March 14, 2023

Helping patients navigate the complexity of long-term conditions with natural language

Ian Wharton
Founder, Chief Executive Officer

We all have our own experiences of long-term conditions, whether first-hand or through a friend or family member. It’s a complex and multifactorial challenge that requires the correct support, content, monitoring, and insight to be delivered in the right way and at the right time.

Natural language processing is a branch of artificial intelligence that enables a computer system to understand text and spoken words like a human. It has fundamentally changed how we interact with computers and is the primary way a patient uses Aide.

Rather than traditional interfaces that require the user to navigate via ‘fixed’ points of interaction (e.g. buttons and prompts), experiences designed using natural language can engage with people in a way that is ‘generative’ and highly individualised at scale. We believe natural language can play a significant role in one of the main challenges of healthcare delivery: an improved patient experience that drives engagement.

Symptoms can be reported in under 10 seconds in Aide.

How we use natural language

A person with chronic disease will spend just 0.01% of their time with the healthcare system. They may come together with their clinician for a matter of minutes over a year as a result of unplanned GP visits or for their annual review.

When combined with research that shows up to 80% of medical information is forgotten immediately after a consultation, there is a measurable communication and information gap in healthcare delivery today. We want to close this gap through the use of natural language.

Aide has short, daily conversations with patients to help them with their medicines, record structured monitoring, and gradually improve their levels of health knowledge. It has the conversations clinicians and patients want (and need) to have that available resources in healthcare systems today don’t easily enable.

Depending on the patient’s medical record and Health Plan in Aide, these conversations can take the form of:

  • Medicine reminders and optimisation
  • Reporting on mental well-being and patient confidence levels
  • Quickly capturing readings for monitoring, such as blood pressure
  • Reviewing and exploring the importance of critical check-ins such as retinal screening for someone with type 2 diabetes (events that are vital to pre-empt exacerbations but typically have low attendance)
  • Helping a patient prepare for a consultation
  • Improving patient access to local services
  • Delivering health content

Improving adherence

A core function of Aide is to support patients in taking their medicines. Through conversation, Aide sends timely and personalised reminders based on the patient's daily routine. It can also capture individual beliefs and in-the-moment challenges relating to that dose. This real-world medicine use is a missing layer of data to help guide clinical decision-making.

Aide can also help someone quickly reschedule a medicine reminder when life gets in the way and capture reasons for skipping – from concerns with side effects to the patient not feeling that they need it. These insights can be discovered within days of using Aide rather than waiting for an annual review.

As an example, a person living with hypertension can be entirely asymptomatic. Once the condition is diagnosed, the patient will likely be placed on antihypertensive medication to bring their blood pressure under control which can cause side effects.

Capturing in-the-moment concerns and challenges

This person can go from feeling fine in themselves to having a lower quality of life due to their treatment. Unfortunately, their clinician may only uncover this in a consultation after several months of non-adherence and a condition suffering from suboptimal treatment. Aide is designed to capture worries about side effects whenever a medicine reminder is delivered. If concerns or non-adherent behaviour is discovered, Aide coaches the patient to speak to their healthcare professional or signposts to trusted medical information at a national or local level.

Uncovering reasons for non-adherence

Discovery with no additional clinical time

At Aide, we want to extend the clinician's reach with progressive design and technology, not replace them. There are, of course, elements to the clinician-patient relationship where technology is not a substitute, but when technology is used to supplement the traditional relationship, there are learnings to be captured that may otherwise have been missed.

A study at the University of Southern California showed that patients are more willing to speak openly and honestly about their health with a virtual clinician than a human. In part, this is due to many patients believing the virtual encounter to be judgement-free.

We are beginning to see similar patterns in Aide. In our NHS England pilot, 41% of patients willingly told Aide that they stop taking their medicine when they feel their condition is under control. For a long-term condition such as asthma, stopping daily preventer medicine when feeling well can often lead to complications. These population-level insights are shared with the practice, which can then take action to support their patients however they choose.

Tested up to age 71

Even today, many health technologies and the associated research stop at age 65. From the beginning, our team has committed not to contribute to this digital exclusion in our design, especially considering it is at this age and beyond when long-term conditions and comorbidity can increase rapidly.

Smartphone adoption by socio-economic group in the UK

So far, Aide has been successfully tested in patients with asthma and type 2 diabetes up to the age of 71, with repeated daily use and marked improvements in medication adherence over the course of six months. Our NHSE pilot was also run in a part of North East England with an area classification of ‘rural, ageing and hard-pressed’, our first step towards showing that regardless of age or socio-economic group, these technologies can be used and are desired. If a person can message their friends and family using WhatsApp, SMS or any other messaging service, they can use and benefit from conversational services like Aide.

'Structured conversation' vs. 'large language models'

Structured conversation systems have been in use for several years, enabling effective and guided interactions between humans and machines. These systems have relied on predetermined rules, templates, and content provided by experts, ensuring the quality and relevance of the information exchanged.

With the recent introduction of large language models (LLMs), including ChatGPT, a new approach to natural language processing has emerged. LLMs learn from vast amounts of data available on the internet, allowing them to generate contextually relevant responses in a more dynamic and flexible manner. While this innovation has significantly expanded the potential applications of language models, it also raises concerns regarding the reliability and safety of the generated content, especially in sensitive domains such as healthcare.

Structured conversation systems, like those used in Aide, provide safe and reliable use of natural language in the healthcare setting. All conversations in Aide are clinically designed and, in the UK, follow NICE guidelines. This ensures that the information provided to patients is accurate, validated, and relevant to their prescribed care. Importantly, these structured conversations do not have the facility to diagnose or advise changes to a patient's prescription.

If you would like to explore how Aide can support your patients or employees, we would be happy to schedule a demo.


  • Patients' memory for medical information. Roy P C Kessels. 2003
  • USC News. Virtual humans inspire patients to open up, USC study suggests. Tanya Abrams. 2014

Other articles you might like