Substance Use Treatment Admissions

Author

Vera Mazeeva, Adam Sultan, Helen Amon, Kevin Wu

Published

April 28, 2025

Data Overview

The TEDS-A tracks every admission from state-licensed substance abuse services and tracks demographic information (age, sex, race, employment status, etc.), substance use history (age of first use, frequency, number of prior admissions, and substance used. etc.), and facility information (type of service, geographic information, etc.). Reporting this information is required, meaning that the data available is robust and spans across the United States and its territories. Each observation is one admission, meaning that certain individuals may be double counted. Since the data is anonymized, this is unavoidable. We will be focusing mainly on the data from 2022, consisting of 1498034 observations in all US states and territories except for Delaware, Oregon, Washington, and West Virginia. Graphs that look at admissions data from prior years will do so in order to compare with and contextualize the 2022 data.

Research Questions

  1. How does education level influence the types of substances reported?
  2. How does employment status influence frequency of substance use?
  3. How do substance use treatment admission rates vary across different regions of the US?
  4. How do individuals of different ages utilize substance counseling services?

Research Question 1

How does education level influence the types of substances reported?

First we used a heatmap to provides a broad, visual overview of how secondary substance use distributes across different education levels. The heatmap allows us to quickly compare proportions of all types of substances within each education group at once.

In the graph above, the x-axis represents secondary substances used, while the y-axis represents education levels. The fill color intensity represents the proportion within each education level (dark red = higher proportion). We see that “None” dominates across all education levels, especially “4 years of college or more.” After “None,” the next visible dark bands are for Alcohol, Cocaine, and Marijuana. We notice that cocaine use fades as education rises, and alcohol remains medium-light across all rows — confirming no strong trend. The intensity for methamphetamine is clearly darker among the lowest education levels, fading with higher education. What is most interesting is that marijuana stays more consistent or slightly darker at higher educations. Substances like PCP, inhalants, OTC medications, hallucinogens, and others have minimal representation across all education levels.

The heatmap gives a broad visual overview — good for showing dominance and absence, but harder for comparing trends across specific substances without a line graph. Higher education groups have simpler substance profiles (dominated by “None” and fewer “dark” tiles for dangerous substances). Lower education groups show more diverse and heavier secondary substance involvement.

Research Question 2

How does employment status influence frequency of substance use?

We also wanted to better understand how employment status at the time of admission affected the frequency of substance use for individuals in the dataset. We wanted to better understand this because employment is often closely related to financial stability, access to health care, mental health, and many more aspects of life. We wanted to investigate how frequency of substance use may vary across differently employed persons, and if there were any interesting findings that would arise. To do this, we created both a stacked bar chart and a facetted bar chart.

The TEDS dataset classifies employment status into four categories: “Full-time” (working 35+ hours per week), “Part-time” (working fewer than 35 hours each week), “Unemployed” (looking for work during the past 30 days or laid off from a job), and “Not in labor force” (not looking for work in the past 30 days, or a student, homemaker, retired person, etc.). As for frequency of use, the TEDS dataset classifies this into three categories: “No use (in the past month),” “Some use,” and “Daily use.” The way it defines “Some use” is not specified. We decided to focus on the frequency of use of the client’s primary substance, but the client’s secondary and tertiary substances, as well as their respective frequencies, were also available in the dataset.

In the above stacked bar chart, it appears that the clients who are not in the labor force have the highest proportion of daily primary substance use, at 54%. This is followed closely by clients who are unemployed, where their proportion of daily primary substance use is 48%. For clients who work full-time or part-time, the proportion of frequencies of use seems to be split rather evenly. Clients who work full-time have the highest proportion of no substance use at 33% (followed very closely by the proportion for clients who work part-time, at 32%). Part-time clients and full-time clients have an equal proportion of some drug use, at 34%.

So, in conclusion, clients who are not in the labor force or unemployed tend to engage in daily primary substance use at higher rates than clients who are employed full-time or part-time.

In the above facetted bar chart, we are able to compare proportions within each employment status group side-by-side. The graph further reinforces the findings we obtained from the previous stacked bar chart. It is again clear that clients who are not in the labor force or are unemployed have higher proportions of daily substance use compared to those employed part-time or full-time. We also see once again that the proportions of frequency of use are rather equally distributed for both the full-time and part-time groups.

We also decided to conduct a Chi-squared test of independence to determine whether employment status and frequency of primary substance use are associated. Since the resulting p-value from this test was less than 2.2 x 10^-16, which is much less than the chosen significance level of 0.05, we can reject the null hypothesis, which is that employment status and frequency of primary substance use are independent. We have sufficient evidence to suggest that there is a significant association between the two variables, which is further supported by the two previous graphs produced.

Research Question 3

How do substance use treatment admission rates vary across different regions of the US?

For the next research question, we aimed to examine how substance use treatment admission rates varied across the United States. To ensure a fair comparison, we adjusted the admission counts by dividing them by the population estimate for each year (population data sourced from US Census Bureau) and multiplying by 100,000. This conversion transformed the raw admission counts into admission rates per 100,000 people.

First, we look at the overall admission rates across US states in 2022:

The map above illustrates significant regional variation in substance use treatment admissions, with some states reporting notably higher rates than others. For example, states in the South and Northeast tend to have higher admission rates, indicating a higher prevalence of substance use or potentially greater awareness and access to treatment services in these regions. Conversely, the West and parts of the South show lower rates of admissions, suggesting that these areas may either have different substance use patterns or, alternatively, more effective prevention and treatment infrastructure. The map also highlights states like South Dakota, Minnesota, Colorado, and Connecticut, which stand out with particularly high admission rates. This may reflect unique regional challenges, such as localized substance use crises or differences in healthcare accessibility, warranting targeted interventions in these areas.

To better understand how substance use treatment admission rates have evolved over time across different US regions, we now turn to time series analysis. These graphs display admission rates from 2010 to 2022, grouped by region and primary substance use reported at the time of admission:

The graphs clearly show regional differences, with certain patterns emerging across the U.S.

The first key observation is that alcohol consistently emerges as the most commonly reported primary substance at the time of admission across the U.S. This trend highlights the continued prominence of alcohol use, suggesting that alcohol-related issues remain a significant public health concern. Another key observation is that Northeast has the highest admission rates for four out the six substances presented (alcohol, cocaine/crack, heroin, other opiates and synthetics), indicating a potential regional concentration of substance use disorders.

We now turn our attention to each substance individually:

Alcohol For alcohol, the Northeast consistently reports higher admission rates and the South consistently reports lower admission rates. There is a general decline across all regions, with the West experiencing the largest absolute drop from 2010 to 2022, and the South being most stable.

Cocaine/Crack Cocaine, overall, is the least commonly reported primary substance upon admission to substance use treatment, which could be due to a combination of factors such as stigma, prevalence / accessability of other substances. The trends for cocaine/crack reveal that all regions have maintained relatively stable rates of admission.

Marijuana/Hashish For marijuana/hashish, all regions have very similar admission rates from 2010 to 2022, and we observe a general decline in admissions for all regions.

Heroin Heroin admissions peak sharply in the Northeast, particularly around 2015-2016, likely due to the opioid crisis, while other regions, including the West and Midwest, show more gradual increases up to 2017-2018. Following these peaks, heroin admissions decrease, returning to approximately the same levels as in 2010 across all regions. However, it is notable that Northeast admissions remain significantly higher, more than double those in the other regions, indicating that the opioid crisis had a more profound and lasting impact in this region

Other opiates and synthetics For other opiates and synthetics (e.g. fentanyl), the West, South, and Midwest all have very similar and relatively stable admission rates from 2010-2022. The Northeast admissions are consistently higher, with an overall decrease from 2010 to 2022.

Methamphetamine/Speed Lastly, the rise in methamphetamine/speed admissions, particularly in the Midwest, aligns with the growing problem of methamphetamine use in states like California. The South and Northeast exhibit more gradual and smaller increases in admissions. The West has a minor peak around 2017, before admissions decrease back to roughly the level they were in 2010.

These findings highlight significant regional disparities in substance use treatment, suggesting that tailored, region-specific interventions and policies are needed to address the varying patterns of substance use and treatment demand across the U.S.

Research Question 4

How do individuals of different ages utilize substance counseling services?

The data included over a million data points of individuals seeking a vast array of counselling services for individualized substance misuse needs. Since age is a factor that is intertwined with other features that shape someone’s life – eg. employment, familial responsibilities, access to services – examining age differences can help us broach these other lifestyle differences that are key factors in understanding how people use drugs, and crucially, how they can get help through addiction services. For the first graph of this research question, we created a mosaic plot to examine the conditional distribution of services given the age of patients. Within the dataset, age was already a categorical variable with twelve levels, with the age levels spanning between 2 years and 8 years. We have collapsed the age levels so that there are fewer levels each spanning a more consistent number of years.

The mosaic plot shows that there is a significant difference between the conditional distributions of services (meaning the type of treatment provided) and the age category. The mosaic plot is colored according to the Pearson residuals so we can see where the observed counts differ significantly from the expected counts, assuming independence between the two variables. From the plot we can see that for the first three age categories (12-34), ambulatory care, – where patients get outpatient treatments for less than two hours per day – represents a larger proportion of services than what would be expected given independence. On the other hand the next three age categories (35-64) display a smaller proportion of ambulatory services and a larger proportion of Detox and Rehab/Residential. Both Detox and Rehab/Residential are impatient programs where patients receive 24 hour care over a longer period of time. The 65+ age group appears to have a similar distribution as the younger ages, with a greater proportion of ambulatory services than Detox and Rehab/Residential.

This analysis shows that there are indeed significant differences between the services used according to age, and highlights an important divide between the 35-64 year olds and the rest of the age groups specifically in regards to the use of ambulatory care. Using this knowledge, ambulatory care facilities can provide resources specifically for the younger and the oldest age groups.

For the next graph of this research question, we wanted to look at the marginal distributions of the age groups and compare these across the last decade. To this end, the time series data shows the number of admissions between 2013 and 2022, colored by the age group.

The graph shows the age group with the highest number of admissions across the decade is 25-34, with 65+ category having the fewest. Looking at the 2013-2022, we can see three groups of age levels that have similar changes in the number of admissions. The number of admissions of individuals between the ages of 25 and 54 have all increased since 2013, hitting their peak in 2018 and decreasing until 2020. Since 2020 the number of admissions have been largely stagnant with much less change from year to year. For the youngest two age levels, we can see a global decrease since 2013, though there still appears to be the same flattening out effect since 2020. For the oldest two age levels (55-64 and 65+), we can see a global (though shallow) increase, though 55-64 shows some peaking in 2018 similar to the middle age group.

Like in the mosaic plot, there is a divide between the middle aged group that makes up the majority of the admissions and the youngest and oldest groups. Since age is such a key factor, the dramatic changes in 2018 and 2020 that are seen in some age groups and not others could be the result of changes in policy, supply, legalization and treatment facilities that are relevant to that age group and not others. From the graph, it is also clear that a major shift in the admissions trend coincided with the COVID-19 pandemic, though given the wide-scale disruption of the coronavirus and the wide ranging policy responses to it, it is difficult to precisely pinpoint why the admissions rate have stayed largely the same since.

Conclusion and Additional Questions

In our report, we have discussed the various demographic and geographic features that affect substance use and engagement with substance treatments. As we have seen, things like education and employment can affect the substances people use and the rate at which they use them. We have also explored the national trends in admission data and seen how different trends emerge when considering location, type of drug and age group. This greater understanding of who and how people use drugs can help policy makers provide greater resources for those seeking substance abuse treatment.

As we analyze the data, we should ensure that our analysis isn’t colored by survivorship bias. Looking at the covariates of drug usage, we should consider individuals who were unable to enter treatment, are receiving treatment in jail or prison, or those who have passed away. This inspires further questions to be formed: are there systematic differences between those who access treatment and those who don’t? Or, does the severity of the substance used and the frequency of use influence whether or not someone seeks professional treatment? These questions could not be answered with the dataset available to us and the tools currently at our disposal, but would provide further insight and build upon findings we discovered during the making of this report.

Citation

Substance Abuse and Mental Health Services Administration. (2024). Treatment Episode Data Set Admissions (TEDS-A) 2022: Public Use File (PUF) Codebook. Rockville, MD: Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. Retrieved from https://www.samhsa.gov/data/.