Reflections on the tech of childbirth

My previous post talked about my overall impressions about the childbirth process from a non-expert's view.  This time I wanted to dive in a bit to a geek's view of the tech involved, which, at the risk of offending, was pretty depressing.  While admitting the serious challenges in developing approvable or clearable medical devices, this seems like an area ripe for improvement.  Caveat - this is closer to my expertise, but I'm not a doctor nor a medical device person;  this is semi-founded opinion. :)

1.  Much medical monitoring is both unnecessarily intrusive and restrictive.

If you end up towards the "high-intervention" spectrum during labor, you may find yourself attached to:  A pulse-oximeter;  a blood pressure cuff;  a fetal heart rate monitor;  a contraction monitor.  Attached to your bed - and in other spots around the room - you'll have devices to use to notify or communicate with the nursing staff.  You'll also be wearing one or more ID bracelets, now bar-coded.

Ouch #1:  Each of these things has its own, separate cable running from the patient to the wall.

Don't imagine going to the bathroom in a hurry.  If you do, one trick involves pulling all of the cables out of the recorders, and fleeing the now-beeping array of unhappy monitors while trailing four long, rubbery cables like a medical octopus as you rush to the bathroom.  Followed shortly thereafter by a visit from a nurse to plug things back in.

Ouch #2:  Almost none of these are wireless, by default.  While there exist wireless fetal & contraction monitors (ask for "telemetry"), they're not the default.  Other than the blood pressure cuff, none of these is a particularly high-power application.  They can be made remote.

A suggested vision #1, the easy part:  When I'm a patient in a hospital, I want to walk in and have the staff hand me a wristwatch.  It will act as a full-time pulse-oximeter (heart rate and blood oxygen saturation).  It will tell the staff at all times where I am in the hospital.  It will be my two-way communication device to the nursing staff if I need something -- maybe serviced by a nursing call center somewhere else, I don't care.  It will be my hospital ID bracelet via NFC or QR code.  If someone's really clever, it might even be able to determine my blood pressure, but that's asking for some advances.  And yes - it will tell the time.

Before yelping about the fetal monitoring devices, let's go to...

2.  Interpretation of contraction data and fetal vital signs is a subjective art.

All medical diagnostics carry a risk of false positives;  a false positive typically results in additional interventions, all of which themselves carry risk.

There are over four million births per year in the U.S.  [CDC].  2.76 million vaginal deliveries, 1.35 million cesarian deliveries.  That is the most delicious big data set I can imagine.  But despite this potentially massive trove of data to obtain correlative relationships between actual fetal HR patterns and outcomes, fetal monitoring is actually a big question.  Monitoring has not decreased fetal deaths from inadequate oxygenation, [more] and is associated with an increased number of cesarian deliveries.

3.  External monitors are fidgety little widgets that glitch a lot.

Move around, roll over to your other side, or just look funny at them, and the EFM devices are likely to slide into a new spot that doesn't work well.  The typical state of the art is a circumferential belt - for each monitor, a big band of stretchy webbing wrapped around the abdomen.  This is, (a) yet another device making you feel like a patient, and (b), when they glitch - you guessed it - another visit from the nursing staff, wasting their time, and interrupting a patient who'd probably rather be focusing elsewhere.

So:  Same request.  The fetal heart rate monitor uses ultrasound, and the transceiver is a disc about the size of a half-height hockey puck.

1)  Why can't we replace this with a small grid of miniaturized ultrasound transceivers that adhere unobtrusively to the patient's abdomen or back?  Heck, while we're at it, why not let them also export medium-quality ultrasound imaging?  We've already seen that multi-sensor approaches can be a good way to go for NMR... and watching the skies.  A 2004 study from Tours university in their Actifoetus project similarly suggests the promise of this approach.  No wiggly monitors, something that lets the patient move freely, more data, fewer glitches.  Please!

2)  Big data analysis applied to EFM data.  Why are humans trying to interpret high-frequency wiggly chart line data?  There's already data to suggest that we can make computers do better ... as we're going to see in nearly every other area of medical diagnostics as time marches on.

3)  Improve the technology to improve the human experience.  Happier patients are healthier patients.  The stress of the hospital environment ("whitecoat hypertension") can raise blood pressure, slow labor, etc.  The advances in small, low-power, wireless sensing over the last decade are a perfect opportunity to drastically improve the patient experience in modern medicine, but these advances are being fielded at a glacial pace.

The catch, of course, is that none of this is entirely simple or we'd be there already.  As Kevin Fu points out,

  1. Interoperability is hard.  It's hard to reason about the safety properties of complex, integrated systems.
  2. Wireless can be tricky.  It's subject to interference (even in the medical bands).  All of these monitors must be designed to "fail-notify" instead of fail-silent.  It's harder to be certain you're even monitoring the right person (no cable...).  Batteries fail, and can themselves pose hazards.
  3. Software can lead to overconfidence.  Software and hardware can and will fail, and the human brain provides for nice redundancy, if we can engineer for it.
But there's a lot of room for improvement even within those constraints.  And perhaps the focus has been too much on the high-pressure, high-profit, high-risk operating theater instead of trying to improve the medical experience of simpler parts of being human - such as giving birth.

[Update:  Some good discussion about this post on google+ - hop over and join the conversation!]

[Update 2: In the discussion, Dr. Beth Prairie suggested reading an excellent New Yorker article by Atul Gawande that provides a lot of context about the medicalization of obstetrics.  It's well worth reading.]


Popular posts from this blog

Reflecting on CS Graduate Admissions

Minting Money with Monero ... and CPU vector intrinsics

Finding Bugs in TensorFlow with LibFuzzer