A Cross-US Exploration of Mental Health in the Tech Workplace

Basic Info

Course: CSE 4507 — Introduction to Visualization

Project Title: A Cross-US Exploration of Mental Health in the Tech Workplace (OSMI 2014 + CDC Data)

Team: Aman Verma, Sayee Sreenivas G B

Emails: aman.v@wustl.edu, g.sayeesreenivas@wustl.edu

Student IDs: 529549, 527548

Table of Contents

1. Introduction

Mental health in the workplace has become a central issue in the tech industry, where long hours, distributed teams, and high expectations are common. In this project, we build an interactive web-based dashboard that combines the OSMI Mental Health in Tech Survey (2014) with CDC PLACES 2024 depression and obesity data to explore how workplace support and regional context shape mental health outcomes.

Our final website consists of six coordinated visualizations, all implemented in D3.js v7, that allow users to move smoothly between: (1) national depression patterns, (2) state-level physical vs mental health trends, (3) workplace experiences from the OSMI survey, and (4) individual-level stories embedded in the data.

The goal is not just to show where depression rates are high, but to ask: how does the tech workplace help or hurt people living with mental health conditions, and how does this vary across the U.S.?

2. Background and Motivation

The OSMI survey has become a widely cited source for understanding mental health in tech. It captures how employees experience stigma, support, and practical barriers when they consider seeking treatment. At the same time, the CDC’s PLACES dataset reveals how depression and other health indicators vary geographically across the U.S.

Our motivation was to bring these two perspectives together:

By pairing workplace-level experiences with regional health patterns, we hoped to identify:

Ultimately, we want the visualizations to help employers, policymakers, and researchers better understand where support is working and where it is clearly not enough.

3. Project Objectives

The main questions we focused on are:

Across all of these, our broader objective is to use visualization to make hidden patterns visible and to give users a flexible way to explore the data rather than just reading a static report.

4. Task Abstraction

We thought about our design using classic visualization task categories: who is using the system, what are they trying to see, and why they are looking.

4.1 Users

4.2 Data and Tasks

High-level tasks:

Why these tasks matter:

These tasks support typical questions like “Is my state doing better or worse than others?”, “Do states with higher obesity also tend to have higher depression?”, and “What workplace factors tend to show up when people say mental health interferes with their work?”.

5. Data

We use three main data sources.

5.1 OSMI Mental Health in Tech Survey (2014)

The OSMI dataset includes roughly 778 U.S. respondents working in tech. For each respondent, we have:

5.2 CDC PLACES 2024 — Depression and Obesity

These variables give us the broader mental health and physical health context for each state where OSMI respondents live.

5.3 Supporting Geospatial Data

6. Data Processing

The raw datasets were not ready to visualize directly, so we spent a significant amount of time on cleaning and preprocessing.

The final processed data flows into six separate D3 visualizations, all loaded from data/ and wired up through:

7. Visualization Design

Our final site integrates six main views. Below we describe the intent, design choices, and example screenshots for each one.

7.1 U.S. Depression Choropleth + County Drill-Down

Figure 1: US Depression Rate by State Choropleth Map
Figure 1. Choropleth showing depression rates by state with drill-down into county-level depression patterns.

This view uses a state-level choropleth as the starting point. Each state is color-coded by depression prevalence. Hovering reveals a tooltip with the state name and exact rate. Clicking on a state smoothly transitions into a state-specific Albers projection that reveals county-level depression rates.

A reset button and the Escape key both take users back to the national view. We used a white background and a fairly restrained color palette so that the color scale remains readable for small states and the overall map feels clean.

7.2 State Obesity vs Depression Scatter Plot

Figure 2: Obesity vs Depression Scatter Plot
Figure 2. Scatter plot comparing obesity and depression rates by state, with regional colors and a fitted trend line.

This scatter plot shows each state as a dot, with:

We add a regression line and display a correlation coefficient to summarize the overall relationship. Median reference lines divide the chart into quadrants, making it easier to spot states that are high/low on both variables.

An interactive legend lets users highlight or isolate regions, and hover tooltips show the exact values for each state. The scatter plot uses a slightly darker background and clear axis labels to keep the points visually distinct.

7.3 Work Interference Bar Chart (State-Level)

Figure 3: Work Interference by State
Figure 3. Horizontal bar chart showing how often mental health interferes with work for respondents in a selected state.

This visualization focuses directly on how respondents feel mental health impacts their work. A dropdown allows users to pick a state. For that state, we show the percentage of respondents in four categories:

Animated transitions make changes across states feel smooth and help the user track the movement of bars. A small hint box explains how to read the chart (“taller bars toward Sometimes/Often generally indicate higher work impact in this state”).

7.4 Radar Chart: Workplace Support Profile (National)

Figure 4: Workplace Mental Health Support Radar Chart
Figure 4. Radar chart summarizing key workplace mental health support dimensions at the national level.

The radar chart summarizes several workplace support dimensions using aggregated OSMI data:

The polygon shape visually highlights which areas are relatively strong versus weak. For example, benefits may be relatively common, while anonymity and ease of leave lag behind. Hover tooltips show exact percentages on each axis. We used a darker background and subtle grid lines to keep the polygon readable.

7.5 State Mental Health Dashboard (Sortable Table)

Figure 5: State Mental Health Profile Dashboard
Figure 5. Sortable dashboard comparing states across key workplace mental health indicators.

This dashboard condenses several state-level OSMI indicators into a single table. For each state, we display:

Users can click any column header to sort the table by that metric (ascending or descending). Small ▲/▼ arrows make the sort state visible. This view is meant to support “which states are highest/lowest?” tasks and makes it easy to spot standout states.

7.6 Parallel Coordinates: Individual Experiences

Figure 6: Parallel Coordinates of Workplace Mental Health Factors
Figure 6. Parallel coordinates plot showing how individual respondents line up across multiple mental health and workplace dimensions.

The parallel coordinates plot is our way of showing that behind every state-level statistic are individual stories. Each line corresponds to one respondent. Axes include:

A dropdown lets users highlight respondents from a particular state. Lines from that state are drawn in a stronger color, while others fade into the background. Hovering reveals tooltip details for individual lines. This view is more complex, so we include an “How to read this view” instruction panel directly below the heading.

8. Visualization Evaluation

Overall, we feel the six views complement each other and support a range of tasks from overview to detailed exploration. Here we summarize what works well and what could be improved.

8.1 What Worked Well

8.2 Limitations and Trade-Offs

9. Milestones and Schedule

9.1 Milestone 1

In the first milestone, we focused on data access, cleaning, and getting basic versions of our core visualizations running.

9.2 Milestone 2

By Milestone 2, we aimed to have all of our core visualizations implemented and wired up with cleaner interactions.

9.3 Final Week

10. Reflection

Working on this project taught us as much about data limitations and design trade-offs as it did about the actual topic of mental health in tech.

On the technical side, we got comfortable juggling multiple datasets and feeding them into coordinated D3 visualizations. Handling GeoJSON/TopoJSON, projections, and drill-down interactions felt challenging at first, but seeing the county-level map animate into place was very satisfying.

On the design side, we repeatedly had to scale back complexity to keep things understandable. For example, we originally considered even more derived indicators and compound scores but realized that simpler, clearly labeled metrics did a better job of communicating the story. The parallel coordinates view was a good reminder that powerful visualizations need guardrails, such as instructions and filtering, to be usable.

Content-wise, it was sobering to see how many respondents reported that mental health “Sometimes” or “Often” interferes with their work, and how many felt unsure about their benefits or worried about consequences. At the same time, we saw that support is not uniformly bad everywhere—some states and workplaces clearly do better.

11. Future Work

If we had more time, there are several directions we would like to explore:

12. References