Do hospitals experience a larger number of patient admissions to the emergency room and/or labor and delivery during full moons? My nurse friend claims that this is a fact.
Brian, 34, San Ramon, California
When there’s a full moon, hospitalization rates do not increase (or decrease for that matter). That pretty definitive conclusion is based on several studies I’ve read this week, all of which tested the hypothesis that the moon affects our health.
Most studies have exonerated the moon. Academics have ruled out an effect of the lunar cycle on psychosis, depression and anxiety; violent behavior and aggression; seizures; suicide; absenteeism rates; coronary failure; in vitro fertilization conception; birth; menstruation; surgery and survival of breast cancer; postoperative outcome (general); renal colic; outpatient admissions (general); and automobile accidents — to name just some of the studies noted in a 2008 literature review.
Digging into that vast literature, I saw how numbers can be misused and misunderstood, and learned a bit about selenology — that’s the scientific study of Selene, aka the Earth’s moon.
To answer your specific question (and to answer a second question of my own: “How can people get the numbers about the moon so wrong?”), I’m using a study published in the May/June 2015 edition of the journal Nursing Research. The paper was written by Jean-Luc Margot, a professor at UCLA who specializes in the formation and the evolution of planets. It turns out, Brian, you have something in common with Margot. When I asked how he had come to this topic, Margot explained that a friend of his who is a midwife told him, swore to it, that there are a larger number of births when there is a full moon.
Margot could find only one study that might be used to support his friend’s claim — a 2004 study by academics in Spain who found a connection between hospital admissions for gastrointestinal bleeding and the moon’s cycle. Margot’s subsequent paper is all about the deep flaws he found in that Spanish research — criticisms that I hope will be of interest to you, too, Brian.
As a professor teaching astronomy, Margot spotted something I definitely wouldn’t have — the authors of the Spanish study (who worked in a Barcelona hospital’s gastroenterology department) had misunderstood the moon.
The researchers classified each day of hospital admission data as either a “full moon day” or a “non full moon day.” To do that, they assumed that the moon worked on a 24-hour basis and over a 29-day cycle. But the moon is a natural phenomenon whose cycle occurs independently of legislated time-zone changes or any perfectly consistent monthly pattern. As Margot points out, a more accurate way to measure the moon’s effect would have been to look at the time in minutes between a hospital admission and the previous or next full moon. Tricky and time-consuming, but far more precise.
There was another problem with the study (one that wasn’t noted by Margot but one I come across all the time): small sample sizes. The Spanish researchers based their analysis on 447 hospital admissions between 1996 and 1998. That’s just not very many data points to start off with.
The study also seemed to confuse a pattern with a meaningful trend. Imagine for a moment, Brian, that you have a sack of balls numbered one to a thousand. Each day you pull a ball out of the bag and note the number written on it. If you were to pick five even-numbered balls in a row, I doubt you’d start looking to the sky for answers — you’d probably (rightly) conclude that the result can be explained by randomness.
Statisticians can use tests that show the range of outcomes you might expect if chance alone were at play. Margot did, and concluded that the patterns the Spanish academics found and attributed to the moon could quite easily be attributed to randomness.
Then there’s the problem of interpreting meaning. Just because hospital admissions vary as the lunar cycle changes doesn’t necessarily mean they do so because of the moon. Let’s say (completely hypothetically) that there was a sudden surge in hospital admissions on Dec. 25, 2007. Would you be more inclined to put the high number down to the fact that 0.96 of the moon was illuminated that night or to Christmas holiday celebrations?
A series of numbers will almost inevitably have some kind of pattern in it — the real trick is figuring out whether that pattern is attributable to the thing you’re investigating (in this case the moon). Margot found that the Spanish hospital data didn’t control for other things — like, for example, the day of the week — that could have affected the fluctuation of hospitalization rates.
Taken together, those research flaws led Margot to reach a conclusion that I can’t help but agree with: “The moon is innocent.” Some of those errors were based on the authors’ limited knowledge of the specific topic at hand. And far be it for a journalist like me to criticize that — better to acknowledge our other limitations as imperfect thinkers.
One example is something called confirmation bias — that’s when we pay better attention to information that confirms our hypotheses. Maybe your friend is less likely to notice a full moon when she hasn’t treated many patients (and maybe she’s less likely to look for it and less likely to remember seeing it). As Margot explained over the phone, “we tend to ignore data points that don’t match our beliefs.”
And the less we know about a topic, the more likely we are to start out with bad hypotheses. Even answering a question as basic as “What is the moon?” isn’t easy, as these presenters on a home shopping channel demonstrate (before you guffaw at them, it turns out that there’s debate within the scientific community about whether the moon is a planet or a satellite).
Here’s the good news: We can try to correct our biases by being more systematic about how we collect and analyze data. The next time your nurse friend has a slow day at work, tell her to take the time to look up at the sky. She might start to notice there’s a full moon on those nights, too.
Hope the numbers help,
CLARIFICATION (June 11, 3:45 p.m.): As two astute readers pointed out, the hypothetical example we used in an earlier version of this story to illustrate the concept of randomness was in fact extremely unlikely to occur by chance (a one in (1/10)^30 chance). We’ve updated the story with a scenario you’d be more likely to conclude is random: pulling five even-numbered balls out of a sack of 1,000 in a row (a 3.1 percent chance).