Tag Archives: health

Home birth is more dangerous. Discuss.

How dangerous is too dangerous?

We don’t prohibit all dangerous behavior, or even behavior that endangers others, including people’s own children.

Question: Is the limit of acceptable risks to which we may subject our own children determined by absolute risks or relative risks?

Case for consideration: Home birth.

Let’s say planning to have your birth at home doubles the risk of some serious complications. Does that mean no one should do it, or be allowed to do it? Other policy options: do nothing, discourage home birth, promote it, regulate it, or educate people about the risks and let them do what they want.

Here is the most recent result from a large study reported on the New York Times Well blog, which looks to me like it was done properly, from the American Journal of Obstetrics & Gynecology. Researchers analyzed about 2 million birth records of live, term (37-43 weeks), singleton, vertex (head-first) births, including 12,000 planned home births (that is, not including those where the home birth was accidental). They also excluded those at freestanding birthing centers.

The planned-home birth mothers were generally relatively privileged, more likely to be White and non-Hispanic, college-educated, married, and not having their first child. However, they were also more likely to be older than 34 and to have waited to see a doctor until their second trimester.

On three measures of birth outcomes, the home-birth infants were more likely to have bad results: low Apgar scores and neonatal seizures. Apgar is the standard for measuring an infant’s wellbeing within 5 minutes of birth, assessing breathing, heart rate, muscle tone, reflex irritability and circulation (blue skin). With up to 2 points on each indicator, the maximum score is 10, but 7 or more is considered normal and under 4 is serious trouble. Low scores are usually caused by some difficulty in the birth process, and babies with low scores usually require medical attention. The score is a good indicator of risk for infant mortality.

These are the unadjusted low-Apgar and seizure rates:

homebirthoutcomesThese are big differences considering the home birth mothers are usually healthier. In the subsequent analysis, the researchers controlled for parity, maternal age, race/ethnicity, education, gestational age at delivery, number of prenatal care visits, cigarette smoking during pregnancy, and medical/obstetric conditions. With those controls, the odds ratios were 1.9 for Apgar<4, 2.4 for Apgar<7, and 3.1 for seizures. Pretty big effects.

Two years  ago I wrote about a British study that found much higher rates of birth complications among home births when the mother was delivering her first child. This is my chart for their findings:

Again, those were the unadjusted rates, but the disparities held with a variety of important controls.

These birth complication rates are low by world historical standards. In New Delhi, India, in the 1980s 10% of 5-minute-olds had Apgar scores of 3 or less. So that’s many-times worse than American home births. On the other hand, a number of big European countries (Germany, France, Italy) have Apgar<7 rates of 1% or less, which is much better.

A large proportional increase on a low risk for a high-consequence event (like nuclear meltdown) can be very serious. A large absolute risk of a common low-consequence event (like having a hangover) can be completely acceptable. Birth complications are somewhere in between. But where?

Seems like a good topic for discussion, and having some real numbers helps. Let me know what you decide.

23 Comments

Filed under Uncategorized

Are White women high school dropouts getting sicker?

My Twitter feed lit up yesterday with this story about how life expectancy is falling for White women who have not finished high school. The story was called, “What’s Killing Poor White Women?“, by Monica Potts.

I have complete sympathy for poor people with health problems and high mortality rates. Things are killing them, and that’s bad. They should have better education, better jobs, better health care and more money.

White women without high school degrees have lost five years of life expectancy. Something must be getting worse. But I don’t quite think so. I could be wrong. But I think that as the category White women without high school degrees shrinks, it is the healthier people who are leaving (or never entering) the group. As a result, the group’s average health is declining.

The first thing to realize is that, according to the Census Bureau [spreadsheet link], 95% of non-Hispanic White women ages 25-29 have completed four years of high school or more. So we’re talking about a very (negatively) select population. And it’s getting more select – it was 92% 20 years ago. (Potts’s story revolves around a woman who died at 38.*)

The article doesn’t give any numbers to show that more people are dying, just that the life expectancy of the group has fallen. If this were a group, like race or gender, whose membership doesn’t change much over time, that would be enough to indicate their health status was getting worse. But an education group isn’t like that. It’s membership changes over time. Neither of the two academic articles Potts cites seem to consider this possibility (here and here).

One take

Here’s a try at it. Since 1996, the Current Population Survey has asked an excellent health status question, asking people to rate their own health as excellent, very good, good, fair, or poor. Let’s treat those whose health is “poor” as the group driving the mortality trend (which seems to fit the narrative in the story).

Here is the scary trend: A sharp rise in the proportion of non-Hispanic White women high school dropouts, ages 20-29, who rate their health as “poor.” (All the figures use three-year averages.)

poorhealthThat looks terrible, and it is, of course. But look at the size of the total group (all health statuses) over the same period:

dropoutsSo, the group has shrunk by about 18%, from about 850,000 to less than 700,000. And here is how the group’s population has changed according to health status, using the two endpoints of the trend, 1996-98 and 2010-12:

drophealthSo, there has been, in effect, no change in the number of non-Hispanic White women high school dropouts ages 20-29 in poor health, for the last decade and a half (the numbers shown are population estimates based on a sample size of only a few hundred women in this category per year, so I discount small shifts). In contrast, there has been a decline of those in good health. Result: the average health of the group has declined, but there are not more sick women.

That’s good news, because in Potts’s telling their problems are very serious, and something should be done about it.

*I (or you) could redo this to include more ages. I used young people because, if they have high mortality rates, they’re going to disappear from the sample at relatively young ages and make the group look healthier.

 

 

 

11 Comments

Filed under In the news

Life expectancy, life disparity

This is a serious post about life expectancy and inequality. But first a short rant.

Quick: Life expectancy in the U.S. is 78.7 Your parents are 85. How much longer are they expected to live? If you were worried about how much time you had left to spend with them, and you asked the helpful site seeyourfolks.com, you would get this:

seeyourfolksThis app, and the Slate piece about it, managed to combined two of my pet peeves: the understandable difficulty with understanding life expectancy, and the inexcusable use of second-person reporting on social science findings, which does more to discredit than to disseminate important research.

The error here (apart from “you”) is the common notion that “life expectancy” is the average age at which people of any current age can expect to die. If we were more rigorous about using the phrase “life expectancy at birth” it would be easier to grasp.

In 2008 the life expectancy at birth in the U.S. was 78.1. That means that if a group children born in 2008 lived every year of their lives exposed to the risks of death observed in 2008, their average lifespan would be 78.1 years. But those who made it to age 60 would live an average of 22.7 more years, for a total of 82.7. And those who live to age 99 would live an average of 2.4 more years, for an average of 101.4.

So “life expectancy” as commonly used is not a prediction of how long today’s babies will live — since we hope the future is better than living 2008 over and over — and it’s not a prediction of how long your elderly loved ones will live.

Life disparity

Life expectancy — for any age — is a measure of central tendency: the average number of years of life remaining. And so there is a dispersion around that mean. That dispersion is inequality. A very nice article in the open-access journal BMJ Open, by James Vaupel, Zhen Zhang and Alyson A van Raalte, describes the measure of life disparity. It’s complicated, but a neat tool.

Life disparity is the average number of years people are expected to live when they die. For example, in the U.S. in 2008 an infant who died on the first day of life died 78.1 years early. And a 78-year-old who died, counterintuitively, died 10 years early (since the life expectancy at 78 is 10). To understand what this measure means, consider that if everyone died at exactly 78.1 years of age, life expectancy would be unchanged but life disparity would be 0. On the other hand, the greatest life disparity would occur if all early occurred at age 0.

Life disparity and life expectancy usually go together. That’s because reducing early deaths has the biggest effect on both measures. Here is the cool figure from that paper:

The association between life disparity in a specific year and life expectancy in that year for males in 40 countries and regions, 1840–2009. The black triangle represents the USA in 2007; the USA had a male life expectancy 3.78 years lower than the international record in 2007 and a life disparity 2.8 years greater. The brown points denote years after 1950, the orange points 1900–1949 and the yellow points 1840–1900. The light blue triangles represent countries with the lowest life disparity but with a life expectancy below the international record in the specific year; the dark blue triangles indicate the life expectancy leaders in a given year, with life disparities greater than the most egalitarian country in that year. The black point at (0,0) marks countries with the lowest life disparity and the highest life expectancy. During the 170 years from 1840 to 2009, 89 holders of record life expectancy also enjoyed the lowest life disparity.

The association between life disparity in a specific year and life expectancy in that year for males in 40 countries and regions, 1840–2009. The black triangle represents the USA in 2007; the USA had a male life expectancy 3.78 years lower than the international record in 2007 and a life disparity 2.8 years greater. The brown points denote years after 1950, the orange points 1900–1949 and the yellow points 1840–1900. The light blue triangles represent countries with the lowest life disparity but with a life expectancy below the international record in the specific year; the dark blue triangles indicate the life expectancy leaders in a given year, with life disparities greater than the most egalitarian country in that year. The black point at (0,0) marks countries with the lowest life disparity and the highest life expectancy. During the 170 years from 1840 to 2009, 89 holders of record life expectancy also enjoyed the lowest life disparity.

Countries at the bottom left (0,0) have both the world’s highest life expectancy and the lowest life disparity in the world for that year, which occurred 89 times over 170 years. Countries below the diagonal have relatively low life disparity given their life expectancy; those above the diagonal (like the U.S.) have higher-than-expected life disparity for their level of life expectancy. In our case that reflects the fact that we do a pretty good job keeping old people alive, but let too many young people die.

U.S. improvement

The good news is that life expectancy is increasing in the U.S. (and most other places), and that the inequality between Blacks and Whites is getting smaller, as reported by the National Center for Health Statistics. That is, the Black-White inequality in average expectation of life at birth has shrunk.

The mixed news is that life disparity is much higher for Blacks than Whites — but that gap is falling as well. Here are those numbers for 1998 and 2008 (I did the life disparity calculations from this and this, and will happily share the spreadsheet). Click to enlarge:

expectancydisparity

So Black deaths are more dispersed than White deaths: 14 and 13 for males and females, compared with 12 and 11. For comparison, the Swedish female life disparity is 9. What does a higher disparity mean? Generally, a larger share of early deaths. That’s why the race gap in life expectancy at birth is greater than the race gap in life expectancy at older ages — average 65-year-old Whites and Blacks have more similar life expectancies than do infants.

Why is life disparity more interesting than life expectancy alone, and how does this help explain Black-White inequality in the U.S.? For one thing, high life disparity indicates either relatively unhealthy or dangerous living conditions at younger ages. So it’s partly a measure of the quality of life. Vaupel et al. add:

Reducing early-life disparities helps people plan their less-uncertain lifetimes. A higher likelihood of surviving to old age makes savings more worthwhile, raises the value of individual and public investments in education and training, and increases the prevalence of long-term relationships. Hence, healthy longevity is a prime driver of a country’s wealth and well-being. While some degree of income inequality might create incentives to work harder, premature deaths bring little benefit and impose major costs. Moreover, equity in the capability to maintain good health is central to any larger concept of societal justice.

I think what they say about differences between countries would apply to differences between groups within a society as well.

2 Comments

Filed under Uncategorized

Off topic: My 5-year cancerversary

This is not what this blog is about.

I didn’t even register it right away. Five years ago this Memorial Day I got my diagnosis of follicular lymphoma, a form of non-Hodgkin’s lymphoma. It was late on the Friday afternoon when the surgeon called with the biopsy results. He never said the word “cancer,” but recommended I see an oncologist. He was a very nice guy, and told me I was going to live to be an old man. Within 15 minutes I had read that follicular lymphoma is usually incurable. (The UpToDate database I used now puts it this way: “most cases of follicular lymphoma are not curable with currently available therapies.”) It was a long long weekend.

Usually follicular lymphoma – a blood cancer – is advanced before it’s first discovered. In the next few weeks, one oncologist told me the median survival was between 10 and 20 years. I was 40 with a wife and 4-year-old daughter. I asked her why she was an oncologist. She said she was interested in end-of-life issues. Also, the nicest people get cancer.

Eventually we determined that I had what apparently was a rare case of Stage I, which may be curable. I had 18 days of painless radiation and didn’t (physically) miss a day of work. Lucky is a funny word for this.

Five years later I don’t have an oncologist anymore. It’s the first line on my medical chart but not a to-do list item. When we moved away, my Bayesian-minded oncologist wrote in his farewell note, using his best handwriting: “Your chance for cure is reasonable: pre-test probability is low. Early detection is not helpful. If you get an enlarged lymph node, get biopsied.” Maybe that’s oncology speak for: “Relax, good luck!”

pretest-probability-is-low

Anyway, there were lots of people I never told, including the chair of my department and some good friends and colleagues. Maybe that’s because it went from incurable (yikes, too much information) to possibly-cured (so stop complaining already) so quickly – before the start of the new semester – so I didn’t know how to bring it up or what to say.

For most people with this disease, the story is different. Thankfully, we’ve had a revolution in lymphoma treatment, and it’s usually a very long story. Most people live many years, and I’m told the new treatments usually aren’t that bad. (Easy for me to say.) Chance of surviving (that is, dying from something else) is pretty good. Experts debate whether the word “cure” should be used more.

Meanwhile, now there are two kinds of people in the world: people with a better prognosis, and people with a worse prognosis. Of course that’s always been true. But this experience sometimes makes me dwell on that, which increases my tendency to draw a sharp resentment/sympathy line according to this criterion. That isn’t healthy because it obscures the more important bases upon which to relentlessly judge people and compare myself to them.

seesawline

I’m writing this because I remembered how lonely and scared I felt back then – when I didn’t even know where on the scale to put myself. Nothing aggravates the modern identity like incalculable risk. Fortunately, I had the greatest family and friend support – and medical care – anyone could ask for. Life got back to normal. We adopted another daughter. There are other risks to worry about.

But I’m thinking that somewhere someone with no idea what to do next is getting news like I did and Googling “follicular lymphoma.” If that’s someone you know, or it is you, maybe it will help to know about one more person who’s still living about as normal a life as I was before. Feel free to drop me a note.

11 Comments

Filed under Me @ work

People with disabilities are more likely to get divorced

File this under things to look into about divorce.

There was a recent paper showing that people who experience the onset of a disability face an increased likelihood of divorce, but that’s about all I found in a quick search. Now that we have the giant American Community Survey, which has both disability status and marital events data, we can ask the simple question: In a given year, are people with a reported disability more likely to report they have been divorced in the previous year? The answer is yes.

Age is a tricky issue with disability, since some risks of disability are cumulative over the life course. To do this quickly I just limited this to people ages 18-49. Otherwise the disability group is dominated by older people who have been married a long time, and who have low divorce rates. Here it is, by type of reported disability for the pooled 2009-2011 ACS:

disability-divorceThose are pretty big effects (odds ratios from 1.4 to 1.9). Over a lifetime these odds would really add up.

Economists would tell you that when a spouse experiences the onset of disability, this is new information for the other spouse, and increases his or her chance of leaving the marriage, since the disability implies a decline in future income. Maybe. But what about people who have disabilities already when they get married, which is presumably the case for most of these people. Is having a difficult life a cause of divorce? Is this related to economic stress, or carework obligations (I checked and found not much gender difference, but men’s disability has slightly stronger effects).

If you are interested in this question, don’t let me stop you from pursuing it – send me your results!

13 Comments

Filed under Uncategorized

Yo, how big is that yogurt bucket?

People don’t know how much they’re eating.

A recent experiment found that people eat more when the container is larger, even when the portion size is not. They gave Belgian college students a container of M&Ms and parked them in front of a TV, with some cover story. The students were randomly assigned to three groups, medium-portion/small-container, medium-portion/large-container, and large-portion/large-container. These were the results: The ones who got the large container ate more, whether it was full or not (the difference between the two wasn’t significant). These kinds of experiments continuously suggest that distractions, distortions and other apparently irrelevant information and events routinely have large effects on people’s eating practices (here’s an extensive review). One infamous study showed that even people served 14-day-stale popcorn at the movies ate 34% more when it was served in a large container. In an earlier popcorn study, researchers found that people given large containers not only ate more, but were less able to report how much they ate. They concluded:

When a food is eaten from a large container, it appears easy to lose track of how much one eats. Even if the food were to taste relatively unfavorable, eating it from a large container may cause one to overeat because they lose track of how much they have consumed.

About that yogurt tub All this occurred to me when I visited one of our many local Frozenyo franchise outlets. It’s a self-serve frozen yogurt place where you pay one price by weight no matter what you put in your bucket. The trick that impressed me is the bucket — there is only one size, and it’s very large. But you can’t judge how big it is because there’s nothing to compare it with — no sizes or prices on the wall, no mini cup for kids — just one stack of identical buckets. So the person who posted this picture on Yelp probably thought she had a reasonable size serving, since the thing is barely half full:

There are three possible ways to judge your self-served serving size. You can go by the tub (“I filled it half way”), you can go by the person next to you (“sheesh!”), or you can look at the cartoon penguins on the wall:

How much is the penguin eating? I took home one of the buckets, and measured the volume of water it holds: 18 ounces. In comparison, a standard kid-sized serving bowl, the kind some people use to give their kids ice cream at home, holds 12 ounces:

An innocent child used to half a bowl of ice cream — in the bowl on the left — might be pretty steamed if you served her this:

According to the serving size information on the back wall of Frozenyo, I think that’s about 1.5 servings, or 150 calories of the nonfat variety, before toppings. The penguin’s overflowing bowl is 5.0 servings. With no toppings that’s 500 calories. If you pile it with M&Ms, sprinkles, hot fudge, Captain Crunch, coconut topping and fresh kiwis, who knows. It’s not really that many calories to consume — the same number as a single slice of banana bread at Starbucks.

But the point is you don’t know how much you’re eating. One Yelp reviewer cautioned that you can get a stomach ache after eating at Frozenyo, because “your eyes are bigger than your stomach.” I think it’s because the dump-truck sized delivery vehicle you eat it out of is bigger than your stomach.

But most reviewers love it for the individual control over serving size and toppings, and the reasonable price ($.39 per ounce by weight, or $5-$6 for a typical load).* I think it’s a winning business model, with low labor costs, because all you need is one person to pour the mix into the machines and another to weigh the tubs and swipe credit cards. According to the company’s ambitious map, there are still 46 states with “territory available.”

If I were them, I would increase the bucket size by 5% per year. I doubt anyone would notice.

* Paging George Ritzer: it’s the irrationality of rationality.

Leave a comment

Filed under Me @ work, Research reports

Incarceration’s contribution to infant mortality

A recent study in the journal Social Problems by sociologist Chistopher Wildeman shows that America’s practice of mass incarceration may be exacerbating both infant mortality in general and stubborn racial inequality in infant mortality in particular.

Drawing on recent literature by himself and others, Wildeman spells out the case for incarceration’s negative effect on family economies, including: lost earnings and financial contributions from fathers, the expensive burden of maintaining the relationship with an incarcerated parent, and the lost value of the incarcerated parent’s unpaid labor. All of those costs may take a toll on mothers’ health, which is the primary cause of infant mortality.

In addition, family members of incarcerated parents may contract infectious diseases, experience significant stress, and lose support networks — all taking an additional health toll.

Sure enough, his analysis of data from the Pregnancy Risk Assessment Monitoring System confirms that children born into families in which a parent has been incarcerated are more likely to die in the first year of life. The association may not be causal, but it holds with a lot of important control variables.

Does this increase racial inequality? Probably, because parental incarceration is so concentrated among Black families, as Wildeman and Bruce Western reported previously (my graph of their numbers):

To make the connection to racial inequality explicit, Wildeman moves to compare states over time, on the suspicion that incarceration could increase infant mortality rates, and racial inequality in infant mortality rates. That could be because concentrated incarceration undermines community support and income, people with felony records often are disenfranchised (so the political system can ignore their needs), and the costs of incarceration crowd out more beneficial spending that could improve community health.

The results of a lot of fancy statistical models comparing states show that:

the imprisonment rate is positively and significantly associated with the total infant mortality rate, the black infant mortality rate, and the black-white gap in the infant mortality rate.

It’s an impressive article on an important subject, one that thankfully is attracting more attention from good scholars.

I previously reported on Wildeman’s work on how the drug war affect families, here.

5 Comments

Filed under Research reports

The disparate lives of fifth graders

A new study of about 5,000 fifth-grade students in the three public school districts shows wide disparities by race/ethnicity in a number of important health practices and outcome measures. The study, published in the New England Journal of Medicine, showed unadjusted disparities and then attempted to account for them statistically with common control variables, such as family socioeconomic status and school characteristics.

Here is a breakdown of some of the indicators (my graph):

On all but alcohol consumption (remember these are fifth graders), the white students showed advantages over Black and Latino students. In the subsequent analysis, the authors showed what amount of the disparity was accounted for by the different control variables. Here is their graph illustrating the findings:

It shows, for example, that about 10 points out of the 20-point difference between Latinos and Whites on the frequency of reporting fair or poor health is accounted for by their control variables. For Black children, about four points out of the eight point difference is accounted for. (These gaps would likely be larger if private school students were included.)

Determining the causal story behind these disparities is interesting and important, however it is most important to realize that at the descriptive level these represent major disparities in the lived experience of young children who are blameless.

It is interesting to note that some of these practices and outcomes speak to parenting practices, which has been the subject of a growing literature in recent years. However, after Annette Lareau reported that parenting practices in her study differed more by social class than they did by race, class has been the focus of much of this research. For example, although I did not see it, a study by Jessica McCrory Calarco at Indiana University, presented at the annual meeting of the American Sociological Association last week, looks very interesting. She used observation and interviews and found stark differences between middle-class and working-class parent-child interactions. From the press release:

Working-class parents, she found, coached their children on how to avoid problems, often through finding a solution on their own and by being polite and deferential to authority figures. Middle-class parents, on the other hand, were more likely to encourage their kids to ask questions or ask for help.

Leave a comment

Filed under Research reports

Study shows home births are not as safe. So?

There’s an interesting example of how to interpret scientific results — and draw policy implications from them — from the world of birth practices and safety.

The subject of the debate is a major new study from the British Medical Journal. The study followed more than 60,000 women in England with uncomplicated pregnancies, excluding those who had planned caesarean sections and caesarean sections before the start of labor. They compared the number of bad outcomes — from death to broken clavicles — for women depending on where they had their births.

One comparison stands out in the results. From the abstract: “For nulliparous women [those having their first birth], the odds of the primary outcome [that is, any of the negative events] were higher for planned home births” than among those planned for delivery in obstetric units. That is, the home births had higher rates of negative events. The difference is large. Here’s a figure to illustrate:

The error bars show 95% confidence intervals, so you can see the difference between home births and obstetric-unit births is statistically significant at that level. These are the raw comparisons, but the home-versus-obstetric comparison was unchanged when the analysts controlled for age, ethnicity, understanding of English, marital or partner status, body mass index, “deprivation score,” previous pregnancies, and weeks of gestation. Further, by restricting the comparison to uncomplicated pregnancies and excluded all but last-minute c-sections, it seems to be a very strong result.

But what to make of it?

In their conclusion, the authors write:

Our results support a policy of offering healthy nulliparous and multiparous women with low risk pregnancies a choice of birth setting. Adverse perinatal outcomes are uncommon in all settings, while interventions during labour and birth are much less common for births planned in non-obstetric unit settings. For nulliparous women, there is some evidence that planning birth at home is associated with a higher risk of an adverse perinatal outcome.

But in what way do the results “support a policy”? The “higher risks” they found for planned home births are still “uncommon,” by comparison, with those in poor countries, for example. But the home birth risk is 2.7-times greater.

The Skeptical OB, who is a reliable proponent of modern medical births, titled her post, “It’s official: homebirth increases the risk of death.” She added some tables from the supplemental material, showing the type of negative events and conditions that occurred. Her conclusion:

“In other words, any way you choose to look at it, no matter how carefully you slice and dice the data, there is simply no getting around the fact that homebirth increases the risk of perinatal death and brain damage.”

I guess the policy options might include include whether home births should be encouraged, more regulated, covered by public and/or private health insurance, banned, penalized or (further) stigmatized.

Home birth seems safer than letting children ride around unrestrained in the back of pickup trucks, which is legal in North Carolina — as long as they’re engaged in agricultural labor. On the other hand, we have helmet laws for kids on bicycles in many places. And if a child is injured in either situation, hopefully an ambulance would take them to the hospital even if the accident were preventable.

In other words, I don’t think policy questions can be resolved by a comparison of risks, however rigorous.

8 Comments

Filed under In the news, Research reports

Warning: What do smokers Google?

If I ran the Federal scary anti-smoking image warning program, I might show smokers the list of health-related terms that show up most in the states with the highest cigarette smoking rates.

The Google Correlate tool is showing the great potential for using Internet search activity to investigate layers of behavior and meaning behind other observable social phenomena, such as race/ethnic composition, health behavior, and family patterns. Today’s example is smoking.

If you take the smoking rates by state, and throw them into the Google Correlate hopper, you can see the 100 search terms that are most highly correlated with that reported smoking behavior. That is, the terms that are most likely to be used in high-smoking states and least likely to be used in the low-smoking states.

Is the result just a lot of noise? Maybe, but I don’t think so. Here are the smoking-related terms in the top 100:

  • camel no 9
  • cigarette coupon
  • cigarette coupons
  • marlboro coupons
  • my time to quit
  • safe cigarettes
  • stopping smoking
  • time to quit
  • fire safe cigarettes
  • ways to stop smoking

So that’s good for face validity — a list of random search terms isn’t likely to have all those smoking terms on it.

But after the smoking terms, the thing that jumps out is the health-related terms. We know from the Google flu tracker that people search for their symptoms. So these caught my eye.

Here is a screen shot of the first page of results:

I selected “stages of copd” as the term to map. The map on the left is the smoking rates; the one on the right is the relative frequency of searches for “stages of copd.” That is, chronic obstructive pulmonary disease, a nasty disease the most common cause of which is smoking.

Here is the complete list of health-related terms among the top-100 correlates with smoking rates:

Lymph node swelling, which is implicated in the jaw and neck searches, most often reflects infection — which smoking causes.

How strong are the connections? They’re not the strongest I’ve seen on Google Correlate. The “stages of copd” search is correlated with smoking rates at .77 on a scale of 0 to 1. It’s not uncommon to find correlations of .93 (which is the relationship between “quiche” and “volvo v70 xc”).

But considering the smoking rates come from a sample survey (the National Survey on Drug Use and Health) which includes random error, and states are somewhat arbitrary geographic units, that correlation seems pretty high to me. Here’s the scatterplot:

What is the correlation causality story here? I can’t say. But the simplest explanation is that these are the terms smokers (and maybe those who know or care for them) are most likely to Google relative to non-smokers — not that they are the most common searches smokers do, of course, but the searches that differentiate them from non-smokers. The simplest explanation is the best place to start.

I like this list of conditions because in my experience smokers sometimes have the attitude of “you have to die of something.” But it’s not just the chance of dying that smoking increases – it’s a lot of possible forms of suffering along the way.

8 Comments

Filed under Me @ work