Healthtech, news, and other cool things. We help over 30k providers get home early.
🇨🇦
Joined June 2023
- Tweets426
- Following197
- Followers70
- Likes1K
Pinned Tweet
Family doctors spend ~19 hours/wk on admin work, Scribeberry saves them ~2 hours/day
blog.scribeberry.com/scribeb…
Parents who'd sat with a remote human scribe said they felt heard and seen.
Asked about an AI scribe, those same 33 parents went to accuracy and security. The AI was hypothetical for all of them, which the authors call the study still missing.
Appl Clin Inform, Oct 29.
Nobody schedules Sunday charting.
It's not on the day sheet and it isn't billable. We still hear about it every Monday.
Quebec's 811 line now has patients enter symptoms online before a nurse picks up. @gosantequebec says average call handling at pilot centres fell from 29:38 to 22:49 over four weeks.
That saving didn't come from a better note. We build notes, so we notice.
Bill S-5 makes interoperability a vendor obligation, ours included, and bans data blocking. Commons second reading was Nov 5.
Section 4: it binds a province only by order in council, and only where provincial rules fall short.
So nothing changes in your clinic yet.
Six real world ambient scribe evaluations. All six measured time saved and satisfaction. None reported checking whether the conversation itself changed.
Vendors don't publish that either, us included.
pmc.ncbi.nlm.nih.gov/article…
Every privacy regulator in Canada has asked health AI vendors to show clinicians how the model was trained and where the training data came from. Joint resolution via @PrivacyPrivee, Oct 29. It's a fair thing to ask us, so ask.
A year of hospital inbox logs in @BMJ_Open: AI draft reply adoption started near 80%, settled at 35%.
Where drafts were barely reused, messages took longer than having none at all. 175s against 155s. Ours to fix.
pmc.ncbi.nlm.nih.gov/article…
3% of core family medicine clerkships let a student use an ambient AI scribe, and of the directors who object, 82.5% named clinical reasoning. We build these tools and don't have a great answer to that yet.
94 US clerkship directors, Family Medicine, Nov 2. @OntarioCollege
Never answered In-Common Laboratories' fax verification request? Your patients' FIT results start arriving by mail.
ICL results begin flowing straight into Telus, Accuro, Wellstar, Alpha Global and CHS charts this month. Everyone else stays on fax.
Via @OntariosDoctors, Nov 3.
The same text in Spanish ran 1.57x the English token count on one OpenAI tokenizer and 1.33x on another, per PLOS Digital Health. Usage billed AI, ours included, charges a bilingual clinic more for the same work.
pmc.ncbi.nlm.nih.gov/article…
If an Ontario EMR loses its OntarioMD certification, HRM, OLIS and DHDR go with it. Hospital reports, labs and drug history stop arriving. We hadn't clocked that the certified list updated this October already carries one suspended offering.
Our favourite line in @UofT_TCAIREM's new hospital AI governance framework is the one aimed at vendors like us: buying the product doesn't transfer responsibility for it to us.
Free, and built with all 15 TAHSN members.
65% of Canadian family doctors can now send referrals electronically. 54% still call coordinating them a major problem, the highest share of any country CIHI surveyed.
@Infoway has been funding the pipes. Going digital didn't make the follow up anyone's job.
An AI grading clinical text against expert panels missed 10.2% of clinical errors and flipped 41.7% of head to head winners.
Six open models, four German corpora. No setup held up across tasks. A lot of our category's quality claims rest on this method.
npj Digital Medicine, Nov 4 2026: nature.com/articles/s41746-0…
Same Spanish text, 1.57x the English token count under one tokenizer, 1.33x under another.
Token billing costs more in a bilingual country. Ask any vendor, us included, what counts as one unit.
PLOS Digital Health, Nov 2: pmc.ncbi.nlm.nih.gov/article…
One note has to work for five readers: you at the next visit, the consultant, the insurer, the disability adjudicator, and maybe a lawyer in four years.
That's why nobody agrees on what a good note looks like.
K077A geriatric telephone support goes from 8 to 12 units per patient per 12 months, backdated to April 1 2026.
A ministry adjustment sweeps the claims already submitted, so there's nothing to resubmit. A billing change where the move is to do nothing.
ontario.ca/document/ohip-inf…
Three pediatric rheumatologists scored an LLM decision aid for juvenile arthritis. 97% of outputs acceptable on safety, 60% on factual consistency. Safe and wrong are not the same thing.
10 cases. PLOS Digital Health, Oct 22 2026.
pmc.ncbi.nlm.nih.gov/article…
Nobody asked for a better note.
Toronto mental health clinicians, asked what AI would help ED to community handoffs, wanted a trusted resource directory, live wait times, and referral tracking after discharge.
Four interviews only. JMIR Formative Research, Nov 2 2026.