6 years later and AI is here: Revisiting Blog Post 1 "What we're missing"
As we announce our Series A raise and the expansion of our team, I’ve been having this urge to tie it all back to where we started. We’ve been doggedly building toward a simple, transformative vision for 9 years. Six years ago, I wrote our first “thought leadership” post - and I think it actually has aged okay! (As for me…the headshots may tell a different story.)
Today, I’ll be taking a look back at “What we’re missing” - a battle cry from 2019 to pay attention and build better health data to power growing machine learning models, before the AI bots grow and take over the world with misinformed decisions on your health!
It’s 2025, and AI is here. Are we ready?
Note: Light orange = 2019 original. Dark orange = 2025 commentary.
➡️ Read the original 2019 post
| 2019 Original | 2025 Reflections |
|
For years, we’ve been hearing the same refrain about the digital transformation of healthcare.
“Patient-centered care”1 will introduce higher quality, individualized, more palatable treatment.
“Personalized medicine,” brought to life by the magic of genomics2, will ensure that every patient receives the right treatment, at the right time.
With “AI and machine learning,” a bot — a la Watson3 — will make rapid diagnoses, removing human error and excruciating guess-and-check from the diagnostic odyssey.
1 “and patient-focused drug development”
2 That’s so 2019, it’s AI now baby.
3 ChatGPT; definitely not Epic AI; maybe Claude, iykyk.
|
1 “and patient-focused drug development”
2 That’s so 2019, it’s AI now baby.
3 ChatGPT; definitely not Epic AI; maybe Claude, iykyk.
|
|
But have you been to the doctor recently?
The digital health community has been so busy dreaming of the possibilities of a fully tech-enabled ecosystem that we’ve forgotten the basics.4
4
Still mostly true, but glimmers of hope abound now. We, the tens of thousands of scientists, clinicians, and businesspeople who collectively make decisions that shape the care our communities receive, are holding ourselves more to account for the things that matter, directly, to people. Maybe it’s because the ‘tried-and-true (?)’ US healthcare system has been under seemingly constant stress and siege for 5 years now - pandemic-era chaos, vaccine hesitancy, GLP-1 inhibitors and the rise of direct-to-consumer drugs, the murder of the United Healthcare CEO, the MAHA movement…even if negative, opposing forces can have the combined effect of forcing a hard look at ourselves and opening a path for transformation.
|
4
Still mostly true, but glimmers of hope abound now. We, the tens of thousands of scientists, clinicians, and businesspeople who collectively make decisions that shape the care our communities receive, are holding ourselves more to account for the things that matter, directly, to people. Maybe it’s because the ‘tried-and-true (?)’ US healthcare system has been under seemingly constant stress and siege for 5 years now - pandemic-era chaos, vaccine hesitancy, GLP-1 inhibitors and the rise of direct-to-consumer drugs, the murder of the United Healthcare CEO, the MAHA movement…even if negative, opposing forces can have the combined effect of forcing a hard look at ourselves and opening a path for transformation.
|
|
We’re hiring consultants to build patient experience presentations, while neglecting to add Keurig machines to our waiting rooms.5
Similarly, we’re scheming to use the latest in machine learning6 frameworks,
while forgetting to consider the quality of the data that we have to feed them.7
5 Hey fair, but there’s nice spa water at OneMedical now.
6 I, for one, left talk of “ML” in a closet with my skinny jeans 2 years ago.
7 New bold for emphasis! Say more!!
|
5 Hey fair, but there’s nice spa water at OneMedical now.
6 I, for one, left talk of “ML” in a closet with my skinny jeans 2 years ago.
7 New bold for emphasis! Say more!!
|
|
Let’s talk about that one, for a moment. There’s no doubt that improved methods in analytics hold immense promise for phase-shift improvement in the quality of care.8
But these new technologies require high-quality data food to work correctly.
This data food must provide a good level of detail on how different types of patients respond to treatments, and how the outcomes that they experience manifest over time.9
8
We can be a lot more specific about this now. As we look toward 2026, we now have broadly-available tools like ChatGPT that are eager to assist with all sorts of complicated questions, literally only requiring 30 seconds of prompt writing and sometimes a few minutes of loading to get to a new answer.
9
Ultimately, we need the insights that come out of our use of these tools to be meaningful and useful!
This has never been more important. Perhaps the part that concerns me the most is the opacity of the sources right now - it’s so hard to clearly understand where the AI that you’re talking to is getting its information from - what are the sources, and what important pieces of info are missing? Would you want all your health decisions to be based solely on what is in the doctor’s notes in your EHR portal?
|
8
We can be a lot more specific about this now. As we look toward 2026, we now have broadly-available tools like ChatGPT that are eager to assist with all sorts of complicated questions, literally only requiring 30 seconds of prompt writing and sometimes a few minutes of loading to get to a new answer.
9
Ultimately, we need the insights that come out of our use of these tools to be meaningful and useful!
This has never been more important. Perhaps the part that concerns me the most is the opacity of the sources right now - it’s so hard to clearly understand where the AI that you’re talking to is getting its information from - what are the sources, and what important pieces of info are missing? Would you want all your health decisions to be based solely on what is in the doctor’s notes in your EHR portal?
|
|
Right now, we’re dependent on incomplete and skewed datasets that focus on a small portion of the patient’s overall experience of his condition and the impact of treatment. Simply put, we don’t yet have the data that we need to realize the potential of advanced analytics in healthcare.10
10
But the cool thing is, we’re starting to see the machines being built that can ingest the right data food, if we have it to give them.
|
10
But the cool thing is, we’re starting to see the machines being built that can ingest the right data food, if we have it to give them.
|
|
Consider a patient with a complex condition, who visits his care team 5–10 days per year. Most of the information that we have on whether his treatment plan is effective is based on those clinical encounters, 15-minute interactions in which clinicians drive the conversation. If the patient is confident, or has an activist caregiver, his observations may also be shared, but generally in an anecdotal way — “I’ve also been noticing this…” As a result, relevant details are often left undiscussed, and unrecorded.11
11
In this way, even the clinical notes are an incomplete picture of what’s actually happening - how this person feels when they wake up on Tuesday, and how they get through their day.
|
11
In this way, even the clinical notes are an incomplete picture of what’s actually happening - how this person feels when they wake up on Tuesday, and how they get through their day.
|
|
|
|
Even the device-measured data that we collect at the appointment is likely to be skewed — the patient is often at the clinic because something different from the usual is going on, so what’s observed on that day is unlikely to be representative of what “usual” is.12
12
And frankly, there is a wild amount of measurement error. Two years ago, we went on an incredibly stressful wild goose chase to understand my son’s blood pressure - at 7 months old, it was flagged as high (>99th %ile). Not the thing you want to hear when you’re trying to let your baby with a pre-existing heart condition learn how to soothe themselves a bit - we literally hardly put him down for the next 2 months. In the meantime, 3 providers each read it as very high, we ended up doing a kidney ultrasound, and I learned way too much about primary hypertension…until finally we learned from a nephrologist that this is a common mistake, small babies need to have their bps read with the assistance of a Doppler to hear properly! A $400 piece of equipment to do the test properly. My son’s blood pressure was borderline low. Meanwhile, months of incorrect readings in his chart - and they’re still sitting there, part of my son’s “medical record”.
For these reasons, and many more, the information that we collect at the point-of-care — clinical and claims data — is insufficient to truly understand the outcomes that patients are experiencing in their daily lives, and so is pretty poor-quality data food for our machines. Meanwhile, patients and their family members spend every day living with the outcomes of treatment. They’re observing changes in themselves and their loved ones, but don’t have a good method to track what they’re seeing. At best, some will quickly jot down notes of their children’s fatigue, or keep logs of a parent’s forgetfulness. Some are tracking their own symptoms, behaviors, moods, side-effects, and questions about treatment in binders, home-grown Excel sheets, and notes on their phones, lost forever to the millions of accumulating records that no one will ever read. |
12
And frankly, there is a wild amount of measurement error. Two years ago, we went on an incredibly stressful wild goose chase to understand my son’s blood pressure - at 7 months old, it was flagged as high (>99th %ile). Not the thing you want to hear when you’re trying to let your baby with a pre-existing heart condition learn how to soothe themselves a bit - we literally hardly put him down for the next 2 months. In the meantime, 3 providers each read it as very high, we ended up doing a kidney ultrasound, and I learned way too much about primary hypertension…until finally we learned from a nephrologist that this is a common mistake, small babies need to have their bps read with the assistance of a Doppler to hear properly! A $400 piece of equipment to do the test properly. My son’s blood pressure was borderline low. Meanwhile, months of incorrect readings in his chart - and they’re still sitting there, part of my son’s “medical record”.
|
|
We need to organize these observations — these home-reported outcomes — to harness the power of the lived experience of complex disease.13
We need to enable individual people to capture what they see, for the advancement of precise and high-quality care — for themselves, and for their communities.
13
And we have! To the tune of 4.5 million datapoints directly captured by our user community, to use themselves and to contribute to research. It’s happening!
|
13
And we have! To the tune of 4.5 million datapoints directly captured by our user community, to use themselves and to contribute to research. It’s happening!
|
| At Folia, we’re already laying the groundwork. We’ve spent the past two years working with the cystic fibrosis community to build a system to collect and translate the most important things that patients and caregivers notice each day, and many have already used their datasets to drive treatment changes. For many of our patients, ours is the most comprehensive set of information on the outcomes they’ve experienced that has ever been collected.14 |
14
Okay wow, there’s so much to say here, let’s see if I can be concise! As we got deeper in CF, we had the amazing opportunity to work alongside the CF Foundation and researchers from the University of Indiana, UPenn, and around the country to design a first-of-its-kind study to understand the impact of using a powerful new therapy, Trikafta, and whether this meant that people were able to pare down to fewer of the daily therapies that required 2+ hours per day historically.
Then, we started to expand to new conditions! Our Folia user community now includes people tracking for more than 500 different conditions, and I had to ask ChatGPT to summarize the research that our contributing users have made possible over the past few years! Check it out: ChatGPT summary
2025
2024
2023
2022
2021
This doesn’t even include the research results that our team & community will be sharing at conferences around the world this fall (see handy graphic below). 
Why does all this research matter? Because it builds the scientific story and credibility to rely upon the real, lived experiences of individuals to better understand chronic conditions and put our collective energy and resources into the therapies and other solutions that actually work. |
|
Together with growing datasets from genetic sequencing and passive tracking,
home-reported outcomes are the food that we need to feed the machine-learning15
beast — and step into the future of healthcare.16
15 AI
16 In the coming months, we’ll be sharing how Folia data is being used to feed the models that will make your health decisions better, faster, and personalized to what’s most important to you. In short - making healthcare make sense. Onward!
|
15 AI
16 In the coming months, we’ll be sharing how Folia data is being used to feed the models that will make your health decisions better, faster, and personalized to what’s most important to you. In short - making healthcare make sense. Onward!
|
