How artificial intelligence could change the fight against AIDS

Africa is working to expand its health capacity, but a complicated patchwork of conflicting and confusing data-sharing laws stands in the way.

It's a problem that “hinders a lot of science,” Aliki Edgcumbe, a legal scholar at South Africa's University of KwaZulu-Natal, told Nature earlier this year. Edgcumbe and a team of experts hope a new artificial intelligence (AI) chatbot could be the answer to untangling the web of privacy rules.

Amid the Covid-19 pandemic, the National Institutes of Health committed approximately $75 million to the Harvesting Data Science for Health Discovery and Innovation in Africa (DS-I Africa) program to advance health innovation and data science on the continent . What came next was a chatbot that scientists hope can bypass data sharing regulations in 12 African countries. Starting this fall, the free resource will help make research more easily shared and accessible, which could lead to major breakthroughs and impact on global health.

The rise of AI will impact every continent on earth, but none will benefit more than Africa. By the middle of this century, one in four people on the planet will be African. AI will play a key role in empowering this population and strengthening health systems – which could be crucial to the global fight against AIDS. Here's how.

Test, test, one, two, three

In healthcare, diagnosis depends on access to reliable diagnostics. Without testing, treatment is guesswork.

AI is already changing the diagnostic landscape. Modules trained on thousands of X-rays now produce sophisticated algorithms that identify abnormalities that appear in X-rays, MRI and CT scans. Companies like Qure.ai have been instrumental in providing life-saving diagnoses to people with tuberculosis in minutes and providing them with rapid access to care and treatment – a vital contribution to stopping the spread of tuberculosis. The same technology can also be used to detect other diseases such as lung cancer. This allows patients to be screened for multiple diseases at the same time, reducing the number of doctor visits and, in some cases, enabling early detection that would otherwise have gone unnoticed. This is an invaluable tool for advancing global healthcare, particularly where fragile health systems and a shortage of trained medical professionals exist.

AI is used specifically in the fight against AIDS to identify potentially at-risk HIV patients. Earlier this year, a study in The Lancet showed that a new Ukrainian AI algorithm led to a 37% increase in HIV detection rates compared to a non-machine learning approach. According to Anna Cherednichenko of the Alliance for Public Health, a leading advocacy and HIV services organization in Ukraine, the algorithm allows testing services to be targeted to people at increased risk of contracting HIV and is more accurate than social workers' assessments which can be subjective. Similar algorithms have been developed in Kenya and Uganda to identify ideal candidates for pre-exposure prophylaxis (PrEP).

HIV attacks white blood cells, weakening the immune system and increasing the risk of infections and certain cancers. Women living with HIV are six times more likely to develop cervical cancer – a disease that is usually treatable if diagnosed early. In 2018, European researchers developed an AI-powered cervical screening that examined 740 HIV-positive women who regularly attended the Kinondo Kwetu Health Center in Kenya. With a very high level of accuracy, the AI ​​analyzed the scans and all women diagnosed with precancerous lesions were cared for.

Change treatment and care

AI is also changing medicine in some profound ways, from helping researchers identify proteins, molecules and compounds for new treatments to shortening the time it takes to develop drugs. Beyond the hope of new HIV treatments, AI has the ability to provide customized treatment plans that can lead to more personalized care with fewer side effects.

Accessibility remains a key barrier to HIV treatment, particularly in resource-limited settings. Nigeria – the largest country in Africa – with over 200 million inhabitants has around 25,000 doctors. The UK, which is a third the size, has over 375,000 doctors. As connectivity improves in Africa, new opportunities for telemedicine and AI are opening up. Amid Covid-19, WhatsApp chatbots in South Africa, Rwanda and Senegal provided timely information on testing and social distancing. Similar tools helped combat misinformation and provided important health updates in Ghana. And that's just the tip of the iceberg.

Imagine a world where someone living alone with HIV who is hesitant to get tested or needs a pep talk before sharing their status with a loved one can communicate in real time with a carefully trained chatbot. This does not mean that bots should ever replace professional healthcare workers – quite the opposite. Bots can complement the incredible service provided by caregivers. In the future, chatbots can encourage more people living with HIV to seek human help for testing, treatment and care. They can help remind patients to take their life-saving medications and monitor adherence. AI can help schedule appointments, navigate health records, and streamline data entry and reporting, giving doctors more time to help patients.

In the future, AI will also play a larger role in the allocation of HIV resources. The more accurate prediction models become, the better experts can identify new trends. In the same way that researchers analyze patterns to determine which flu vaccine to produce each year, experts can more accurately predict which countries, regions and communities are at increased risk of HIV and prioritize resources accordingly to address the threat reduce.

The risks and future of AI

Humans are inherently flawed and therefore any technology developed by humans will incorporate our biases. In medicine, bias can mean the difference between life and death. That's why, as global health experts develop new technologies and AI-powered tools, they also need to educate themselves about people and data from low- and middle-income countries, not just Silicon Valley. Innovation must be carried out in partnership with Africa and other places that will benefit most from these exciting breakthroughs but are most vulnerable to the negative consequences of bias.

Once these technologies are developed, experts warn that they must be shared with low- and middle-income countries and not hoarded by the world's richest. We've seen this act before. By the early 2000s, AIDS was no longer a death sentence in the United States and Europe, thanks to widespread access to inexpensive antiretroviral therapies. But elsewhere, particularly in sub-Saharan Africa, millions of people died needlessly as new developments in testing, treatment and care were only available in Western countries. To end global health inequities, we must harness new technologies while learning from the mistakes of the past to ensure history does not repeat itself.

In an article published in The Lancet, a group of professors from Harvard University and the University of Massachusetts Amherst (disclosure: this author is a proud graduate of UMASS Amherst) issued a similar warning: “The extent to which AI approaches will be useful for HIV prevention will depend on their ability to catalyze the expansion of antiretroviral therapy and PrEP in the communities most likely to benefit.”