608 Final Project: Opioid Overdoses in the State of Connecticut

This project explores data containing the details regarding individuals who overdosed on opioids between the years of 2012 and 2017 (inclusive) in the state of Connecticut. The opioid epidemic is one that is taking over the country, but has increasingly plagued Connecticut. For this reason, I chose to focus on this state for my analysis. When many people think of opioids, they think of heroin. While this is one example, a more common example is the prescription drugs sitting in the cabinets of families worldwide. The important thing to note here is that opioid overdoses and related deaths are not exclusive to a degenerate subset of our population. This issue plagues mostly every race, gender, age, and socioeconomic status, with some slight differences between. We will discuss these further.

For reference, the data can be found here: https://catalog.data.gov/dataset/accidental-drug-related-deaths-january-2012-sept-2015

## Warning: Expected 2 pieces. Missing pieces filled with `NA` in 153 rows
## [93, 292, 310, 320, 337, 363, 398, 423, 466, 505, 513, 534, 632, 663, 717,
## 746, 754, 820, 849, 943, ...].

## Warning: package 'bindrcpp' was built under R version 3.3.3

Firstly, we look at where these individuals call home. As we can see in the map of the US below, these victims lived as far as the West Coast.

Zooming in on Connecticut, we see a large cluster of individuals lived in the state in which they died. The blue circles represent the location of residence and the red circles represent the location of death.

Location Type

We know geographically where these individuals were located when they died, but what type of location was it? The chart below indicates that overdoses are predominantly happening in residences.

## Warning: Ignoring unknown parameters: binwidth, bins, pad

County

By county, we can see that Hartford is the leader, with New Haven a close second. This follows the population distribution of the counties, so we can conclude that the increased deaths in Hartford and New Haven counties do not necessarily suggest increased opioid usage and subsequent death, rather, it is likely a factor of increased population size.

Race

In the state of Connecticut, 77.6% of its occupants are White, followed by 13.4% Hispanic, and 10.1% Black. When it comes to the race breakdown of our overdose subset - we see a similar trend: 79.2% White, 10.8% Hispanic, 8.2% Black. It seems the racial distribution of the overdose population follows the racial distribution of the overall Connecticut population.

County and Race Combined

We’ve looked at county and race separately. How do they behave when combined? We see our breakdown below. As we know, Hartford and New Haven are our top counties, but their racial breakdown is a bit different. Hartford seems to have more Hispanic individuals and less Black individuals than New Haven.

Sex

The state of Connecticut is comprised of 51.3% females, 48.7% males. However, 73.4% of overdoses in our data set occurred in males, with the remaining 26.6% in females. This pattern overall does follow the national numbers regarding illicit drug use (which does include types of drugs not in the scope of this analysis, e.g. cocaine), with more males reporting usage than females. The spread is not as large, however. On a nationwide level, males are only slightly ahead of females.

The national study referenced can be found here:

https://www.samhsa.gov/data/sites/default/files/NSDUH-DetTabs-2016/NSDUH-DetTabs-2016.pdf

Age

Interestingly, we can see a bimodal distribution of age for our overdose victim population. This means we have two different groups, one centering around age 30, and the other centering around age 50.

## Warning: Ignoring unknown parameters: binwidth, bins, pad

Age and Race

We saw our racial distribution vary from county to county, so I was curious if there were variations in race throughout the age distribution as well. In looking at the graph below, we see an interesting pattern in race over “time”. As the individuals get older, there is an increase in the Hispanic and Black representations in the population. However, looking at these differences is somewhat difficult on this stacked bar chart. Let’s move on to a boxplot.

Here, the differences are much clearer. There is a significant difference in age across the races. As we can see more specifically in the table below the boxplot, the median age by race varies drastically across races. For example, the median age for Asian Indian victims is 31, while the median age for Black victims is 48. Analyzing the reasons for these differences are likely cultural and outside the scope of this project.

Median Age by Race

Race	Age
Asian Indian	31
Asian, Other	30
Black	48
Chinese	23
Hispanic, Black	40
Hispanic, White	44
Native American, Other	44
Other	41
White	41