[ad_1]
IIt is quite difficult to develop artificial intelligence models using the health data generated in the most prestigious medical centers in the world. Today, a group of humanitarian organizations faces an even greater challenge: using data collected in developing countries to improve public health abroad.
The Precision Public Health initiative, led by the Rockefeller Foundation and unveiled Wednesday on the sidelines of the UN General Assembly, hopes to provide advanced technologies to those parts of the world that have been slow to benefit. The idea is that AI and data science could provide health workers with important information that they might not have otherwise – for example, a suggestion, transmitted via http://www.youtube.com/watch a tablet notification, homes to visit to check mothers and children in need of care.
Initially, the initiative will be funded to the tune of $ 100 million. The Rockefeller Foundation will provide $ 60 million, UNICEF $ 15 million and the remaining $ 30 million will come from other partners in the initiative – the World Health Organization, the Global Fund, the Global Financing Facility supported by the World Health Organization. the World Bank and the public. public-private known as Gavi, the Alliance for Immunization.
publicity
The initiative will be launched in India and Uganda, where donors hope to introduce technological tools to prevent the death of mothers and children. It will be extended by 2030 to eight additional countries, which have not yet been announced. The ambitious goal of the initiative is to prevent six million deaths in these countries during the decade.
The tools will be built using existing health databases collected by front-line health workers or governments, and supplemented by databases not traditionally associated with health, said David Mitchell , Director General of Innovation for Health at the Rockefeller Foundation.
Mitchell cited data on telecommunications, weather and satellites as examples of non-health data that could be used. Genetic data will not be used, confirmed a spokesman for the initiative.
The initiative will work in partnership with organizations such as Medic Mobile, a non-profit organization that builds software for front-line health workers and, in doing so, has collected potentially valuable health data in Uganda and Uganda. in other countries, said Mitchell.
One of the major concerns of the Rockefeller Foundation as it embarks on the effort? The quality of the data with which he works, said Dr. Naveen Rao, executive vice president of health of the foundation. Rao said the initiative solicited the help of data scientists to examine questions such as, "How can we ensure that our models are based on quality data that we actually believe? And how can we test it in real time? "
Health data is already notoriously fragmented and incomplete when stored in state-of-the-art electronic medical records systems, not to mention places where many health records are written by hand. According to Mitchell, at least initially, the initiative will not attempt to "translate a lot of paper-based information that is still dominant in many of the geographic areas we are looking at".
Instead, Mitchell said, "We are trying to focus on data that is already reassuringly high quality, and then we will confirm that the data inputs make sense, that there are no data sets. 39, an anomaly, that there are not many whites, that there are not a lot of strange and repetitive metrics in these areas. "
It will not be an easy task, as Mitchell acknowledged.
Mitchell said the initiative would be based on training data for its algorithms and models "from countries". That would mean that there would be no data from the developed world, "nor even from adjacent geographical areas where they might not be relevant and aligned with the reality of the geographical areas we are aiming at," he said. Mitchell.
This is an approach that could help the initiative avoid some of Watson's challenges for oncology, IBM's product to recommend cancer treatments to doctors around the world. This system was formed by doctors at the Memorial Sloan Kettering Cancer Center in New York – which has created problems because the treatment recommendations of these doctors often do not match the practices of doctors in other parts of the world.
The initiative is wary of the difficulty of obtaining such predictive models, Mitchell said. To this end, he said, "we focus on use cases that are not controversial and do not seek to replace the decision-making process of doctors, nurses, or health experts." and attention, such as the prospect of a high-risk pregnancy or a child with high needs, Mitchell said.
In addition to potential data quality issues, it will not be easy for the initiative to translate ideas into health outcomes.
"We already know what causes a lot of preventable deaths, and it's not necessary for high-tech machines to know it," said Merlin Chowkwanyun, a public health historian at Columbia University, who criticized marketing and rhetoric around accuracy. public health.
With the Rockefeller Foundation's initiative, "we could perhaps discover a little more texture about what we might be able to do more effectively in terms of intervention design," said Chowkwanyun. "But the fundamental reasons why people on a large scale die prematurely or where there are inequalities between groups in terms of morbidity and mortality are very hard ground."
[ad_2]
Source link