Antibiotic Buddy: Better prescribing, Better treatment adherence.

Antibiotic Buddy, a WhatsApp Chatbot utilizes existing surveillance and antibiotic use data to strengthen NDoH’s responsiveness to the community drivers of AMR. It will interact with both prescriber and patient, driving behavioral changes to improve treatment compliance, prescriber adherence to guidelines and informed adjustments thereof based on community resistance profiles.

Kim Faure – Team Lead,  SECURE Project Lead, a collaborative initiative with the WHO and DNDi GARDP Southern Africa. With the support of the National Department of Health. 

An estimated 230,710 people died of sepsis related infections in South Africa in 2019[1] with 38,989 deaths associated with resistance to bacterial infections.

 South Africa’s well established AMR surveillance program shows increasing prevalence of ESBL and carbapenem-resistance amongst Gram negative pathogens.  

Variations in resistance and antibiotic use occurs across the provinces, highlighting differences in empiric treatment and appropriate prescribing. In some provinces, increasing “Watch” antibiotic use has resulted in growing resistance to last resort medicines. This limits treatment to a single older drug, as newer antibiotics are too expensive.

There has been a 50% annual growth in outpatient and primary healthcare antibiotic classes use. In the public sector the AwaRe ratio has shifted from "Access" to predominantly "Watch" (52%) drugs; a dramatic change to the prior reporting period.

A sudden increase in empiric azithromycin use during COVID-19 pandemic has continued indicating that correct prescribing practices have not resumed.

Additionally, poor patient knowledge of the need for, how to take an antibiotic course, impacts in effectively clearing infections and fuels resistance further.

[1] Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis, (2022), Lancet, 399, 629 - 655

We are dedicated to a collaborative approach in co-developing our solution, emphasizing iterative testing, improvement, and adaptation. Through close collaboration with the Affordable Medicines Directorate, AMR focal point at the NDOH, and participation of provincial DOHs, we aim to pinpoint pilot hot spots and enhancements to the NSC monitoring mechanisms. The Patient Chatbot design process will draw insights from patient forums in English and subsequently be expanded to include other local languages. For the HCP Chatbot, an advisory panel comprising doctors, prescribing nurses, clinical pharmacists, will be consulted to identify official, evidence-based guidelines to serve as guardrails for our AI tool and ensure a comprehensive practitioner-focused design.

Our target audience will benefit by:

• Patients: education and guidance in judicious use of antibiotics, fostering Antibiotic Champions within communities to reduce antibiotic misuse.
• Healthcare Professionals (HCPs): being supported in making informed treatment choices by integrating standard treatment guidelines and providing easy to understand alerts on regional resistance profiles.
• Provincial DOHs, Hospital Clinical Managers, AMR Committees, NDoH: will benefit from the enhanced NSC dashboards, facilitating tracking and improvement of inappropriate prescribing, identification of potential outbreaks, adjustment of antibiotic treatment policies, and adaptation of procurement processes to seasonal fluctuations in antibiotic needs.

Antibiotic Buddy generates a substantial public good by offering an innovative and accessible tool to combat Antimicrobial Resistance (AMR). The solution contributes valuable knowledge and data to the public domain, uncovering gaps in knowledge of both the patient and the HCP to enhancing awareness, understanding, and management of AMR. Antibiotic Buddy's open-source model ensures that its advanced language models can be built and scaled to take into account various languages and when integrated with evidence-based information, are globally accessible on fair, reasonable, and non-discriminatory terms.

The solution provides a public good in the form of reducing healthcare related cost related to increasing AMR, as its primary goal is to improve healthcare outcomes and reduce the societal burden of AMR.

By leveraging widely used platforms like WhatsApp, Antibiotic Buddy facilitates cost-effective and scalable interventions. The data generated, including patient adherence rates, resistance profiles, and prescribing patterns, becomes a shared public resource through the NSC, aiding national, regional and global efforts to monitor, analyze, and combat AMR. In summary, Antibiotic Buddy not only addresses the immediate healthcare challenge but also contributes to the global public good by fostering collaborative, data-driven solutions to the complex issue of Antimicrobial Resistance.

Antibiotic Buddy creates tangible impact by directly engaging with patients and HCPs.

For patients, the solution provides accessible and personalized information, improving adherence and empowering Antibiotic Champions within communities. The LLMs employed in AB have already shown improved medication adherence and impacted on patient outcomes. [1]

By enhancing HCPs' decision-making through simple to read evidence-based guidelines, specific real-time resistance profiles AB drives appropriate prescribing.

The logical link between these activities and outcomes is supported by the solution's integration with the NSC, leveraging rich longitudinal data for immediate interventions and long-term policy adjustments. The NCS has already demonstrated its impact during COVID-19 pandemic to  provide visibility of critical drug supplies.  [2]

Pilot testing in AMR "hot spots" will demonstrate if:

·       patient education can improve patient adherence and the gaps in awareness to address educational needs,

·       by prompting HCPs  we can change inappropriate prescribing behaviors and what the education needs are.

By focusing on hotspot areas and collaborating closely with the Affordable Medicines Directorate and provincial health departments, AB strategically targets areas most in need of additional direct support to create ensuring tangible and equitable impact in the ongoing fight against AMR.

[1] (Ref: 10.1089/tmj.2019.0305)

[2] https://www.ghsupplychain.net/news/forecasting-crisis-using-demand-planning-prevent-medicine-stock-outs-south-africas-public

https://www.ghsupplychain.net/news/reducing-wastage-and-saving-money-using-national-surveillance-center-improve-medicine-supply

 https://www.ghsupplychain.net/news/supporting-supply-chain-aiding-resilience-and-adherence

 https://www.ghsupplychain.net/national-surveillance-centre-technical-brief

 https://www.ghsupplychain.net/news/now-you-see-it-improving-medicine-availability-south-africas-provinces

Year 1: Initial Implementation and feasibility

Focuses on implementation in the initial hotspot areas. This involves deploying the Chatbots, enhancing the NSC capabilities, and closely monitoring the impact on AMR. The objective is to gather substantial data to assess the feasibility and effectiveness of the solution in hot spots.

Year 2: Expansion and Country Customization

Building on the insights gained in year 1, the second year involves expanding the deployment to additional hotspot areas within country whilst monitoring impact on AMR and use. This expansion allows us to validate the scalability and adaptability of the solution in diverse settings whilst creating a team of Roll out Champions from within the provincial structures – creating a sustainable a solution.

Year 3: Customization and Collaboration

Focusing on the customization of the entire concept for deployment in another country. This phase includes evaluating the AMR landscape, adapting the NSC and Chatbots to local needs. This phase emphasizes knowledge transfer, capacity building, and establishing partnerships. Our goal is a scalable model that can contribute significantly to the global fight against AMR.

Success in achieving impact goals will be systematically measured through a combination of quantitative and qualitative indicators and tracked on the NCS through the existing dashboards on consumption and AMR and new “hotspot” monitoring dashboards. Key performance indicators include:

• Patient Adherence Rates: Tracking the percentage of patients completing prescribed antibiotic courses through the submission of daily photos or text confirmation.
• Prescriber Adherence to Guidelines: Monitoring HCPs' alignment with evidence-based treatment guidelines and their response to resistance trends changing.
• Reduction in Inappropriate Antibiotic Use: Quantifying the decrease in the prescription of "Watch" category antibiotics, particularly in identified AMR hot spots.
• AMR Outcomes: Assessing the impact on AMR-related deaths and infections by analyzing longitudinal data from the National Surveillance Centre and how it has been modelled by the GRAM Study.
• Stewardship Interventions: Evaluating the effectiveness of interventions triggered by Antibiotic Buddy in improving prescribing practices and reducing AMR.

Performance against these metrics will be continually assessed, building on insights from pilot testing. The success of Antibiotic Buddy will be evident through the positive trends in these indicators, demonstrating its efficacy in improving antibiotic use, reducing resistance, and ultimately contributing to better public health outcomes.

Several barriers may impact our goals for scale up:

• Infrastructure Challenges:  Limited internet access or outdated devices may hinder patient and HCP engagement. To overcome this, we will optimize AB for low-bandwidth conditions and pilot in urban areas to start.
• Data availability and financial constraints: when we expand AB into other countries, may impact the development of their NSC. Our country selection will therefore include those with some existing data or partners supporting this acquisition.
• Cultural, Educational and Technology use Gaps: Misconceptions about antibiotics and a new technology use may impede patient adoption. Leveraging patient forums to refine AB’s design to make the tool engaging and simple to use from the start will be our plan.
• Regulatory Compliance: Adhering to evolving personal data protection is crucial. Our design team will need to align consent measures to data protection laws.
• Market Resistance: Resistance from HCP’s to use or behavioural change gaps may pose challenges. We will design the solution using well respected peers, implement peer to peer training programs and showcase the benefits.

Our experience with AMR interventions in South Africa has highlighted the need for new approaches beyond traditional education and awareness to address inappropriate prescribing and use.

Antibiotic Buddy (AB) targets primary healthcare, general practitioner rooms, and district-level hospitals, aiming to address inappropriate prescribing. The Trinity Challenge will be pivotal in helping grow the NSC and innovate by helping fund the pilot of the Antibiotic Buddy as a potential solution to overcome the current critical barriers:

• Enhanced Data Impact at ground level: Leveraging the robust data and visualization tools within the NSC, our solution will amplify impact on HCPs by delivering powerful data at the ground level in a format they can use. Feeding in patient compliance data and their knowledge gaps will enhance policy makers responses to address awareness.
• Digital Innovation Reach: AB's novel use of AI and WhatsApp aligns with the Challenge's emphasis on digital innovation and will allow us to broaden the solution's reach and impact by testing this technology as a tool to effectively gather citizen data and drive behavior change. Optimized Hotspot Interventions: Deploying in identified "hotspots," AB benefits from The Challenge to test out the impact at the frontline in areas with the most challenges.

• Jhpiego – it would be valuable to work with Jhpiego to help develop the training and implementation processes for our solution so that a train the trainer approach may be used with a focus on making the technology design simple and user friendly.
• Wellcome – leveraging their ability to analyze large data sources, their modelling capabilities and health impact insights to analyze the AMR and use data on the NSC and create visual displays to improve decision making.
• AWS and Microsoft. For the Chatbot to be integrated in a cloud environment with sufficient computational resources, it will be advantageous to be in partnership with either AWS or Microsoft.  We have partnership status at both cloud providers and will be keen to explore the advantages of such partnerships for this project.  

I was involved in the South Africa Hackathon with Prof Mendelson. I helped to facilitate a part of the event and then became aware of the amazing potential we had to use this call to help solve a problem we have been experiencing in SA with AMR. Thereafter a series of meetings led to the development of the collaborative team who is submitting this proposal.