KenyaSurv360: A 360 degree view of antibiotic use in Kenya

Community health workers will collect household data on antibiotic use and garbage collectors will capture images of medication packaging discarded in waste which will then be identified using computer vision technology. We will also collect environmental and animal waste samples to detect antibiotic residue and resistance in isolated bacteria.

The team lead for the solution is Dr Loice Ombajo a physician and infectious disease specialist who leads antimicrobial resistance surveillance across a network of health facilities in Kenya. 

Regulatory frameworks on antibiotic usage in sub-Saharan countries are often not enforced leading to over-the-counter dispensing of antibiotics without prescriptions or clear indication. Globally, the prevalence of community pharmacy non-prescription antibiotic dispensing is estimated at 63%, with higher prevalence in low-and-middle-income countries (LMICs). In Kenya, pharmacies are often the first point of contact when seeking care for common illnesses but there are over 6000 unregistered pharmacies and 1 in 5 pharmacies do not comply with Good Distribution Practices which explains the high rates of antibiotic use without clear indication. In addition, the administration of veterinary antibiotics remains largely unregulated in LMICs. Recent Kenyan studies reveal alarming levels of drug residues in animal products, which can contribute to the proliferation of antibiotic-resistant bacteria, posing significant risks to human health. Antibiotic-resistant bacteria and antibiotic residues, primarily originating from human and animal waste, pose a significant environmental threat. Yet despite this knowledge, inadequate coordination and use of data across human health, animal and environmental sectors has resulted in underestimates of the burden of AMR. In order to successfully tackle the scourge of antibiotic resistance, it is important to adopt a one health approach to measure antibiotic use and resistance in the community. 

Our solution empowers the Kenya National Antimicrobial Stewardship Interagency Committee (NASIC) to effectively undertake its role in developing guidelines, rules and regulations related to antimicrobial use. NASIC is the highest policy and governance body responsible for AMR activities in Kenya and is led by the ministries of health and agriculture with representation from the ministries of environment, fisheries and trade. 

While the ongoing hospital-based AMR surveillance in Kenya yields some data, a significant portion of antimicrobials are utilised outside hospital settings. This proposal enhances our understanding of erratic antimicrobial usage and its impact on health outcomes by focussing on antibiotic use outside the hospital setting. The limited hospital based perspective on antimicrobial use, resistance rates, drivers, and trends in LMICs impedes policy development, depriving health experts of crucial insights needed to create relevant and impactful local policies.

Representatives from the technical committee of NASIC, the community department of the Ministry of Health and the Pharmacy and Poisons Board are integral members of our solution team, having contributed to identifying our proposed  solution. They will also take part in the implementation of the solution and monitoring and evaluation of the solution as part of our advisory committee.

Our solution will improve current AMR surveillance practices, enabling timely insights into community antibiotic consumption and resistance patterns. We will bridge the gap between community-level data and existing hospital-based surveillance datasets for a comprehensive understanding of antibiotic trends across healthcare and non-healthcare settings. Through rigorous analysis and collaboration with key stakeholders, our initiative seeks to inform evidence-based policy decisions and contribute to peer-reviewed publications, thereby advancing the field of antibiotic surveillance in Kenya. The specific public goods related to our solution include: 

• development of a user-friendly, free to use, real-time dashboard to visualize community-level antibiotic use and resistance data
• insights from analysis to identify correlations between community surveillance findings and hospital-based datasets of antibiotic use and resistance.
• dissemination of project outcomes through peer-reviewed publications and knowledge-sharing platforms to contribute to global efforts in antibiotic stewardship.

Ultimately this is expected to improved understanding of antibiotic usage and resistance dynamics within Kenya, enhance capacity for evidence-based policymaking and targeted interventions to address antibiotic resistance, strengthen collaboration between community and hospital-based surveillance systems to facilitate holistic one health antimicrobial stewardship efforts and contribute to scientific literature through high-quality research outputs.

Our solution will have tangible impact on antimicrobial stewardship guidelines and the next iteration of the Kenya national action plan for AMR. 

Our existing AMR surveillance efforts have a proven track record of informing the development of facility and national antibiograms, as well as guiding the formulation of antibiotic usage guidelines at both facility and national levels. We have observed widespread use of broad spectrum 3rd generation WHO Watch antibiotics and high rates of Escherichia coli resistance to these antibiotics by Escherichia coli. These antibiograms serve as invaluable tools for clinicians, aiding them in making informed decisions regarding empiric antibiotic therapy, particularly in regions lacking specific empiric antibiotic guidelines. Additionally, they assist healthcare institutions in planning for their antibiotic formularies.

To date, our antibiograms have solely relied on data from hospital settings, predominantly focusing on hospital-acquired infections and in-patient treatment. Our proposed solution aims to expand this scope by incorporating data from community sources. By integrating community data, we aim to develop comprehensive community antibiograms and subsequently craft guidelines specifically tailored for managing community-acquired infections. This advancement will empower clinicians to make judicious decisions regarding the appropriate antibiotics for patients with community-acquired infections, particularly those who require out-patient care.

In the initial phase of the project, spanning the first half of the first year, our primary focus will be on project initiation. From the second half of the first year and extending into the second year, our main activities will revolve around data collection and analysis. It is during this period that we anticipate unveiling our dashboard and antibiogram for public use.  The third year of the grant will be dedicated to leveraging insights derived from our data to implement tangible improvements in practice and AMR guidelines.

In the third year, our findings on reasons for antibiotic use at the household level have the potential to  inform the type of point of care tests that may be placed within primary care settings to discourage unwarranted antibiotic use. Moreover, insights into antibiotic usage in animal husbandry, particularly for non-treatment purposes, could guide the development of alternative strategies to meet farmers' needs. The antibiotics commonly detected in the community can help inform the antibiotics for selection in resistance testing and influence first line treatment options within health facilities. 

We will utilise a comprehensive Monitoring & Evaluation framework that encompasses three key components:

1. Process indicators: These indicators will track the activities integral to our project's success, such as the number of training sessions conducted for CHWs, the execution of the pilot phase etc. 

2. Output indicators: These metrics will gauge tangible outputs generated by our activities, including the number of households enrolled, CHWs trained, and medications identified through our initiative. To measure the performance of the machine algorithm we will use metrics such as accuracy, precision, recall, confusion metrics and F1 score to measure how well the algorithm classifies medication and provides accurate inference.

3. Outcome indicators: Aligned with the overarching objectives of our challenge, these indicators will assess the broader impact and success of our project. Specifically, they will measure:

   1. The quantity and significance of data insights derived from the collection of antibiotic data within the community.

   2. The frequency of utilization of the publicly accessible Antimicrobial Resistance (AMR) dashboard, serving as a gauge of its effectiveness and relevance.

   3. The integration of our project's findings into future iterations of the National Antimicrobial Plan or other relevant regulations and guidelines, demonstrating our project's influence on policy and practice at a national level.

Computer vision technology is not widely used in Kenya. As such, there is a risk that we may not be able to obtain enough local expertise to launch the product within the project timeline.  To overcome this challenge, we are working with a tech company that has experience in providing digital health solutions. To further support their work, we have mapped out organisations with expertise in computer vision technology globally and prioritised organisations that are affiliated to the Trinity Challenge and where we have had cross-institutional collaborations on previous health and/or technology projects. We will enlist individuals from these institutions to work with our team to help us reach our goal as quickly as possible.

The Trinity Challenge offers us a unique opportunity to pioneer a new one health approach to antibiotic surveillance within the community, an avenue that remains largely unexplored. Traditional funding sources available to us tend to favor more established methods. Through this solution, we aim to gather data that demonstrates the effectiveness of integrating one health surveillance into antimicrobial surveillance efforts, along with the added value of monitoring waste, especially in regions where local pharmacies and informal drug sellers operate with limited regulatory oversight which means that the antibiotics they hold and dispense are not known to the national regulator.

If this approach proves successful, community-based antibiotic surveillance could easily become a standard component of data collection conducted by Community Health Workers (CHWs). Furthermore, integrating waste surveillance to monitor antimicrobial usage within the community could be adopted as part of public health surveillance.

Our solution team would greatly benefit from mentorship in the development of computer vision technology. With this extra support, we can accelerate completion of this project deliverable. By achieving this deliverable more quickly, we would gain additional time to broaden the scope of the technology. This expansion could involve using the technology to identify other relevant antimicrobials within our setting, such as anti-malarials.

Potential collaborators include the Machine Learning Initiative at Imperial College London and the Computer Vision & Robotics Group at the Machine Intelligence Laboratory of the University of Cambridge.