CATCH: Community Antibiotic Tracker for Children

CATCH is both a mobile and a web-based health application designed to collect community antibiotic data in children up to 3 years of age, a critical age for microbiome development, from physicians practicing in primary healthcare centers and private clinics, community pharmacists, daycare centers, and parents across Lebanon.

Dr. Khalid Yunis, MD, FAAP, Professor of Pediatrics, Head Division of Neonatology, Director National Collaborative Perinatal Neonatal Network, American University of Beirut

Overuse of antibiotics at the community level appears to have reached an alarming level in Lebanon hosting the largest per capita population of refugees in the world, providing shelter to close to 2 million. In a study published in 2009, the team lead of this solution recruited 117 pediatricians in the greater Beirut area to collect community Antibiotic usage data on 1320 infants during the first year of life [1]. Results showed that antibiotics were prescribed for common cold and acute bronchiolitis, both of viral etiologies. Lower maternal education and pediatricians working in dispensaries were associated with increased antibiotics overuse. A recent study by the American University of Beirut found that 37 % of pharmacists surveyed supported giving antibiotics without a physician’s prescription, and 43% of patients reported getting antibiotics without prescription [2]. 

Globally, 1 child dies every 3 minutes due to sepsis attributed to multidrug resistant organisms [3]. In Lebanon, unregulated antibiotic availability contributes to resistance within the community [4]. CATCH aims to collect essential data on antibiotic usage in the community and storing these data in a knowledge graph that facilitates exploring patterns in the data, allowing for link analysis to effectively develop awareness campaigns and surveillance operations. 

The target audience includes parents of children up to 3 years of age who are taking antibiotics, nurses in daycares, physicians in private clinics and primary healthcare centers, and pharmacists. The graph analytics that will be provided with the help of the knowledge graph representation will help in understanding their needs and will take us further into implementing solutions. This will serve to examine patterns in antibiotics prescription, dispensing and use in a manner that allows for longer-term adaptability to the problem, especially in the presence of spatial or demographic disparities across how and where antibiotics prescriptions are impacting the population, and at a later stage to guide the allocation of proper resources and interventions.

Parents, nurses, pharmacists, and physicians who engage in this application will receive enrollment certificates. This not only contributes essential data on community-level antibiotic usage but also equips them with educational insights from data analysis. Empowered as influential figures, they become community role models, encouraging active participation in understanding antibiotic practices in children up to 3 years of age.

One public good that will come out as a direct output is the collected data about the different major stakeholders involved in antibiotic usage in the community for the first time. A report detailing the outcomes of the surveillance system will be disseminated among the physicians and pharmacists in the community, providing valuable insights to identify gaps in knowledge and spotlight areas of improvement. We anticipate publishing a paper outlining the results of the surveillance system, aiming at sharing our experience with wider scientific community in LMICs. Parents will receive educational materials through the application, tailored to address the identified gaps.

The knowledge graph proposed aims to rationalize all information gathered from the data, to help understand the connections between actors involved in the antibiotic use, and to aid investigations into the behavioral associations behind them. As the graph gets enriched with more data, it becomes a unified and accessible source for shareable data that our team can query without requiring expert knowledge in database management and helps them transition from mere collection to effective exploration. Once the graph’s knowledge is established, we could introduce predictive measures around the antibiotic usage to identify intervention targets and make informed prevention insights.

Enhanced awareness and education on proper antibiotic use within the community, especially among parents and caregivers, can improve adherence to treatments, minimize overuse, and enhance children's overall health. Our solution's data offers valuable insights for evidence-based policies, efficient resource allocation, and public health campaigns to combat antibiotic overuse.

Knowledge graphs integrate data of different formats and qualities from diverse and disconnected sources. This integration results in a unified view of intelligence that analysts can readily and visually query by extracting subgraphs of relevance. When nodes in the graph represent individuals (physicians, pharmacists, parents, children), edges represent a connection between those individuals, and metadata associated with nodes or edges capture additional details (type of prescription, reason behind prescription, etc.), analysts can explore patterns of antibiotic usage, identify networks of physicians/pharmacists engaging in excessive antibiotic usage, whether the network is geographically connected or demographically, identify networks of children being subjected to excessive antibiotic usage and explore the leading factors, and use this knowledge to guide awareness campaigns, surveillance plans, and interventions. These networks also contain spatio- temporal data, allowing for time-based evaluations of antibiotic usage, and analyzing the place and time associated with this activity.

The first year will focus on designing and developing a first viable product behind CATCH application and on preparing the materials for data collection. The first prototype knowledge graph will also be built, and matching complex network analysis modules also developed, using toy data that mimics the real data that will be collected. During the second year, the application will be disseminated to target stakeholders to initiate data collection, with regular monitoring and evaluation of the progress. During the third year, the team will be able to analyze and interpret the collected data, with a focus on the insights needed to guide the three major deliverables (awareness campaigns, surveillance programs, and interventional strategies, including but not limited to, law enforcement and amendments to existing policies). In parallel community data collection will continue through the app.  The application will also incorporate educational material to educate the public about the harms of antibiotic overuse in children. We envision that the success in collecting community data through CATCH at the national level is scalable for replication in different LMICs through collaboration with WHO and other stakeholders.

With the help of the adaptive knowledge graph representation whereby data is continuously fed into the system, graph visualization techniques that show the evolution of the antibiotic usage network can help reveal whether the network is growing in size and in density. A network whose nodes are no longer active indicates that less physicians or pharmacists are dispensing antibiotics. A network whose structure is becoming less dense indicates lesser children are being prescribed antibiotics. Other graph metrics capturing the level of graph connectivity can reveal whether the entire system of antibiotic usage being dismantled with time. 

Indicators:

• Number of pharmacist, physician, and caregivers using the application
• Number of children who use antibiotics (with and without prescription)
• Number of antibiotics prescriptions
• Number of antibiotics dispensed (with and without prescription) 

The team has accumulated experience in working with different segments of the population, including students, workers, families and physicians in the community. This enables the team to address the following expected barriers:

Barrier 1: Implementing two different systems for disseminating the application, covering both iOS and Android platforms may introduce delays in the review process and incur additional costs associated with maintaining both software environments. 

Barrier 2: Lebanon is culturally diverse, with people speaking English, Arabic, and French. To ensure the application suits everyone's preferences, its content will be translated into these three languages. 

Barrier 3: Literacy rates among the community, taking into consideration a population of approximately 2 million refugees, might hinder the reach and use of the application. Therefore, the application will feature audio-visuals to facilitate engagement of the addressed population.

Barrier 4: Encouraging the usage of the application requires the establishment of a system offering incentives. Without such motivation, the application might face challenges gaining acceptance and widespread usage, especially considering the competing priorities individuals face daily in this context. 

Barrier 5: Reporting bias could emerge since participants might under-report the usage of antibiotics or modify their behavior due to their involvement in this study.