Part 3 - What can AI say?
Problem
Living with a chronic disease can be incredibly challenging. Around the world, doctors don’t have the capacity to deliver the outpatient care that these diseases necessitate. Patients' needs go unattended, their conditions worsen, and doctors don’t have the ability to properly monitor and treat conditions as they progress.
In the Peruvian context, there is a deep-rooted centralisation in healthcare services around the capital which means patients outside of Lima don’t receive the care they need. Only 14% of primary care entities are properly equipped, and there is an estimated deficit of 73,345 medical workers (Frontier Tech Hub, n.d.).
To address this problem, Inavya have been collaborating with FCDO and the FT Hub to answer two key questions:
How can we ease the burden of chronic diseases for patients and let them lead their lives on their terms?
How can we help doctors to service a demand for care as efficiently as possible?
The idea
In collaboration with the Frontier Tech Hub and British embassy – Lima, Inavya, a software company, have piloted the use of Avatr to improve healthcare access for at-risk patients in the Peru health ecosystem.
When a patient is discharged from hospital, their doctor creates a care plan which explains the different medicines they need to take, when to take them, and how they should adapt their lifestyle to manage their disease. This is logged into the Avatr system which is connected to an app downloaded by the patient.
Through the app, the patient can interact with a chatbot which can prompt them to take their medicine and answer questions they have around following their care plan. Suppose a patient is suffering with diabetes and wants advice to plan their diet. They can ask the chatbot specific questions around different diets, and have it respond based on verified medical information, which is tailored to their specific needs.
The data generated through these interactions is presented on a dashboard for doctors to make informed decisions about how to manage ongoing care for the patient. For example, the doctor may receive an alert to say that a patient hasn’t been taking a specific medication on a regular basis, and they can intercept this decline in adherence to the care plan and adapt it to suit the needs of the patient.
The team are currently exploring how LLMs can enhance their solution for this use-case, and we’ll focus on that aspect of their work throughout this case study.
Assessment of AI’s fit to the problem
Scoping whether AI is a good fit to help address a problem is a multi-stage, iterative process, which requires more than an abstract consideration of the technology's capability. For Inavya, it all starts with a conversation leading to co-development with healthcare providers and patients For Inavya, understanding the needs of patients and clinicians in Peru was crucial before implementing Avatr in Project EmpatIA. Clinicians in Lima are well-positioned to identify the primary causes of non-adherence to Care Plans, providing rich and descriptive insight into social, economic, behavioural, environmental and medical factors. During the project, exploratory discussions occurred with EsSalud, the Ministry of Health, private hospital groups, pharmaceutical companies, and universities based in Peru.
By starting first with the problems both clinicians and doctors face on a daily basis, the team had a strong foundation to assess whether AI could genuinely add value. They surfaced three key areas for improvement:
-
When patients have left hospital and don’t have access to doctors, they rely on medical information online to address their queries. A lot of this information is of poor quality, which ultimately leads to them making bad decisions about how to manage their own care.
-
Patients needs to be able to get the information they need at a time that’s right for them, so they can lead as normal a life as possible
-
Patients need information which is tailored to their unique condition