https://wonder.cdc.gov/wonder/help/mcd.html

There were many variables apart of this data set but the ones I was particularly interested in are the following:

• State
• Year
• Deaths
• Population

https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/Medicare-Provider-Charge-Data/OpioidMap.html

I had acess to 4 different data sets for 4 different years from 2013-2016. Again, there were several variables apart of this data set as well, however I was only interested in a few.

• State
• Opioid Prescribing Rate
• Extended Release Opioid Prescribing Rate

Opioids are a class of drugs that include (but are not limitted to):

• Oxycodone
• Hydrocodone
• Codeine
• Morphine
• Methadone

Opioids also consist of the illegal drug Heroin and synthetic opioids such as Fentanyl.

Some of the above drugs are available in extended release form. Extended release simply means that the pill is made so that the drug can slowly release over time. This means the patient could take the pill less often, and sometimes means less side-effects.

• Compare the two data sets to look for correlations
• Is there a higher death precentage rate it states that have a higher opioid or extended release opioid prescribing rates?

I first wanted to tidy up the two data sets and select for the variables that I am interested in for both data sets.

## # A tibble: 900 x 5
##    State          Year Deaths Population deathPerc
##    <chr>         <int>  <int>      <int>     <dbl>
##  1 West Virginia  2016    973    1831102    0.0531
##  2 West Virginia  2015    806    1844128    0.0437
##  3 Ohio           2016   4544   11614373    0.0391
##  4 West Virginia  2011    723    1855364    0.0390
##  5 New Hampshire  2016    502    1334795    0.0376
##  6 West Virginia  2014    688    1850326    0.0372
##  7 Pennsylvania   2016   4746   12784227    0.0371
##  8 West Virginia  2012    666    1855413    0.0359
##  9 West Virginia  2013    648    1854304    0.0349
## 10 Maryland       2016   2099    6016447    0.0349
## # ... with 890 more rows

To compare data sets I must innerjoin them so that I can create graphics of their relationship. I plan to first compare the Drug Related Deaths data to the Opioid Prescribing Data for each year from 2013-2016. I then plan to compare the data all together.

Now that all the data has been joined together in four different tables, for the four different years I would like to create some graphics to look for trends or correlations in the data.

##              Estimate Std. Error t value Pr(>|t|)  
## (Intercept) 0.0072349  0.0040970   1.766   0.0838 .
## opPreRate   0.0014204  0.0006592   2.155   0.0362 *
## 
## Residual standard error: 0.005269 on 48 degrees of freedom
## Multiple R-squared:  0.0882, Adjusted R-squared:  0.0692 
## F-statistic: 4.643 on 1 and 48 DF,  p-value: 0.03623

The regular opioid prescribing rate coefficient is 0.0014204. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0014204. With a p-value less than 0.05, we can assume this relationship to be significant.

##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 0.0139117  0.0026272   5.295 2.93e-06 ***
## xrOpPreRate 0.0003078  0.0003855   0.798    0.429    
## 
## Residual standard error: 0.005482 on 48 degrees of freedom
## Multiple R-squared:  0.0131, Adjusted R-squared:  -0.007458 
## F-statistic: 0.6373 on 1 and 48 DF,  p-value: 0.4286

The extended release opioid prescribing rate coefficient is 0.0003078. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0003078. However, the p-value is not less than 0.05, therefore this finding is not statistically significant.

##              Estimate Std. Error t value Pr(>|t|)  
## (Intercept) 0.0090726  0.0043440   2.089   0.0421 *
## opPreRate   0.0013162  0.0007065   1.863   0.0686 .
## 
## Residual standard error: 0.00563 on 48 degrees of freedom
## Multiple R-squared:  0.06744,    Adjusted R-squared:  0.04801 
## F-statistic: 3.471 on 1 and 48 DF,  p-value: 0.06858

The regular opioid prescribing rate coefficient is 0.0013162. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0013162. However, the p-value is not less than 0.05, therefore this finding is not statistically significant.

##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 0.0147337  0.0028163   5.232 3.65e-06 ***
## xrOpPreRate 0.0003457  0.0004059   0.852    0.399    
## 
## Residual standard error: 0.005786 on 48 degrees of freedom
## Multiple R-squared:  0.01489,    Adjusted R-squared:  -0.005638 
## F-statistic: 0.7253 on 1 and 48 DF,  p-value: 0.3986

The extended release opioid prescribing rate coefficient is 0.0003457. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0003457. However, the p-vaule is not less than 0.05, therefore this finding is not statistically significant.

##              Estimate Std. Error t value Pr(>|t|)   
## (Intercept) 0.0155881  0.0053518   2.913  0.00542 **
## opPreRate   0.0005588  0.0009030   0.619  0.53898   
## 
## Residual standard error: 0.006964 on 48 degrees of freedom
## Multiple R-squared:  0.007914,   Adjusted R-squared:  -0.01275 
## F-statistic: 0.3829 on 1 and 48 DF,  p-value: 0.539

The regular opioid prescribing rate coefficient is 0.0005588. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0005588. However, the p-value is not less than 0.05, therefore this finding is not statistically significant.

##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 0.0174169  0.0036590   4.760 1.82e-05 ***
## xrOpPreRate 0.0002023  0.0004998   0.405    0.687    
## 
## Residual standard error: 0.00698 on 48 degrees of freedom
## Multiple R-squared:  0.003403,   Adjusted R-squared:  -0.01736 
## F-statistic: 0.1639 on 1 and 48 DF,  p-value: 0.6874

The extended release opioid prescribing rate coefficient is 0.0002023. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0002023. However, the p-value is not less than 0.05, therefore this finding is not statistically significant.

##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  0.0268748  0.0067421   3.986 0.000228 ***
## opPreRate   -0.0008814  0.0011784  -0.748 0.458157    
## 
## Residual standard error: 0.00912 on 48 degrees of freedom
## Multiple R-squared:  0.01152,    Adjusted R-squared:  -0.009074 
## F-statistic: 0.5594 on 1 and 48 DF,  p-value: 0.4582

The regular opioid prescribing rate coefficient is -0.0008814. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to decrease by 0.0008814. This would suggest a negative correlation. However, the p-value is not less than 0.05, therefore this finding is not statistically significant.

##              Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 0.0201099  0.0048178   4.174 0.000125 ***
## xrOpPreRate 0.0002494  0.0006376   0.391 0.697363    
## 
## Residual standard error: 0.009159 on 48 degrees of freedom
## Multiple R-squared:  0.003179,   Adjusted R-squared:  -0.01759 
## F-statistic: 0.1531 on 1 and 48 DF,  p-value: 0.6974

The extended release opioid prescribing rate coefficient is 0.0002494. This means that if the extended release opioid prescribing rate increases, we can expected the death percentage to increase by 0.0002494. However, the p-value is not less than 0.05, therefore this finding is not statistically significant.

## # A tibble: 200 x 5
##    State          Year Deaths Population deathPerc
##    <chr>         <int>  <int>      <int>     <dbl>
##  1 West Virginia  2016    973    1831102    0.0531
##  2 West Virginia  2015    806    1844128    0.0437
##  3 Ohio           2016   4544   11614373    0.0391
##  4 New Hampshire  2016    502    1334795    0.0376
##  5 West Virginia  2014    688    1850326    0.0372
##  6 Pennsylvania   2016   4746   12784227    0.0371
##  7 West Virginia  2013    648    1854304    0.0349
##  8 Maryland       2016   2099    6016447    0.0349
##  9 Kentucky       2016   1490    4436974    0.0336
## 10 Massachusetts  2016   2286    6811779    0.0336
## # ... with 190 more rows

## # A tibble: 200 x 7
##    State          Year Deaths Population deathPerc opPreRate xrOpPreRate
##    <chr>         <dbl>  <int>      <int>     <dbl>     <dbl>       <dbl>
##  1 West Virginia  2016    973    1831102    0.0531      5.27        4.16
##  2 West Virginia  2015    806    1844128    0.0437      5.81        4.39
##  3 Ohio           2016   4544   11614373    0.0391      5.04        5.31
##  4 New Hampshire  2016    502    1334795    0.0376      5.11       10.1 
##  5 West Virginia  2014    688    1850326    0.0372      6.35        4.23
##  6 Pennsylvania   2016   4746   12784227    0.0371      4.69        7.2 
##  7 West Virginia  2013    648    1854304    0.0349      6.54        4.06
##  8 Maryland       2016   2099    6016447    0.0349      5.71        9.82
##  9 Kentucky       2016   1490    4436974    0.0336      5.49        4.49
## 10 Massachusetts  2016   2286    6811779    0.0336      3.84        7.69
## # ... with 190 more rows

Conclusions:

• Based on the indiviudal graphs for each year, there was only a significant correlation between regular opioid prescribing rate and death percentage for the year 2013.
• Though the other graphs were not significant, all but one showed a positive correlation.
• Based on the combinded data, there was a negative correlation between opioid prescribing rate and death percentage. However, there was a positive relationship between extended release opioid prescribing data and death percentage, but the correlation is not statistically significant.

I had hoped to see a stronger correlation between the two data sets, however I do have some thoughts as to why I did not.

• The CDC set deals with all drug realted deaths, not just opioid related deaths. Other drug related deaths could include: Alcohol, Amphetamines, Cocaine.
• Often times, patients might get addicted from a prescription, but then continue to use illeaglly; this would mean they would not be receiving a prescription for their further drug use.
• Illegal drug use often puts people at a higher risk for death.

1. "Medicare Part D Opioid Prescribing Mapping Tool" CMS
2. "Multiple Causes of Death 1999-2016" CDC