EPCON supports high TB burden countries in Asia and Africa using artificial intelligence. We were invited to share our experience at the Asia Pacific Tuberculosis Forum on 7th December 2021. Dr Sumbul Hashmi, EPCON’s Global Public Health Manager, represented EPCON and shared her views on “Data optimization in TB patient finding”. The forum brought together leaders, policymakers and clinicians in the Asia Pacific Region to collaborate in an attempt to meet the United Nations Sustainable Development Goal of ending TB by 2030.
An estimated 10 million people develop TB worldwide every year. Devastatingly, around 40% go undiagnosed and unreported for a variety of reasons. Such pandemic-led derailment has hit India, Indonesia and the Philippines particularly hard. At the Asia Pacific Tuberculosis Forum, Hashmi discussed how data can be optimized to assist in finding these missing TB patients.
Who is missing and Where ?
While we know TB affects almost all age groups, a huge portion of people in the age group of 15-44 years, especially men remain missing. Children are also difficult to diagnose. In addition, vulnerable populations - for example, those living in economically-deprived areas - as well as high-risk groups, such as those suffering from other conditions including diabetes and HIV, are frequently left undiagnosed. Though it is tricky to find these missing TB patients , there are ways to optimize data in order to aid with the identification of this devastating illness.
Data optimization to find the missing millions
Here are four key ways to help find missing TB patients:
Digital capturing of data
Paper based reporting can be replaced by digital data collection, which allows for much quicker data analysis.
Geo referencing of program data
Frequently, the geographic location of newly found patients is not recorded. Instead, locations often only include a region or a city, too broad a boundary to draw insights, whereas the exact location coordinates help greatly with data optimization. Health facilities, laboratories, active cases and contact tracing also require mapping. In resource-constrained areas, the lack of smartphones and a viable internet connection present obstacles to geotracking, but efforts are being made to overcome these challenges.
Real time data monitoring
The adage “a stitch in time saves nine” certainly holds true when it comes to machine learning-based modeling for disease risk estimation. If a disease is changing rapidly or worsening suddenly, disease analysis works best in real time.
Incorporating high resolution, open source contextual data assists greatly with predictive modelling for risk estimation . Such data may include levels of poverty, malnutrition, access to health care, sanitation, and more.
Capturing and organising data is at the heart of data optimization. By combining real time data with contextual factors, valuable insights are generated by predictive models. Bayesian network models are capable of predicting risk at local community level, so that efforts can be channelised in the right direction. In this way, we come closer to a key UN Sustainable Development Goal: ending the TB scourge by 2030.