Online diabetes service

[dpkingma]

Hello, I'm new to this forum. I"m a student from the University of Utrecht and me and some other students are thinking about making a free diabetes web application. We have some experience with development of diabetes software.

This is our system proposal:

Free Diabetes Web Application

Features:
Users can register to the system. Upon login, a web application starts.
Diabetics can:
- Update their online diabetes diary, by filling in new dairy entries (insulin/food usage, activities, glucose measurements) by:
* A interactive web interface;
* Or sending an email with the information;
* (maybe even the option to send SMS with glucose measurements)
- Get Information:
* Graphical view of the diary. This way users can learn about their diabetes by analyzing this graph, and for example induce what the impact was of certain food.
* Information about diabetes related things, ‘wiki’-style. People can add/edit information like wikipedia does.

Extra feautures:
* Forums

Long-term features:
I’m working on an computer program that can statistically analyse dairies, and make conclusions about the effect of certain factors. So it calculates the personal effect on glucose level of apple, types of insulin, stress maybe even weather etc. The problem is that since the problem is so complex, the system needs quite a lot of data.
Eventually, if enough users use the system, this system should be able to tell users what the effect is of certain food/insulin etc. Using this information it could even make short-term predictions of future glucose level.


Could people be so kind to give some feedback on this system? We want it to be useful and worthwhile to make!
- Do you think such a program would be usefull?
- What would you like to have in this program?
- What of the above mentioned features would not be usefull?

Thanks for your feedback!

[melissabeth]

It sounds like a really interesting project! The only drawback that i can see is that, as you mentioned, there are so many factors at play. Would users have to manually enter their insulin dosages, blood glucose, all food eaten every day and times, as well as other factors like exercise /stress? Even though we keep all of these things in mind on a day-to-day basis, it sounds like it would be time-consuming to enter all of this on a web page, so for me I guess a lot would depend on the benefits and convenience of it. How much useful information would you get right away? Will the system rememer your "normal" schedule and dosages so you don't have to re-enter everything and can just adjust it as you go along? Would it be compatible with existing programs for tracking blood glucose so people who use those could just upload the information?
Let us know when it's up and running, it sounds like it will be really useful.

Good luck,
Melissa

[dpkingma]

Melissabeth, thanks for the quick reaction!

To answer your questions:
- Yes, it would definately be some more work to fill in all these data every day. That's why we want to make it as easy as possible by offering email and maybe SMS-service (depending on your country). We also experienced that some diabetics, like pregnancy-diabetics, are willing to do this every day.
- Importing data and spreadsheets from other software is high on our list.
- Exporting data to a spreadsheet is also important, because people want to
- The option to fill in your 'default' daily insulin/food is absolutely high on our priority list. Whenever you don't fill in anything, the system will assume you use/consume your default amount. You would be able to fill in some fixed default amount, and if you don't, it will take the 'mean' of entries for the timeslot.

One thing I forgot to mention, is that it would theoretically be possible to give your doctor/dietist temporary access to your online data. I imagine that this would enable some easier communication.

[JasonJayhawk]

Holy cow, You're describing my current Master's thesis. Cool!

You mentioned going a statistical route; I'm using a rough set route. Like you said, data collection is a very difficult thing to obtain. Dealing with missing data is like pulling a rabbit out of a hat -- one can predict, but it's not complete. Therefore, one assumption of my project is having a CGMS.

I like the idea of having a complete portal system with a wiki!

The U. of Utrecht has published some other telemedicine papers on diabetes starting back around 2001. Are you working with the advisor that worked on that project?

One of the important keys for the data entry system is ease of use. People don't like to log everything. Perhaps a SOA service that allows multiple methods of writing data to a backend database.

How much time will you have for this project?

[dpkingma]

Wow, you're actually doing a master thesis of it? I was thinking about using that myself too, but since I"m still in my bachelors, I will have to wait 1 or 2 years.

I'm not aware of these telemedicine papers, but it sounds interesting and I will do some research.

This online diabetes software idea was born from one of the computer science bachelor projects. We have been working on software called 'Cheetah' (http://cheetah.aardrock.com/download/). Cheetah is a Java program that was supposed to let diabetics connect with each other. The communication was supposed to run via a P2P network (JXTA). During the 6 months of development though we spend most of the time building an installer, automatic updater, the beginning of a user interface. We just didnt have enough time to implement all the P2P services.

I was the one responsible for the machine learning part. The program manager 'envisioned' aself-learning and advisory system, using the user-diary as its learning set. It turned out that the idea was very optimistic, as the problem is extremely complex and (as you also mentioned) you're dealing with incomplete data.

I've been working on a bayesian learning theory that copes with incomplete data. The problem is that, since the data is so incomplete, you need LOTS of them. Using lots of data/diaries and the bayesian statistical theories, you can learn a lot of individual responses. This would be very helpfull for diabetics. Its still inclear however if it will be possible to give glucose predictions. Even physicians have a very hard time interpreting glucose diaries...

Speaking about this new software, we could use help :) Since I'll be on exchange to Rome until next februari, dont expect any progress until then.

Would you be interested in helping?