We are very lucky to have a guest post today from Gabriel Belfort MD/PhD. Here are some of his thoughts from the last Hacking Medicine – written when he was “in the moment”. Enjoy.
As an M.D./Ph.D. who has tended toward the basic sciences I was initially nervous about the foreign worlds of engineering and business that seemed to be the core of Hacking Medicine (held at the MIT Media Lab on 2/25 and 2/26/2012).
When I arrived I had several ideas for problems in medicine, but I couldn’t have predicted how the forces which had initially made me nervous could provide such interesting and viable solutions.
The problem I posed was one of scheduling.
My wife works in a pediatrics clinic. Due to a 40% “no show” rate and no penalty for not showing, my wife and her colleagues are forced to overbook their schedule so that they have enough patients who do show up to bill for.
This results in an unbalanced system which results in many days with far too many patients, rare days with the right number of patients, and still days when too few people show up.
My initial idea was to shift some of the financial burden of a “no show” to the insurance companies and the patients so that these parties are incentivised to get the patient to the clinic.
Enter Hacking Medicine.
After positing my problem I was approached with an entirely MIT solution: Use machine learning trained on patient characteristics and prior no show rates to assign each patient a likelihood to show value between 0.1 and 1 (A full 1 is for steady person who always shows up). Then each provider can be assigned a schedule that is more likely to add up to 20 patients actually showing. So for a 1 you wouldn’t overbook their spot – they are going to be there. For a 0.5 you would overbook a second 0.5 so that each slot adds up to 1.
On average this system promises to normalize Dr’s schedules and save money for the practice by filling wholes in the schedule and eliminating the tendency of patients to leave when they are frustrated by waiting too long for their doctor.
Creating such a system over such a short time frame has been really interesting. It has been eye opening to have a team of MIT computer savvy undergraduates who are able to code these ideas into a prototype so quickly. Along the same lines having MBA types envision a way to make this actually work as a company makes the whole exercise feel “not like an exercise”, but like starting a company.
More to come. Another component we are adding is adding a “Doodle” form for rescheduling so that patients who previously said they could like an earlier appointment so that when such a slot opens up we only contact them if they could make it on that day.
Today is 2/26/2012. I am very excited to get into the media lab and keep Hacking. Thanks to all the participants and organizers.
Gabriel M. Belfort, M.D./Ph.D.