One patient, one record — but are we ready?

Singapore has long understood something that most healthcare systems learn only after costly mistakes: fragmented patient data kills. When a patient visits a specialist clinic, then a polyclinic, then ends up in an emergency room, the information about their allergies, their current medications, their recent test results — if it lives only in paper files or in a clinic management system that does not talk to anyone else's — every subsequent clinician is flying partly blind. Duplicated tests. Missed contraindications. Decisions made without context that exists, just somewhere no one can reach it in time.

The National Electronic Health Record was introduced in 2011 to address exactly this. Fifteen years later, the vision has remained largely intact but the adoption has not. As late as 2024, only around 30 per cent of private healthcare providers were contributing to NEHR. The public system was largely on board. Private specialist clinics, clinical laboratories, radiological services, and dental clinics — the parts of the healthcare system where many Singaporeans receive regular care — remained on the outside looking in.

01The law that changed the picture

On 12 January 2026, Singapore's Parliament passed the Health Information Bill. When enacted, the Health Information Act will make it compulsory for all licensed healthcare providers to contribute key health information — allergies, vaccinations, diagnoses, medications, laboratory test results, radiological images and discharge summaries — to the NEHR. The aim is explicit and ambitious: One Patient, One Health Summary, One Care Journey.

This is no longer voluntary. Starting in early 2027, every licensed healthcare provider in Singapore — from major private hospitals to neighbourhood general practitioner clinics — must contribute key patient health information to the NEHR system. Patients cannot opt out. Their consent is not required. The Ministry of Health has determined that comprehensive data sharing is simply necessary for proper clinical care in the modern era.

The legislation also introduces stringent cybersecurity standards aligned with the Cyber Security Agency's Cyber Essentials framework — multi-factor authentication, patching, monitoring, incident reporting — requirements that apply equally to a major hospital and a two-room GP clinic. Severe, systemic failures to protect data carry fines of up to S$1 million. The law is serious. The timeline is firm. The Act is planned to take effect from early 2027, making 2026 the critical window for healthcare providers to align their operations.

02The problem the law cannot solve

Legislation can mandate what data must be shared. It cannot conjure the infrastructure to share it from. And for a significant number of Singapore's healthcare providers — particularly the older, smaller private clinics that have operated the same way for decades — the infrastructure simply does not exist yet.

Many healthcare organisations face significant challenges in adapting to NEHR requirements: legacy systems that cannot communicate with external platforms, a lack of standardised data formats, manual processes that increase the risk of errors, data security and privacy concerns, and limited internal IT resources. These are not technical edge cases. They are the operating reality of a large portion of Singapore's private healthcare sector.

For clinics that have kept paper records for twenty or thirty years, the challenge is more fundamental than upgrading software. If you are still using pen-and-paper or a legacy Clinic Management System that is not linked to the cloud, you need to take urgent and immediate action. Your CMS must be NEHR-ready, which means you must be able to contribute key health information online. For that to happen, you must first be able to digitalise your patient records.

That sentence — you must first be able to digitalise your patient records — contains a decade of work for some providers. A patient who has been seeing the same GP for twenty years has two decades of handwritten notes, physical test results, referral letters, and prescription histories that exist only in a folder in a filing cabinet. Before any of that can flow to NEHR, before any AI can reason about it, before any clinician at another facility can access it — it has to be converted from paper into structured, searchable, machine-readable form. That conversion is not automatic. It is not quick. And it is not free.

03Why interoperability is not just a compliance checkbox

The frustration with healthcare data fragmentation is not new and not unique to Singapore. Globally, interoperability — the ability of different clinical systems to exchange and use data meaningfully — has been promised and partially delivered for decades. The gap between what technology can theoretically do and what healthcare systems actually do with it remains wide, for reasons that are as much organisational and cultural as they are technical.

But the stakes of closing that gap are now much higher than they were even five years ago. The reason is AI. A clinician reading a patient's record manually can work around incompleteness — they ask follow-up questions, they make educated guesses, they build a picture over time. An AI model reading that same record to generate a recommendation, flag a drug interaction, or predict a deterioration risk has no such flexibility. It works with what it is given. And what it is given has to be structured, verified, and semantically meaningful — not a folder of scanned images that a computer cannot read or relate to anything else.

This is where the concept of vector embeddings becomes critical, and where the gap between a compliance exercise and a genuine healthcare AI capability becomes most visible. A vector embedding is a mathematical representation of a document or piece of data — a way of encoding not just the content of a record but its semantic meaning, its relationship to other records, its context within a patient's full health history. When medical records are converted into vector embeddings and stored in a governed knowledge environment, AI systems can:

• Find semantically similar cases across thousands of patient records — not just keyword matches but meaning matches
• Surface relevant precedents when a clinician is making a treatment decision for a rare or complex presentation
• Detect patterns across a patient population that would be invisible to any individual clinician reviewing individual records
• Power conversational AI that lets a doctor ask "what do we know about this patient's cardiac history" and receive a synthesised answer grounded in verified records, not a generic response
• Flag potential drug interactions by reasoning across a patient's full medication history, not just the medications recorded in the current encounter

None of this is possible if the underlying records are still paper. None of it is possible if the digital records are in formats that different systems cannot read. And none of it is possible if the data has not been enriched — indexed, structured, and converted into the vector representations that AI actually operates on.

"The NEHR mandate creates the legal obligation. The data quality determines whether it becomes a genuine clinical intelligence system or an expensive compliance exercise."

04Legacy data is not the enemy — it is the asset

There is a temptation, when confronting the scale of the digitisation challenge, to treat legacy records as a problem to be managed rather than an asset to be unlocked. That framing is wrong, and it matters.

A GP clinic that has been operating for thirty years holds thirty years of patient data. That data — if properly digitised, enriched, and made accessible — is extraordinarily valuable. It captures longitudinal health trajectories that no algorithm trained on short-term data can replicate. It contains the kind of granular, real-world clinical context that academic medical datasets rarely include. For an AI system trying to understand how a patient's chronic conditions have evolved over two decades, or what treatments have and have not worked for them historically, that legacy record is irreplaceable.

The challenge is not that the data is bad. The challenge is that it is locked — in formats that computers cannot read, in systems that do not talk to each other, in filing cabinets that have never been opened by an algorithm. Unlocking it requires a digitisation process that is not just technically sound but legally defensible. Records that will feed clinical AI systems and contribute to NEHR must be admissible evidence if their provenance is ever challenged. That is why Evidence Act certification — ensuring that digitised records meet the legal admissibility standards of Section 116A — is not a nice-to-have for healthcare providers. It is a foundation requirement.

05What this means for Singapore's healthcare providers right now

The Health Information Act comes into force in early 2027. That is less than a year away. For providers who are already operating on modern, cloud-connected clinic management systems, the compliance pathway is relatively clear. For providers who are not — and that is a substantial portion of the private sector — 2026 is the year the work has to happen.

The practical steps are not optional and they are not sequential. They have to happen in parallel: legacy records need to be digitised and certified, clinical management systems need to be upgraded or replaced, staff need to be trained on new data handling procedures, and cybersecurity controls need to be implemented to meet the HIB's specific requirements. Any one of these steps takes months. All of them together, for a small clinic with limited IT resources and no dedicated technology team, is an enormous undertaking.

Trusted Hub understands this. We have spent 25 years working with organisations that hold legacy data in physical form — insurance companies, legal firms, government agencies — helping them move from paper to governed, AI-ready digital knowledge. The healthcare sector faces exactly the same structural challenge, with the added complexity that the data is sensitive, the regulatory environment is strict, and the consequences of getting it wrong affect patient safety, not just business efficiency.

Our approach is not to replace legacy systems — that path is often impractical, prohibitively expensive, and risks losing historical continuity. It is to unlock the value in them: digitising physical records through our Evidence Act–certified process, converting them into vector-embedded, AI-readable knowledge in the Trusted KnowledgeVerse, and enabling the clinical intelligence that the NEHR mandate was always intended to power. The records are already there. The knowledge is already there. It just needs to be made accessible to the AI and the clinicians who can put it to use.

The One Patient, One Health Summary vision is the right one. It will improve care coordination, reduce duplicated tests, catch interactions that individual clinicians miss, and lay the foundation for a healthcare AI capability that could meaningfully improve outcomes for Singapore's ageing population. But the vision is only as useful as the data that feeds it. And right now, a significant portion of that data is in a filing cabinet, waiting to be unlocked.