Community Surveillance and AI Solutions for AMR Reduction in Kenya

Living Goods and Pendulum Systems’ solution strengthens Kenya’s community-level AMR response by empowering community health workers with enhanced data collection tools to inform targeted interventions; leveraging AI-driven supply chain optimization solutions to improve antibiotics access; and partnering with government to explore the integration of AMR data solutions into government information systems.

Erick Yegon, Global Director, Performance, Evidence & Insights, Living Goods

Antimicrobial resistance (AMR) is a leading public health challenge, with Sub-Saharan Africa experiencing the greatest mortality rate. In 2019, it caused 1.3 million deaths globally and was associated with 37,300 deaths in Kenya. Increasingly seen in pathogens associated with lower respiratory infections in Kenya, AMR threatens the effective treatment of sepsis, possible serious bacterial infections, and pneumonia as leading causes of child mortality. 

Top AMR drivers include misuse and overuse of antimicrobials (with approximately 50% of antibiotics consumed unnecessarily); poor access to quality, affordable medicines, vaccines, and diagnostics; lack of awareness and knowledge; lack of access to clean water, sanitation, and hygiene for humans and animals; poor infection and disease prevention and control in health facilities and farms; lack of legislation enforcement; and rampant use of antibiotics in agricultural settings, where they are often used to promote animal growth rather than treat illness.

Kenya’s Ministry of Health (MoH) has highlighted the need for community-level AMR surveillance, a national antimicrobial use (AMU) and consumption (AMC) surveillance system/platform, and a stockout monitoring system. While Kenya has increased AMR awareness and is building facility-level capacity, gaps remain in leveraging community health workers (CHWs) and community-level AMC/AMR-related data and analytics.

Our solution creates an immediate, lifesaving impact for hard-to-reach populations by connecting them with high-quality health services. This solution will improve visibility into antimicrobial need, use, access, and quality at the community level and leverage this data for amplified impact. By supporting CHWs in delivering an optimized package of interventions to improve rational AMU and ensuring AMA via supply chain optimization strategies, we will improve individual and population health in the short term and strengthen Kenya’s AMR response via community health for long-term sustainability. LG engages CHWs as part of a human-centered design process, including formative research, co-design, solution development, and testing.

Our solution also builds government capacity to better manage, strengthen, and own these health systems themselves, laying the groundwork for longer-term impact and sustainability. This solution will strengthen the government’s ability to meet several needs outlined in the 2017-2022 National Action Plan on AMR, including those pertaining to national and community-level AMR surveillance and stockout monitoring. LG has strong national and county-level government partnerships and will engage county government as part of co-design, adaptation, and scaling recommendations.

LG is committed to ensuring our proven digital solutions and innovations are open-sourced and available for others to use and replicate; the CHT is a public platform, available globally. Specifically, we will:

• Blueprint the AMC/AMU community surveillance workflows and document operational best practices to support scaled deployment by government and other actors.

• Share technical knowledge on the development and deployment of predictive analytics for community-level targeting of AMR drivers.

• Develop AMC/AMU dashboards for government adoption.

• Design additional digital innovations, such as the chatbot, on open-source technology.

LG will also leverage our existing membership of the eight-member Community Health for UHC (CH4UHC) advocacy group and the Community Health Impact Coalition (CHIC) to disseminate and share findings, evidence, and blueprint models for scaled adoption across the sector.

By analyzing historical data on AMU, resistance patterns, and other relevant contextual factors, predictive models can help generate insights to anticipate the emergence and spread of resistant pathogens. This enables healthcare providers and policymakers to implement proactive measures, such as optimizing antimicrobial stewardship programs and enhancing infection control practices. Additionally, predictive analytics and the combination of datasets can aid in the development of early warning systems, enabling timely responses to drivers of AMR and informing public policy and programming to mitigate their impact.

Ultimately, this solution will enhance the quality of care CHWs can provide by increasing AMA and promoting responsible AMU in these communities, improving health outcomes in the short term and reducing AMR in the long term. Enhanced CHW AMR surveillance activities will enable CHWs to better target interventions to those at risk, the chatbot will aid CHWs in providing more accurate diagnoses, and household-level data will enable optimized supply chain data to help increase the availability of quality antimicrobials.

LG is well-positioned to leverage Kenya’s current momentum for community health to drive system-level change for AMR. In addition to our longstanding government relationships, the MoH’s selection of LG as a lead technical partner in the design and national rollout of eCHIS validates the strength of this partnership and depth of our expertise in community health system development.  

Our scaling approach entails:

• Year 1: Short inception phase and co-design with government. Launch AMC/AMU community surveillance, AI-driven forecasting, and LLM chatbot for 300 CHWs in Busia County (reaching 134,300 people), with interactive design adaptation and evaluation. 

• Year 2: Evaluate 12-month pilot. In consultation with the government, position for phased scaling across Busia. Scale solution to 635 CHWs while developing dashboards and advocating for evidence-informed action. Work with the government to support Busia as a gold standard model for AMR surveillance and stewardship. 

• Year 3: Scale solution across Busia. Test model for integration in national eCHIS. Disseminate evidence to MoH and government in Burkina Faso, Uganda, and other Kenyan counties to inform replication. Identify where conditions exist to expand to new markets and prioritize new markets.

Our daily data syncing and monthly analysis ensure a responsive approach to monitoring impact:

Comparative Analysis with Control Sites

Indicators:

• Intervention and control site indicator comparisons (AMR detection rates, stockout rates, accuracy of diagnoses).

Measurement plan:

• Conduct baseline and periodic assessments.

• Use statistical methods to analyze outcomes, adjusting for confounding variables.

Community-Based Surveillance

Indicators:

• Increase in AMR early detection rates.

• SmartHealth coverage/utilization for AMR-related surveillance.

Measurement plan:

• Monitor community transmission and AMR driver trends via daily data syncing.

• Assess targeted interventions’ effectiveness through predictive analytics.

Supply Chain Optimization

Indicators:

• Reduced antimicrobial stockouts.

• Improved supply chain forecasting accuracy, aligned with actual consumption/stock levels.

Measurement plan:

• Integrate and analyze DHIS2 and LG community surveillance data.

• Assess supply chain optimization impact on AMA reliability using monthly analyzed data.

Accuracy of Pneumonia Diagnoses

Indicators:

• Improved accuracy of pneumonia diagnoses.

• Reduced inappropriate household-level AMU for pneumonia.

Measurement plan:

• Monitor diagnoses against confirmed cases using SmartHealth data.

• Evaluate changes in AMC through monthly data analysis.

Performance Metrics

Indicators:

• Evaluation of performance against objectives (i.e. user adoption rates, AMR surveillance improvement).

• CHW and stakeholder feedback on SmartHealth and AI-driven tools.

Measurement plan:

• Conduct surveys, interviews, and focus groups.

• Use performance data to identify areas for improvement/scale-up.

Policy environment: LG-supported CHWs have previously issued antibiotics to certain populations for iCCM. A new Public Health Act may restrict this, limiting Pendulum’s supply chain optimizations to the facility level and shifting some of LG's targeted interventions (i.e. predictive analytics would not be used to identify CHWs likely to mis-prescribe). Facility-level supply of antibiotics, however, remains critical to community-level AMA regardless of whether CHWs treat or refer patients. LG is in conversation with government and remains cognizant that stronger supply chain forecasting is necessary. 

Government dataset access: Because eCHIS and DHIS2 data is managed by government, ease of access to critical datasets is reduced. As a national technical working group (TWG) on eCHIS member, however, LG has agreements with government that allows LG to access DHIS2 data. LG will use this opportunity to access all data required by the project while ensuring compliance with the Data Protection Act of Kenya.

CHW training & performance management: LG’s CHW network is not currently trained in AMR. LG will complement the solution by deploying AMR awareness training for CHWs and their supervisors and integrating AMR within our digitized performance management systems. LG will complement this with e-learning AMR modules for refresher trainings. 

LG has proven that our approach to supporting digitally empowered CHWs excels at reaching vulnerable populations with essential health services at scale. We have demonstrated the need to invest in health systems strengthening starting at the grassroots, community level. Meanwhile, Pendulum’s solutions have successfully increased the effectiveness and efficiency of supply chains across industries and geographies. LG and Pendulum are driven by the significant potential this unique partnership will hold. Not only will this work to improve community health supply systems expand access to quality, lifesaving treatments and early interventions in the context of AMR in Kenya, but we see potential adaptations and applications of this solution for a range of essential health products.

We see this as an opportunity to innovatively scale our impact further through your support. We also want to apply and learn from the expertise of the Trinity Challenge members as we continue to assess how we can further strengthen our approach to ensuring resilient health systems and building scalable solutions to far-reaching health challenges.

In the short term, LG and Pendulum will collaborate with Busia County government and any other dataset owners whose data could be integrated to support the supply chain forecasting solution. In the medium term, LG will collaborate with Medic to build the predictive analytics solution into the CHT for broader reach across the multiple countries and community health systems they serve, as well as the MoH to test solution integration into eCHIS. Other medium-term partners include other county governments in Kenya and KEMSA (as the changes in forecasts will ideally inform the stocks they deliver to each county and health facility). Long-term partners include national and county-level governments in LG’s other countries of operation as we explore future scaling.