Something Not Unlike Research

Inequality? Oh we'd love to help, but...

post by Bill Gardner

Arnold Kling and Nick Schulz argue that the growth in US inequality derives from two factors:

• Technology that allows a few very able persons to achieve extraordinary things (think Sergey Brin), and
• Defective families and assortative mating based on ability that result in large groups of people at the bottom who lack the talent to produce much of value.

These are important causes of inequality. Kling and Schulz would do something about these things if they could, they really would, but 

there may be less that government can do than one might hope. Research from Heckman suggests that education is a relatively feeble remedy for the effects of family background (although Heckman believes that early intervention, in preschool or even before, shows promise). In order to make a dramatic impact on inequality, government would have to do something about the fundamental causes: technology and marriage patterns. However, putting a brake on technological progress seems hardly feasible or desirable. And forcing people to select mates at random rather than on the basis of similar backgrounds and tastes seems similarly unlikely. As much as inequality may be a problem, no real solution is in sight.

OMG, if the only alternatives are freezing technological progress or making people marry random partners (on second thought, this is the best explanation for my first marriage, and the kids turned out great anyway), then I guess there really is nothing to be done.

Or maybe we could try the following:

• We could give every mother good prenatal care, and institute programs to reduce premature delivery. 
• We could provide every child with universal, high-quality preschool education. Please, read Heckman for yourself. He isn't saying that early interventions "show promise." He's saying that "Early interventions promote schooling, reduce crime, foster workplace productivity, and reduce teenage pregnancy." 
• We could reform our criminal justice system. The US has 751 persons in prison per 100,000 population. Canada has 117. England's rate is 151; Germany's is 88; and Japan's is 63. Mexico has 204. The global median is 125. Are Americans really six times as violent and criminal as the rest of the world? The burden of US incarceration falls on the poor and above all on people of color. It destroys families. But it could be fixed.

America could do these things and it would still have considerable inequality. But these are solutions that would help and, contrary to Kling and Schulz, they are in plain sight.

'Health Shock': How childhood chronic illness affects children's schooling and adult employment

Pediatrics is sometimes treated as the least important branch of medicine. This kind of makes sense, because children are mostly healthy. Focusing just on health outcomes, however, is terribly short-sighted. As Jason Fletcher and Michael Richards show, chronic diseases that begin in childhood have powerful effects not only on children's adult health, but also on their non-medical development and achievements. 

Fletcher and Richards looked at the association between the 'health shock' of  getting diabetes in childhood and achievements in schooling and work for more than 15,000 children. They found

effects in several measures of educational attainment, including a high school dropout rate that was six percentage points higher than among young adults without the disease. We also found lower employment and wages: A person with diabetes can conservatively expect to lose more than $160,000 over his or her working life, compared to a peer without the disease. 

This is an observational study, so whether diabetes causes these effects is a matter of inference and interpretation. Fletcher and Richards used a data set that included detailed information on the children's family backgrounds, the health of parents, and other diseases that children had; and their estimates include statistical controls for such factors. We cannot necessarily infer that an intervention that prevented diabetes in a child would increase the years of schooling they would achieve, or how much they would earn as adults. But based on these data, it would be worth a shot.

Prohibiting Anti-Smoker Hiring Policies is Not "Leveling Down"

post by Paul Kelleher

Bill responds today to my view that anti-smoker hiring policies are unfair to those who smoke. I have argued that many smokers are victims of social arrangements that confine them to highly constrained socioeconomic situations. These situations are highly correlated with smoking uptake, and since smoking (and/or just feeding an addiction) relieves stress, many smokers are doing nothing irrational or particularly worthy of condemnation. I have made the point that anti-smoker hiring practices promise to make many already disadvantaged smokers even worse off by erecting further barriers to scarce employment opportunities. I would rather see a more compassionate approach that actively helps smokers quit without making them worse off economically. (I do not doubt that Bill also wants to put compassion at the fore, although his last paragraph here does sound a bit harsh to me given that smoking is an addiction that most smokers acquire before they are considered wise enough to vote).

Against my view, Bill claims that the policies I favor amount to "leveling down": my preferred policy is said to sacrifice improvements in average and overall economic well-being for the sake of a narrower variance in social outcomes (assuming, that is, that non-smokers are more productive). Indeed, Bill argues that I favor less variance in social outcomes at the cost of both fewer economic and fewer health improvements, since an increasingly anti-smoker society would contribute to less smoking overall. Bill concedes that less inequality can sometimes justify a loss in aggregate or average outcomes, but not in this case. 

Bill hasn't got the dynamics of my argument quite right. I am not in favor of leveling down. I am not, for instance, in favor of intentionally harming the better off in order to reduce the inequality between them and those worse off. Nor am I arguing here for policies that harm the better off as a side-effect of improving the situation of those worse off (although I am in favor of other policies fitting this mold, as I assume Bill is). Each of these sorts of policy would bring the top down and would thereby decrease inequality. But I am advocating for neither. Rather, I am arguing that it is unjust to make the worst off even worse off than they already are. I do concede that this might lead to lower overall health or welfare, at least in the short to medium term. How much lower aggregate health of welfare? Who knows.

My argument type is, then, the third on this list:

1. Worsen the better off in order to reduce inequality.

2. Worsen the better off as a side-effect of policies designed to improve the worst off. 

3. Prevent worsening the worst off, even if that interferes with potential improvements for the better off.

In my view, (1) is the hardest to justify and (3) the easiest; I think most would agree with me on at least that. If the worst off referred to in (3) are indeed absolutely badly off for reasons of injustice, then we see that the issue is not as clear cut as simply "mean vs. variance." Mine isn't an aesthetic thesis about the virtue of flat distributions. Nor am I wholly insensitive to foregone improvements for the better off. Instead, I am focusing on a third factor: what we owe to those who already experience the brunt of injustice.

I am therefore against anti-smoker hiring policies for both reasons of compassion and reasons of justice. That does not mean that I think smoking should be "a lifestyle choice that we want to protect" (to quote Bill). I am fine with less draconian attempts to turn people off smoking, such as higher cigarette taxes (used in turn to fund smoking cessation programs) and even greater health care cost sharing for smokers. Erecting greater barriers to employment is, however, going too far in my opinion.

Health services research: After the money's gone

post by Bill Gardner

My recent post on the reduction of grantmaking by the Agency for Health Research and Quality must have hit a nerve, based on how frequently the link was retweeted. Many health services researchers make their living by winning research funding competitions. As a PhD, you cannot aspire to a tenured position in a school of medicine or public health without some victories. But you can't win a competition that isn't being held.

So if the money's almost gone, and you are an American health researcher, and you do not yet have a secure position, what should you do?

Consider working for  an accountable care organization. They need your skills. As Elliott Fisher, Mark McClelland, and Dana Safran write in the current NEJM,

Even integrated [ACOs], however, are finding that they need substantial time and resources to develop the informational, technical, financial, and professional capabilities required to provide and reward coordinated, longitudinal, population-based care. 

The emergence of the ACO is the best attempt yet by Americans to achieve the goals of population health. Implementing better care might be a better job than the one you thought you were training for.

When the high price of health care is a good thing

post by Paul Kelleher

If an [Implantable Cardiac Defibrillator] were a buck and a quarter, would we have gone through this entire process of reviewing the evidence at all? Maybe not. But because it's a bit more than a buck and a quarter we wanted to make sure the evidence was clear and that it was a benefit. We don't use cost to decide the evidence issue, but we do use cost to decide if this is important enough to address.

That is Dr Steve Phurrough, then-acting director of the Centers for Medicare and Medicaid Services Coverage and Analysis Group, speaking in 2003. I find this quotation shocking, and I wonder if others do as well.

I have no idea if things at CMS have changed since 2003, but this strikes me as illustrating one reason why it can be good for health care to be pricey. I do not claim that this reason makes it true, all things considered, that health care is so expensive. But its just not clear that we would be better off if health care grew on trees. 

The morality of health-care priority-setting

post by Paul Kelleher

I'd bet a tidy sum that the most cited thought experiment in contemporary moral and political philosophy--aside from Rawls's Original Position, perhaps--is T.M. Scanlon's Pious Aid Recipient example. Google Scholar says the source paper has only 310 citations, which I believe raises pressing questions about the reliability of Google Scholar. In any case, here it is:

The strength of a stranger's claim on us for aid in the fulfillment of some interest depends upon what that interest is and need not be proportional to the importance he attaches to it. The fact that someone would be willing to forego a decent diet in order to build a monument to his god does not mean that his claim on others for aid in his project has the same strength as a claim for aid in obtaining enough to eat (even assuming that the sacrifices required of others would be the same).

Scanlon states his point succinctly, but let me repeat it: to determine which duties we have to others, we should look at the claims they have on us, and not merely at the strengths of their preferences. Someone may want help building their religious monument more than they want to be adequately nourished, but my main duty is to respond to their claims, not to their preferences. This does not mean that there are no claims to have one's preferences satisfied, but note that this would still give priority to the underlying claims and not to the preference themselves.

Scanlon's thought experiment came to mind when reading this post by Leo Katz. Katz asks us to imagine Al, a pianist, who has a finger badly injured in a traffic accident. The only doctor around must choose between helping Al, or helping Bea, who faces paralysis in one of her legs if not helped immediately. Many would say that Bea should receive priority, since losing the use of a leg is in general worse than losing the use of a finger. Of course, to Al, this isn't so. But the doctor's question is who has the stronger claim on the doctor's help, and many would say that Bea does. (If you doubt this, ask yourself: should the doctor interrogate Bea to determine whether her leg is as important to her as Al's finger is to him?)

But there is one more piece to Katz's story. Just before making his decision about whom to help, the doctor sees that Al has another injury that will lead to paralysis in both his legs if not addressed immediately. Now the doctor must choose between three courses of action: (1) save Al's finger; (2) save Al's legs; or (3) save Bea's leg. Here is Katz:

How can he [the doctor] treat Al’s finger when Bea’s more serious leg injury needs attention? How can he treat Bea’s leg injury when Al’s far more serious leg injury isn’t getting treated? How can he treat Al’s leg injury when Al clamors to have his finger treated instead? We are in a sort of cycle.

How is it to be broken? At least one plausible way of breaking it, my own guess is that it is the most plausible, is to not let Al opt for getting his finger treated. He retains his priority over Bea only if he has is leg injury treated, not otherwise. But that means that we are rejecting a win-win transaction. For that is what the doctor’s treating Al’s finger rather than his legs surely is...The reason, in brief, is that Al has a strong claim on the doctor’s attention for his serious injury and a much weaker claim for his less serious injury.

I think Katz is correct to invoke the Scanlonian notion of claims here. I am less sure, however, that treating Al's finger really should be considered a "win-win" that creates the problematic "cycle" Katz describes. Fixing Al's finger is a win-win only if Bea is no longer in the picture. For if Al is the only relevant patient, then it might well be true that the doctor should help Al in whatever way Al prefers (unless fixing his finger is much more costly than fixing his legs). But Bea is never "out of the picture." It is Al's legs, and not Al or Al's preferences, that generate the claim that overrides Bea's claim. If Al's legs-based claim is ignored, then Bea's leg-based claim is the strongest claim remaining. Within the Scanlonian framework of claims, fixing Al's finger while Bea suffers paralysis in her legs is in no sense a "win-win."

The morality of claims and claims-adjudication is essential to the morality of priority-setting in the health care setting. And things can get philosophically hairy very quickly. I will be posting a good deal on these issues in the coming weeks.

Social microstructure and health behavior change

post by Bill Gardner

Fighting chronic illnesses and curbing rising health care costs will require that we change health behavior on a population scale. How do we do this? There is lots of smart work on how to use incentives and persuasive health communications to induce people to change. But we also need to understand how change diffuses through a population. 

The importance of the microstructure of your social network for health behavior change was demonstrated in a brilliant set of experiments by the sociologist Damon Centola, reported in Science. By all means, read the whole thing.

What Centola did was create artificial social networks. In each network, each participant is connected to 6 other participants. The network on the left consists of neighborhoods of 6 persons with many connections to each other, and fewer connections to others. On the right, people also have 6 connections, but they are distributed across the population so as not to create clusters. People in the left are close to a small group, but far away from everyone else. People are the right are closer to a randomly chosen population member, but they do not belong to a close group. 

Strangers were recruited to serve at nodes in the network. What they were asked to do was to decide whether to perform a health behavior, specifically to sign up for an online health forum. If a person decided to join, messages were sent to his 6 neighbors informing them of the decision. The question is, given that one person decides to join the health forum, in which network did the behavior spread faster? What he found was

Individual adoption was much more likely when participants received social reinforcement from multiple neighbors in the social network. The behavior spread farther and faster across clustered-lattice networks than across corresponding random networks... My findings show that, not only is individual adoption improved by reinforcing signals that come from clustered social ties but this individual-level effect also translates into a system-level phenomenon whereby large-scale diffusion can reach more people and spread more quickly in clustered networks than in random networks. 

So, changing health behavior is not just a matter of incentives and persuasion. These tools work better if their power is focused, when all your friends change together.

Centola suggests that public health programs target "clustered residential networks," but is this going to work if we are increasingly bowling alone? To my mind the obvious target is the workplace. Workplace wellness programs may work not just because the boss holds the levers, but because workplaces are often the locus of our most important social networks. But where will we be if workplace-based insurance ends, and these programs go away?

Not all gradients are created the same

post by Paul Kelleher

Earlier this week I contrasted a recent quote of Michael Marmot's about the source of the social gradient in health with findings from a 2010 study that conflict with his appraisal. To review, Marmot was quoted as saying that "[Lifestyle factors] accounted for about a third of the gradient." Here Marmot seems to refer to the gradient found in the original Whitehall Study (Whitehall Cohort I), where those at the top of the occupational hierarchy had the best health, while those in the occupations just below the top had the second-best health, and so on down the line. Yet 2010 research by Stringhini et al., which analyzed data from the follow-up Whitehall II study, seemed to show that when health behaviors are assessed more accurately, they can explain the vast majority of the gradient that Marmot found and attributed to other causes.

In a comment on that post, Daniel Goldberg drew my attention to a follow-up article by Stringhini et al. from February 2011. I'll explain in a moment why this second article will likely be very important to the research on the health gradient. But the thing that excited me most was that Marmot himself is listed as a co-author of this second article. This suggested to me that the conversation about the social gradient is unfolding as it should, with the original pioneer joining forces with the new iconoclasts to push the debate forward. That is a noteworthy development.*

From a research perspective, the February 2011 article is exciting because it claims to show that not all health gradients are created the same--that is, not all pronounced social gradients in health are caused by the same factors. As we have seen, Whitehall II appears to show that the gradient in health between individuals of different occupational status is mediated by a gradient in risky healthy behaviors. Those at the bottom of the occupational hierarchy engage in more risky behavior, and this explains 75% of the association between occupational position and all-cause mortality. After first reading Stringhini et al.'s report of these findings, I said that "the question" must now become, Why are risky health behaviors themselves socially patterned?

But the latest Stringhini et al. article may prove me wrong. That article also reports on a second study, the GAZEL study of French national gas and electricity employees from 1989 to around 2005. In GAZEL, there was a pronounced occupational gradient in health, with those in the lowest occupations having the worst health and those at the top having the best. In fact, GAZEL workers in the lowest status occupations actually had a 1.94 times higher risk of death than workers at the top, while in Whitehall II that hazard ratio was 1.62. And yet when the researchers crunched the data to determine the role that health behaviors played in creating the GAZEL gradient, they found that such factors could explain only 19% of it. This doesn't mean that the GAZEL cohort doesn't engage in risky behaviors. In fact they do, and to a higher degree than members of the Whitehall II cohort. It's just that everyone, from the top of the occupational hierarchy to bottom, engages in risky health behaviors to a similar degree in GAZEL. So something else must be causing the pronounced GAZEL gradient in health between the top and the bottom. But what? The authors offer a list of hypotheses that echoes the original list Marmot made when he thought the Whitehall gradient couldn't be explained by the distribution of health behaviors:

In this study, we show that, not surprisingly, health behaviours  contribute little to socioeconomic inequalities in mortality when  their social patterning is weak. Other factors are likely to play an  important role in the SES-mortality association in both cohorts.  Material deprivation or financial insecurity, work stress and work  environment, psychosocial factors such as job control or social  support, or differential access to health care may be other possible  mechanisms through which SES influences health [57–70].  Moreover, GAZEL includes both blue and white-collar workers;  it is therefore possible that physical occupational hazards and  working conditions contribute more to socioeconomic differences  in health in this cohort [67,71]. However, restricting the analysis in  GAZEL to only the white-collar workers did not substantially change the results. 

At this point I simply quote and register agreement with the authors' own driving motivation: " A better understanding of these mechanisms is essential in order to identify targets for intervention aimed at reducing social inequalities in health." Since not all gradients are created in the same way, we cannot assume that they will be eliminated in the same way.

*The follow-up article by Stringhini, et al. does not withdraw the claim that health behaviors account for the majority of the gradient found in Whitehall II. This leaves it curious why Marmot--a co-author on the new article--evidently reiterated to the reporter recently that lifestyle factors can explain only one-third of the gradient. I won't speculate as to what was going on there.

The Health Gradient and Health Behaviors

post by Paul Kelleher

In an article purporting to be on the adverse effects of income inequality, but which also spends a great deal of time on status inequality, Sir Michael Marmot is quoted as saying:

When Marmot and his colleagues controlled for lifestyle factors like smoking and lack of exercise in the lower socioeconomic groups, the gradient remained. "Those things accounted for about a third of the gradient," he says, noting that you have to look for the "causes of the causes" — the reasons that the lower classes might be driven to smoke, drink, take drugs or indulge in sweet, fatty foods. "It's now been described the world over," Marmot says.

Many who have read Marmot's work in the past know that it is often reported that the Whitehall Study of British civil servants "controlled for health behaviors," and that the health gradient remained. But follow-up research on Whitehall employees that was published last year seems to conflict with Marmot's original findings. And the conflict appears to trace to different (improved?) ways of assessing the impact of health behaviors on the gradient:

Design, Setting, and Participants Established in 1985, the British Whitehall II longitudinal cohort study includes 10 308 civil servants, aged 35 to 55 years, living in London, England. Analyses are based on 9590 men and women followed up for mortality until April 30, 2009. Socioeconomic position was derived from civil service employment grade (high, intermediate, and low) at baseline. Smoking, alcohol consumption, diet, and physical activity were assessed 4 times during the follow-up period.

Main Outcome Measures All-cause and cause-specific mortality.

Results A total of 654 participants died during the follow-up period. In the analyses adjusted for sex and year of birth, those with the lowest socioeconomic position had 1.60 times higher risk of death from all causes than those with the highest socioeconomic position (a rate difference of 1.94/1000 person-years). This association was attenuated by 42% (95% confidence interval [CI], 21%-94%) when health behaviors assessed at baseline were entered into the model and by 72% (95% CI, 42%-154%) when they were entered as time-dependent covariates. The corresponding attenuations were 29% (95% CI, 11%-54%) and 45% (95% CI, 24%-79%) for cardiovascular mortality and 61% (95% CI, 16%-425%) and 94% (95% CI, 35%-595%) for noncancer and noncardiovascular mortality. The difference between the baseline only and repeated assessments of health behaviors was mostly due to an increased explanatory power of diet (from 7% to 17% for all-cause mortality, respectively), physical activity (from 5% to 21% for all-cause mortality), and alcohol consumption (from 3% to 12% for all-cause mortality). The role of smoking, the strongest mediator in these analyses, did not change when using baseline or repeat assessments (from 32% to 35% for all-cause mortality).

Conclusion In a civil service population in London, England, there was an association between socioeconomic position and mortality that was substantially accounted for by adjustment for health behaviors, particularly when the behaviors were assessed repeatedly.

As Marmot notes, he always thought we need to ask why health behaviors themselves may be distributed in a graded fashion. But this new research seems to support the claim that this is the question. If that's true, it means we should focus much more on health behaviors than the summaries of the original Whitehall studies would have us do.

I would be grateful for pointers to additional research and/or commentary on these points.  

 

Corporate coercion and health behavior: Is it really coercion?

post by Bill Gardner

Yesterday, I summarized a meta-analysis that suggested that corporate wellness programs provide a positive return on investment by controlling employee's health care costs. Today I want to look at whether these programs are coercive.

Ezra Klein raised this concern in his post on the successful Cleveland Clinic employee wellness program:

[C]onsider what [changing health behaviors] actually entails: Changing habits. Breaking addictions. Getting people to the gym. Who wants to hear about any of that from their employer? The clinic, however, didn’t give employees a choice. “First thing we said was we had to make our institution toxin free,” [Cleveland Clinic Chief Wellness Officer] Roizen said. “The biggest toxin we have in the U.S. is tobacco. So we began offering free tobacco-cessation programs to our employees. Then we banned smoking on campus. You can’t even smoke in the parking lot in your car. The first offense you get a warning, and the second you get fired. We fired two high- profile physicians who refused to quit. Then they knew we were serious.”

Klein (see also here and here) noted that

Whether [the Cleveland Clinic's] success is a model for American health care or a preview of a dystopian surveillance state is an open question.

Let's step back, though, and ask what 'coercive' really means. Suppose someone confronts you in the dark with a pistol and says, "Your money or your life." This is coercion. Now suppose you are diabetic and need insulin, but your pharmacist won't give it to you unless you pay for it. In both cases, you are in life-threatening circumstances, unless you give someone money. But in most cases we feel that even though the pharmacist may be pressuring the diabetic, he isn't coercing her. The reason is that we view a threat with a gun as an illegitimate form of pressure, whereas barring dire emergency we believe that pharmacists have the same right to require payment as any other merchant. We distinguish coercive situations from legitimate pressures based on our moral judgments. 

So is what the Cleveland Clinic does coercive? Americans have complex views about what norms employers can legitimately require you to conform to, but in general your employer has wide scope to regulate your behavior. Your boss can require you to be punctual, if that is essential to the functioning of the business. But you can't be discriminated against on the basis of your religion. Your employer can, however, require you to meet a dress code -- unless it's a matter of dress or grooming required by your religion. And so forth. How do we feel about pressure to conform to norms of healthy behavior? My sense is that as people come to understand that unhealthy behavior is a significant business cost, they will increasingly accept employers' monitoring and enforcement of norms about healthy behavior.  

But although I believe Americans would accept Cleveland Clinic style carrots and sticks as part of voluntary employment contracts, they would not accept them if they were imposed by the government. Citizenship isn't voluntary in the way that employment is.