Machine learning & mobile health to improve TB treatment | @Dimagi, South Africa & Lesotho
Marta Milkowska
DESIGN & DIGITAL & STRATEGY
During the summer of 2016, I worked with Dimagi, a social enterprise focusing on mobile health, on a mobile-based tool leveraging machine learning to predict default risk of HIV and TB patients. High default rates are a key challenge in TB and HIV treatment, resulting in thousands of multi-drug resistance cases. The multi-drug resistant TB, for example, causes over 440,000 deaths annually. In South Africa, it causes a drop in cure rates from 90% to 5-35% and increases the cost of treatment 26 (MDR-TB) or 103 (XDR-TB) times. A risk prediction tool could help healthcare service providers tailor its intervention to each patient based on the risk rate.
I spent that summer living in South Africa and Lesotho to research, prototype, and design a strategy for a mobile-based default risk prediction tool. When I originally arrived in South Africa, I was armed with only a broad concept of a tool and the support of Dimagi’s leadership. After an intensive summer of research, customer and partner interviews, field empathy work, and synthesizing data, I run early experiments with patients and community health workers in the main TB clinics in Lesotho. Based on the insights, I refined the strategy and advised on early design of the tool. The tool was approved by the global senior leadership in Dimagi as a new investment line with a potential to save millions of patients globally.