How Singapore’s CHAMP chatbot became a scalable population health tool

During the Covid-19 pandemic, a healthcare team in Singapore was brainstorming ideas to improve and scale chronic disease patient care remotely.

This was one of the key problem statements that the team at the National University Health System (NUHS) sought to address during that time, shares Professor Ngiam Kee Yuan, who is also the Lead of the Artificial Intelligence (AI) Office at NUHS.

Speaking to GovInsider, he says the team tapped into AI to automate patient support to reach more people without needing much internal manpower.

After a year and a half of designing, testing, and building the system, the programme went live in late 2022 under the name of CHAMP, says Ngiam.

“Since then, we have been progressively onboarding our patients who have chronic diseases on CHAMP, and we now have over 28,000 people on the programme,” he notes. 

Currently, CHAMP is the only system that integrates directly with the electronic medical record (EMR) system, using AI to stratify and nudge patients to look after their chronic diseases better.

CHAMP is designed to support patients with hypertension, diabetes, and hypercholesterolemia, and plans to cover other chronic diseases like asthma in the future.  

The CHAMP Team was awarded the Best in AI-Driven Healthcare Innovation award at GovInsider’s Healthcare Day 2025 for successfully applying AI to improve clinical outcomes in healthcare. 

Prof Ngiam shares more about the process and impact of CHAMP in achieving real-world healthcare outcomes at population scale.  

Breaking barriers for better patient compliance

One of the key aspects of CHAMP is the use of WhatsApp to reach out to patients on the programme, which has been an intentional choice since the start, says Ngiam.

Given that most chronic disease patients are seniors, the self-tracking tool prioritised a user-friendly design.

“WhatsApp is the app that everyone has,” notes Ngiam. “So, we took time and technical effort to enable it as the main chat interface with our patients”. 

The CHAMP team held interviews and user engagement sessions to ensure the usability of the user interface (UI), user experience (UX) and overall functionality of the tool, notes Ngiam. 

He adds that another benefit of the chatbot is that it doesn’t need new features or buttons that a conventional application would require. 

The designing of a simple interface has been an essential choice to enhance patient compliance as other blood pressure management systems require separate applications or devices that elderly patients struggled with, and as a result, do not upload their readings. 

All that patients need to enrol to CHAMP are the NUHS App and WhatsApp installed on their mobile phones, and a functional blood pressure monitor.  

“A lot of our patients already have access to blood pressure machines; they don’t want to buy another machine just to integrate it to their phones. [CHAMP] is simple enough for patients to just take their blood pressure reading and then key in that value into the WhatsApp chat,” explains Ngiam.  

He adds that the simple interface has helped to increase compliance in submitting results, with some patients submitting results daily or more than once a day.  

With more convenient home monitoring of blood pressure, patients enrolled in CHAMP have experienced significant improvements, with over 81 per cent achieving blood pressure targets set by their healthcare providers. 

The role of EMR integration and AI nudges 

One of the features of CHAMP is the AI-driven nudge based on patient risk segmentation.  

The risk segmentation follows a scoring system based on local and international guidelines, supported by validation studies to ascertain its usefulness, explains Ngiam.  

For this to work, the integration with EMR was a step done even before CHAMP was created.  

“We built a platform called Endeavour AI which is the basis for this integration with the EMR system, and is also the basis for all of our AI products that we have today,” says Ngiam.  

The risk scoring consists of 27 variables and is integrated to the backend of EMR systems on an event-driven basis.  

Whenever a patient does a blood sugar test in any one of the public clinics, the risk scoring algorithm is automatically triggered and nudges the patient.  

If the blood sugar test shows a high value, the outcome for that patient will be different than that of a regular value, notes Ngiam.  

“This gives us the ability to get the patient’s data and let AI do the calculations, and for this to flow back into the EMR system,” he adds. This step enables providers to access chatbot-collected data in real time and support patient care.  

To protect this data, there are security appliances at every level that help to adhere to ISP compliances for cybersecurity and patient’s information, adds Ngiam.  

Expanding care to other conditions 

Beyond the three active programmes in diabetes, hypertension, and hypercholesteremia, the CHAMP team is working to include asthma, heart failure, and eye conditions moving forward. 

Improvements in the large language model (LLM) are critical to enable these scaling up projects, says Ngiam.  

Currently, CHAMP does not use an LLM to directly generate responses due to the risk of hallucinations. Instead, the messages are fixed based on medical expertise and rule-based systems.  

The LLM does intermediate level processing to help interpret patient’s inputs, questions, or responses.  

“With the improvements in LLM technology, I think we are moving toward a state where the LLMs may be able to give more accurate responses, and when that happens, we will be able to provide more programmes and services,” says Ngiam.  

Patients who have a record with the public health system can register and get onboarded onto CHAMP. 

Currently, CHAMP is distributed throughout National University polyclinics and partner General Practitioners (GPs) and is scaling up to reach more patients. 

“Instead of spending money on different programmes or buying machines, this is one way we are providing free care to our users, to help them better manage their disease,” adds Ngiam.