DHRUVI MEHTA

Hi! I'm Dhruvi, a data analyst and visualisation specialist with a background in graphic design based in New York City.


SELECTED WORK
2025 by Dhruvi Mehta

Beyond Access: Understanding Mental Health Disparities

According to the National Institute of Mental Health, over 1 in 5 adults in the United States experience mental illness each year. However, access to mental health treatment still varies drastically depending on where you live. Data from the National Alliance on Mental Illness (NAMI) reveals that less than 50% of adults who suffer from mental illnesses were able to get care in 2021; this number was even lower among marginalised groups and rural populations. Unfortunately, there are wider gaps in treatment, specifically amongst members of the LGBTQ+ community, communities of color, and people residing in rural areas, many of whom encounter difficulties such as shortages, affordability issues and even limited access to these facilities. These inequalities lead to impacts like higher rates of substance use, dropping out of school, unemployment and imprisonment—for example, 33.5% of adults with mental illness were also known to have tendencies of substance use, and 70% of youth in juvenile detention have been diagnosed with mental health conditions.

These inequalities motivated this project, which focuses on exploring the critical question: Does the number of mental health treatment facilities in a state improve mental health outcomes for its residents?

As we constantly face increasing mental health concerns nationally, it’s easy to presume that building more facilities is the solution. By further studying the national datasets from 2023, the project focuses on revealing whether availability is the solution to improving mental health outcomes or if there are deeper structural issues that also need to be addressed to ensure meaningful mental health support and care for all.


Research Question and Hypothesis:

Research Question: How does the geographic distribution of mental health treatment facilities impact community mental health outcomes in the U.S? 

Hypotheses:
H0: There is no statistically significant relationship between facility access and mental health outcomes 
H1: There is a statistically significant relationship between the availability of mental health treatment facilities and mental health-related outcomes. 


Data Source and Variables:

National Substance Use and Mental Health Services Survey (NSUMHSS)
  1. Source: Substance Abuse and Mental Health Services Administration (SAMHSA)
  2. Unit of aggregation: Facility-level data, which has been aggregated to the state level
  3. Time: 2023
  4. Key Variables:
    MHFLAG - If the facility offers mental health services
    SERVMHOSP, SERVMOUT, SEERVMRES, SERVMPHP - Types of mental health care provided (inpatient, outpatient, residential, partial hospitalisation)
    INSURMCAID, INSURMCARE, INSURPRIV - Accepted insurance types
    OWNERSHIP - Ownership model ( e.g, public, nonprofit, for-profit)

 Behavioural Risk Factor Surveillance System (BRFSS) 2023    
  1. Source: Centres for Disease Control and Prevention (CDC) 
  2. Unit of aggregation: State level 
  3. Time: 2023
  4. Key Variables:
    _MENTI14D - Number of days of poor mental health in the past 30 days (continuous outcome) 
    ADDEPEV3 - Depression diagnosis status (binary outcome) 
    EDUCA - Education Level
    INCOME2 - Income bracket
    HLTHPLN1 - Insurance status
    _STATE - FIPS code for state-level merging


How I Approached the Analysis:

For this project, I merged and cleaned the state-level data from the two sources: the 2023 Behavioural Risk Factor Surveillance System (BRFSS) and the 2023 National Substance Use and Mental Health Services Survey (NSUMHSS).

Data Preparation:
I started by filtering the NSUMHSS data to include only mental health facilities, then counted the number of facilities in each state. Using state population data helped me calculate the facility density per 100,000 residents, which later helped compare access to care across states of varying sizes.

For the BRFSS dataset, I was able to select variables which were related to mental health and socioeconomic status: 
  • Average number of poor mental health days 
  • Prevalence of depression
  • Income level, education, and health insurance coverage

Once the data was cleaned (missing or invalid responses were removed), I aggregated the dataset at the state level.

Creating a Socioeconomic Index:
I used a Principal Component Analysis (PCA) to combine income, education, and insurance coverage into a single socioeconomic status (SES) index to grasp the broader structural conditions. This method aimed to reduce multicollinearity and simplify the regression model.

Statistical Analysis:
I then conducted a correlation analysis, further exploring the relationship between facility access, mental health outcomes, and socioeconomic status. I then built a multiple linear regression model to examine the impact of facility density and SES index outcomes on poor mental health days and depression rates.

Logistic regression was considered as a future extension but was not implemented in this version of the analysis.


Results and Interpretation:
Going into the analysis, it seemed reasonable to think that more mental health treatment facilities would be the solution to better mental health outcomes. However, the study illustrated a more refined image. States like Maine and Vermont were seen to have higher facility density, but the relationship between facility density and the number of poor mental health days reported (r = -0.19, p > 0.05).

A comparable pattern emerged when looking at the depression diagnosis rates, although that did have a slight positive correlation with facility density (r = 0.19), which again was not statistically significant. These findings propose that simply increasing the number of facilities/ treatment centres does not directly correlate with improved mental health at the population level. 
This choropleth map visualizes the density of mental health facilities per 100,000 people across the United States utilizing the National Substance Use and Mental Health Services Survey (NSUMHSS). A color gradient ranges from bright yellow (low density) to deep purple ( high density). It effectively conveys the variation in mental health infrastructure across states. For example, based on the map, states like Maine and Vermont stand out with the highest facility density. At the same time, Texas, Florida, and Georgia, which are seen in yellow, signify limited access relative to their population. Through this visual, we can identify critical regional disparities that could mirror broader systemic issues, such as public health investment, mental health policy priorities, and social stigma around treatment. Some areas, such as the Southeast and some parts of the Midwest, show alarmingly low densities, which could affect treatment access and outcomes.  

To dig deeper into what truly makes a difference, I created a Socioeconomic Status (SES) index with the help of Principal Component Analysis. The created index is a combination of education, income and insurance. It was interesting to see that the SES index came out to be a much stronger predictor of mental health. States that had a higher SES score suggested that they had better access to economic and education resources, which also reflected in their logs of fewer poor mental health days. In a multiple linear regression, the SES index was the only statistically significant predictor of mental health outcomes (p < 0.001), while facility density remained insignificant. 

The findings uncover an essential point: although the number of facilities plays a role in access, the broader socioeconomic conditions that truly shape mental health outcomes.  States that have enough treatment centres still struggle with poverty, lack of education, and not enough insurance coverage, and they still manage to face higher levels of mental distress. Unfortunately, this tells us that the problem isn't just about physical proximity to care, but it is about whether people have the resources, means, knowledge and support to get to the care in the first place. 

From this study, we learn that access is not enough. We must know that expanding the mental health safety net in the U.S must go beyond increasing facility counts. We as a community need to think about the potential policies we could develop that address these structural inequalities, such as improving education, raising incomes, and creating stronger insurance coverage, to see improvements in mental well-being.


Conclusion:

This project focuses on exploring whether the number of mental health treatment facilities in a state significantly improves mental health outcomes. Through the integration of national datasets from NSUMHSS and BRFSS and the development of the socioeconomic status (SES) index using Principal Component Analysis, the results show that even though the number of facilities was different across the states, it was not a significant predictor of mental health outcomes such as poor mental health days or depression diagnoses. Socioeconomic factors such as income, education, and insurance coverage turned out to be more influential, highlighting that we need to dig deeper and go past the surface-level access and toward deeper, systematic conditions that shape mental health. 

That being said, the study had various limitations. A large part of the analysis relied on self-reported data, which again can be skewed due to bias or underreporting, specifically regarding mental health conditions. This was also a cross-sectional study using data from a single year (2023), which reduced the capability to asses the changes over time to infer casuality. The data was aggregated at the state level, meaning important local or community-level nuances, such as urban-rural differences, were likely missed and could not be further explored. 

If extended beyond the scope of this course, I would like to explore this concept further by incorporating multi-year data to track trends over time and applying logistic regression or multi-level modelling to learn about the interaction of regional and individual factors. Including qualitative data or community-based surveys to understand the difficulties beyond numbers would be interesting. For example, this could range from cultural attributes to stigma and how the community perceives quality of care. Another dimension worth exploring is the role of telehealth and digital mental health services, which have rapidly expanded, especially after the COVID-19 pandemic. Incorporating data on telehealth access and usage could reveal how virtual care is reshaping mental health support and whether it offers a viable solution for underserved regions. Ultimately, the project lays the groundwork for reconsidering how we measure and improve mental health infrastructure, moving the focus from quantity to one of equity and impact.

References:

  1. Centers for Disease Control and Prevention (CDC). 2023. Behavioural Risk Factor Surveillance System (BRFSS). U.S. Department of Health and Human Services. https://www.cdc.gov/brfss/index.html.

  2. National Alliance on Mental Illness (NAMI). 2023. Mental Health by the Numbers. https://www.nami.org/about-mental-illness/mental-health-by-the-numbers/

  3. National Institute of Mental Health (NIMH). 2023. Mental Illness. U.S. Department of Health and Human Services. https://www.nimh.nih.gov/health/statistics/mental-illness.

  4. Substance Abuse and Mental Health Services Administration (SAMHSA). 2023. National Substance Use and Mental Health Services Survey (NSUMHSS). U.S. Department of Health and Human Services. 
    https://www.samhsa.gov/

  5. U.S. Census Bureau. 2023. Population Estimates Program. https://www.census.gov/programs-surveys/popest.html