Forecasting field hospital consumption with Médecins Sans Frontières

Markéta and Jakub Tomek, Oxford University doctors shared their experience with forecasting field hospital consumption with  Médecins Sans Frontières

How we joined the project

For both of us, it’s of paramount importance to do research that aims to contribute to wellbeing of humankind. However, at some stage during our studies, we realized we’d like to add a more “typical” volunteering to our studies/research. Consequently, we mentioned our interest to Oxfam and soon went to the first event. The experience was “interesting” and consisted of standing at a race checkpoint, jumping and yelling at near-dead racing runners how amazing they are, that the race is almost over (at 50 km out of 100), and how amazing they are again. Of course, a totally fulfilling and responsible role that was perfectly suited for two fairly introverted computer scientists…

We quickly realized this is not fulfilling us indeed and started thinking of what might work better. We agreed that the ideal scenario would be to put our computer science skills to good use in an efficient way, rather than doing whatever for a trendy CV element. How surprised we were when we found that there is a webpage for people like us. It is named SocialCoder and it puts together two sets of people: {computer scientists, programmers, designers, analytics, etc. with interest in charity work} and charity organizations seeking such volunteers. We set up our profiles and joined the system. We expected two possible scenarios – a bad one, where nobody would be interested in us, and a good one, where we’d get to something hopefully useful, but not too interesting, such as making webpages for community groups, or processing excel tables. It can be consequently clearly understood how surprised we were when we got an e-mail with invitation for a preliminary discussion from Médecins Sans Frontières (MSF), which we consider to be one of the most reputable and venerable organizations in the world.

What is it about?

MSF run hospitals all over the world and these hospitals of course require supply. Several times per year, each mission places an order of medical supplies (drugs, bandages, tools, …), that are delivered by a convoy after a certain time. Given the acute lack of high-quality reliable prophets, an obvious question arises: how many supplies to order? When the ordered amount is too small, required products are missing or have to be restocked in an emergency way, which is clearly not good. When the ordered amount is excessively high, the supplies take storage space and may expire, causing monetary loss.

The forecasting of how much to order is carried out by field staff, taking their valuable time which they’d much rather give to their patients. The question of our “contact” in MSF, Rob, was, if the process of forecasting could be automated and done via statistical methods based on consumption history, aiming to save time of field staff, or even improve the quality of forecasts. This sounds like a classical “school” task on machine learning and Big Data (in case you don’t know, Big Data is a really popular term, ranging between “does not fit a single Excel screen” to “does not fit a room full of hard drives”). We eagerly started working, cleaning and pre-processing data, and started trying a myriad of possible approaches. At the start, it was fantastically motivational that even sort of lame and simpleton statistical methods seemed promising, compared to the human-made forecasts. Slightly less motivational was that when we tried a wide range of “sexy” approaches of machine learning (to show we’re proper computer scientists; of course), they didn’t improve the lame and simpleton methods at all. In the end, a truce was made; Gaussian processes, an elegant method of regression, useful for temporal forecasting, seem to generally outperform the more elementary methods and they offer several other advantages. The current cost analysis seems that their application could considerably reduce both the expenses and the medical impact of understocking, which is great.

We were thrilled when the project has been recently deemed promising enough to be continued and it gained the support of MSF medical directors. Currently, a prototype of the application to-be-created is made by the Dutch organization HumanityX, with the plan to eventually develop a full software tool to be distributed to the field hospitals.

What is it like to work with MSF?

Fantastic, of course. ☺ The difference of this work to our ordinary scientific life is in the basic approach to problem solving: problems are not to be overly dissected, but solved instead. This means, among other things, that it does not matter how novel, elegant, or complex method gets the result. MSF members also seem incredibly efficient, motivated, and hard-working. We get stuck at something on Saturday at midnight? We write an e-mail and there is a chance we may Skype about the issue with Rob in ten minutes (this actually happened). If not, probably the next day (this of course seems to work both ways; if we’re to prepare materials for a presentation and/or process additional data, we do this with highest possible priority). There is no “can’t do” because of laziness or bureaucracy we face so often in the academic world of Oxford. We also really enjoy the bidirectional discussion, where MSF provide us with data, we process them, which then inspires further questions and development, which we can discuss again, and decide the next steps.

It is great that Rob has background in logistics and knows a lot about data analysis and uses tools we like (I mean, as much as you can like R); this made our interactions much easier than if he was, e.g., a pure clinician. That said, we had to learn lots and lots about the area and it took us a while to understand the terminology, how missions work, what are practical risks and pitfalls, what is possible and what not, etc.

It also would be a shame not to mention how kind and welcoming the people in the organisation are. We were invited to the Amsterdam operational centre for a visit and it was lovely. It was great to see how the single centre supports many missions, be it matters of disease treatment, logistics, security, psychology, or other management. The number of problems to be tackled seems high indeed. One often admires the field surgeons who risk their lives in dangerous areas (for Czech readers, e.g., Tomáš Šebek is great, see TV programmes with him or his books), but also the Amsterdam operational centre is full of heroes!

At the end, we can but thank providence (and SocialCoder) that we can work on such a fun and motivating project that might actually reach the stage of practical applications and helping life/money saving. All that was needed was a little bit of activity – we recommend checking out SocialCoder to the readers, it’s worth it!

Jakub and Markéta

Baraka 100 beds pharmacy tent
Jakub and Markéta
on the road to mobile clinic
Health centre pharmacy DRC

Sdílejte tento článek:

Podobné články

27.01.2019

Romanian Presidency: January – June 2019

The Presidency of the Council of the European Union rotates among the EU Member States…

27.01.2019

Romanian Presidency: January – June 2019

The Presidency of the Council of the European Union rotates among the EU Member States…

22.11.2018

Technologies in drug discovery and development

The process of launching a new drug generally comprises two major stages: drug discovery in…

22.11.2018

Technologies in drug discovery and development

The process of launching a new drug generally comprises two major stages: drug discovery in…

22.08.2018

Medicine lifecycle

Nearly 60 thousand registered medicinal products are currently available on the Czech pharmaceutical market (State…

22.08.2018

Medicine lifecycle

Nearly 60 thousand registered medicinal products are currently available on the Czech pharmaceutical market (State…

07.07.2018

Austrian Presidency: July – December 2018

The Presidency of the Council of the European Union rotates among the EU member states…

07.07.2018

Austrian Presidency: July – December 2018

The Presidency of the Council of the European Union rotates among the EU member states…