.png)
8 hours ago
1
I.
After continuously feeling fatigued and not knowing what else to suggest, my primary doctor referred me to a sleep clinic.
I went to the clinic with many questions but received no guidance. Did it matter what position I fell asleep in? If I woke up in the night, should I try to vary my position to get more data? The staff offered no answers. I remember being told by the staff that it was a huge issue when patients couldn't get enough sleep, as it rendered their stay and any collected data useless for a meaningful diagnosis.
On top of the stress of sleeping in a new place with equipment strapped to me, the clinic did little to make falling asleep easier. Bright, hospital-style light from the hallway seeped into my room, where no effort had been made to effectively block it. While not as bright as the outdoors, it was brighter than any room one would consider fit for sleeping. Throughout the night, I could clearly hear other visitors watching TV. Each time someone needed to use the bathroom, they had to alert the staff to walk them to the bathroom, which led to loud conversations that permeated my room and woke me up multiple times.
In short, the sleep clinic did not seem to care about the quality of the patient experience or, more critically, whether the environment was conducive to collecting good data. Their job, it appeared, was simply to meet the minimum criteria to charge the medical system for a sleep test.
Given that I'm young, thin, and don't snore, the results were surprising: moderate sleep apnea. They based this on my Apnea-Hypopnea Index (AHI)—the number of times I stopped breathing per hour. My score was 16 AHI while sleeping on my back (measured over five hours) and 7 AHI on my side (measured over 25 minutes of sleep), putting me just over the official threshold of 15.
II.
The sleep doctor wrote me a prescription for a CPAP machine. In Ontario, where I was living, a prescribed CPAP machine is eligible for a 75% reimbursement of its cost, but not for necessary components like the mask or hose.
About an hour after my appointment, I received a call from a CPAP supply store trying to sell me a machine. They quoted me a price of over $2,000—significantly more than I knew the machines cost. When I asked how they got my number, they immediately hung up, leaving me with the inescapable conclusion that the clinic had illegally sold my personal health information.
I then started researching how one buys a CPAP machine. You can't just buy them at a normal store; you must go to a specialized CPAP supply store. At these stores, you don't just buy a machine; you buy their "CPAP expertise," along with a package of all the necessary supplies. They are meant to be your CPAP gurus—telling you what to buy, helping you refine your treatment, and navigating the health bureaucracy. Realistically, because government insurance pays part of the fee and private insurance often covers another portion, this system inflates the price because the patient, insulated from the true cost, is less price-sensitive. Without insurance, you would likely just buy each item at its standalone cost without any of these additional services bundled.
After researching the best place to buy a CPAP—no easy feat, given how confusing the pricing models are—I was told that to actually get the machine, I needed my sleep doctor to sign an additional form beyond the prescription. I contacted the sleep clinic's office and was told they didn't have the doctor's contact information and couldn't help.
For context, the clinic that organized the sleep study apparently contracted with different "gig" sleep doctors. The doctor overseeing my file was only there for a set number of hours and wasn't a permanent part of the clinic.
For weeks, I called the clinic and was told, "Oh, this is so weird and unfortunate, this has never happened before. Of course, we will try to follow up with the doctor." Each time I called, they’d say, "We're so sorry, we don't know what happened, but we will definitely get you an answer by next week."
They never followed up. Each time I called, it was like speaking to a different person, even when I recognized their voice and name from a previous call. I asked if there was another way to get the device or have a different doctor sign the form. I was told no; it had to be the doctor who oversaw my sleep study and wrote the initial prescription.
After months of waiting, I had enough and contacted the physician complaints body. I explained that I had an unusual request: I didn't want to discipline the doctor—in fact, I was confident he didn't even know a request had been made. Rather, I suspected the clinic staff couldn't contact him and didn't care enough to solve the problem. I just needed to get his attention so he could sign a form for me.
The next day, the form was signed.
III.
When I first got the CPAP, I was told it was programmed so the sleep doctor and the guru at the CPAP supply store could analyze my data to assess my treatment's effectiveness. The machine itself only shows basic data: your AHI per hour, whether your mask is leaking, and how long you use the device each day. I presumed the data being shared with my doctor and the store was far more extensive.
After using the CPAP, I felt much better. Not perfect, not cured, but noticeably better. I had follow-ups with the sleep doctor and the CPAP supply store. After reviewing my data, both told me the treatment was a smashing success, pointing to my low AHI numbers as proof that, with time, I would feel much better.
Life was busy. I felt better, and the "expert" advice I received confirmed things were working as hoped. I didn't feel the need to research or optimize any further.
IV.
Flash forward one year. I was frustrated that despite the improvements, I still felt notable fatigue in the mornings and wondered if the treatment was truly working.
On a whim, I asked an AI for help. It suggested I download an open-source program called OSCAR, use it to analyze my CPAP data, and share the results. I then tried to find the detailed CPAP data that was supposedly shared with my doctor and the supply store. I quickly learned they never had any meaningful data to review.
For a CPAP machine to record useful, detailed data, you need to install a $5 SD card. In other words, despite using the machine for over a year, I had no data history. The doctor and the supply store that had assured me the treatment was going well had never reviewed anything meaningful. This machine cost over $1,000 and could record all kinds of useful data, yet it wouldn't without a cheap SD card. Why didn't the manufacturer provide one? Why didn't the doctor or the store that sold me the device tell me I needed one? An entire year of "data-driven" medical monitoring was based on a single, misleading metric.
A few days after installing the SD card, I uploaded the data from OSCAR to the AI. I asked it to assess the data and tell me if the user's treatment was likely effective.
The AI's response was unequivocal: this person's CPAP therapy was not working. The data showed a huge, glaring problem called Respiratory Effort-Related Arousals (RERAs). The minimum pressure on my machine was set so low that every time I started to have a breathing event, the machine had to slowly ramp up its pressure to react. This process alone caused numerous micro-arousals that, while too small to be counted in my official AHI score, were still enough to damage my sleep quality. It created the perfect illusion: a "wonderful" sleep score on the machine, despite a terrible night's sleep. Not only was this problem immediately obvious from the detailed data, but the solution—raising the minimum pressure—was also apparently obvious. I followed the AI's advice, and the next day, I woke up feeling more refreshed than I had in recent memory. Successive days brought the same results.
V.
So why am I sharing all of this?
Because so much of the medical system seems designed not to solve a patient's problem, but to create a structure where goods and services can be sold.
Why doesn't ResMed (the company that makes the CPAP machine) include a $5 SD card with their $1,000+ machines? Because they sell through CPAP supply stores who make their money convincing you that you need their ongoing expertise to interpret your data. Why doesn't the sleep clinic care if you can actually sleep there? Because they get paid the same whether the data is good or garbage—they just need to check the boxes that insurance requires.
The medical care itself—the diagnosis, the advice—often feels like the pretext for the transaction. It is the necessary component that allows a bill to be issued, but the intention feels less about providing an opportunity to help you and more about an opportunity to bill someone. The entire structure is optimized for the metrics of commerce (how can we reduce the cost of a new patient at the sleep clinic, or make more profit per cpap machine sold etc), not the quality of care.
In contrast, the AI is completely detached from this ecosystem. It has no supply store to partner with, no insurance forms to process, and no revenue targets to meet. It isn't a vehicle for anything else. Its sole function is to analyze information and provide advice. And this is why I think AI is such a valuable addition to the medical system: it's there merely to help, with no misaligned incentives or commercial structures to appease.
