An exploration over mental health on 2013 BRFSS

Research question 1

Based on BRFSS data, can we say that in USA the use of legal drugs is a phenomenon associated with mental disorders occurrence?

My first question comes from the hypothesis that consumption of legal drugs and mental disorders manifestation are related. Specifically, I will investigate the BRFSS data to seek evidences regarding smoking and risky alcohol consumption in relation to these manifestations.

rate.mental <- mean(qdata$menthlth > 0)
rate.smoker <- mean(qdata$smoker, na.rm = TRUE)
rate.drinker <- mean(qdata$drinker, na.rm = TRUE)

It’s given that in 2013, in USA, smoking rate were 16.1% and 24.2% were the risky drink rate. If smoking and alcohol consumption are independent of mental disorders, there should be no significant difference in smokers and drinkers proportion between mentally healthy and the unhealthy groups. Similarly, both proportions shouldn’t vary greatly depending on the severity of the disorder.

mental.disorder <- qdata %>% 
  group_by(mental.disorder = ifelse(menthlth > 0, "Yes", "No")) %>% 
  summarise("risky drink rate" = mean(drinker, na.rm = TRUE),
            "smoking rate" = mean(smoker, na.rm = TRUE))

mental.disorder %>%
  gather(key, prop, -mental.disorder) %>% 
  ggplot(aes(x = mental.disorder, y = prop)) +
  geom_bar(stat = "identity") +
  geom_text(aes(label = round(prop, 3)), vjust = 1.5, color = "white") +
  geom_hline(aes(yintercept = c(rep(rate.drinker, 2), rep(rate.smoker, 2))), lty = "dashed") +
  labs(title = "Smoking and risky drink rates by mental disorder occurence",
       x = "mental disorder occurrence") +
  facet_grid(~key)

Inicially we found differences in proportions between healthy and unhealthy groups, specifically 9.20% on smoking rate and 6.10% on risky drink rate. In addition, dashed lines are indicating that a person’s likelihood of smoking or consuming alcohol regularly is greater than the average population when they are mentally disturbed.

This result shows some evidence that psychic suffering is positively associated smoking and risky drink rates. We do not know, however, whether disorder severity is related to these behaviors.

mental.disorder.drugs <- qdata %>% 
  group_by(disorder.level) %>% 
  summarise("risky drink rate" = mean(drinker, na.rm = TRUE),
            "smoking rate" = mean(smoker, na.rm = TRUE))

mental.disorder.drugs %>%
  gather(key, mean, -disorder.level) %>% 
  ggplot(aes(x = disorder.level, y = mean)) +
  geom_bar(stat = "identity") +
  geom_text(aes(label = round(mean, 3)), vjust = 1.5, color = "white") +
  geom_hline(aes(yintercept = c(rep(rate.drinker, 5), rep(rate.smoker, 5))), lty = "dashed") +
  labs(title = "Smoking and risky drink rates by mental disorder severity", x = NULL) +
  facet_grid(~key)

The chart above shows that both smoking and risky drink rates varies according to the mental illness duration severity. Among psychically healthy people, the survey registered 13.2% of smokers and 22,3% of risky drinkers. On the other hand, smoking rate is 34.8% and risk drinkers are 34.7% among those experiencing chronic mental disorders (26 to 30 days).

It is particularly noteworthy that smokers and drinkers proportions are less than population rate among people with mild disorders but is higher in groups with moderate to chronic disorders. Thus, these groups are more likely to smoke or drink regularly than people in general.

We can say this result reinforces the evidence of positive association previously verified. Moreover, it suggests that behavior related to the use of licit drugs is also associated with the severity of the mental disorder.

By the inclination of the bars it is possible to suppose that the relation is more intense in the case of smoking than in the consumption of alcohol. I will plot a scatterplot to analyze how smoking and risky drink rates are correlate to the mental distress duration

disorder.level.drugs <- qdata %>% 
  select(menthlth, smoker, drinker) %>% 
  gather(key, value, -menthlth) %>% 
  group_by(menthlth, key) %>% 
  summarise(mean = mean(value, na.rm = TRUE))

ggplot(disorder.level.drugs, aes(x = menthlth, y = mean, color = key)) +
  geom_point() +
  stat_smooth(method = "loess", se = FALSE, size = .5) +
  scale_color_discrete(NULL, label = c("risky drink", "smoking")) +
  labs(title = "Smoking and risky drink rates by mental disorder days", x = "days")

The positive correlation is clear. Regarding tobacco, it can be considered a linear correlation with mental illness duration. In the case of alcohol also positive correlation is verified, but the trend changes from something around 20 days (severe mental disorder), when the risky drink rate begins to fall.

Conclusions

This exploratory analysis generated some interesting insights about the relationships between legal drugs consumption and mental disorders in USA. They’re listed below:

• The BRFSS 2013 data brings evidences that a good mental health status is related to lower propention to smoking or to drink risky amounts of alcohol.

• Mental disorder level is positively associated with smoking and a risky alcohol consumption

• While people whose have light or no mental disorders are less likely to smoke or drink a lot than people in general, those who have moderate, severe or chronic mental disorders are more likely to present this risky behavior

• Particularly alcohol consumption shows a less linear variation over population mean, decreasing in higher mental disorder periods. This could makes sense if we consider that more serious mental disordes should be related to higher psychological counseling and medication use rates, leading to interruption of alcohol consumption due to potential drug interactions.

Research question 2

Is income related to mental disorders occurrence? Among people who smoke or drink, given the association between use of licit drugs and mental disorders, is the occurrence of these disorders more common in the low- or high-income population? That is, smoking and risky drink rates for people with mental disturbs show some variation according to income level?

There are studies that point out income and mental disorders relations, in general negative. So development of mental disorders is less prevalent in the population with higher income level.

mental.income <- qdata %>% 
  filter(!is.na(income2)) %>% 
  group_by(income2) %>% 
  summarise(mean = mean(mental.disorder))

p <- list(
  qdata %>%
    filter(!is.na(income2)) %>%
    count(income2) %>% 
    mutate(prop = prop.table(n)) %>% 
    ggplot(aes(x = income2, y = n, fill = income2)) +
    geom_bar(stat = "identity", show.legend = FALSE) +
    geom_text(aes(label = round(prop, 3)), vjust = 1.5, color = "black") +
    scale_fill_brewer(palette = "YlOrRd", "Income level") +
    labs(title = "Income distribution in US", x = NULL, y = NULL) +
    theme(axis.text.x = element_text(angle = 90)),
  mental.income %>% 
    ggplot(aes(x = income2, y = mean, fill = income2)) +
    geom_bar(stat = "identity", show.legend = FALSE) +
    geom_hline(yintercept = rate.mental, lty = "dashed") +
    geom_text(aes(label = round(mean, 3)), vjust = 1.5, color = "black") +
    scale_fill_brewer(palette = "YlOrRd", "Income level") +
    labs(title = "Mental disorder rate by income level",
         x = NULL, y = NULL) +
    theme(axis.text.x = element_text(angle = 90))
)

do.call(grid.arrange, c(p, ncol = 2))

In BRFSS data we found evidences that this relation proceed for american residents.

Nearly 58% of the population has an annual household income of U$ 35,000. This is an important marking point in terms of mental health. Household income bigger than this is related to mental disorder rates below the population mean, according to mental disorder rate by income level chart on the right.

On the other hand, prevalence of higher mental disorder rates for household incomes below U$ 35,000 per year suggest that poor people are more likely to develop disorders. In practice, the group with annual household income less than U$ 10,000 is almost twice as likely to develop mental disorders as the population group with annual household income above U$ 75,000.

This trend persists when only smokers and risky drinkers are included.

drinkers.income <- qdata %>% 
  filter(!is.na(drinker), !is.na(income2)) %>% 
  group_by(income2, act = drinker) %>% 
  summarise(mental.rate = mean(mental.disorder)) %>% 
  mutate(behavior = "risky drink")

smokers.income <- qdata %>% 
  filter(!is.na(smoker), !is.na(income2)) %>% 
  group_by(income2, act = smoker) %>% 
  summarise(mental.rate = mean(mental.disorder)) %>% 
  mutate(behavior = "smoking")

rbind(drinkers.income, smokers.income) %>% 
  ggplot(aes(x = income2, y = mental.rate, fill = ifelse(act == 1, "Yes", "No"))) +
  geom_bar(stat = "identity", position = position_dodge(width = .9)) +
  geom_text(aes(label = round(mental.rate, 3)), hjust = 1.2, angle = 90,
            color = "white", position = position_dodge(width = .9)) +
  scale_fill_manual(values = c("grey40", "coral"), NULL) +
  labs(title = "Mental disorder rate for smokers and risk drinkers",
       subtitle = "by income level", x = NULL, y = "prop") +
  facet_grid(~behavior) +
  theme(legend.position = "top",
        axis.text.x = element_text(angle = 90))

Among them, mental disorders occurrences was reported by 56.6% of drinkers and by 60.7% of smokers belonging to the lower income group, whereas among the highest income group, rates were 29.8% and 32.5%, respectively.

As I’m not searching for causality, response and explanatory variables could be inverted to take another look in this relationship.

disorder.income.drugs <- qdata %>% 
  filter(!is.na(income2)) %>% 
  select(menthlth, smoker, drinker, income2) %>% 
  group_by(menthlth, income2) %>% 
  summarise(smoking = mean(smoker, na.rm = TRUE),
            "risky drink" = mean(drinker, na.rm = TRUE))

disorder.income.drugs %>% 
  gather(key, value, -menthlth, -income2) %>% 
  ggplot(aes(x = menthlth, y = value, color = key)) +
  geom_point() +
  stat_smooth(method = "loess", se = FALSE, size = .5) +
  scale_color_discrete(NULL) +
  labs(title = "Smoking and risky drink rates by mental disorder days and income level",
       x = "days", y = NULL) +
  facet_wrap(~income2, ncol = 4, strip.position = "left") +
  theme(legend.position = "top")

And as we can see smokers and risk drinkers amounts increase as long disorders are, but decrease as higher the income is. It’s interesting to see, either, that there’s difference in variability according income level. I will take a look at standard deviations of smoking and risky drink status to analyse this better.

qdata %>% 
  filter(!is.na(income2)) %>% 
  group_by(income2) %>% 
  summarise(smoking = sd(smoker, na.rm = TRUE),
            "risky drink" = sd(drinker, na.rm = TRUE))

## # A tibble: 8 x 3
##   income2           smoking `risky drink`
##   <fct>               <dbl>         <dbl>
## 1 Less than $10,000   0.461         0.491
## 2 Less than $15,000   0.444         0.477
## 3 Less than $20,000   0.428         0.473
## 4 Less than $25,000   0.410         0.456
## 5 Less than $35,000   0.387         0.445
## 6 Less than $50,000   0.367         0.432
## 7 Less than $75,000   0.337         0.423
## 8 $75,000 or more     0.278         0.400

Conclusions

• We found a negative association between income and mental disorder

• Smaller household incomes are related with more severe mental disorders

• Variability in risky drink and smoking rates is smaller for the higher income group, suggesting a more predictable behavior

Research question 3

What is the sleep impact on mental illness occurrence? People who smoke and drink and do not sleep properly are more likely to present severe cases of psychic disorders than people who smoke or drink but sleep properly?

Sleep is extremely important for a good performance of psychic, physical and physiological functions. It is proven that sleeping less than necessary can cause health damage such as cardiovascular problems, and strongly affect cognitive performance. It is hoped, therefore, that sleep has some type of relation with mental problems like stress, mental fatigue, difficulty of concentration and memory, emotional imbalance, anxiety, depression.

rate.sleep <- mean(qdata$sleptim1, na.rm = TRUE)

On USA, people aged 18 years old or over sleep on average 7.05 hours per day.

The National Sleep Foundation² recommends between 7 and 9 sleeping hours a night for adults. The majority of american population, about 80%, sleep between 6 and 8 hours. The distribution of relative frequencies for daily sleep time also shows that amount of people sleeping less than necessary is bigger than who sleep above 9 hours per day..

p1 <- qdata %>% 
  filter(!is.na(sleptim1)) %>% 
  ggplot(aes(x = sleptim1, y = ..prop..)) +
  geom_bar(fill = c(rep("grey40", 6), rep("coral", 3), rep("grey40", 15))) +
  scale_x_continuous(breaks = seq(0, 24, 2)) +
  labs(title = "Sleeping time distribution", x = "hours")

p2 <- qdata %>% 
  filter(!is.na(sleep.status)) %>% 
  count(sleep.status) %>% 
  mutate(prop = prop.table(n)) %>% 
  ggplot(aes(x = sleep.status, y = prop)) +
  geom_bar(stat = "identity", fill = c("grey40", "coral", "grey40")) +
  geom_text(aes(label = round(prop, 2)), vjust = 1.3, color = "white") +
  labs(title = "% population by sleeping status", x = "sleeping status")

grid.arrange(grobs = list(p1, p2), ncol = 2, widths = c(2, 1.2))

More precisely, healthy sleep rate is 64%, with an average of 7.6 sleeping hours. The 33% of people who sleep less than ideal have an average of 5.5 hours. On the other hand, the smallest group, made up of 4% of people which sleep above the recommended, spend in average 11 hours sleeping per day.

qdata %>% 
  filter(!is.na(sleep.status)) %>% 
  group_by(sleep.status) %>% 
  summarise(rate.hours = round(mean(sleptim1), 1))

## # A tibble: 3 x 2
##   sleep.status rate.hours
##   <fct>             <dbl>
## 1 Less                5.5
## 2 Enough              7.6
## 3 More               11

Ok. But I’m focused on seek for relations between sleeping status, and as the chart below shows us, there’s sense consider a negative association between a healty sleep and mental disorders occurrence and severity.

p <- list(
  qdata %>% 
    filter(!is.na(sleep.status), smoker == 1 | drinker == 1) %>%
    group_by(disorder.level) %>% 
    summarise(prop = mean(sleep.status == "Enough")) %>% 
    ggplot(aes(x = disorder.level, y = prop)) +
    geom_bar(stat = "identity") +
    geom_text(aes(label = round(prop, 2)), vjust = 1.5, color = "white") +
    labs(title = "Helthy sleep rate by mental disorder level", x = NULL),
  qdata %>% 
    filter(!is.na(sleep.status), smoker == 1 | drinker == 1) %>%
    group_by(sleep.status) %>% 
    summarise(prop = mean(mental.disorder))  %>% 
    ggplot(aes(x = sleep.status, y = prop)) +
    geom_bar(stat = "identity") +
    geom_text(aes(label = round(prop, 2)), vjust = 1.5, color = "white") +
    labs(title = "Mental disorder rate by sleeping status", x = NULL)
  )

grid.arrange(grobs = p, ncol = 2, widths = c(.55, .45))

Based on this results we can say that BRFSS data shows evidences that healthy sleeping people are mentally healthy people.