3 ms·

Since this is such a mission critical piece of hardware that would severely impair the user if a bug or malfunction developed, how can you completely trust the

by sizzle 8y ago

Since this is such a mission critical piece of hardware that would severely impair the user if a bug or malfunction developed, how can you completely trust the device will continue to work 99.9% of the time? I really want to recommend this to loved ones but hesitate with the liability of over/underdosing insulin and them not being as tech saavy as the HN crowd...

aarondf 8y ago
To answer the question you're asking, I trust that the _device_ is basically never going to fail because it's FDA approved and manufactured by Medtronic, who has a loooot on the line to make their devices fail-safe. To answer the question I _think_ you're getting at... I trust the open source software that runs the FDA approved device because it's designed in such a way as to be as careful as possible (temp basals vs boluses). It's also open source and a lot of people contribute to it. I also only update it once a year or so when my Apple Developer certificate needs renewal. All of that ^ helps, but honestly, it's not a guarantee. I still wear an off the shelf, FDA approved CGM with their native 1st party app installed so I still get their alerts on low/high blood sugar. To me, it's a calculated risk and one that (so far) has been extremely worth it. My A1C went from ~7.5 to ~5.6 after using it.
dekhn 8y ago
In the business this trust comes from the fact that they are regulated, validated devices (lots of testing and process; any change involves a lot of work). I don't think the DIY movement will address regulation or validation (and I don't know that regulation and validation is guaranteed to make a safer product than DIY can).
bigiain 8y ago
And keep in mind, "99.9% reliable" might mean it only kills you once every three years...
hasahmed 8y ago
A valid concern, but also humans are guaranteed to make mistakes and I'm sure a high percentage of them, so anything that lowers that error rate is an improvement even if it's not perfect. My thoughts anyway
- sizzle 8y ago
  Similar to self driving cars improving car accident death rates over human drivers. Great point.
DavideNL 8y ago
> how can you completely trust the device I feel the same, however, people not having access to such an automated insulin delivery system may also suffer/die because of the consequences of a worse regulated sugar.
imtringued 8y ago
Both Humans and Software can malfunction. The hardware is identical. Also if the algorithm does small injections every few minutes then the risk of overdosing/underdosing is extremely low compared to a single big injection at midnight.
larryjounce 8y ago
Of course it is a calculated risk, however as a T1D myself, you can't ignore the fact that the current FDA approved pumps are also very dangerous. Pumps that don't respond to dropping blood sugars (most of them on the market today) and keep delivering insulin when you don't need it, compound low blood sugars, which has personally got me into serious trouble many times. Picking between a FDA approved "dumb" pump running on 10 year old software versus an open source option that explicitly has code to stop insulin delivery when blood sugars fall, seems to be an obvious choice. I've had diabetes for 31 years and this is by far, hands down, the greatest advancement I've experienced.

by sizzle 8y ago

aarondf 8y ago

To answer the question you're asking, I trust that the _device_ is basically never going to fail because it's FDA approved and manufactured by Medtronic, who has a loooot on the line to make their devices fail-safe. To answer the question I _think_ you're getting at... I trust the open source software that runs the FDA approved device because it's designed in such a way as to be as careful as possible (temp basals vs boluses). It's also open source and a lot of people contribute to it. I also only update it once a year or so when my Apple Developer certificate needs renewal. All of that ^ helps, but honestly, it's not a guarantee. I still wear an off the shelf, FDA approved CGM with their native 1st party app installed so I still get their alerts on low/high blood sugar. To me, it's a calculated risk and one that (so far) has been extremely worth it. My A1C went from ~7.5 to ~5.6 after using it.

dekhn 8y ago

In the business this trust comes from the fact that they are regulated, validated devices (lots of testing and process; any change involves a lot of work). I don't think the DIY movement will address regulation or validation (and I don't know that regulation and validation is guaranteed to make a safer product than DIY can).

bigiain 8y ago

And keep in mind, "99.9% reliable" might mean it only kills you once every three years...

hasahmed 8y ago

A valid concern, but also humans are guaranteed to make mistakes and I'm sure a high percentage of them, so anything that lowers that error rate is an improvement even if it's not perfect. My thoughts anyway

sizzle 8y ago
Similar to self driving cars improving car accident death rates over human drivers. Great point.

DavideNL 8y ago

> how can you completely trust the device I feel the same, however, people not having access to such an automated insulin delivery system may also suffer/die because of the consequences of a worse regulated sugar.

imtringued 8y ago

Both Humans and Software can malfunction. The hardware is identical. Also if the algorithm does small injections every few minutes then the risk of overdosing/underdosing is extremely low compared to a single big injection at midnight.

larryjounce 8y ago

Of course it is a calculated risk, however as a T1D myself, you can't ignore the fact that the current FDA approved pumps are also very dangerous. Pumps that don't respond to dropping blood sugars (most of them on the market today) and keep delivering insulin when you don't need it, compound low blood sugars, which has personally got me into serious trouble many times. Picking between a FDA approved "dumb" pump running on 10 year old software versus an open source option that explicitly has code to stop insulin delivery when blood sugars fall, seems to be an obvious choice. I've had diabetes for 31 years and this is by far, hands down, the greatest advancement I've experienced.