/diabetes

hacking diabetes: raising fidelity of care

Primary LanguageJavaScript

Hacking diabetes

Fidelity of Care ################ Circa 2011 Despite many novel applications of technology to the medical field, therapies for chronic disease lacks enough transparency and credibility to humanely empower users to manage therapy.

Diabetes Data Bus

There's a rising notion of a diabetes data bus. A system which integrates data collected from a variety of systems, and communicates that data to authorized users. In addition, this infrastructure would support agents from an expert systems presenting analyses and simulations of expected results.

Diabetics own multiple mobile computers that record biometric data on a regular basis. This typically includes a menagerie of glucometers, of which I own at least 5, 2 of which are in active rotation at any given time. I also use an insulin pump, like many diabetics, and it keeps logs of insulin given, as well as performs opaque simulations on expected results. In addition, there are ancillary devices that measure interstitial glucose levels on a real-time basis, as well as pedometers, sleep monitors, and the list goes on ad nauseum.

With so many sources of data critical to managing medical therapy, it is impossible to predict the new sources of data that will arise. It's also impossible to replace all the existing devices with new devices that are designed to cooperate with one another. However, all existing devices have a serial port with which an authorized agent can communicate with the device in order to audit therapeutic details. Therefore, it's much easier to adapt existing devices into a common framework that knows how to present data to expert systems, knows how to store data over time, and knows how to keep the user connected to that data in ways that allow better decision making.

Despite all the data currently logged by devices, how much of it is leveraged to drive ongoing decisions? The proprietary software offered by medical industry offers snapshots of interesting data from the past, and then asks the user to manually fill in any missing data. Each manufacturer offers a perspective that their software knows everything about managing diabetes, and in so doing fails to offer a holistic perspective on therapy.

Instead, a data bus accepts input from a variety of sources, aggregates it with other available sources, and makes it available to the user at any time and any place. The user can choose which applications can subscribe to data, as well as re-route and transform data into those applications. Indivo already provides the container for aggregating a user's data with customizable schema types. Cube offers a great presentation engine for arbitrary data. When the two are tweaked to manage the data from diabetic therapy, we have a diabetic data bus.

Many parts, loosely coupled.

Help wanted

Join one of our mailing lists:

  • medevice For anyone wanting to increase the fidelity of their therapy using their skills and resources. We have many projects, one of them probably your kind of project.

  • medical-device-users Participate in developing advocacy to help share what and why we are doing.

  • insulaudit - for hacking insulaudit: discussion of python and features

Projects

We need people of all stripes, from linux kernel hacking, to graphic design.

Visualizations

Add your visualization here!

If you've created a visualization, add it to the list. Include some sort of self-attribution.

Firmware

For new devices, and to audit old devices.

Hardware

List your hardware designs here. Reference designs for pumps, sensors, meters, connectors, etc...

Backend

Assume you have a bunch of decentralized "agents," implemented in many languages, and all have data to report.

We'll need bits of the following types of software tailored for use by a diabetic data bus:

If you've got software to keep records, allow read/write access to them, add it to the list.

  • asset management/inventory

  • basic auditing/accounting/record keeping

    • opendatabetes - an open standard for recording/sharing diabetes data - very early days
    • sanguine - WIP glucose/insulin/pump tracking app. In prototype, to be rebuilt using web technologies

  • stats/data/mining

  • authorization and authentication

  • transport-independence - orchestrate which sockets, ports to use, etc.

  • manageBGL pump simulator and data visualization with API

Vendor transports

Bus consumers.

Adapters for other web services: glucosurfer, sugarstats, VISTA OSEHRA, INDIVO/SMART (CHIP), [@netspective], others?

Also see ABBI: http://wiki.siframework.org/ABBI+Pull+Workgroup for OAUTH based transport.

  • connect
  • Sanguine is an app already "in the wild", it should be able to either poll one of our compatible servers and fetch/store medical records and annotations.
  • opendatabetes - an open standard for recording/sharing diabetes data - very early days
  • indivo python/django friendly

Math

Stats, predictive algorithms (so we can actually measure what we are always implicitly predicting), the works.

If you've got a project that analyzes glucose, or insulin or similar data, add it to the list:

Device IO

The data on these devices belongs to us. Let's get audit the data:

  • insulaudit - this project consumes many libraries to re-format data in a prefered format
Decoder rings

These projects collect packet captures, provide tests suites, and tools to communicate with medical devices. Application authors can use these tools to analyze and debug devices, exercise their decoding skills, or contribute to language-independent test suites, or even help implement the protocols.

Unit tests, docs, protocol schematics, ports to other languages.

Help with the "decoder rings," some have data available, ready to analyze, some need captures!

Git based medical record?

Would solve transport and versioning issues.

  • git-phr - if only!

Web Development

  • gallery of visualizations, gist-based? proposed:
databetics.com/ -> here (medevice/diabetes)
databetics.com/visual/ -> here (medevice/diabetes/visual)
databetics.com/formats/ -> damon's databetes?
databetics.com/projects/ -> better list of projects we actually commit on
short term projects:
  iPancreas: re-usable, blog-posting tool
  visualizations: need to make it easy for more people to make
    graphs and charts, to iterate on UI, etc... this is essentially UI
    development broken down, feature by feature
  insulaudit: pumping data out so that sanguine, ipancreas and others
              can consume, pushing support for more devices,
              implementing in other languages?
  • gallery of advocacy, or maybe a "planet" style aggregator, with diabetesmine at the helm?
  • docs, and technical writing
  • templating for nice reports such as Jana's, we need to make it easy to produce documents to share like that.

other projects

Mostly personal loggers.

Documentation