Human-in-the-Loop Machine Learning and Improved Immunization Data

Benjamin Fels and Suvrit Sra of Macro-Eyes, Inc. in the U.S. will engage with frontline health workers in immunization centers and combine their knowledge with existing supply chain and immunization data using machine learning to better predict vaccine demand and thereby improve immunization coverage. Vaccine supply levels in Ethiopia are predicted using data that may be inaccurate or outdated. These low-confidence data could be enhanced with the unique insights of frontline health workers by using machine learning, which is a valuable statistical method for increasing the accuracy of predictions. They will test this at three health centers in Ethiopia by exploring approaches such as WhatsApp to engage health workers and collect relevant information on vaccine stocks and demand in the clinics. These data, along with available supply data, will be used to train so-called classifiers, or algorithms, that transform the input data into more accurate predictions of monthly vaccine use. They will test whether their method improves the accuracy of predictions compared to the original methods.