Thinking Out Loud: A Qualitative Study of Health Information User Experience in People with Disabilities

This qualitative study identifies specific usability barriers and "pain points" faced by people with disabilities when seeking COVID-19 health information online, highlighting challenges related to distracting layouts, physical fatigue, and low trust in AI-generated content, while recommending that future digital health design incorporate direct user feedback from this community.

The Gist

Imagine you are trying to find a specific recipe in a giant, chaotic library. Now, imagine that library is the internet, and you are looking for health advice about COVID-19. For most people, this might be a bit annoying. But for people with disabilities, this library is often designed with broken ladders, confusing maps, and walls that move when you try to climb them.

This research paper is like a "behind-the-scenes" tour of that library, but instead of just asking people how they feel about it later, the researchers asked ten people with disabilities to talk out loud while they were actually searching. It's like having a tour guide narrate their thoughts in real-time as they navigate the maze.

Here is what they found, broken down into simple ideas:

1. The "Distracting Layout" Problem (The Cluttered Kitchen)

For people with attention issues (like ADHD), the internet is like walking into a kitchen where the lights are flashing, the radio is playing three different songs, and the ingredients are scattered everywhere.

• The Finding: Even if the information is there, the way the webpage is designed is so messy and "noisy" that it's hard to focus. It's not just about seeing the text; it's about the brain getting overwhelmed by the visual clutter.
• The Result: People with physical disabilities also reported getting "tired" just from the effort of clicking and scrolling. It's like trying to run a marathon while carrying a heavy backpack; the search itself becomes exhausting.

2. The "Robot Librarian" (AI)

The internet now has a new helper: Artificial Intelligence (AI). It's like a robot librarian that instantly summarizes books for you.

• The Finding: Almost everyone saw this robot summary first. But here's the catch: only one person trusted it. Most people were suspicious, like, "Is this robot making things up? Did it read the whole book, or just the cover?"
• The Problem: The robot often didn't say where it got its information, or it hid that fact in tiny, hard-to-read print at the bottom of the page.

3. The "Old News" vs. "Fresh Facts"

When people looked for info, they found a mix of sources:

• The Good Guys: Government sites (like the CDC) and hospitals were like the reliable, serious librarians. People trusted them, but sometimes the information was old (like a map from 2020 when the city has changed since then).
• The Confusing Guys: There were also news articles and ads. For the "COVID-19 vaccine" search, the top results were often ads for pharmacies saying, "Come buy this here!" or news stories about political fights.
• The Metaphor: Imagine you are looking for a weather report, but the first thing you see is an ad for an umbrella store and a news story about a storm in a different country. It's hard to find the actual weather forecast.

4. The "Trust" Issue

The study found that people with disabilities are very wary of misinformation.

• The Fear: Some people were terrified they would see fake news (like "vaccines cause autism" or "a parasite cure works for a virus"). One person said, "If I see that, I'm going to lose it."
• The Reality: They knew the internet was full of traps. They wanted clear, honest answers from trusted sources, but the "noise" of ads and fake news made it hard to find the truth.

The Big Picture: Why This Matters

Think of the internet as a public park. If the park has broken paths, confusing signs, and no benches for people who get tired easily, then people with disabilities can't enjoy the park. They can't get the "fresh air" (health information) they need to stay safe.

The Main Takeaway:
The internet isn't just "hard to use" for people with disabilities; it can be exhausting and dangerous because it's hard to tell what is true and what is fake. If we want to protect everyone from diseases like COVID-19, we need to redesign the "park." We need to make the paths smoother, the signs clearer, and the "robot librarians" more honest, so that everyone can find the information they need without getting lost or tired.

In short: The researchers are saying, "We need to listen to the people who are struggling in the maze and fix the maze for them, not just for us."

Technical Summary

1. Problem Statement

People with disabilities face significant barriers when seeking health information online, particularly regarding COVID-19. While quantitative data suggests lower satisfaction with online health content among this demographic, there is a lack of granular, real-time insight into the specific "pain points" (usability and accessibility frustrations) they encounter.

• The Context: The integration of Artificial Intelligence (AI) into search engines (e.g., Google's AI overviews) has altered how users access information. AI is known to harbor biases against people with disabilities.
• The Gap: Existing research often relies on retrospective surveys or studies conducted by non-disabled researchers without centering the voices of the disability community ("nothing about us without us"). Furthermore, few studies examine how people with disabilities interact with AI-generated health information in real-time, specifically regarding critical topics like vaccines, symptoms, and misinformation (e.g., ivermectin).
• The Risk: Inaccessible or confusing information can lead to delayed health decisions, reduced vaccination rates, and increased morbidity/mortality for a population already at higher risk for severe COVID-19 outcomes.

2. Methodology

The study employed a qualitative, concurrent think-aloud methodology to capture real-time user experiences.

• Participants: 10 adults with self-reported disabilities recruited from a prior quantitative survey.
  • Demographics: Predominantly White (70%), female (70%), and highly educated (50% held graduate/professional degrees).
  • Disability Types: 50% reported developmental/mental health disabilities (e.g., ADHD, autism), and 50% reported physical disabilities (e.g., Type 1 Diabetes, POTS, MS).
• Procedure:
  • Duration: April 1 – May 1, 2025.
  • Task: Participants performed four specific search tasks using their own devices and assistive technologies:
    1. "COVID-19 infection"
    2. "COVID-19 symptoms"
    3. "COVID-19 ivermectin"
    4. "COVID-19 vaccine"
  • Protocol: Participants used the concurrent think-aloud method, verbalizing their thoughts, frustrations, and decision-making processes in real-time while navigating search engines (9 used Google, 1 used DuckDuckGo). Sessions were recorded via Zoom and transcribed.
• Analysis:
  • Transcripts were analyzed using Thematic Analysis with both deductive (pre-defined) and inductive (emergent) coding.
  • Deductive Codes: AI-generated content, accessibility/usability, information sources, COVID-19 relevance.
  • Inductive Codes: Advertising, news, trust, sentiment.
  • Tools: Taguette (open-source qualitative data analysis software) was used for coding. Peer debriefing was utilized to ensure validity and reduce bias.

3. Key Contributions

• Real-Time UX Data: Unlike retrospective studies, this research captures the immediate cognitive and physical load experienced by users with disabilities as they navigate complex search results.
• AI Interaction Insights: It provides specific evidence of how AI-generated search summaries (overviews) are perceived, trusted, and navigated by users with diverse disabilities.
• Disability-Centered Design: The study adheres to the "nothing about us without us" principle, prioritizing the direct narration of the disability community to inform Universal Design principles.
• Identification of Specific Pain Points: It moves beyond general "accessibility" to identify specific friction points such as "distracting layouts" for neurodivergent users and "physical fatigue" for those with chronic physical conditions.

4. Key Results

The analysis yielded six primary themes and associated pain points:

A. Accessibility and Usability

• Neurodivergent Challenges: Participants with ADHD reported that webpage layouts were "jumbled," "distracting," and difficult to process, leading to cognitive overload.
• Physical Fatigue: Participants with physical disabilities (e.g., POTS, Type 1 Diabetes) reported that the time required to navigate search results caused physical exhaustion, limiting their ability to persist in finding accurate information.
• Assistive Tech Failure: Some screen readers failed to parse certain websites, forcing users to download or print PDFs, adding unnecessary steps.

B. AI-Generated Information

• Prevalence: AI-generated results appeared as the top result for the first three search prompts for all Google users.
• Trust Issues: Only one participant expressed trust in AI-generated summaries. Most viewed them with skepticism, noting that sources were often unlisted or the content was a "mixture" of sources without clear attribution.
• Labeling: AI labels were often in "very small print" at the bottom of the results, making them easy to miss.

C. Information Sources

• Trusted Sources: Government (CDC, FDA, NIH) and hospital (Mayo Clinic) websites were highly trusted. However, participants noted issues with outdated dates (e.g., 2020/2021 content appearing alongside 2024 content).
• Confusion: Some participants could not identify the origin of certain government sites (e.g., UK's NHS).

D. Peer-Reviewed Articles

• Accessibility Barriers: While open-access articles (e.g., New England Journal of Medicine) were found, many participants struggled to distinguish between editorial pieces, systematic reviews, and original research.
• Currency: Many articles found were from 2020, predating the end of the public health emergency, leading to potential misinformation regarding current protocols.

E. News and Advertising

• Irrelevance: Search results for "COVID-19 vaccine" were heavily populated with sponsored ads (pharmacies like CVS, Walgreens) and news articles.
• Misinformation Risk: Participants expressed anxiety about encountering anti-vaccine propaganda or news sources they did not trust (e.g., Fox News affiliates) in their results.
• Geographic Mismatch: Ads and news often targeted locations irrelevant to the user.

F. Sentiment and Trust

• High Distrust: There was significant distrust regarding ivermectin as a treatment and the link between vaccines and autism.
• Emotional Toll: The search process induced frustration and anxiety, particularly when users feared encountering misinformation that could harm their health decisions.

5. Significance and Implications

• Public Health Impact: The study concludes that the "usability gap" in digital health information directly threatens public health equity. If users with disabilities cannot easily find or trust accurate COVID-19 information, they are less likely to get vaccinated, exacerbating their risk of severe illness and death.
• Design Recommendations:
  • Co-Design: Technology and health information platforms must be co-designed with neurodivergent and physically disabled users to reduce cognitive load and physical fatigue.
  • AI Transparency: AI search summaries must clearly label sources, dates, and the AI nature of the content to build trust.
  • Content Curation: Search algorithms should prioritize up-to-date, verified medical information over news or advertising, especially for health-related queries.
• Future Research: The authors call for larger studies with more diverse racial/ethnic representation and broader inclusion of mental health disabilities, utilizing methods like focus groups and Community-Based Participatory Research (CBPR).

In summary, the paper argues that accessibility is not just about screen readers; it is about the cognitive and physical endurance required to navigate a cluttered, AI-saturated, and often outdated information ecosystem. Addressing these pain points is essential for achieving health equity.