Colleagues and I are advertising for research data analysts. If that’s you, this is an opportunity to work with us at the Partnered Evidence-based Policy Resource Center (PEPReC). Though PEPReC is a center in the Veterans Health Administration, the position will be filled through Boston University.
This isn’t just bragging. It’s objectively true. You see, my machine comes with an app that gamifies use. I automatically get scored every night on hours of use, mask seal, apnea/hypopnea events per hour, and mask removals.* I routinely get 98-100 points out of 100. The last time I scored below that was in the first week of use, when I got a shameful 96.
Killing it. (I also feel a lot better, which matters too.)
So, that’s not interesting. What is interesting is how variable the apnea/hypopnea index (AHI) is. Some nights I have literally zero events. Others, I have a few per night (or, like 0.3 per hour), with little discernible rhyme or reason. I don’t vary my diet (much), drinking (zero alcohol), other medication, bedtime, etc. Yet, my AHI varies. (I should point out that an AHI well below 1 per hour is off-the-charts awesome. Recall that 5 per hour is the threshold for mild sleep apnea. I suspect a lot of people I know without sleep apnea have higher AHIs than I do on the machine.)
It’s not just me. I’ve watched instructional videos of people demonstrating how to interpret CPAP data. They show their own data and their AHI varies by night too. I am sure this is fairly common.
Here’s my theory for the variation: our pathetic human airways are always very close to obstruction. The high prevalence of snoring is an indication of this. (Snoring is on the spectrum of sleep apnea.) Other evidence: It takes very little pressure to keep the airway open. I’m at about 6 cmH2O, which is like 0.09 PSI. It’s almost nothing. Even “high” CPAP pressures of 16 cmH2O isn’t much. Just imagine being half a foot under water and that’s the pressure. It hardly seems noticeable.
But it’s enough to push collapsing airways open. Same goes for mandibular advancement devices. They thrust the lower jaw out by a whopping half a centimeter or so, typically. That changes the airway topology ever so slightly, but it’s enough to keep it open a lot more.
With tolerance so tight, probably very slight changes in sleeping position or inflammation or breathing rate or whatever can tip one from zero AHI to a few points of AHI. That would explain the observed variability.
Why humans evolved to be so close to not being able to breathe at night is a good question. Got answers? Send them my way.
* Minor gripe: The mask removal scoring frustrates me, because it considers removing the mask at the end of the night a “removal.” You’re allowed only one more removal (like, to go to the bathroom) before your score gets lowered. It’s not enough! Sometimes one has to scratch one’s nose, chase a kid back to bed, or open a window. There are plenty of reasons to remove a mask more than once in a night. I would either allow three removals before point deductions or, better, not count the end-of-night removal as a removal. I mean, it’s not like you can avoid that removal.
Including Santa Fe High School, I count 32 deaths from school shootings so far in 2018. It’s a shocking number. Nevertheless, school shootings are an uncommon cause of death. What has received less attention are the high overall mortality rates for US children and youth. I want to examine these deaths and then comment on the light they shed on US population health.
Here’s why I think that too many US children die. In Health Affairs, Ashish Thakrar and colleagues analysed US infant (birth to the first birthday) and child (1-19 years old) mortality rates (hereafter, I’ll say pediatric to refer to both age groups). They compared these mortality rates to 19 peer countries* in the Organisation for Economic Co-operation and Development (OECD). Thakrar et al. estimated age-specific mortality rates from the pooled OECD data. They then applied these age-specific OECD mortality rates to the age-structure of the US pediatric population, for the years 1960-2010. This allowed them to calculate the number of deaths that would have occurred in the US during these years if, counterfactually, the US had had the mortality rates of the OECD. The counterfactual death counts based on the OECD mortality rates were lower which, conversely, meant that the US had an excess of pediatric deaths. Thakrar:
The United States has poorer child health outcomes than other wealthy nations despite greater per capita spending on health care for children… While child mortality progressively declined across all countries [from 1960 to 2010], mortality in the US has been higher than in peer nations since the 1980s. From 2001 to 2010 the risk of death in the US was 76 percent greater for infants and 57 percent greater for children ages 1–19.
Childhood mortality has been falling in the US and the OECD throughout this period. Epidemiologists saw this and celebrated the progress. What they didn’t notice was that, as the graph shows, mortality rates weren’t falling as fast in the US as in the rest of the OECD.
Changes in US and OECD infant and child mortality rates.
To appreciate what this difference means, it’s helpful to convert rates into counts. We’ll focus on the most recent period, from 2001 to 2010. US infants had a mortality rate of 68.8 deaths/10,000 infants during this decade, but the OECD infant mortality rate was only 39.0 deaths/10,000 infants.§ So,
Likewise, the US mortality rate for children aged 1-19 years, 3.1 deaths/10,000 children, was higher than the 2.0 deaths/10,000 children that would have occurred if the US had had the OECD child mortality rate. This means there were 1.1 Excess Child Deaths per 10,000 children. There are 76 million children aged 1-19 years during any given year, so
Excess US Post-Infant Deaths/Year = 0.00011 × 76 million Children 1-19
= 8,360 Excess Child Deaths/Year.
Excess Pediatric Deaths/Year = 11,920 Infants + 8,360 Children
= 20,280 Pediatric Deaths/Year.¶
This is greater than the number of AIDS deaths each year. It’s about half the number of motor vehicle deaths (for all ages). Hurricane Katrina caused 1,833 deaths. If we scale excess pediatric deaths in Katrina units, then the annual excess US pediatric deaths comprise just over 11 Katrinas.
However, the comparison to Katrina understates the problem. This is because the children who suffered excess deaths were about five years old when they died, on average. At that age, the Social Security Administration estimates that they had 74 years of life expectancy remaining. Call this Years of Life Lost/Death or YLL. The Katrina victims were on average 69 years old when they died. At that age, Social Security estimates that they lost 16 years of life.
If so, then the Total Years of Life Lost (TYLL) for Katrina was
If you want to recall one number from this, it might be that there are 1.5 million years of life lost each year due to excess pediatric deaths. Or, roughly, 70 Katrinas.
Now, how should we interpret these calculations? They are dependent on our choice of a benchmark and several assumptions. No one went out and determined whether a given dead child was an expected or excess death. You should attach wide error bars to these numbers. Nevertheless, the magnitude of the number of excess US pediatric deaths is huge and will remain large even if we substantially shrink these numbers to reflect our uncertainty.
There is no single cause for this. The US fell behind the OECD in reducing pediatric mortality rates in the 80s and has stayed there since, across many administrations, so there is no reason to make this a partisan issue. We should just study the causes of infant and child mortality and address them. Thakrar:
Policy interventions should focus on infants and on children ages 15–19, the two age groups with the greatest disparities [between the US and the OECD in cause-specific mortality rates], by addressing perinatal causes of death, automobile accidents, and assaults by firearm.
We could approach this as litigation for or against the United States, debating whether Americans have, collectively, neglected their children. But this makes little sense. Excess deaths are inferred, not witnessed. Let’s view this as an opportunity: We have discovered that each year there are 20,000 lives that could be saved.
*The OECD comparison countries were Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Iceland, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, and the United Kingdom. Canada, by the way, also performs worse than the OECD average, although better than the US.
§Unfortunately, the deaths of premature infants are counted differently in the US and some other countries. Some countries do not count the deaths of infants with very low gestational age (e.g., 24 weeks or earlier) as infant deaths, whereas the US does. This problem will tend to inflate the estimated number of excess infant deaths. Thakrar et al. acknowledge this problem, but they could not correct it using the available data. However, CDC analyses show that the US has higher infant mortality even after adjusting for this. There would be substantial numbers of US excess infant deaths, even if this problem could be corrected.
¶Many analysts argue that calculations of YLL should not use the raw life expectancy that a person had at the time of their premature death, but rather a life-expectancy that discounts the value of future years of life, conventionally at a rate of 3%/Year. For discount rate r and undiscounted life expectancy K, the formula for calculating this is
Discounted(YLL) = ∑ (1 + r)–k,
where the summation is from k = 0 to K – 1. Using this, Discounted YLL(Katrina) = 12.9 Years/Death. This means that
Discounted TYLL(Katrina) = 23,751 Years.
Carrying out a similar calculation for US children, we have
Discounted TYLL(US Pediatric) = 618,418 Years.
Then the annual total years of life lost due to excess US child deaths, in discounted years, is 26 Katrinas.
The following originally appeared on The Upshot (copyright 2018, The New York Times Company).
Heart disease is the most common killer of men in the United States, and high blood pressure is one of the greatest risk factors for heart disease. Despite knowing this for some time, we have had a hard time getting patients to comply with recommendations and medications.
A recent study shows that the means of communication may be as important as the message itself, maybe even more so. Also, it suggests that health care need not take place in a doctor’s office — or be provided by a physician — to be effective.
It might, as in this study, take place in a barbershop, an institution that has long played a significant social, economic and cultural role in African-American life. A setting that fosters both confidentiality and camaraderie seems like a good place to try reaching men to talk about hypertension.
Years ago, researchers ran an experiment in which they trained barbers to check blood pressure and refer people with high levels to physicians. One group received this intervention; a control group received pamphlets handed out by barbers. Blood pressure values were only minimally improved in the intervention group. This was thought to be because even when patients were referred to primary care physicians, those doctors rarely treated their condition appropriately.
The more recent study went further, removing physicians almost entirely from the process. The control group consisted of barbers who encouraged lifestyle modification or referred customers with high blood pressure to physicians. In the intervention group, barbers screened patients, then handed them off to pharmacists who met with customers in the barbershops. They treated patients with medications and lifestyle changes according to set protocols, then updated physicians on what they had done.
The results were impressive. Six months into the trial, systolic blood pressure (the higher of the two blood pressure measures) in the control group had dropped about 9 mm Hg (millimeters of mercury) to 145.4, which is still high.
In the intervention group, though, blood pressure had dropped 27 mm Hg to 125.8, which is close to “normal.” If we define the goal of blood pressure management to be less than 130/80, more than 63 percent of the intervention group achieved it, compared with less than 12 percent of the control group.
It gets better. The rate of cohort retention — measuring how many of the patients remained plugged into the study and care throughout the entire process — was 95 percent.
The barbershop customers were part of a population that is traditionally hard to reach. More than half of participants lived in households earning less than $50,000 a year, and more than 40 percent in households earning less than $25,000. On average, they were overweight or obese, about a third smoked, and more than a fifth had diabetes. Yet the improvement in blood pressure was more than three times that of the average of previous pharmacist-based interventions seeking to improve blood pressure, and many of those had focused on populations easier to reach.
One reason this trial succeeded where others failed is that it adapted its intervention to overcome barriers. When barbers weren’t consistently screening customers by measuring their blood pressure, pharmacists stepped in to do that. When labs slowed things down, pharmacists brought measuring tests to the barbershops.
The larger implications of this study shouldn’t be ignored. Getting barbers involved meant health messages came from trusted members of the community. Locating the intervention in barbershops meant patients could receive care without inconvenience, with peer support. Using pharmacists meant that care could be delivered more efficiently.
Of course, this study is limited by the usual sorts of questions. Who will pay for this in the real world? Who would do the training necessary to scale it up? Who would be responsible?
But those concerns reflect the shortcomings of our current health care system, not those of the study. Health care reimbursement in the United States usually focuses on the clinical encounter, at a physician office or hospital. This reflects a belief that care is best offered there, even when evidence says otherwise. Coverage and payment focus on the individual patient, not on the community, even when research shows that the latter is more effective. Care often requires the participation of a physician, even when studies prove that it can be delivered well in many cases by midlevel practitioners.
It’s important to remember that we have the health care system we do because of history and economics, not because of studies that show it’s optimally designed. Changes are most often made within the current framework; those that buck the system are usually met with more resistance.
If we really want to improve health on a large scale, especially with populations distrustful of the health care system, it seems we need to go to where they are; to use people they trust to deliver messages; and to allow care to occur without much of the infrastructure usually demanded for billing. Such efforts may not be traditional, but they may deliver much better results.
A judge in California recently ruled that coffee would be required to carry a carcinogen warning label, since it contains acrylamide. Well, have the barista make you a double espresso with a shot of evidence, because that coffee probably isn’t going to give you cancer.
This episode was adapted from a column I wrote for The Upshot. Links to sources can be found there.
This audio summary reviews a cohort study that uses Nurses’ Health Study data to investigate associations between diethylstilbestrol (DES) use in pregnancy and self-reported development of ADHD in grandchildren.
That hasn’t always been the case. America was in the realm of other countries in per-capita health spending through about 1980. Then it diverged.
It’s the same story with health spending as a fraction of gross domestic product. Likewise, life expectancy. In 1980, the U.S. was right in the middle of the pack of peer nations in life expectancy at birth. But by the mid-2000s, we were at the bottom of the pack.
Health spending and life expectancy are not necessarily closely related, so it’s helpful to consider them separately.
“Medical care is one of the less important determinants of life expectancy,” said Joseph Newhouse, a health economist at Harvard. “Socioeconomic status and other social factors exert larger influences on longevity.”
For spending, many experts point to differences in public policy on health care financing. “Other countries have been able to put limits on health care prices and spending” with government policies, said Paul Starr, professor of sociology and public affairs at Princeton. The United States has relied more on market forces, which have been less effective.
“Confronted with fiscal pressures, as the share of G.D.P. absorbed by health care spending began to get serious, other nations had mechanisms to hold down spending,” said Henry Aaron, a health economist with the Brookings Institution. “We didn’t.”
One result: Prices for health care goods and services are much higher in the United States. Gerard Anderson, a professor at Johns Hopkins and a lead author of a Health Affairs study on the subject, emphasized this point. “The differential between what the U.S. and other industrialized countries pay for prescriptions and for hospital and physician services continues to widen over time,” he said. Other studies also support this idea. However, by some measures, growth in the amount of health care consumed has also been a factor.
The degree of competition, or lack thereof, in the American health system plays a role. A recent study by economists at the University of Miami found that periods of rapid growth in U.S. health care spending coincide with rapid growth in markups of health care prices. This is what one would expect in markets with low levels of competition.
Although American health care markets are highly consolidated, which contributes to higher prices, there are also enough players to impose administrative drag. Rising administrative costs — like billing and price negotiations across many insures — may also explain part of the problem.
“We have big pharma vs. big insurance vs. big hospital networks, and the patient and employers and also the government end up paying the bills,” said Janet Currie, a Princeton health economist. Though we have some large public health care programs, they are not able to keep a lid on prices. Medicare, for example, is forbidden to negotiate as a whole for drug prices, as Ms. Currie pointed out.
But none of this explains the timing of the spending divergence. Why did it start around 1980?
Mr. Starr suggests that the high inflation of the late 1970s contributed to growth in health care spending, which other countries had more systems in place to control. Likewise, Mr. Cutler points to related economic events before 1980 as contributing factors. The oil price shocks of the 1970s hurt economic growth, straining countries’ ability to afford health care. “Thus, all across the world, one sees constraints on payment, technology, etc., in the 1970s and 1980s,” he said. The United States is not different in kind, only degree; our constraints were weaker.
Later on, once those spending constraints eased, “suppliers of medical inputs marketed very costly technological innovations with gusto,” Mr. Aaron said. They “found ready customers in hospitals, medical practices and other entities eager to keep up with rivals in the medical arms race.”
The last third of the 20th century or so was a fertile time for expensive health care innovation. Sherry Glied, an economist and a dean at New York University, offered a few examples: “Coronary artery bypass grafting took off in the mid-to late 1970s. Later, we saw innovations like drug treatments for H.I.V. and premature babies.”
These are all highly valuable, but they came at very high prices. This willingness to pay more has in turn made the United States an attractive market for innovation in health care.
Yet being an engine for innovation doesn’t necessarily translate into better outcomes. Almost no matter how it’s measured, longevity in the United States has not kept pace with that of other nations. Again, the inflection point is around 1980. Why?
Some have speculated that slower American life expectancy improvements are a result of a more diverse population. But Ms. Glied and Mr. Muennig found that life expectancy growth has been higher in minority groups in the United States. Another study, published in JAMA, found that even accounting for motor vehicle traffic crashes, firearm-related injuries and drug poisonings, the United States has higher mortality rates than comparably wealthy countries.
The lack of universal health coverage and less safety net support for low-income populations could have something to do with it, Ms. Glied speculated. “The most efficient way to improve population health is to focus on those at the bottom,” she said. “But we don’t do as much for them as other countries.”
The effectiveness of focusing on low-income populations is evident from large expansions of public health insurance for pregnant women and children in the 1980s. There were large reductions in child mortalityassociated with these expansions. “Those reductions were much larger for poor children than for richer children,” Ms. Currie said.
A report by RAND shows that in 1980 the United States spent 11 percent of its G.D.P. on social programs, excluding health care, while members of the European Union spent an average of about 15 percent. In 2011 the gap had widened to 16 percent versus 22 percent.
Although this is a modest divergence over time, Mr. Anderson says it could be significant nonetheless. “Social underfunding probably has more long-term implications than underinvestment in medical care,” he said. For example, “if the underspending is on early childhood education — one of the key socioeconomic determinants of health — then there are long-term implications.”
Slow income growth could also play a role because poorer health is associated with lower incomes. “It’s notable that, apart from the richest of Americans, income growth stagnated starting in the late 1970s,” Mr. Cutler said.
Even if we can’t fully explain why the United States diverged in terms of health care spending and outcomes after 1980, one thing is clear: History demonstrates that it is possible for the U.S. health system to perform on par with other wealthy countries. That doesn’t mean it’s a simple matter to return to international parity. A lot has changed in 40 years. What began as small gaps in performance are now yawning chasms. And, to the extent greater American health spending has spurred development of valuable health care technologies, we may not want to trade away all of our additional spending.
Nevertheless, Ashish Jha, a physician with the Harvard T.H. Chan School of Public Health and the director of the Harvard Global Health Institute, is hopeful: “For starters, we could have a lot more competition in health care. And government programs should often pay less than they do.” He added that if savings could be reaped from these approaches, and others — and reinvested in improving the welfare of lower-income Americans — we might close both the spending and longevity gaps.
Rear-facing car seats do keep kids safer, but they can be hard to use for larger children. In Sweden, kids stay in backward facing seats a lot longer, and have fewer injuries. The study that encouraged American parents to turn their car seats around turned out to be flawed. But that doesn’t mean it’s not a good idea!
I’m very proud of my Upshot post today, which explores a US health care mystery: our levels of spending and outcomes were among those of peer nations until the early 1980s. Then things went kaflooey (to put it technically). Why?
You can see it in the charts in the piece and read why Joe Newhouse, Paul Starr, Henry Aaron, Garard Anderson, Janet Currie, Sherry Glied, and Ashish Jha think things went sideways around then.
This piece was a ton of work, and I thank Stephanie Caty at the Harvard T.H. Chan School of Public Health for chasing down the data for the charts. Lest you think I was cherry picking the data, the two charts in the piece are far from the only ones that illustrate the phenomenon. Below is one more on outcomes, that appeared in Nicholas Kristof’s piece last summer. It shows that reductions in the rate of maternal deaths during childbirth ceased in the US in the 1980s, and then started to increase in the 1990s. That did not happen in Japan, Italy, Germany, Britain, Sweden, Ireland, or France (Canada has a pattern most similar to ours).
Below is another chart that makes one think about the mechanism. It shows that the growth in annual income of the top 1% took off in the 1980s. Other quantiles of income have either held steady or increased less rapidly.
Now, this is pure speculation, but maybe our health system caters to the wealthy. As their incomes grow, so does their demand for ever more expensive, high-tech care that is only marginally better than what came before. Political and social influence being what it is, they get it, but we all pay for it. The share of our economy going to health care grows. But outcomes for the vast majority of the population with lower incomes don’t improve as much, because more high-tech, expensive, low value health care isn’t what they need as badly as they need higher wages, better education, better housing — things provided by other social programs that the health care budget is consuming.
Again, pure speculation, suggested by the charts but not at all proved by them. Anyway, there are tons more charts that show things about the health care system went off the rails in the 1980s. I could do this all day. Go read my Upshot post.
Research for this piece and The Upshot post it promotes was supported by the Laura and John Arnold Foundation.
Austin and Aaron are participants in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to amazon.com.