How often do you see headlines like ‘Dogs May Help You Live Longer’ , as if you needed another reason to have one, ‘Lack of Sleep May Shrink Your Brain’ , ‘ Tall People More Prone to Cancer’ , or ‘Eating brown rice reduces the risk of diabetes’ ? Do you understand the relationship between the mentioned factors? Do you accept the headlines as established truths?
It may surprise you to know, dear reader, that in the media we are more likely to find content derived from observational studies than from randomized controlled trials , which are generally always more accurate and reliable. Most likely, these two terms sound Chinese to you. Let me explain.
“Life is a mistery
The scientific method , the way of asking questions and establishing strategies to find the truth that confirm and reject our hypotheses, or that help us formulate new questions to continue expanding our body of knowledge, is only the best method we have to approach the truth. It is not infallible, but it is the most reliable .
When we pose a question, for example, ‘does doing puzzles help prevent Alzheimer’s?’ , we need to consider how we design a study that can give us an answer. We could, for example, search for people with Alzheimer’s and ask them if they did puzzles and how many. This has its limitations, of course. For example, many will not remember, others will exaggerate or not be very precise in their answers. Therefore, this would not be a good study design to answer this question. Another option would be to select people who do puzzles and people who don’t, follow them for 10 years and see if more cases of Alzheimer’s develop in one group than in another. This approach has other limitations, for example, it is more expensive, perhaps 10 years is not enough, and we would have to differentiate which differences between groups are significant and not due to chance. The two mentioned approaches are types of observational studies (cases and controls and cohorts) . As you can see, they are not ideal, but many more options cannot always be proposed to get closer to answering some questions.
a better design
There are questions that give us more possibilities. For example, if we ask if taking a multivitamin supplement can reduce urine infections in women between 50 and 60 years old without other diseases, we could make two groups of people in this age range, one group we would give the supplement and another group we would give him a placebo with the same external appearance. After 2 years, for example, we would assess whether the supplement group had significantly fewer urinary tract infections. In this case, we, the researchers, are actively introducing an element. In the previous puzzle cases, we were selecting people who already do, or don’t do, puzzles. In this case, we are the ones who are modifying elements in people’s lives to determine their impact on our question. This is a clinical trial. It often serves to assess new treatments. For example, will this new antihypertensive be better for controlling blood pressure, or the one that has been used up to now? and more secure? What about omega-3 supplements vs eating fish 3 times a week? Etc.
The responsibility of the media and… of the population?
Some time ago, Harvard researchers examined studies echoed by several major US media. Next, they examined the trials covered in the top five magazines in the country. You won’t be surprised to learn that they found that the media were more likely to cover observational studies and less likely to cover randomized controlled trials (well-designed clinical trials) than high-impact scientific journals. That is, what is important to the media, and often to the population they serve, is often not as important to the scientific community. Furthermore, when the media covers observational studies, they choose articles of lower quality . Newspapers mainly cover medical research with a weaker methodology.
The authors also recognize that much of what newspapers and other media choose to publish is based on reader interest. However, they add that when there is a significant difference between observational and randomized controlled trials, the media has no justification for not correcting this practice. Regardless of the topics covered by the media, they must be able to understand the difference between the two .
Let’s explain it in more detail.
Correlation vs Causality
Science is often concerned with measuring the relationship between two or more factors. For example, scientists may want to know whether solving logic puzzles reduces the risk of dementia or whether frequent television viewing has a negative effect on vision.
Correlation is when two factors are related, but one is not necessarily the cause of the other.
Causality is when one factor causes another (for example, heat causes burns on our skin).
In the case of television, research may find that people who spend a lot of time watching television are more likely to develop vision problems. In that case, we can conclude that watching TV frequently causes vision problems. However, it may also be the case that people who watch a lot of television are people whose work involves several hours of computer use. Therefore, we can say that there is a correlation between watching television and vision problems, but we do not know if watching television is the cause of vision problems.
In the case of dementia, research may reveal that people who often solve logic puzzles have a lower risk of developing dementia. Therefore, we conclude that the frequent resolution of logical tests causes improvements in brain functions. However, there may be people with a high IQ in the study. Therefore, we can say that there is a correlation between test resolution and a lower risk of dementia, but we do not know if it is the cause of the improvement in brain function.
It is easier to show a correlation between two things than that one causes the other . When changes in one factor cause a change in another factor, it is described as a causal relationship. Correlation and causation are not the same thing . Sometimes two things can be related without one causing the other.
For example, the more the police are called, the more likely it is that thieves will rob a bank . In this case, the bank robbery is not the result of calling the police. Both variables (number of police officers and bank robbery) were caused by the presence of robbers.
Although there is a causal relationship between variables, it can sometimes be difficult to determine which variable causes another to change. For example, there may be a correlation between people’s moods and their physical health. But it is not clear which variable affects the other. Does a good state of mind improve physical health, or does good physical health improve people’s state of mind?
Certain types of studies can show us a cause and effect relationship between two things. Other types can only help us find correlations. For example, randomized controlled trials can provide good evidence of a cause and effect relationship. Cross-sectional studies , like point surveys , cannot do this.
In the media, article titles often completely misrepresent the research they are based on. Many headlines suggest cause and effect relationships . However, reading the article more carefully, we will discover that the research was correlational, and in that case, the article does not justify the title. When you read health research, you need to distinguish between correlations and causation, and understand which of these two things the research demonstrates. I am aware that not everyone can do this, and you don’t have to be a science expert to discern what information to trust or not. The important thing is that you maintain a certain degree of skepticism, that when in doubt about the veracity of any data your skepticism is even greater, and that you choose your sources of information well, based on their rigor and credibility (beyond the numbers on your audience). Let’s not fall into the mistake of thinking that “if it’s on TV” or “if an expert from Germany says so”, it has to be true.
References:
Bhandari, P. (2021). Correlation vs. Causation | Differences, Designs & Examples. Scribbr. https://www.scribbr.com/methodology/correlation-vs-causation/
Correlation and causation. (s.f.). Understanding Health Research. https://www.understandinghealthresearch.org/useful-information/correlation-and-causation-15?fbclid=IwAR0h3eeH54Ht9rtntzwkE6ZyecZgwxL_6trtJaO4kUMIIbS-447bzWzKJfY
Correlation of Physical Sciences. (1856). The Buffalo medical journal and monthly review of medical and surgical science, 12(1), 34–36.
Correlation versus causation: The media typically gets it wrong. (2014). American Council on Science and Health. https://www.acsh.org/news/2014/01/22/correlation-versus-causation-media-typically-gets-wrong?fbclid=IwAR3HarTRJsH1j5gH1XS1BUvY0v0ZNMglIb6skWnhp1QJMWQT31-4nkFtJms
Papana A. (2021). Connectivity Analysis for Multivariate Time Series: Correlation vs. Causality. Entropy (Basel, Switzerland), 23(12), 1570. https://doi.org/10.3390/e23121570
Statistical Language – Correlation and Causation (s.f.). Australian Bureau of Statistics.. https://www.abs.gov.au/websitedbs/D3310114.nsf/home/statistical+language+-+correlation+and+causation
