​​​​​

Every parent’s nightmare is when their children fall sick. The anxiety is exacerbated for first-time parents, who may have limited experience in caring for sick children. Unsure if they require critical medical attention, they may rush them to the hospitals for emergency care.

The Opportunity

The Children’s Emergency at KK Women’s and Children’s Hospital (KKH) is crowded every day. However, many children are attended to for less serious medical conditions. When doctors and nurses have to deal with so many less serious cases, children with real emergencies may risk getting less immediate attention.

The Solution

Coming to the rescue is the KKH Urgent Paediatric Advice Line, also known as U-PAL. By providing virtual paediatric advice, parents can better understand if their childen’s condition requires immediate medical care, and this can potentially help KKH to attend to patients who require more immediate attention even more promptly and right-site the patients.

First of its kind in Singapore

Believed to be the first of its kind in Singapore, the KKH U-PAL is a pilot online service for parents and caregivers to seek advice for common paediatric conditions for children below the age of 17 years old. Some of these include fever, cough, diarrhoea, vomiting, and common injuries.

KKH, with support from IHiS, launched U-PAL with the aim of reviewing the number of less complex and less difficult cases seen at Children’s Emergency – particularly the Priority Level 3 cases.

Emergency Priority Level 3 cases refer to patients with acute symptoms but are in stable condition and are able to walk. These patients may be treated by their General Practitioners or Family Physicians with acute care resources. Some examples include sprains, minor injuries, minor abdominal pain, vomiting, fever, rashes and mild headaches.

Building up a solid database through a 2-phased pilot

This year-long pilot is being implemented in 2 phases, where the responses from the live chat in the first phase are used to train the AI-enabled chatbot that will be implemented in the second phase:

Launched in June 2019

• Text-based live chat platform where the healthcare team responds in real-time to public queries.
• To prepare the responses for the upcoming AI-enabled chatbot, all text messages are collected and analysed to build up paediatric medical knowledge and FAQs that can be refined iteratively with the expertise of KKH Children’s Emergency care team.
• Chatbot is enhanced by supervised machine learning – where the algorithm learns based on a labelled dataset – to constantly increase its efficiency and accuracy. This gradually reduces the need for the healthcare team to complement the service in the long run.

To be launched in first quarter of 2020

• Together with a triage engine, the chatbot should be capable of performing basic medical triaging and providing appropriate responses to FAQs for four types of commons symptoms, estimated to cover majority of common acute concerns:
1. Respiratory
2. Gastrointestinal
3. Accident and trauma
4. Febrile illness
• Ongoing efforts to review and improve the chatbot’s capabilities in understanding human language in the local context and providing appropriate responses, as well as keeping the content relevant and updated.
• If U-PAL is unable to answer the question, it will be brought to the attention of the healthcare team.

Facilitating KKH’s adoption of a chatbot solution

Recognising the potential of chatbots in enhancing the delivery of healthcare, IHiS worked with various public healthcare professionals from the three clusters to map out 18 possible use cases and shortlist six chatbot solution providers that can be tapped on.

To support the clusters, IHiS manages the Master Service Agreement with the shortlisted providers and negotiated a slew of terms and conditions with them e.g. their man-day rates. Users can then save the hassle of negotiating the contract with these providers. IHiS would advise on the appropriate category of provider, help organise demo if necessary, and assist our users by coevaluating and co-negotiating with the selected provider.

In addition, IHiS would also conduct security assessments to ensure the chatbot platform is secured, and support our users in scaling up the chatbot solution, where necessary.

KeyReply – the chatbot provider driving KKH U-PAL – is one of the providers on the Health Chatbot Accredited Panel. KKH U-PAL is enhanced by supervised machine learning that is powered by KeyReply’s proprietary Natural Language Processing (NLP) engine. Parents and caregivers can rest assured that there is still a human touch to this online chat service where the more complex queries will be brought to the attention of the healthcare team for follow-up.

Key Challenge: Closing the ‘language barrier’ between a robot and its user

It is no easy feat getting the NLP engine to understand medical queries from users. For instance, the chatbot might not be able to interpret lengthy conversations as users tend to type a long paragraph with multiple contexts, such as having two or more chief complaints in a single message.

As part of the journey to train the chatbot and close this ‘language barrier’, the second phase of the project will see the healthcare team at KKH and IHiS enhancing the knowledge base for a chatbot engine that can provide relevant answers to users.

Since its launch, KKH U-PAL has been managing an average of 37 chats daily, with each chat averaging about 22 minutes. These add up to more than 1,100 chat sessions and about 420 hours of chat time every month!

As of 4 August, the public rated their experience with U-PAL an average of 4.33 out of 5. There is a high satisfaction level with the service as it helps to address parents and caregivers’ queries on common paediatric conditions. In the long term, it can potentially nudge the public to use the service as the first port of call.