Join Monika Wahi for an in-depth discussion in this video Dealing with scientific plausibility, part of Healthcare Analytics: Regression in R.
- [Narrator] As we continue to design our research, I wanted to make a point that, in order to have a good hypotheses, we need to be able to address scientific plausibility. This time, we are going to use a different paper about BRFSS. By Wiener and Sambamoorthi, to demonstrate this point. That's because they base their analysis on kind of a famous theory that has scientific plausibility. I call it the icky mouth, icky heart theory. Weiner and Sambamoorthi published a paper in 2014 about the association between the number of missing teeth and cardiovascular disease among adults age 50 or older.
Why would they think someone's oral health has anything to do with someone's cardiovascular health? Let's try to figure out where they got their ideas to write this manuscript. In their abstract, they say, "The relationship between oral health "and cardiovascular disease is an "emerging area of research." They are right about this. If you scroll down through the Introduction, you'll find that some people believe in what I call the icky mouth, icky heart hypothesis, which is that if your mouth harbors certain bacteria, it affects your whole cardiovascular system and may be one cause of cardiovascular disease.
This explains why the authors are trying to quote, "understand the association between "number of missing teeth and risk "of cardiovascular disease," unquote, as missing teeth are a marker of poor oral health. So Weiner and Sambamoorthi wanted to see if poor oral health was associated with cardiovascular disease in pursuit of adding evidence to what I call the icky mouth, icky heart hypothesis. I like to use the term scientific plausibility, rather than the more common term, biologic plausibility because that term has come under fire in recent years.
Remember plausible is like the word possible. And who knows what is biologically plausible anyway? We don't know everything. Also, we know that there are social determinants of health. People who experience more racism have worse health. How do we come up with a biologic plausibility when the relationship might not be biologic? Also, early on it became clear people were just making up how they thought the biology worked. It's like, if you could tell a good story and have a good imagination, then you can come up with some biological way to relate the exposure and the outcome.
I say it's better to just look at the science and use the science, all the science, biolic, social, psychological, to make a case that the exposure and the outcome are connected. But I also wanted to point out that Weiner and Sambamoorthi have the same problem we have and every BRFSS researcher has. We have cross-sectional data so it's hard to gather evidence for causality. The reason I bring that up is, we are going to try to use strength of association and dose response association to provide evidence of causality when we do our analysis.
But they also point to using another piece of evidence and that is biologic plausibility. In other words, can the authors explain a biological way that poor oral health can actually cause cardiovascular disease? And can we apply this concept to our paper? In our paper, can we show a biological mechanism by which drinking can cause sleep duration to be affected? How about a biological mechanism between drinking and asthma? Sleep apnea has been linked to drinking and this can cause asthma.
So I think we will be able to make a case for a biologic plausibility with our hypotheses just as these authors do with theirs. So in this movie, I wanted you to recognize that you need to have a plausible explanation as to why your exposure and your disease are connected. Epidemiology has always leaned towards explaining a biologic plausibility but people found we could just make up anything and it wasn't helpful to be just making things up. But you do have to have a scientific plausibility as to how the exposure and outcome are connected.
These can come in the form of biologic plausibility as I demonstrated with the icky mouth, icky heart hypothesis or you can argue that they can be connected for social reasons or even psychological reasons. You just have to have an explanation of how you think they are connected and be able to support that with scientific literature.
- Dealing with scientific plausibility
- Selecting a hypothesis
- Interpreting diagnostic plots
- Working with indexes and model metadata
- Working with quartiles and ranking
- Making a working model
- Improving model fit
- Performing linear regression modeling
- Performing logistic regression modeling
- Performing forward stepwise regression
- Estimating parameters
- Interpreting an odds ratio
- Adding odds ratios to models
- Comparing nested models
- Presenting and interpreting the final model