My research question is whether social factors or biological factors are a better predictor of mental health. This is inspired by the different perspectives of psychology, and specifically about debating the perspective between biopsychology and social psychology. I grouped the data into biological factors (gender, body mass index, how many sedentary minutes they have a day, whether they have high blood pressure or cholesterol, whether they have anemia, whether they are overweight, and how many hours of sleep they receive each night) and social factors (education, family income to poverty ratio, race, self image, monthly poverty index, and the number of hours they work). The population of the survey is adults over the age of 20 living in the U.S., and I used the data collected between the years 2013-2020. The result was that the models trained on social factors had a significantly higher accuracy than the models trained on biological factors. This suggests that the problem with mental health is not based on the environment around us individually, but the social environment imposed on us by the people around us. Rather than look at each individual separately, we should take into account the entire population as a whole.
My research question is determining whether social factors, like their education, family income to poverty ratio, race, self image, monthly poverty index, and the number of hours they work, or biological factors, like their gender, body mass index, how many sedentary minutes they have a day, whether they have high blood pressure or cholesterol, whether they have anemia, whether they are overweight, and how many hours of sleep they receive each night, are a better predictor of their overall mental health. My hypothesis is that there is a difference between the two factors, and one set of factors will be better at predicting one's mental health than the other set.
This is important to see because it allows us to narrow down on a set of factors and focus our resources and attention that would make the most impact on people. It would allow the process to be more time efficient as well as more cost efficient for the overall population, and serve as a starting point. The data is from the National Health and Nutrition Examination Survey from 2013-2020, conducted by the Centers of Disease Control and Prevention on people living in the United States.
The overall accuracy of the models trained on social factors is higher than the models trained on biological factors. This is important to note, because it suggests that the root of mental health problems comes from the structure of society in the United States. It supports my original hypothesis, that models trained on different sets of factors will produce different accuracy percentages. However, a caveat is that though models trained on social factors were more accurate, the accuracy in which they predicted whether someone had a mental health issue was similar to the other models. They were only slightly better at predicting whether if someone didn't have any mental health issues.
However, it is important to mention that this is merely a suggestion, and does not determine a causal relationship between social factors and mental health issues. The method I used, of training different models to determine the accuracy, is also flawed in that these models do not conclude a definite relationship between the two factors. The data is also flawed because it only accounts for people interviewed between 2013-2020 and older than 21. Also, the way that I defined a mental health issue was how they answered the question, "[Over the last 2 weeks, how often have you been bothered by the following problems:] feeling down, depressed, or hopeless?" Of course, two weeks is in no way a long enough period of time to determine the extent of one's mental health, and has the potential to be impacted by extraneous factors. Another notable factor missing from biological factors is their family history of depression, which would've added a new insight into an individual.
The next logical step of this research is to investigate the best combination of both biological factors and social factors that would lend to the highest accuracy score of its trained models.
You can see the code here: https://github.com/the-codingschool/DSRP-2023-DiDonato/tree/main/Laurel
All the data comes from the CDC website:
-
2017-2020: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?Cycle=2017-2020
-
2015-2016: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2015
-
2013-2014: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx?BeginYear=2013
I would like to acknowledge Yijia Wang and AJ DiDonato for providing feedback on my project, as well as Sarah Parker for teaching me all I know about R right now and Delaney Rice for being the organizer of this program.