From May to August 2019, the KIT Royal Tropical Institute’s Centre for Spatial Epidemiology (CASE), together with Pakistan’s National TB Control Program and the Stop TB Partnership, hosted a virtual hackathon which brought together research groups interested in TB modelling to collaborate on an exercise to estimate the sub-national TB burden in Pakistan in 2018.
One group wanted to remain anonymous, until now…
A blog by Anny Yuanfei Huang
If you were to overhear some of the conversations I have with my colleagues, you’d be forgiven for thinking that we have nothing to do with public health. It’s become part of our banter now, trying to make it sound like we’re part of the more glamorous fashion industry: “Oh, I went to Europe to do some modelling,” and “Want to see my modelling pics?” and of course the punchline comes when we whip out a map showing the spatial distribution of TB.
So who are we?
In some ways, this is how it feels to write a post revealing our identities. We’re not quite sure what if anything, the community is expecting. Hopefully, not anticlimactically, this is us: Anabelle, a pharmacist from Hong Kong, Andrea, a nutritionist from Mexico, and me (Anny), a general practitioner from Australia. We are a little cluster of recently-hatched-from-grad-school juniors in the epidemiology world, who were impatient to stretch their wings before we had even completely finished our studies.
We had genuine doubts about our chances of coming up with a model that gave some roughly sensible numbers, however simple, and when we saw some of the intimidatingly illustrious names among the other teams (including those of the researchers at the University of Sheffield, where we completed the first year of our MPH), we asked to remain anonymous. Thankfully the organisers at KIT Royal Tropical Institute agreed, as it allowed us to focus on Hack TB as a learning opportunity rather than just a competition.
We were excited when we saw the competition advertised and decided to enter. It was not until a couple of days later when we actually stopped to think, that we realised they could say no! After all, with our lack of skills and experience, could we actually offer anything useful? We may know very little, and we may not have done this before, but we wanted to figure out how much we still needed to learn (a lot, it turns out), and what specific gaps in our knowledge needed filling. And besides, it’s fun to model infectious diseases, right?
Keeping it simple
Early on, we decided that we would keep things as simple as we could, by only modelling TB in one province, and by only picking a small number of variables. But even so, we had not fully accounted for the fact that we also needed to do a lot more background reading, to be up to speed with the current state of TB modelling. And then we had to figure out which statistical packages were out there that we could learn to use within a short time.
Biting off more than we could chew
If you already think that we had bitten off more than we could chew, this was only half the story. The competition process started during the final days of our Master of Public Health at EHESP France, when we were busying ourselves with finishing our dissertations, defending them, graduating and then moving to different cities and countries (we were international students, after all).
Hack TB also became oddly comforting during this time of transition. The three of us could still meet regularly over video calls and laugh at the absurdity of trying to explain our R codes to each other over flaky cafe WIFI connections and background noise. We’d have a general chat about life before getting down to business. This included political updates. Anabelle had returned to Hong Kong during a particularly tumultuous time. The way she spoke about Hack TB to us always gave me the feeling that it was a kind of an escape. It’s funny how in the field of modelling, we always talk about all the different kinds of uncertainty that we need to account for. But when faced with societal conflicts and questions of safety (or in Andrea’s and my cases, just generally being adrift), it is a relief to be able to assign numbers to the upper and lower confidence limits of one’s calculations.
What did we learn?
So, what did we learn? Firstly, that modelling TB is hard, but fun. We had always known that it was challenging, but we were now able to appreciate this for ourselves. Despite having lectures about this, we were only now able to grasp the sheer enormity of undertaking a national prevalence survey. We’ve developed greater admiration for the frontline workers who collected each piece of data. We started connecting the dots (and the DOTS!) between the individual TB patients whom we had met, to potential observations in the survey, and this human aspect gave us extra motivation. We’ve become more self-reliant in coding. We’ve become acquainted with the innovative global TB modelling community, and even learned to spell Muzaffargarh (a city in Punjab).
And how do we feel? To be completely honest, truly satisfied and grateful, and a little bit incredulous and relieved that we made it to the end. We really had so much to gain from the experience that we couldn’t possibly lose. We are truly indebted to the wonderful organisers at KIT and the National TB Program in Pakistan. Hopefully, more national bodies around the world will follow suit and throw open their datasets to the modelling community. If they do, please let us know, we’ll be the first to put up our hands.
KIT will award the winner of the hackathon at the 50th Union World Conference on Lung Health in Hyderabad.