Feeds:
Posts
Comments

Posts Tagged ‘correlational studies’

Here at Applied Behavioral Strategies, the mission is to improve the quality of life through effective intervention. One way we hope to do that is by reviewing research articles for our readers. Today’s article has actually been the topic of a lot of social media. See these headlines here and here and a reader actually wrote in about it on Monday.

Do pregnant women who get the flu or a fever actually increase the likelihood that their child will get autism? Let’s find out by actually reading the research.

The journal called Pediatrics published this study and Hjördis Ósk Atladóttir, Tine Brink Henriksen, Diana E. Schendel and Erik T. Parner authored the study. A quick search on Medline reveals that Dr. Atladóttir is chasing this topic of autism trying to find something to explain how it develops. He published a similar paper in 2010 in the Journal of Autism and Developmental Disorders. In that paper, he discussed pregnant women who had been hospitalized and later had a child with autism. He has published on cytokine levels and autism, patterns of contact with hospitals and autism, and family history of autoimmune disease and autism.

Purpose of the Study

The authors set out to “assess the association between self-reported common infections and autism in the child“. The authors clearly state that they estimated an association. Yet, when this study hit the news, reporters and scientists discussing the study omitted these little details.

Methods

The sample. The authors used an existing data base to gather their data (Danish National Birth Cohort). The authors selected 31% of the cohort for their data analysis.

The data collection. All the interview questions asked to the mothers occurred during the initial cohort recruitment completed by different researchers. These authors did not have contact with the mothers. Interestingly, the authors actually reported that “there was no specific question regarding respiratory disease and influenza”. It should make you wonder how they “estimated” the results of their highly disseminated “study”. In fact, the researchers actually asked the mothers, “did you take an antibiotic?” The authors clarified further, “The questionnaire did not include a question concerning the direct disease indication for the antibiotic use”. Wow! Yet all the media around this paper specifically said “flu”.

Data facts. Only 1% of the sample actually reported having the flu. Compare that to the percent of women with other issues: fever (24%), antibiotic use (19%), yeast infection (19%), cystitis (12%) and urinary tract infection (12%). Another interesting fact is that the researchers compared maternal responses during interviews with data from hospital records (e.g., diagnosis at discharge). The authors state, “The overall agreement between maternal reports of infection episodes and a corresponding hospital contact record was fairly good for most infections” (e.g., cystitis, pyelonephritis, and vaginal yeast infection). However, the authors also noted that “there was a very low agreement between maternal-reported infection and hospital-registered infection when the self-reported information was retrieved from open-ended questions” (e.g., flu). Thus, it seems that the likelihood the mothers really had the flu when they reported that they did, is actually quite low.

Data analysis. The authors used statistical analysis to determine if any relationships between the variables existed. What the media did not cover in reporting this study, is the important fact that the authors examined relationships between illnesses and any form of autism spectrum disorder as well as any relationship between diseases and infantile autism.

Results

The authors reported a number of results, most of which had no statistical significance. The authors noted that a statistically significant difference was found among mothers who self-report the flu (be sure to see the note above regarding the accuracy of reporting) and went on to have a child with autism. Specifically, out of the entire sample, only about 800 mothers reported having the flu. Of those, only 9 went on to be diagnosed with an autism spectrum disorder. This is hardly reason for alarm especially since we are having autism diagnosed at a rate of 1 in 86!

The authors noted that another statistically significant association was found between mothers who had a fever longer than 7 days. The number of women with a fever episode was quite high 23, 027). The number of them who went on to have a child with infantile autism was 101. Again, this hardly seems reason for alarm given the staggering rate of autism. Finally, the number of women who had a fever lasting longer than 7 days was 1361. Of those, only 14 went on to have a child diagnosed with infantile autism.

The authors found similar associations with antibiotic use. Again, the numbers are not alarming given the overwhelming rate of autism.

Discussion

The key statement in the discussion section should be highlighted: “There was little evidence that self-reported common infections during pregnancy are risk
factors for ASD in the child”

Can someone explain how the media complete twisted this in to a “flu during pregnancy increases the risk of autism” headline?

The authors did go on to talk about their previous work on this topic, ” We reported in our previous study that viral infection during the first trimester gave rise to an almost threefold increased risk of ASD“.

Side note: We all know that the flu is a virus. But isn’t the vaccine for the flu a live virus? Let’s see what the CDC has to say about it. Well, I’ll be darned, it appears that the nasal spray is a live virus. “Live, attenuated influenza vaccine (LAIV) contains live but attenuated (weakened) influenza virus. It is sprayed into the nostrils“. The CDC goes on to say that pregnant women should not take the live virus spray.

Other Thoughts

This study is full of methodological errors. Yet, Pediatrics continue to publish it and the media continue to twist the findings. Please, before you believe the “latest medical study”, you might find it more helpful to actually read the study rather than believe what someone tells you about the study.

The Elephant in the Room

So, if the researchers had access to all this data, why didn’t they ask better research questions? Why didn’t they look for associations between women who got the flu vaccine and still got the flu? Or how about this one: “does getting the flu shot increase the likelihood of your child getting autism?” There is so much more that could be asked, yet these researchers did not seem interested. Maybe it wasn’t the “politically incorrect” thing to do.

 

 

 

 

Read Full Post »

Hi! and welcome to What Works Wednesdays where historically a success story from clinical files is shared. With all the buzz about the latest “research” on getting the flu while pregnant and the supposed link to autism, it seems logical to help readers better understand research so they can interpret findings themselves. If readers know how to read research, then they are better able to know if an intervention works (or if the conclusions from a study are flawed or misinterpreted).

What is Research?

  • “work undertaken systematically to increase the stock of knowledge” (Wikipedia.org)
  • “diligent and systematic inquiry or investigation into a subject in order to discover or revise facts, theories, applications, etc.” (dictionary.com)

Most scientists conduct research by utilizing the scientific method. The scientific method requires the development of a hypothesis (which is usually formed from observation or reading other research), conducting the experiment, gathering results, and determining if the results support the original hypothesis.

Different Types of Research

Using the scientific method, scientists design different types of studies. These study types include:

  1. Experiments. In experimental studies, researchers recruit participants and assign them to treatment groups. Researchers can study one or more treatments and participants may receive some treatments or they may receive a placebo or no treatment at all. Usually, researchers measure one or more important variables before the study and they measure the variable(s) again after the study.
  2. Single Subject Experimental Studies. In these studies (most often conducted by behavior analysts), researchers recruit participants who are observed and measured carefully for a period of time before receiving treatment. Researchers then implement treatment while continuing to observe and measure carefully.
  3. Correlational studies. In these studies, researchers use existing data sets (e.g., collected for some other purpose) or they recruit participants. Researchers gather a wide range of information on each participant (e.g., age, SES, education, health history). Participants do not generally receive treatments or interventions of any kind.
  4. Qualitative studies. In qualitative studies, researchers occasional recruit participants but at times they enroll participants with whom they are already familiar. In qualitative studies, researchers study one or more individuals or one or more groups (e.g., one class). Researchers carefully study the participant and take copious notes. Researchers may interview the participants and they may use focus groups to better understand some of the issues. If a treatment is provided, the researcher continues to carefully study the participants to document the participants’ responses to the treatment.

Conclusions Based on Study Type

Researchers must use caution when drawing conclusions about their studies. Researchers who use well-designed experimental designs can draw cause-effect conclusions. For example, a researcher can enroll a bunch of smokers in a study. Some of the smokers receive a behavioral treatment, some of the smokers receive nicotine patches, and other participants receive both. At the end of the study (if the researchers have conducted the study carefully), the researchers will be able to say that one or more methods is successful at helping smokers quit.

Similarly, in a single subject experimental study, researchers can demonstrate if a treatment changes behavior. Again, the study must be carefully designed and conducted but it is possible to draw cause-effect conclusions. For example, a researcher could study 3 smokers. The researcher would observe the smokers and collect data. One smoker could receive treatment. While she is being studied, the other smokers would still be studied. After the first smoker quits successfully, the next smoker would receive treatment. He would continue to be studied as would the non-treated smoker. Finally, when the last smoker receives treatment, researchers continue to observe him. If the researchers successfully help all 3 participants quit smoking (and the study is carefully designed and carried out), they will be able to say that the treatment caused the behavior change.

Correlational versus Causal

Correlational studies are designed to determine if any relationships exist between variables. Researchers could gather data on 1,000 people from an existing data base. They could sort the data into smokers and non-smokers. They could run a simple data analysis to see if smokers have other tendencies (e.g., like to go to race car events, like to drink socially, and so forth). Researchers may not conclude causal relationships from their studies. They are only able to conclude that a relationship exists. Of more importance is the strength of the relationship. For example, if researchers ran an analysis on the relationship between giving birth to a child and gender, they would find a very strong (almost perfect) relationship between giving birth and being a female. If a weak relationship exists between variables it is more likely due to chance.

Go Forth and Read

In these days of social media, spin rooms, and media crazed talk shows, very poorly designed studies are being presented to the public without appropriate interpretation of the study or its results. If you are interested in reading a few examples of this, check previous posts here and here.

In summary, don’t believe everything you read about the “latest scientific study” unless you read the study itself. When you read the actual study, what you find may actually surprise you.

Read Full Post »

%d bloggers like this: