Medical insurance premiums are rising sharply in Malaysia, and many families are asking a difficult question: how long can they continue to afford private medical protection?
The usual explanation is that medical claims are increasing.
That is true, but it is not sufficient.
We also need to ask why claims are increasing so much, and whether every part of that increase reflects necessary and clearly explained care.
A recent World Bank review of Malaysia’s medical insurance and takaful claims data found that claims rose substantially between 2022 and 2024.
It also found that rising costs were driven mainly by the use of more services, rather than prices alone.
For inpatient claims, hospital supplies and services made up more than 70% of claim amounts.
In simple terms, the issue is not only that medical items are becoming more expensive, it is also that more services, procedures, tests, supplies and hospital items are being used and billed.
Unexpected costs
Too often, Malaysia’s medical insurance debate is framed only as an insurance problem.
Premiums go up, policyholders complain, insurers explain that claims have increased, and the discussion turns to how much premiums should rise.
But private healthcare billing is also a healthcare governance issue.
A recent family experience at a private hospital in Petaling Jaya, Selangor, made this problem very real to me.
An initial estimate of around RM18,000 eventually became a bill closer to RM28,000.
The concern was not only the final amount, but also how difficult it was for the family to understand what had changed, why certain charges were incurred, and whether costs had been clearly explained before they were incurred.
When someone in your family is sick, elderly, anxious or recovering, your focus is on the patient.
You are thinking about pain, test results, surgery risks, discharge planning and recovery.
You are not thinking like an auditor.
Yet, hospital bills often require that level of scrutiny.
Families are expected to understand doctor fees, ward rounds, procedure charges, investigations, consumables, medication, supplies and insurance approvals, often while under emotional stress and with limited medical knowledge.
This becomes more complicated when a medical card is used.
Many patients assume that “insurance is paying”.
But insurance is not free money.
The cost returns later through higher premiums, co-payments, exclusions, reduced coverage or cancelled policies.
This is where agentic artificial intelligence (AI) can help, but only if we explain it properly.
Who should deploy it?
I am not suggesting that a patient should open a free chatbot and ask it to judge whether or not a hospital bill is fair.
That would be unsafe and unfair to the patient.
Patients usually do not have access to the full claims data, clinical records, hospital billing patterns or comparable cases needed to make such a judgment.
The most realistic deployer of agentic AI is the insurer or the third-party administrator (TPA) that processes medical claims.
They already receive the claim, itemised bill, diagnosis, procedure details, approval records and discharge documents.
They are also in the best position to compare one claim with similar cases, flag unusual patterns and send the case to a human claims or clinical reviewer.
Hospitals can also deploy a related version of the system.
For example, before discharge, a hospital could use it to explain why the final bill is much higher than the initial estimate, which items caused the increase, and whether the patient or family was told before those costs were incurred.
Regulators can play a different role.
They do not need to see individual patient disputes in the same way.
Instead, they can use aggregated and de-identified claims patterns to monitor whether certain types of charges, supplies, tests or procedures are rising unusually across the system.
This is not a free, ready-made app that patients can simply download.
The basic building blocks already exist in enterprise AI, claims automation and data analytics.
But a Malaysian AI medical claims agent would still need to be built or configured for local use.
It would need secure access to claims and billing data, clear rules on what to flag, comparison data from similar cases, privacy safeguards, audit logs and human reviewers.
In practice, it would most likely begin as a paid enterprise tool used by insurers, takaful operators or TPAs.
Larger insurers may build it internally.
Others may use a vendor platform that is configured for their claims workflow.
The patient should benefit from clearer explanations and fairer review, but should not carry the burden of deploying the technology.
How would it work?
First, an estimate variance agent would compare the initial hospital estimate with the final bill.
If the final bill is much higher, it identifies which items caused the increase, when they were added, and whether the patient, family or insurer was informed before the cost was incurred.
This does not automatically mean that the bill is wrong.
It simply forces the increase to be explained.
Second, a consumables agent would review items such as gloves, syringes, swabs, dressings, implants and other supplies.
It would compare these against similar procedures, hospital stays and patient profiles.
If usage is far above what is normally seen for comparable cases, the bill is flagged for review.
Third, a repeated-tests-and-visits agent would check whether blood tests, scans, ward rounds or specialist visits were repeated more often than expected for a similar diagnosis, procedure, age group and length of stay.
Some repeats may be medically necessary.
But if the pattern is unusual, the system should ask for a clear explanation.
Fourth, a provider-benchmarking agent would compare similar cases across hospitals or doctors.
If one provider consistently records higher supplies, more repeated tests, longer stays or more frequent visits than comparable providers, the pattern can be reviewed by insurers, TPAs or regulators.
The purpose is not to accuse a provider automatically.
It is to ask for an explanation when the pattern is unusual.
Fifth, an audit trail agent would summarise the case in plain language:
- What was estimated
- What was finally billed
- What changed
- Which items were unusual
- What comparison group was used
- What explanation is missing, and
- Who should review the case next.
This is what makes agentic AI different from ordinary claims analytics.
Ordinary analytics may flag a suspicious item or a high-cost claim.
Agentic AI can follow a sequence of checks, gather the relevant information, reconstruct the billing story, explain why the case was flagged and route it to the right human reviewer.
This does not mean AI should approve or reject claims on its own.
Healthcare decisions need clinical judgment, context and accountability.
The AI should flag, explain and escalate.
Human reviewers should decide.
It should also not delay urgent treatment.
The system should not be used to block emergency care or pressure doctors while a patient is unstable.
Its role is to support claims review, pre-discharge explanation, post-billing audit and system-level monitoring.
Privacy must also be protected.
Claims data contains sensitive medical information.
Any such system must follow proper data access controls, use only the information needed for review, keep audit logs, and use de-identified data when patterns are analysed at industry or regulatory level.
Allowing proper explanations
Hospitals may worry that this sounds like an accusation.
It should not be framed that way.
The aim is not to assume wrongdoing.
The aim is to make unusual cost increases easier to see, explain and review.
If the treatment was necessary and properly explained, the record should show that.
Malaysia already uses data analytics in banking and financial regulation.
Customers are not expected to detect suspicious transactions or money-laundering patterns by themselves.
Systems monitor transactions, compare patterns, generate alerts, escalate high-risk cases and keep audit trails.
Healthcare claims need a similar mindset.
Patients should not have to become auditors while their loved ones are in hospital.
If unusual billing patterns exist, they should be detected by systems designed for that purpose, not by tired families reading itemised bills at discharge.
Malaysia is already moving towards better claims data, price transparency and healthcare cost reform.
The next step should be to turn that data into practical oversight.
Agentic AI can help identify where costs are rising, where billing patterns are unusual and where explanations are missing.
It can help insurers challenge unreasonable claims more effectively.
It can help hospitals explain bill changes more clearly.
It can help regulators see repeated patterns across providers.
Most importantly, it can reduce the burden on patients and families.
Medical insurance will remain sustainable only if patients trust that bills are fair, insurers trust that claims are reasonable, and providers know that unusual patterns can be reviewed fairly.
The medical card should protect patients; it should not become a blank cheque.
Dr Manjeevan Singh Seera is an associate professor of business analytics at Monash University Malaysia. For more information, email starhealth@thestar.com.my. The information provided is for educational and communication purposes only, and should not be considered as medical advice. The Star does not give any warranty on accuracy, completeness, functionality, usefulness or other assurances as to the content appearing in this article. The Star disclaims all responsibility for any losses, damage to property or personal injury suffered directly or indirectly from reliance on such information.
