Skip to main content

Allocation of Funds for Addiction in the United States

Overview

The number of Americans suffering from addiction is steadily increasing and many agencies and departments throughout the federal government work to distribute funds across the nation to assist those Americans who are addicted as well as to prevent others from becoming addicts. SAMHSA is one of these agencies, housed under the U.S. Department of Health and Human Service, whose mission is "to reduce the impact of substance abuse and mental illness on America's communities" [source]. That said, like every government agency, there are limited resources and time that SAMHSA has and we aim to assist in SAMHSA making the greatest impact in the shortest time possible.


Last year, SAMHSA distributed its' grant funding based on the percent of the US population that resides in a particular state; however, our analysis found that this metric does not line up with states that have high crude rates (deaths per 100,000) indiciating that this vital funding is not reaching the most at risk audience. In addition to crude rate, we found a few other factors that can predict a high crude rate. These include: % Non-Hispanic White, % Black, % Hispanic, % Food Insecure, Life Expectancy, % Asian, % Enrolled in Free or Reduced Lunch, High School Graduation Rate, % Less Than 18 Years of Age, % Rural, % Adults with Diabetes, % Female, % Broadband Access, % Homeowners, % Limited Access to Healthy Foods, % Frequent Physical Distress, % Frequent Mental Distress, Homeowners, Median Household Income, % Severe Housing Cost Burden, % 65 and Over, and the Age-Adjusted Death Rate.

With this information in hand, we recommend that SAMHSA consider distributing future grants based on factors such as the state's drug and addiction-related crude rate, and the other factors that were previously mentioned. This will ensure those communities that have greater numbers of drug and addiction related issues are getting funding to prevent and treat those in the areas.

Background:

Substance abuse and addiction have been steadily rising in America for decades and the increase in overdose deaths doesn't seem to be slowing down. During this time, we have seen the American government declare a ‘War on Drugs’, appointed Drug Czar’s to be overseers of anti-drug efforts, and declared public health emergencies such as the Opioid Crisis Epidemic, and yet drug overdoses have only continued to go up. The issue here doesn't seem to be not getting enough funding to drive effective policy, seeing as the Federal Drug Control Funding's 2020 budget totaled $34.6 billion, an increase of 1,090% over the past 39 years when adjusted for inflation [source]. The current White House administration believes there have been systemic inequities in our nation's past policies for addressing criminal justice, prevention, and treatment, and that "people should not be incarcerated for drug use but should be offered treatment instead" [source]. We believe there is truth to this and will be looking into grants from the Substance Abuse and Mental Health Services Administration (SAMHSA) to see if there are inequities or inefficiencies in where resources for treatment and research concerning substance abuse go around the country.

In SAMHSA's own words, they are "the agency within the U.S. Department of Health and Human Services that leads public health efforts to advance the behavioral health of the nation. SAMHSA's mission is to reduce the impact of substance abuse and mental illness on America's communities." [source]. They were established by Congress in 1992 with the goal of disseminating the latest information, research, and services to the general public. The department is currently being led by Miriam Delphin-Rittmon, Ph.D., Assistant Secretary for Mental Health and Substance Use as well as a team of Directors and Regional Administrators. One item of focus for the current leadership's Strategic Plan FY2019-2023 is on "Improving Data Collection, Analysis, Dissemination, and Program and Policy Evaluation." and we believe this project will be a great starting point to achieving just that.

Problem Statement:

This project aims to challenge and investigate the assumption that the Substance Abuse and Mental Health Services Administration (SAMHSA) grants made by distributing grant award amounts closely mirroring the population distribution of the United States. We aim to analyze deaths caused by drug and alcohol use in order to allocate funds based on areas of high need, under the assumption that states that have a higher target audience can put greater amounts of funds to better use. High need is determined based on a state's crude rate (deaths per 100,000 people) and unmet treatment needs based on answers to the National Survey on Drug Use and Health (NSDUH). In addition, we aim to use the acquired data to provide SAMHSA with other possible factors to consider when distributing funding to high-need areas other than crude rates.

Analysis Overview


Drug-Induced Death Data

The Drug Induced Death data was collected from the CDC Wonder database. The main query consisted of number deaths caused by drug or alcohol related health issues along with crude rate (deaths/population * 10,000) broken up by State and County for the years 2010-2019. The data included a notes section with the details of the query that needed to be removed before I could work with it. Some of the crude rate data was incomplete, as the CDC would not calculate crude rate for any county with a death count of under 20 total. This was a small enough number of rows that I felt it wouldn't hurt the integrity of the data to impute those missing data points using the same formula as the CDC given the dataset contained both the population size of the county and death count. The main take away with this data was understanding the distribution of deaths related to drug overdoses among the United States and how that differed from the distribution of crude rate. Along with this, we could see which States were most effective at reducing death counts and which States were trending up at the highest rates.

Substance Use Data

The NSDUH 2019 census data was found through the SAMHSA website here. The data itself came in the form of 33 individual csvs. For this project, 8 tables were taken specifically to deal with subjects of total drug use, treatment (or lack of) for drug problems at specialized facilities, and the treatment/reporting of mental illness. The data was divided by age group and further sectioned into estimates and 95% confidence intervals. The age groups in the data were 12-17, 18-25, 26+, and combinations such as all people over the age of 12. The average estimates were used, and each cleaned table was exported into individual csvs for further use. An aggregated csv was created by taking the total values from each table and dividing them by the estimated state population in order to get % population values. The whole population estimate of each table by state was divided by estimated population found through the census website here. Some data was combined such as taking the summation of the different types of specialty facilities. The main takeaway from this data was to identify who and how many people in each state use drugs and get treatment in order to see how these issues factor into grant data and crude rate. Along with this, data on mental illness and treatment was gathered in order to further identify possible issues and correlations.

County Health Assessment Data

While looking for data related to demographics of the United States broken up into counties, I came across the County Health Rankings & Roadmaps, a program of the University of Wisconsin Population Health Institute. I was able to download individual Excel files for each of the 50 states and the District of Columbia. I wrote a script to gather the data into one dataframe for cleaning and EDA. We decided to leave out any columns that had more than 20% of its data missing, as imputing would be inappropriate. For the columns we did choose to impute, we filled most of those values with the median value from that specific state. There were times when we had no data for a given state and we researched the state values to impute (for instance the state of Utah had no data on High School Graduation Rates. We looked up the state reported average and imputed that.) We also examined the summary statistics and the columns themselves were explored in more detail in our modeling notebook.

Grant Data

The grant data was scraped from the SAMHSA grant data archives and was structured as one row per grant and included the grant id, city, state, and grant amount award for the year 2021. The web scraping was completed in three sections due to repeated connection timeout errors when it was attempted in one sitting. Data cleaning for the grant data included checking for null values and replacing these nulls, represented by ‘-’, with zeros and then dropping the one row that had no grant data. In addition, our analysis stuck with the traditional 50 US states so we dropped data where the grant was awarded to places such as Puerto Rico, Guam, etc. Lastly, in order to further our analysis, we created a few columns that could be inferred from the data we already had including calculating the grant averages and totals by state and region, after we mapped each state to its corresponding US region. Additionally, population data was used in tandem with the grant data to draw further conclusions on how it was being distributed. Thus, the population data was cleaned by mapping the state’s abbreviation to its full name. The big takeaways from the grant data included that SAMHSA tends to distribute more grant money to more populous states than those that have smaller populations. Additionally, the money that more populous states receive tend to come in larger total amounts.

Conclusions and Recommendations

Conclusions

From our analysis, it seemed apparent that funds are distributed to states in percentages that closely mirror the population percentage of each state. We found that there was no correlation with other measures we would consider to be important components of determining need, and we focused on crude death rate (deaths per 100k people).

Recommendations

We recommend that SAMHSA distribute future grants based not only on the population of each state but rather by incorporating other features, such as crude death rate attributed to drugs and alcohol use. Our argument is that there should be a balance to the way in which funds are allocated to states. It should be partially designated based on population, but crude death rate also gets to the severity of the problem in any size region. By using both, the funds will be more appropriately allocated.

Future Work

We recommend that future work in this area focus on drug and alcohol-specific grants or pools of money as this analysis looked at grants that included funds allocated for both substance abuse and mental health services. In addition, future work should use drug and alcohol use, demographic, socioeconomic data in the model to see if these factors also have an influence on the crude rate by county. This analysis did not include these features as data are broken down by county on these features was time-consuming to track down.


* Much of this blog post comes from a group project write-up that was co-authored by me, Andrea O. Clifford C., and Alex G. from my General Assembly cohort. 

Comments

Popular posts from this blog

From the Classroom to Data Science

Ever since I was in 10th grade, I’ve wanted to teach high school math. I was one of the lucky few who knew exactly what my major would be upon entering college. And it was exactly the right path for me. After college, I taught for five years in a public school in the county where I grew up. Then, wanting to leave Michigan and begin life in a new place, I scored a job at a private school in Colorado. Seven years later and my family (created in Colorado) signed up for a new adventure teaching and living at a boarding school in rural New York. Little did I know that last school year in New York would be the hardest of my career.  As I struggled with the decision of whether to stay or leave and completely change my career, there were so many “what ifs” that ran through my mind.  Would this decision be on my mind if I hadn’t struggled through a year of teaching during a pandemic?  Would I want so desperately to move back to Colorado if I had formed a community or felt fully welcomed into th

Coding Challenges as a way to Level Up

 In my pre-course work for General Assembly's Data Science Immersive program, we were introduced to the website CodeWars .  The first few times I practiced my coding, I was frustrated but engaged. The way the website is set up allows you to level up as you continuously progress - like a way to see where you rank among others.  Here's what I love about it:  You can get extra practice by choosing your level of difficulty for each challenge. If you want to practice, stick with challenges at your current level. If you are looking to level up, choose something at a level above your current one. There are options for either path! CodeWars felt like a fun way to push me out of my comfort zone with coding. There is no risk at all to trying something that is just beyond my current level.  There are so many programming languages available! Want to dabble in Haskell? Ruby? SQL? They've got you covered. (From what I counted, there are 29 core languages and 26 beta languages currently s

Fail Your Way to Success

As a recovering perfectionist, I struggle with anything less than perfect. Coming into an immersive Data Science program was like being blind-sided. Python coding is essentially failing over and over again and learning from those small mistakes.  Forget a colon? Didn't indent? That will throw an error.  In many ways, I have learned about changing my approach to life through my experience learning Python. Not an hour goes by where I don't throw some kind of error or have to stop and debug my code. Or, since I'm still learning so much daily, I come across a need to do something that I don't quite know how to do. I've learned to strategically Google (as any coder/engineer can understand and appreciate!)  Isn't this just like our lives? I've been on a healthy person journey now for some time. Prior to working with a coach, I had mastered the 'all or nothing' approach. Instead, as a healthy person, I pause regularly and evaluate my errors. Where did I go