Inspiration

In 2024, 41% of the 1,791 new HIV diagnoses in New York City were Latino. Latino men were diagnosed at 46 per 100,000, white men at 8 (NYC DOHMH 2024 report). Among Hispanic people tested at CDC-funded sites, 1 in 4 knew about PrEP (CDC MMWR, 2019 data).

Finding a clinic is part of the gap. The CDC locator cannot filter for Spanish-speaking clinics. In a 2023 study of Latino men looking for PrEP, only 8 of 47 clinics on the CDC list met their two main needs. The barriers they named were information in Spanish, cost, and fear that being undocumented blocks care.

PEP has to start within 72 hours of an exposure. Someone in that window has no time to call ten clinics and ask who speaks Spanish.

What it does

VidaSegura is a web app that matches a person to NYC clinics for PrEP, PEP or HIV testing. It runs in English and Spanish (usted), with a language toggle.

The welcome screen says Gratis, Privado, Sin estatus migratorio, and offers three choices: prevent HIV, get tested, or "possible exposure in the last 72 hours."

A short questionnaire asks language, borough, ZIP code (optional) and insurance, plus PrEP, PEP or testing if the person has not already picked one. Then a dashboard lists 40 real NYC clinics. The best 3 matches show first: clinics with the service confirmed, then the same ZIP, then the nearest. A ZIP code wins over the borough answer. Each card shows PrEP, PEP and free testing labels, or "Call to ask" when the service is not confirmed. It also shows whether Spanish is spoken, cost, whether no insurance is OK, hours, and a tap-to-call button. Staten Island has no listed clinics, so those users see the nearest ones.

A chat ("Quiero chatear"), run by Claude Sonnet 5, changes the filters when someone types "Necesito algo en Queens" or "¿Es gratis?", and answers in the language they wrote in. If someone writes that they were exposed yesterday, it switches the list to PEP and tells them to call today. It never names a clinic and never gives medical advice. The clinic list stays on the cards.

There is no account. Answers stay in memory and never appear in the web address. The only thing we count is anonymous: service, language, borough and date.

How we built it

We built the app on Base44. The clinic list is a Base44 entity with 22 fields, loaded from a 40-record clinic spreadsheet we assembled beforehand. The welcome screen, questionnaire, dashboard and chat were generated and edited in Base44 from a written build prompt.

The chat sits on Claude Sonnet 5. It can set filters (service, borough, ZIP, insurance, Spanish) and nothing else. Clinic names, phone numbers and hours come only from the database.

The research behind the design is in the repo: a dashboard with a short chat beside it, because long chats lose people and typing is the hard part for low-literacy users.

Challenges

  • Unknown is not "no." In our data, a TRUE flag means documented evidence and FALSE means not confirmed. A card never says "No PrEP." It says "Call to ask about PrEP."
  • Base44 shows no BAA that we could find, so the app stores no health data. That is why it has no account, no name and no saved searches.
  • Keeping the chat from acting like a doctor or a directory. It only changes filters.
  • Staten Island has no clinics in our list. The app falls back to the nearest ones instead of an empty page.

Accomplishments that we're proud of

  • A person can go from the first screen to a call button in a short questionnaire and no sign-up.
  • 40 real clinics, no invented results, with uncertain information shown as uncertain.
  • The same flow works in Spanish and English, including the 72-hour PEP path.

What we learned

  • Privacy decided the design. In a New York study of people living with HIV, half Latino, people worried that "anybody could get into the Internet and go into your information." So the app asks for nothing it does not need.
  • A chat alone is the wrong shape. NN/g found people get lost scrolling a chat. A dashboard holds the answers and the chat adjusts them.
  • A list people can tap worked in earlier work: in a 2025 follow-up of the SaludFindr app, 10 of 15 men who tested at a clinic used its clinic list.

What's next

  • Voice input and output.
  • A caseworker: the chat sends the request to a clinic caseworker, who calls back. This needs a HIPAA setup and a BAA before any real patient detail is stored.
  • Clinic data that updates itself. Today the 40 records are fixed and each carries a last-verified date.
  • More clinics, starting with Staten Island, and clinics that ask to be listed.

Built With

  • base44
  • claude
  • elevenlabs
  • react
  • tailwind
Share this project:

Updates

Submission history