AI-Powered Strategies for Combating AMR in Tanzania (ASCAT)

Community-level data on evidence-based Access to Antimicrobial Agents (AMA), Antimicrobial Use (AMU) in humans and animals is limited. Also, information on Antimicrobial Resistance (AMR) challenges in both interfaces is limited, especially in the rural communities. Empowering community health workers (CHW) with One Health Artificial Intelligence solutions would address this challenge.

Professor Robinson Hammerthon Mdegela is the Primary Investigator and the Solution Lead under investigation.

Misinformation and knowledge gaps on drivers, magnitude and solutions for AMR as well as the limitations of accessing the real time diagnostic services at the community level, represent main specific problems that the proposed project seeks to solve. 

In the United Republic of Tanzania alone, in 2019, there were 12,500 attributable and 54,000 associated deaths with AMR [1]. Inappropriate use of antibiotics, inadequate knowledge about antibiotics, weak compliance to legislations, lack of reliable laboratory diagnostic services, compounded with misinformation are the main drivers of AMR. These altogether counteract the existing efforts to combat this global calamity that either directly or indirectly drive our country into extreme poverty.

Shortage of CHWs, both for human and animal health services as well as distant locations for consultation are among serious challenges that drive community members to opt for self-medication or treatment of their animals without medical/veterinary prescription. Small scale and rural farmers often rely on informal sources of information on use of antimicrobials due to unavailability of reliable systems to guide them in decision making. Thus, development and application of Artificial Intelligence compatible and adaptable with rural and urban settings may offer a promising solution. 

[1]  Murray et al 2022.

General public: The proposed solution serves all Tanzanians suffering from challenges linked to irrational AMU and subsequent AMR. The target audience are the populations in the study areas affected by AMR. The project will support them through provision of correct and reliable information on AMU and AMR using ICT based tool for management of infodemic; linking them to two centralised hubs (for human and animals health) to provide responsive actions in real time, while equally remaining sensitives to societal values to ensure local ownership and inclusiveness. 

Patients (clients): have vested interest in getting the right AMU for their  correctly diagnosed disease(s), a solution that will be provided by the AI linked to the two centralised hubs for authenticating the solutions. The platform will also provide linkage of patients to geo-mapped nearby health care facilities for further guidance on their health conditions/diseases

Livestock owners:  have a vested interest because the AI improved animal health and a lessened risk of disease transmission. The platform will also provide linkage of livestock owners  to geo-mapped nearby veterinary facilities for further guidance on their livestock’s conditions/diseases.

 Local government Authority: has vested interest in obtaining true data that represent a true picture in the community. 

The proposed solution will provide most of the public good in the form of Knowledge generation. Such goods will be in the form of Publications of articles in International Peer Reviewed Journals, and in Proceedings of Scientific Conferences; Policy briefs and fact sheets for policy makers; Popular publications in the form of leaflets, Newsletters,  and feature articles in newspapers to be disseminated through papers, websites and social media as well as materials to be aired through television and radio. The MSc students will produce Dissertation and scientific reports. 

The proposed solution will also provide a product in the form of services through an open sourced model for generation of sentences with language understandable by the community members and CHWs; model for detecting and blocking distorting (fake) information that would lead to infodemic; and a mobile up integrated to a developed system to curb infodemic as well as solutions for AMR.  

The main challenge for AMR in Tanzania is linked to rigidity in behavior change as a result of limited evidence through data at community level.  Using AI through HWS and Crowdsourcing, as well as mobile phone and chatbot technologies, the entire target population including the underserved or vulnerable population will be researched. It will include communities in rural areas, children and women and people with disabilities.  

 The logical links between elements of the result frame work will include:

(i)  Activities: Collect citizen data on the magnitude and trends of diseases and infections leading to high AMU; gather data on the magnitude of misinformation, practices related to irrational use of antimicrobial agents and failure to comply with biosecurity principles;  development and validations HWS, Electronic Network system for AMA, AMU and AMR using block chain, machine learning and chatbot system.

  (ii)Outputs: Community based data on priority diseases and risk factors; community based evidence and magnitude for misinformation;  developed and validated models for collection of citizen data using AI linked to mobile based chatbot system.

(iii). Outcomes: Improved knowledge and practices on AMA and AMU; Changed behavior towards rational use of Antimicrobial agents; Evidence based policy change on AMU and AMU at community level.

We are proposing sequential strategies for scaling up our impacts. Firstly, in the next year, we will leverage our findings by rolling out knowledge translation packages relevant to different audiences to the local government authorities in health and veterinary sectors in all tiers from districts to the village levels in Morogoro region. We will specifically target the frontiers of changes including Community Animal Health Workers,  CHWs, community members, secondary and primary school students; and religious and opinion leaders. We will use banners, brochures, audiovisual media, and social media platforms to ensure wide coverage.

Secondly, in the next 3 years, we will scale out the project to additional two regions (Mwanza and Dodoma) to capture varying geographical and epidemiological predispositions in different settings, using similar approaches deployed in the Morogoro region.  We will conduct consultative meetings with stakeholders from local government authorities (Morogoro, Mwanza and Dodoma), professional councils, professional associations, public-private partnership (PPP) representatives, information and communication technocrats within mobile phone companies and regulatory authorities like Tanzania Medicines and Medical Devices Authority.

We will also expand our scope beyond Tanzania, by forming the East Africa Regional Consortium of AI solutions on AMR.

Measuring success against impact goals will be achieved through Monitoring and Evaluation using the framework under the NAP-AMR for Tanzania.

This will involve periodic monitoring and evaluation while implementing the plan or activities, communicating progress and capturing lessons learned. Thus, we will monitor and evaluate impact from the beginning of the project. Whereas monitoring will be done on daily basis, Evaluation will be done at the end because the project period is short. 

Progress of the project will be measured using Monitoring indicators at output and outcome levels. We will have several indicators such as Number of collected of sentences and words, Number of webs surveyed, Number of people participated in questionnaires, Amount of data compiled, Number of model trained, Size of the model developed and Number of people using the system.

At the goal level, the project will collect and use baseline and the end-line data to measure change in behavior and improved knowledge and practices. Indicators for Evaluation will include, increased seeking behaviour for diagnostic services before treatment, increased demand for prescription before accessing Antimicrobials; increase demand and use of AI in reducing  misinformation while ensuring access to correct information; acceptance and rate of usage of innovation.

The barrier that currently exist include:

1. Financial barriers because Tanzania is among developing countries with limited funding allocation for research, making her depended on financial support from development partners. 

2. Technical barriers are on limited access and usage of AI supported with smartphones, chatbots and other electronic gadgets on solution for global health challenges including AMR. Additional technical barrier is on language as Kishwahili is the National language.

3. Legal: On the proposed solution, there are a number of legal barriers for considerations. They include confidentiality of information directly from the individuals as well as on data to be mined from mobile phone Companies.

4. Cultural: AI technologies can raise complex ethical and societal questions, such as concerns about bias, discrimination, privacy, and autonomy. Addressing these issues requires a deep understanding of both the technology itself and its broader implications for society.

5. Policy: This barrier is based on the fact that, as a county it lacks policy, guidelines and regulations for AI developing and Implementation in the  health sector.

6. Market: This barrier is foreseen because we target a solution at local context in Kishwahili language which may limit the market to non-Kiswahili speaking countries. 

Tanzania is among countries with organised system in addressing AMR using One Health Approach. It is evidenced from from the National Action Plan of AMR (NAP-AMR)  2017 - 2022 and the NAP - AMR 2023 - 2028. 

This challenge is relevant because is in line with NAP - AMR 2023 - 2028 that with emphasis on initiatives at Sub-National level where data and actions at citizen levels are missing

The AMR challenge is accelerating parallel to increased misinformation with rapid expansion in technology. We can harness and take this advantage to accelerate the solution of increasing AMR challenge. 

For the first three years there will be none.