The data scientists were tasked to find a way to use routine survey data to predict who is at risk of depression
NAIROBI, Kenya, Nov 22 – 70 participants competed in teams to find the best statistical model to predict who is likely to be suffering from depression from a sample of thousands of households in rural Nyanza
The data hackathon, organized by Busara Center for Behavioral Economics, opened the challenge to the Nairobi data science community to apply their skills address the problem of mental health.
The data scientists were tasked to find a way to use routine survey data to predict who is at risk of depression enabling clinicians to accurately identify susceptible patients saving time and energy by going straight to those cases.
After hours of analysis and testing, Allan Chepkoy beat the odds to create the best statistical model to predict who is likely to be suffering from depression.
“It feels great especially after long hours of sitting in the house teaching myself how to be a machine learning engineer,” said Chepkoy.
“The learning curve has been very steep and winning justifies the hard work and perseverance. It validates the adage that anyone can do anything.”
Allan has won three hackathons this year including BBC’s ‘Beyond Fake News’ hackathon.
Participants used data from a 2015 study conducted by the Busara Center in Siaya County, in which over 2800 individuals from 1440 households were surveyed about their family composition, economic activity, financial position and health.
“The information and data we get from such hackathons is invaluable. Going forward we look forward to incorporate more stakeholders in such events so that the learnings we get can have direct impact on the ground,” said Alfred Ongere from AI Kenya.
The participants were also asked to take a psychological instrument used to screen for depression.
While household surveys are routine and relatively cheap to collect, the screening process for depression is sensitive and needs to be carried out with protocols in place for referring serious cases.
The World Health Organisation estimates that 1.3m Kenyans are suffering from depression, and that Africa has the highest rate of untreated depression of any continent.
Unfortunately, mental health resources are scarce. Using machine learning to better target these resources could allocate treatment where it is needed most and improve an untold number of lives.
“Calling the tech community’s attention to mental health is important. There are a lot of areas where data science can help alleviate problems in health care and mental well-being, so hopefully, this event will plant ideas in people’s mind for future projects,” said Daniel Mellow, Busara Center’s Data Specialist.
