Elizabeth Glaser Pediatric AIDS Foundation speeds prevention and treatment with Azure AI solution

When she was 18, Dee Mphafi Tanka tested positive for HIV. A doctor bluntly told her she would get sick and die of AIDS—a terrifyingly real prospect in her home country of Lesotho, where HIV is still the leading cause of death. Tanka found support, treatment, and hope from experts at the Elizabeth Glaser Pediatric AIDS Foundation (EGPAF). The US-based nonprofit works to prevent mother-to-child transmission of HIV and advocates for policies to end HIV and AIDS in children and families. 

Thanks to the medical and psychosocial support she received from EGPAF, Tanka not only defied the virtual death sentence issued nine years ago. She also gave birth to an HIV-free baby girl and now works with young women in Lesotho impacted by HIV and AIDS. "HIV doesn’t define me, nor does it end my dreams for a prosperous future,” Tanka says. “Both my daughter and I represent that achieving an end to pediatric AIDS is possible.”

EGPAF has long been a lifeline for people like Tanka. The nonprofit envisions the end of HIV and AIDS, starting with children, but the world has a long way to go. Disparities between adult and pediatric access to HIV care are widening in countries with high rates of infection. 

Now the nonprofit is turning to AI to help accelerate its mission. To close persistent gaps, offer more effective care, and reach more people, EGPAF is leveraging Azure AI and the Microsoft Cloud ecosystem. 

With the support from Microsoft Hack4Good hackathon and Microsoft partner Squadra Digital, EGPAF has created glAIser, an Azure AI-powered smart assistant. The AI solution mines petabytes of organizational data to surface insights, generate data visualizations, and support evidence-backed decision-making. 

“Microsoft is helping us get to the next stage to reach our mission,” says Dr. Shabbir Ismail Abbas, Vice President of Strategic Information, Evaluation, and Research at EGPAF. “This technology improves the efficiency and effectiveness of our solutions to maximize our data for better programming and higher quality care. This is a new era of up-to-date technology and expertise.”

Accelerating technological breakthroughs

As EGPAF actively works in 12 countries highly impacted by HIV, it uses disparate electronic health information systems, some of which it developed for government health agencies in Africa. Each of these systems collects a variety of information, from demographics to medication specifics. This data used to be siloed, resulting in long delays to compare measurements, such as reduction in transmission cases. That is why EGPAF has established a unified data infrastructure to consolidate program data within the Glaser 360 data lakehouse.

Within Glaser 360, Azure Data Factory pipelines pull information from individual data systems. This aggregated, anonymized data is centrally and securely stored in Azure Data Lake. The scalability of Azure Data Lake enables the nonprofit to expand whenever needed, especially as time passes and they increase the number of people they serve. Azure Databricks streamline Extract, Transform, Load (ETL) processes for raw program data in the data lake, enabling advanced analytics and seamless last-mile reporting as well as visualization with Power BI. 

While Glaser 360 improved EGPAF’s data stance, the nonprofit still did not have adequate human resource capacity for advanced data mining. A longstanding collaboration with Microsoft led EGPAF to join a hackathon run by Microsoft, where the organization connected with Squadra Digital. A month of pre-hackathon efforts laid the groundwork for three weeks of intensive work that developed glAIser, a new Azure AI-based data mining assistant. 

Mining data for transformative insights

The collaboration combined expertise from EGPAF, Squadra Digital, and Microsoft. The solution integrated deep knowledge of global health, compute and orchestration capabilities, AI and machine learning, data integration, security and identity management, and storage. “Everything behind the scenes of glAIser is complex and resource intensive. By teaming up, we divided tasks that resulted in a technical solution that mines our data and produces accurate, relevant, and timely information for decision makers. We couldn’t have done this on our own,” says Mike Bitok, Digital Health and Data Analytics Implementation Manager at EGPAF.

The glAIser solution relies on supervised retrieval augmented generation (RAG) in Azure AI Search. Azure OpenAI Service interprets user prompts and translates them into a structure it can use to match with the correct data sets, global health indicators, and formulas used to calculate specific measurements. The solution selects one of a suite of agents to query the aggregated data, depending on the intent and context of the prompt. Azure OpenAI Service provides grounding guardrails to prevent hallucinations, then generates a response with interactive and exportable charts and graphs, a written summary, and key takeaways that identify trends. For example, when prompted for the number of children tested for HIV by country and sex, the response may compare the results of different countries or point out testing patterns over time.

“With this AI solution, users are not just getting data. They’re getting interpretation,” says Bitok. 

As staff of varying technical expertise can query organizational data using conversational prompts, glAIser will enable every department and function to make use of the solution’s data-based insights. “Facilitating the availability of data will significantly improve our data-use culture,” Ismail Abbas says. “It empowers our people to critically question everything before they make decisions.” 

Communications, fundraising, advocacy teams, and more will base strategy and daily tasks alike on the most up-to-date information. “The AI solution puts data, analysis, and insights at the fingertips of decision makers,” says Charlie Maere, Global Digital Health and Data Analytics Director at EGPAF. “It will be accessible to everyone, from the president down to individual operations staff to clinicians. This is revolutionary.”

EGPAF plans to embed a version of glAIser into the electronic health systems of its governmental partners. "Policymakers, clinicians, and others will be able to know their own data and get insights,” Maere says. The nonprofit anticipates this extensive access to instantaneous data analysis will empower government agencies to fight HIV and AIDS proactively and more effectively.

Improving patient care

The sheer scope of EGPAF’s data—billions of data points from the more than 18 million patients the organization has served—is a treasure trove for insights, but it is virtually impossible for humans to manually sift through that volume. This is now possible with the efficiency, speed, and scale of AI. As Ismail Abbas says, “Generative and predictive AI are helping us improve the quality of service we provide at our health facilities.”

The glAIser solution mines aggregated information pooled from disparate health record systems, using Azure Machine Learning to make data-based predictions. For example, glAIser can identify factors associated with individuals who do not consistently adhere to medication plans, which can lead to severe illness. The solution can then flag current patients who are at high risk of lapses in treatment. “That way, clinicians can intervene and change the course to prevent treatment failure and advanced HIV,” Maere says. “That impact will save lives.”

It used to take data engineers several months to generate reports about important factors, such as babies born to HIV-positive mothers. “We need to analyze our data to get informed decisions faster,” Maere says. Now within seconds, a query to glAIser can generate a list of families that need follow-up. Medical professionals can use this information to prioritize outreach, testing, and treatment—and prevent transmission. 

Designing stronger health programs

EGPAF has contributed to enormous progress in reducing maternal HIV transmission, which has dropped to 50% worldwide. The intervention window to prevent a newborn from contracting the virus is just six to eight weeks, a window that can define a child’s future health and survival. “If we miss that time period, we might lose that child,” Maere says. AI is accelerating EGPAF’s proactive efforts and responsiveness because people with HIV and those at risk of transmission simply cannot wait.

Azure AI tools that underlie glAIser are speeding EGPAF’s work to continually improve programming. "Now we are moving from manual monitoring and evaluation to AI-powered digital M and E,” Ismail Abbas says. Thanks to AI-fueled analysis of information aggregated in Azure Data Lake storage, monitoring and evaluation teams can understand and compare the effectiveness of programs at the global, country, and even individual clinic levels. EGPAF uses Power BI to visualize outcomes and other measurements, further empowering teams to improve health services. 

The glAIser solution will also be able to identify trends, such as HIV infection hotspots, “that we can catch at an early stage,” Bitok says. This insight will help EGPAF step in sooner for the “whole cascade of care and treatment,” he adds. For example, they will be able to boost outreach and direct more testing supplies to highly impacted communities, ramp up advocacy and sensitivity exercises, and direct additional funding and staffing for treatment monitoring to ensure viral load suppression. Shifting resources to where they are most needed can help the nonprofit proactively address public health challenges before they become crises. 

EGPAF’s AI solution will also empower the nonprofit to refine programs, making them even more effective. Using glAIser, the organization will be able to pinpoint areas where patients might otherwise miss out on critical services and care. For instance, a positive HIV test triggers a cascade of actions, such as the immediate start of ART. The glAIser solution can identify potential patterns, such as an upward tick in people not starting treatment, so staff can find new ways to engage this group. “The system can look at trends and raise red flags. Then we can create interventions to help in a timely manner,” Bitok says.

Technology, including the Azure AI-powered glAIser, empowers EGPAF to improve the client’s experience. “No child should have HIV,” Ismail Abbas says. “AI-powered, clinical decision-making support helps our providers to be more proficient and to provide the highest standard of care, in the effort to realize that vision.” 

Discover more about Elizabeth Glaser Pediatric AIDS Foundation on Facebook, Instagram, LinkedIn, and X/Twitter