our spending on prevention on disease B from 0 to 70. We never killed people, we never make DALYs get bigger but we do make costs get higher every time we spent more on prevention. We save more people but we might save them at a very high costs. So that's what we've learn from the baseline settings. So this was holding DA fixed and seeing what happens as we rachet up spending on prevention. We found some attractive options and we found that there's a general sense in which there can be diminishing returns to spending on prevention. Now let's move ahead to this model. And when we do these experiments holding DB fixed, And running DA up from 0 to 70, something really surprising happens. This doesn't look like the slide before, now as we run our curative spending up above about 15 or 20, there's this reversal and people start getting sicker and sicker. If you look at the red curve, right about 15:0, right after we start spending $15 per case of A and start spending $20 per case of A, the red line takes a turn to the right, and more people are dying. They're getting DALYs, and the more we spend on cure, the more people die. And the more we spend on cure, the more people die. So after this threshold of about $20 per case of A, spending more on cure makes people die. That's pretty scary, because we always wanted to do the good thing, and we saw these people were suffering. They were having this curable disease, and we thought, wouldn't it be good to spend more on their curable disease? And then in our simulation there's this massive die-out of the people from spending more on cure. So what could possibly be going on here, this was an emerging property that. We were surprised to see, we thought it might happen, but what was happening? The next slide talks about this emergent and surprising property. Too much money on curative care was hurting the people, and why? That didn't happen for prevention, and the real explanation for it, and after we've gone through the model. We can be confident that in the simulation, these were perverse effects that were driven by the doctor's lobby and the zero-sum budget. What was happening is as the donations to curative care were getting bigger, the doctors were getting more and more lobbying power. And as they were going to their government to lobby they were eating away the prevention budget. And they would say very important things, they would say like doctors would. Look at all the sick people, they're suffering. You parliamentarians need to save them. Look, they're having pain, you'd better give us more money. And it doesn't have to be as blatant as that. It could simply be building a few new medical schools. Building some medication procurement programs. It's all very good, warm glow stuff that goes on. But if it's coming at the expense of the prevention budget, what was happening in the model was the hygiene workers were just getting laid off and not stopping Disease B from killing people. And the system didn't respond to that epidemic of B, and could that have possibly happened? Well I think it could, I think prevention budgets around the world can sometimes be put into a state of benign neglect. Now we want to make sure that the effects that we saw these reversal effects where the death rate starts to get higher after a threshold. That it wasn't driven by us just setting a set of payment levels wrong, and so in this slide we show that whether we set the payment to hygienist high or low, the doctors payment high or low. That same general pattern was still happening, and as we worked with the model we found that that emergent property would happen regardless of how much we weighted the DALY weights on disease A or B. The only thing that would make it go away is if we set DALY weights for one of the diseases equal to zero. So disease A didn't matter or disease B didn't matter. The same with the lobbying power, if the lobbying power parameters were set differently, we would still get this phenomenon, except if we put the weights set at zero. So we got a lot of understanding, and since it's a simulation, we could turn on and turn off parts of the system to try to eliminate the effects of pathological NGO spending. And we found that by just stopping that entire loop of doctor lobbying and just turning it off, we could fix that problem of the perverse high death rate from curative spending. The other thing that we found we could do is if we could ring-fence, meaning protect public health budgets and say they can't ever be used up for cure, that would also fix the problem. In that case the doctor lobby would say we can't pull money from prevention, we have to pull money from the Defense Department or something. Well that would also fix the problem. The other fix that would work is if the health planners would actually notice that they were having this epidemic of B and find a way to fix it. And give back the money to the hygiene workers that would obviously work as well. So that's what we got out of the model. And I want to emphasize that we didn't try to make a real model of a real disease, or a real country. We named it disease A, and disease B to try to give us the freedom to think more conceptually. We wanted to use this a time to demonstrate the harmfulness of zero sum budgeting. Meaning this basic asymmetric economics of health where curing people gives you more money than preventing, rising to a powerful class interest of the doctors. Now it's true that there are places where zero-sum budgeting occurs. There are many, many governments that put their prevention and their cure, in the same budget. So we're not wasting our time modeling a non-real disease in a non-real country. We hope that this is a cautionary tale that will help real governments realize the folly of putting their prevention budget in jeopardy by putting it in the sights of hungry interest groups that are looking for extra public resources. And there are real places where there is a cure lobby being driven by disease exceptionalism, where this disease is more important than that disease. And once one cuts the world up into diseases, one forgets about keeping people healthy in general and system dynamics modelling is one way to bring that focus back on and to just promising ways for policy. So I hope you got something from this discussion of the model, and we're making this Vensim model available to you. So that you can try out different settings and see what happens when you play with it. And experiment with different policy experiments in this simulation model. Thank you for your attention. [MUSIC]
