You know the old saw: “If you don’t like the the weather in New England, just wait a minute.”
We tend to see our rapid shifts in weather as a benign local quirk, but there’s a darker side to them as well, and it may grow as the climate shifts: Heat waves and cold snaps are linked to little spikes in death rates, and those spikes add up to long-term effects, according to a paper just out in the journal Nature.
It charted temperature and death rates among the Medicare population for all of New England, zip-code by zip-code, from 2000 to 2008. Among its findings: It’s not just heat or cold that kills, it’s the sudden shifts in weather.
I asked Joel D. Schwartz, professor of environmental health at the Harvard Chan School of Public Health and senior author on the paper, how he would sum up its findings. Our conversation, lightly edited:
J.S.: There are hundreds of studies that have shown that when it gets hot, more people die, and when it gets cold, more people die, on that day or in the next couple of days. But these are all studies that look at what happens in a day or two; they’re not telling you anything about the long-term effects of temperature on people’s life expectancy. Maybe all the people who die when it gets hot would have died in the next month anyway — they’re sick and something is going to cause them to die in the near future.
What we did is a cohort study: We took all the people in New England who are on Medicare — about 2.9 million people — and we followed them over time. And we asked the question: Does their annual mortality rate change when weather changes?
This way, we avoid having to worry about whether these are really just the short-term effects, and we can address the question: Is there really an impact on life expectancy of temperature? This is the first study to do that in a general population study; we studied all of New England.
The second new thing is that all of the studies heretofore that have looked at the effects of temperature on the risk of dying have been done in cities, because a) that’s where the weather stations are, and b) in smaller towns there aren’t enough people to really be able to see anything. But we used satellite remote sensing and we calibrated it, and were able to get the temperature for every day for every square kilometer of New England — that’s about 6/10 of a mile on a side.
We had a separate measurement of temperature for every day for every zip-code from 2000 to 2008. And so we could calculate, then, for each person, the mean temperature in the summer for each year and the mean temperature in the winter, over multiple years. And we could ask, ‘Well, if mean temperature is higher one summer than the previous year in a particular zip-code, were people in that zip-code more likely to die?’ And the same for winter temperature.
What we found was that indeed, if the temperature in the summer was higher, the annual mortality rate went up. If the temperature in winter was higher, the annual mortality rate went down. The summer effect was bigger than the winter effect, so if it went up by the same number of degrees in the summer and the winter, then more people would die.
The next thing we did is we asked the question, ‘Well, since we have this fine geographic scale of temperature, did temperature differences across zip-codes affect mortality rates?’ And what we found was that there was lower mortality rates in zip-codes that had warmer temperatures in the winter.
So even within zip-codes in New England, if you lived in a zip-code with a warmer winter, mortality rates were lower in that zip-code – and it wasn’t that richer people live in warmer zip-codes because we controlled for income and education and value of housing and things like that. This was really pretty solid.
On the other hand, we didn’t see that for the summer effects. Instead, if this year’s summer temperature was higher than average for your zip-code, then more people died. If it was lower than average, then fewer people died. So whatever the average was in your zip-code, whenever you went above or below it, the mortality rate went up or down.
CG: So it’s not just the heat, it’s the variability?
That’s right. That’s the last part: Year-to-year variability really mattered, but also within the year. I told you what happened if the mean summer and winter temperature changed, but we also captured the variation within the winter and within the summer mean for each year. If the variation in temperature went up, more people died, and that was true for winter and summer.
So while warmer winters reduce mortality rates, more variability from day to day in the temperature within the winter increases mortality rates. People do not adapt well to changes in temperature. And so if it gets more variable, more people die.
What’s the size of this effect? How to bring it home, how many more people die?
I think the best way to say it is that a one-degree increase of the standard deviation of temperature– I’m afraid i have to use that term! — was associated with about a 1.3 percent increase in deaths in the summer and over a 4% increase in deaths in the winter.
If you think about it — AIDS is responsible for 1 percent of the deaths in the United States; something that increases total deaths by 1 percent is something to be concerned about.
To get more concrete, let’s say that you live in a zip-code where the summer temperature normally varies from about 70 to 90. Can you spell out how hot it would have to get to bring this several-percentage increase in deaths?
If 70 to 90 encompassed the range of, say, 95 percent of the range of the days in the summer — there were a few extreme days hotter and colder, but 95 percent of the days were between 70 and 90 — that would mean that the standard deviation of the temperature would be about 5 degrees. So if the standard deviation was typically 5 degrees in a typical summer, and this summer was more variable in temperature and the standard deviation went up to 7 degrees, then that would be associated with 1.3 percent more deaths that year. Not just deaths that summer but in the annual deaths.
What could be the physiological mechanisms that would be responding like this to weather changes?
We’ve done studies in a cohort that I’ve been studying for a long time, called the Normative Aging Study, which is a cohort of people living in the greater Boston area, and what we find with these people is that temperature can change their blood pressure; that it can increase inflammatory markers in their blood that might increase risk of having a heart attack or some other cardiovascular disease; that it affects how regular their heartbeat is; that it affects their lung function. So there are a lot of physiological changes that temperature causes. Now, people can adapt — people who live in Hong Kong have adapted to living in a hotter place than Boston — but they don’t adapt right away. It’s a process that actually takes about a month, to fully physiologically adapt.
There was a study done of exercise tolerance, in which they had two groups of people exercise. One group exercised every day in a room where it was hot, and they had to exercise until they were exhausted. They did it for two weeks, and each day they could go for longer and longer as they were adapting to that heat. The other group — one day would be hot and one day it would be comfortable, and then it would be hot again and then it would be comfortable again, and their performance on the hot days never changed, it never got better. So if the climate keeps bouncing around, going from 90 one day to 70 and then back up to 87 — that is what puts the most stress on people’s bodies because they just don’t have time to adapt to the new temperature before it changes again.
So what are the implications for the effects of climate change?
The implication is that understanding what climate change is going to do to the within-season variability of temperature is at least as important as understanding what it’s going to do the mean temperature in terms of human health impact.
Does climate change affect within-season variability of temperature?
Probably, but it’s easier to have any model predict the mean than predict the variability, and that’s true for climate models also. I think that the climate scientists would probably say we’re not sure, and I would say that it’s time to put some effort into nailing that down because that turns out to be really important.
Readers, thoughts? Have you noticed how the New England changes in weather affect you and your health? Further reading: