Introduction
In recent years, many concerns have risen about the dangers behind continuous and addictive phone usage. High schools across the US have begun banning phones in school as a result of the continued debate about technology in the classroom. One of the largest concerns about phone usage is its relationship with mental health. In spite of this debate, I wanted to study the potential effects of phone usage and mental health, which extends to sleep quality and anxiety or depression levels. Additionally, Lab’s ongoing talk about banning phones in schools is why I wanted to research this topic. I hypothesize that phone usage negatively affects mental health and sleep quality and that there is a correlation between phone usage and physiological and behavioral factors.
Dataset Background
My analysis draws on two datasets: a University of Maryland social media and mental health dataset and a Dartmouth StudentLife dataset. Starting with the University of Maryland, this dataset includes variables on demographic, health, and mental information of students from 48 different states in the US. The data spans from 1971 to 2003, and while this data spans further back, the study still uses interesting mental health technology, like the PHQ-9 Depression Screening Tool and the GAD-7 Anxiety Assessment Pool. There are around 120 columns and 579 observations, so there are many data points to analyze with the different mental health screening technologies.
The other dataset I used was the Dartmouth StudentLife dataset, which is a massive, interactive dataset that measures many different variables about students at Dartmouth, their mental health, and many other metrics about their lives. There are many different sectors of this dataset, but some of the metrics included are behavioral data, mental health data, academic performance, smartphone and wearable sensor data, social interactions, physical activity data, sleep patterns, mood, and stress levels. This 2013 study spans a 10-week term at Dartmouth University and involves a large population. For metrics of depression and anxiety, they used tools like the PHQ-9 Depression Scale and the Perceived Stress Scale (PSS). Because the dataset is so big, it was difficult to manage on my computer and in RStudio, so I analyzed the “test” dataset that was included in the source files of the dataset. Additionally, because the dataset is interactive, I’m able to pick particular sub-datasets within this larger study, analyze them individually, and then merge them to discover hidden relationships.
Analysis
This first visualization from the Dartmouth dataset examines the relationship between locked and unlocked phone states, revealing a stark contrast between the usage patterns. The dataset provided me with a phone lock sub-dataset that included a start and end time for when students’ phones were locked. Since I was given the start and end time for lock time, I mutated the variables and did the math to calculate the time when the phone was unlocked. The data shows approximately \(6\times 10^6\) minutes of unlocked phone time compared to the \(2\times 10^6\) minutes of locked phone time. We observe a pretty significant difference between the time spent on phones versus not on phones, so hypothesis testing is not necessarily needed. However, for the remaining graphs, the interpretation of the p-values help us determine whether the relationship between variables are statistically significant. When the phone is unlocked, that most likely means you’re actively using the phone, especially since this study takes place at Dartmouth University where students are studying for the majority of the time. With the increased use of their phones, a higher level of stress can be induced by procrastination of work, comparison with others, or any other stress-inducing factors. This increased stress can be affected and affect many other factors of mental health as well, so phone usage is crucial to the rest of the arguments and factors analyzed in this paper.
This visualization shows the PHQ-9 (Patient Health Questionnaire-9) Depression Scores over time from the PHQ-9 Depression Tool. These depression scores are calculated by assigning number values to each of the answers from the 9 questions of the depression test and then adding up individuals’ total depression scores. The higher the depression score, the higher you’re considered depressed. Beginning in 1970, depression scores showed a steady increase through 1995 when it reached a maximum medium score of 11. Following this peak, a slight decline occurred, followed by a stabilization at a medium score of 9 at the end of 2005. The increase in depression scores can be attributed to the advancements made in society during the rise of the Third Industrial Revolution. With the rise of advanced technology and materials, the US also encountered many economic and political problems alongside new government policies. The slight dip in depression scores in 1995-2000 can be attributed to the increased attention to mental health around this time. However, even with this small dip, the depression scores slightly increased in 2000-2005. With this overall increase from 1970-2005 from a median depression score of 2 to 9, it is safe to assume that these depression scores continue to increase. After 2005, the US experienced even more economic troubles, like the 2008 Stock Market Crash, and concerns about technology taking the place of jobs, which still happen to this day, resulting in higher depression levels among the population.
Now, I aimed to see whether the increase of depression scores over time is statistically significant. My null hypothesis was that the depression scores from 1970-1975 are equal than those from 2000-2005, while my alternative hypothesis was that depression scores from 1970-1975 are less than those from 2000-2005. To test this, I conducted a Two-Sample (One-Sided) T-Test, which returned a p-value of 0.01856. Since this p-value is less than 0.05, we can reject the null hypothesis in favor of the alternative hypothesis, indicating a statistically significant difference between the time periods and affirming that depression scores from 1970-1975 are lower than those from 2000-2005.
##
## Welch Two Sample t-test
##
## data: group_1970_1975 and group_2000_2005
## t = -2.9546, df = 4.4004, p-value = 0.01856
## alternative hypothesis: true difference in means is less than 0
## 95 percent confidence interval:
## -Inf -1.680944
## sample estimates:
## mean of x mean of y
## 4.000000 9.661088Along with depression levels, stress levels are another significant factor in mental health. This visualization examines the relationship between social activity and stress levels. Social activity serves as a good measure of mental health, as isolation is often a sign of mental health issues. The data shows that when you’re stressed the most, your social activity levels are the lowest (approximately 1.2), and conversely when you’re feeling great, your social activity levels are at the highest(approximately 1.7). This suggests that higher levels of social engagement may help reduce stress levels, while individuals experiencing less stress may feel more inclined to participate in social activities. Connecting to the phone usage data from the first visualization, which shows that people tend to spend more time actively using their phones than having them locked, we can draw an important association: increased phone usage may lead to social isolation, which in turn contributes to higher stress levels.
I wanted to determine whether there is a relationship between social activity and stress levels. My null hypothesis was that there is no relationship between the social activity levels and stress levels (meaning the mean social activity would be the same across all stress level categories), while my alternative hypothesis was that there is a significant relationship between the social activity and stress levels. From this, I ran an ANOVA test to understand the relationship across the means of all stress levels, and it returned a p-value of 0.00521. Since this p-value is less than 0.05, we can reject the null hypothesis in favor of the alternative hypothesis, supporting a significant relationship. The data shows that social activity levels decrease as stress increases, dropping from 1.7 when “Feeling Great” to 1.2 when “Definitely Stressed.” Moreover, this reinforces the idea that phone-induced isolation increases stress.
## Df Sum Sq Mean Sq F value Pr(>F)
## factor(level) 4 9.1 2.2852 3.708 0.00521 **
## Residuals 1385 853.5 0.6162
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1After analyzing stress, phone usage, depression, and social activity, the last variable to analyze is sleep. This visualization explores the relationship between sleep quality and depression levels, using the PHQ-9 depression score metric, as discussed before. The relationship between sleep and depression doesn’t follow a simple linear pattern. The sleep rating is measured by participants rating their sleep on a scale of 1-4, where 1 is very bad and 4 is very good. In the line plot, for mild depression scores (0-5), we see a sleep rating of 1.9. This rises to a maximum peak of approximately 2.1 for scores in the (5-10) range. However, there’s a large decline in sleep quality for those scoring (10-15), dropping to a sleep rating of 1.6. Following this, there’s a temporary improvement in sleep ratings for the (15-20) score range, rising back to nearly 2.0, followed by the absolute minimum sleep rating score of approximately 1.3 with the highest depression scores (20-25). This visualization suggests that while individuals with mild and minimal depressive symptoms maintain relatively normal sleep quality, those with higher depression scores experience poorer sleep quality. Connecting this back to earlier findings about phone usage, higher phone usage, particularly during evening hours, could contribute to poor sleep quality, which in turn may increase depressive symptoms. This relationship also shows that individuals with higher depression scores tend to spend more time on their phones, enforcing the ideas of isolation and how that contributes to higher levels of phone usage, stress, and anxiety.
I wanted to explore whether there is a relationship between sleep ratings and depression scores. My null hypothesis was that there is no relationship between sleep ratings and depression scores (meaning the mean sleep ratings would be equal across all depression score groups), while my alternative hypothesis was that there is a relationship between sleep ratings and depression scores. From this, I ran an ANOVA test, which returned a p-value of 0.0906. Since this p-value is greater than 0.05, we cannot reject the null hypothesis, meaning there is not a statistically significant relationship between sleep ratings and depression scores from this data. While there is not a statistically significant relationship, we can still see a relationship between other mental health factors and a relatively low p-value.
## Df Sum Sq Mean Sq F value Pr(>F)
## phqGroup 4 4.42 1.105 2.085 0.0906 .
## Residuals 79 41.87 0.530
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Conclusion
Overall, this analysis supports the hypothesis that increased phone usage is related to negative mental health factors, like sleep, stress, and social activity, and that by decreasing phone usage, you can improve factors like sleep quality and mental well-being. The Dartmouth dataset highlighted a significant imbalance in phone usage, showing that people spend a lot more time on their phones than not. Then, with the statistical testing of depression scores over time with scores from 1970-1975 being significantly lower than those from 2000-2005, the adoption of technology in today’s society demonstrates an increase in depression signals. Similarly, a significant relationship came up between stress levels and social activity, where higher stress levels correspond to reduced social interaction. Lastly, while the connection between the sleep ratings and depression scores did not meet the statistical significance levels, the observed trend still shows declining sleep quality from higher depression levels.
With the imbalance between active and inactive phone states, increased depression scores, stress, and social activity levels, and sleep ratings, these trends suggest a relationship between phone usage and mental well-being. For future analysis, I would like to explore the causal relationships between the factors analyzed in this paper and performance in school. Specifically, I’m interested in modeling students’ performance in classes based on their personal habits and health conditions, like depression, so I can better understand how mental health and daily behaviors impact academic success. This could provide insight into support systems for students experiencing mental health challenges and provide additional support regarding the status of phones in high schools, which I find intriguing.