Hona (YC W24)

Two major problems with diagnoses codes / problems lists: 1. They’re not always kept up to date. 2. They often lack context/explanations. We’ve tried to address this by having our LLM maintain the diagnoses based on what it knows about acute / chronic conditions combined with data from the patient’s record. Goal here is to minimize false positive diagnoses so that clinicians can actually trust the list and code accordingly. We aimed to enhance understanding by detailing the reasons a patient sought care when a diagnosis was made, including observations and linking back to the original sources. This method has greatly improved the accuracy and reliability of risk scores. Most groups we work with don’t have a full, longitudinal view of how sick / complex their patients are, and we’re fixing this.

We've been pretty heads down since my last update in 2023, but I wanted to poke my head out and share that we've begun YC and have launched in their directory! https://lnkd.in/gq9Khyug At a high-level, we're happy about where we're at in terms of product right now. We're being used by some leading institutions in urology as well as oncology and Hona has worked reliably as a summarization tool for highly complex patients. So now we have a tool that's really good at pulling and summarizing, but our next step is to identify the most valuable things to pull and summarize then bake them into the product. We think a lot of diagnoses get missed when patients switch between different parts of the healthcare system. So, we're building tools to find all those hidden diagnoses in clinical notes and turn them into risk scores. The goal here is to help providers get the full picture of how complex a patient's needs are, decide where to focus their efforts, and increase reimbursement. Further, we want Hona to offer providers a bird's eye view of patient populations. It should highlight patients with the highest risk scores and allow filtering by various factors—like what kind of diagnoses they have, which doctors they've visited, their location, and other details. This way, providers can quickly see who might need more attention and understand their patient population better. So, for Q1, that's our focus—building the above features, promoting it at conferences, and continuing to integrate Hona into new EHRs.

We're excited to announce our partnership with Particle Health 😃 Particle helps us get a full view of the patient and has been awesome to work with. They're clearly committed to holistically improving care, whether that's through working with giant provider organizations or a little startups like us. More info can be found here:

Hona (YC W24) 🤝 Y Combinator We're ecstatic to announce that we'll be joining the next batch!!! Big thanks to everyone who has believed in us thus far, we'll make you proud.

After 4 months of building, our team went on a bit of a roadshow over the last week demoing Hona to investors / corporations including Stanford University, Plug and Play Tech Center, Google, INSIDE.COM and more. The best part about building this is doing it with my cofounders Shuying (Sofia) Zhang and Danielle J. Yoesep. We often divide and conquer different aspects of the business, but we had a great time showing off our collaboration and how our work has come together. The rest of our year is pretty simple: 1. Embed Hona into as many common workflows as possible. 2. Make the product / brand beautiful. 3. Provide remarkable customer service to our clients. 4. Publish our White Paper. 5. Spread the word.

Now that we’ve built our core tech and have some validation with customers, we’re in what I would consider the second major phase of hona.ai. We’re going to be spending our next sprint focused on making our product beautiful and seamless. Our thesis is that healthcare tools are lacking here and that design must be a core principle of our company, even though our end product is primarily text. We want to bring what we learned in big tech to a new industry - meaning that we have to make some unconventional tradeoffs + investments upfront to ensure that design isn’t an afterthought, but rather, a lever for us to win. The end goal here is to make a providers’ interaction with hona.ai as easy as the interactions we were building for consumers in our past lives.

A few months ago, I left my dream job PMing at Meta to build hona.ai with Danielle J. Yoesep and Shuying (Sofia) Zhang. I'm extremely excited to announce that our web app is live on desktop. We use AI to make sense of medical history. Here's why this means so much to me: Two years ago, I broke my lower leg while snowboarding. The whole experience with healthcare systems across multiple states was a nightmare. Here's what happened: 1. Friends called ski patrol, who took me down to urgent care on a sled. 2. At urgent care, they put me to sleep to saw off my boot. 3. I went by ambulance to a hospital, had surgery, stayed overnight, and did occupational therapy. 4. I went back to SF, then to the emergency room, where the wait for more medication was so bad I left without getting help. 5. I had to find and follow up with specialists, physical therapists, and imaging centers for months. Ultimately, more than 20 people attended to me, and with so many faces involved, things often felt impersonal, resulting in details getting lost along the way. I found myself bringing a picture of my X-Ray (that I wasn't even supposed to have taken) along with printed medical records to each institution. The process felt archaic, causing me unnecessary suffering. Not only were the various institutions I visited relying on my ability to fetch my own medical records, but each provider was burdened with the task of making sense of 300+ pages of notes from people they didn't know. We started working with care providers to learn more about this issue. They're seeing up to 60 patients a day, so understanding each person's medical history is nearly impossible. By doing some quick math, we found that they're sifting through over 4200 years of medical history every week. To fix this, we made Hona - Whether you're a provider, attorney, or someone else who has to interpret medical history, you can upload medical records along with a fully customizable template, and we'll generate a summary. Our technology safeguards patient privacy, works with any word count, and can turn 100s or even 1000s of pages of medical history into a crisp one-pager. Plus, we're integrating Hona with electronic health record systems for automatic uploads, and fetching documents from external health systems. This enables a comprehensive TL;DR without the need to sift through cluttered or incomplete files. If we get this right, it will transform electronic health records the way TikTok transformed video consumption and ChatGPT transformed information retrieval. Sifting through information is outdated and inefficient. I prefer TikTok because I don't have to sift through video thumbnails, and ChatGPT because I don't have to sift through search results. Healthcare providers, like everyone else in 2023, should get only information they need precisely when they need it. By using Hona's custom patient overviews, instead of manually sifting through thousands of health records, this can become a reality.