Reaction after Dr Leite's lecture

Here’s an improved version of your speech:

Henry Sy Hall

UP College of Medicine

December 2, 2024

I’d like to extend my gratitude to Dr. Leite for visiting us today and sharing his insights on ethical AI. His perspectives remind us how critical it is to ensure that AI in healthcare is both useful and safe.

Someone once said, “Governance is like the brakes of the racecar called innovation.”

If AI is our racecar, and we want to drive it fast and confidently, then we must first ensure that it has a reliable braking system.

When we push racecars to go faster—through better fuels, optimized engines, and advanced aerodynamics—we simultaneously enhance safety with innovations like intelligent brakes, airbags, and helmets.

This is why we set speed limits, enforce seatbelt laws, and require safety features in vehicles. For example, the speed limit is often set at 65 mph because crash tests show that impacts are more survivable at this threshold. Beyond this, safety mechanisms start to fail.

In the same way, health data governance serves as the speed limit for AI.

AI thrives on data. To harness its potential while minimizing harm, we must establish and adhere to sound data governance principles.

These principles must be:

1. Understood

2. Accepted

3. Embedded into the daily practices of everyone in the healthcare ecosystem.

Here at the College of Medicine, we are preparing for a data governance workshop that will help our faculty, students, and community navigate the opportunities and challenges of AI responsibly.

Guided by the principles outlined in healthdataprinciples.org, we aim to:

- Protect people

- Promote health value

- Prioritize equity

These principles are universal—whether you're treating an emergency patient, counseling for a transplant, or integrating AI into your clinical workflows.

Let’s work together to ensure that as we innovate, we also safeguard the health, equity, and safety of everyone we serve.

Thank you.

Frameworks for Ethical AI in Healthcare

Sharing some frameworks that are emerging:

10 Commandments

https://www.computer.org/csdl/magazine/co/2021/07/09473208/1uUtEAWNFOE

7 Deadly Sins

https://human-centered.ai/2024/07/16/seven-sins-of-medical-ai/

3 Health Data Governance Principles

healthdataprinciples.org

Nothing is Important (or The Importance of Nothing)

"Nothing" is a very complex concept that can be simply explained as "nothing". A concatenation of "not a thing", nothing is probably the most abused word of all (if abuse can be measured as the range of all possible assignments of meaning to a word). For some, "nothing" might mean there is something. To others, "nothing" might mean everything. One could say that the spectrum of "not a thing" to "every thing" would cover all possible concepts of anything.

So rather than explaining what is nothing, this blog will explain instead the importance of "nothing" in the context of health.

In traditional health information systems (where reporting was still on paper), there was a concept of "all-or-none" reporting which compromised the quality of the whole national health information system. This was how it worked (or did not work):

If there were ten barangays in a municipality, and ten municipalities in a province, then the all-or-none data reporting required that all barangays must submit complete, accurate data on time to their municipalities. In turn, each of these ten municipalities, after reviewing the barangay data, must submit to their province. Only then can the provinces, after reviewing the reports, submit to their regions. But the reality in the 1990s is that at least one barangay was not able to submit on time which means the municipality, province, and region won't be able to submit on time as well. 

In summary, all we needed was for one barangay to submit nothing and this will affect the quality of the whole country's health information system.

This was the case in the early 2000s when the Field Health Surveillance Information System or FHSIS suffered from late consolidation.

Fast forward 2022, the Internet is on everyone's phone and instant messaging is the norm. There is in fact an easy and cost-effective way for barangays to submit data straight to national, down to the granularity of patient visit. How can we harness this power?

Governance. First there should be a clear structure and decision-making process on how the power is  wielded. Why? Because there are so many ways to skin the cat and there are serious privacy issues.  The governance body will take accountability for the strategy, for the investment, and for taking risks on behalf of the rest of the stakeholders.   

Architecture. Once the governance is formalized, the first thing the decision-makers commission is a blueprint of what health information system will be built. This blueprint serves as a guide to all stakeholders on what the country needs in terms of data management.

But despite this hyperconnected world, there will still be barangays who will submit nothing. What do we do with them? We view them as our underserved stakeholders using the lens of equity. In these barangays, nothing means something - that they lack the minimum capacity to even just send an instant message to the nearby health system and ask for help. As mentioned earlier on, "nothing" is not always nothing -- In data reporting, when nothing is submitted, that could mean that health services are not available in that area, which in itself, requires attention from the authorities. What may be more alarming is when the regional and national health offices receive messages from patients in a barangay and there are no responses coming from any health worker assigned to that barangay. The first thing to do for access is to ensure there are competent and knowledgeable health workers for each Filipino.

So let us not belittle "nothing" because it actually means a lot. Especially if after knowing there is no report emanating from certain barangays, government still does nothing.

The Philhealth Electronic Claims System

 (This is a preamble to another blog where I gave suggestions on improvements that can be made to Philhealth's IT system)

Philhealth is a health insurance corporation created by law in 1995 to purchase health services for its members - the Filipino people.

How does health insurance work?

There are at least three entities in a health insurance system - the payor, the provider, and the patient (sometimes also called a "member").

The concept is risk pooling. Without health insurance, the patient has to pay for his/her healthcare expenses out of his/her own pocket. For many Filipinos, getting hospitalized can easily bring them to bankruptcy when the hospital costs exceed their incomes. But in a given year, only a portion of a population actually needs hospitalization. The majority of the citizens remain healthy and do not get sick enough to be confined. 

Enter health insurance.

By remitting a regular premium to an insurance agency like Philhealth, members can now avail of benefits (such as coverage of their expenses during hospitalization). But this will only work if: all Filipinos regularly pay a premium and the cost of hospital care is controlled. Philhealth is the agency that collects these premiums and also pays for the hospital costs. By making sure the premium payments are collected, Philhealth is able to maintain a shared "insurance fund". By making sure the hospital costs are controlled and not excessive, Philhealth is able to distribute the fund to those who need it. By making sure the hospitals and the health professionals (collectively known as providers) are competent, Philhealth is able to ensure the best outcomes for the members.

In an ideal environment, all citizens regularly pay a premium and all their health expenses are covered. This is traditionally called "universal health coverage".

How will the hospitals and doctors get paid?

First, the hospitals need to be accredited by Philhealth. In this accreditation process, Philhealth makes sure the hospitals are properly equipped to provide care. Doctors also undergo their own accreditation. Aside from being licensed by the Philippine Regulatory Commission (PRC), doctors also pay their membership fees and also provider accreditation fees. 

Second, the providers need to file a claim. When a member gets hospitalized, the providers make sure the best possible care is given to the patient. And then upon discharge when the care is complete, the hospital files a claim.

What is inside a claim?

Quite a number of data but essentially a claim contains details about the patient, about the confinement such as diagnosis and interventions or procedures, and other data to justify the reimbursement of costs to the providers. Note the word "reimbursement" -- this is because the hospital has "advanced" the cost of care (room rate, doctors professional fees, tests, medicines, etc) and are now asking for payment after the discharge.

For Philhealth, a claim contains: CF1 (patient's proof of premium payments), CF2 (summary of the care provided to the patient), CF3 (a simple record of the mother and baby's care). Prior to 2012, these claims were submitted on paper and were processed by human adjudicators.

What is electronic claims?

At the outset, Philhealth already had an internal computerized system for processing the claims it receives. But since the claims submitted by hospitals were on paper, an army of encoders had to type in the details of CF1, CF2, and CF3 into this internal claims system. By 2009, Philhealth Information Technology Management Department began defining the standards needed for hospitals to submit the claims in electronic format. This was called electronic claims or eClaims. The benefits of eClaims were: efficiency, transparency, and auditability. With eClaims, it will be possible to perform three major functions - checking eligibility, submitting the claim, and tracking the status of the claim.

What is CF4? 

In 2019, Philhealth required an additional form called CF4. CF4 contained more detailed data such as courses in the wards, laboratory tests and results, radiology tests and results, medications, and their prices. For hospitals with electronic systems, these data were available but required transformation to fit the Philhealth standard. For hospitals without software, they had to hire encoders and acquire the services of a health information technology provider (HITP) to encode their data and submit it to Philhealth.

 

The 4S Reforms Philhealth can quickly adopt in support of UHC (1st 100 days of the new administration)

(Note: this article was written for those already familiar with Philhealth's electronic claims process (there should be quite a number already given that there are around 2000 hospitals and more than 20 health information technology providers). For those unfamiliar with Philhealth's electronic claims system, pls check this blog or read the circulars on the Philhealth website.)

I was asked for ideas on how to improve Philhealth's information technology (IT). Here are a few suggestions (see disclosure at the end of this article):

1) Simplify

In 2011, Philhealth formed a team to begin developing its electronic claims system (eClaims). Prior to this, all claims received by Philhealth were on paper and an army of encoders had to be hired in each Philhealth regional office to enter the data into the homegrown claims processing system. The vision was to take advantage of automation knowing that once data was digital, the Corporation can  improve efficiency by decreasing the administrative workload of hospitals and Philhealth.

A year later, the eClaims system was deemed ready for scaling nationwide. Back then, eClaims only had three forms (CF1, CF2, and CF3). It was originally designed to just contain simple files (called XML or eXtensible Markup Language) as these were relatively small and were therefore easier to process.  To help hospitals with no IT capabilities, Philhealth accredited Health Information Technology Providers (HITPs) (circ038-2012) who have the requisite knowledge, tools,  and skills to assist them. 

In 2018, Philhealth added claims form 4 (CF4) which required substantially more data (laboratory results, medications, their prices, the course in the wards, etc). In addition, CF4 also required the submission of the encrypted scanned paper medical chart with wet signatures of the doctors. The wet signature requirement was really painful for many hospitals. For some hospitals that already had computerized systems, they had to print on paper, have doctors sign them, re-scan paper back into electronic (PDF or Portable Document Format) and then finally encrypt. At the end of this tedious process, each of these encrypted scanned files would have grown in size which in turn created tremendous technical and administrative burdens on both hospitals and Philhealth. Not surprisingly, soon enough, Philhealth's IT system was strained to its limits in storage and in processing power.

A quick win is to release policy that removes the requirement to submit the encrypted scanned documents. This will alleviate hospitals from the time-consuming process of printing, signing, re-scanning, and encrypting. It will also decrease the file size burden since there are less to upload. Philhealth can just ask hospitals to send the XML files (encrypted of course) which are smaller in size. An even better solution, and if Philhealth leadership is up to it, is to remove CF4 altogether (XML and scanned files) and allow hospitals to revert back to the simpler CF1, 2, 3. 

A further simplification is for Philhealth to accept digital signatures. The Department of Information and Communications Technology (DICT) already released the rules and the digital certificates needed for this (there were no such rules yet when we launched in 2012 thus the requirement for wet signatures back then).

2) Standardize

While Philhealth added more data requirements in CF4, they did not impose standards on terminology. Therefore Hospital A might use "acetaminophen" in the medication section while another might opt for "paracetamol" (acetaminophen and paracetamol are synonymous).  The lack of terminology standards means computers cannot process the claims automatically and each one has to be opened by a human processor to make an evaluation or judgment (thus the term "adjudicator"). Having electronic files pass through a human process is counter-productive. The eClaims files are already digital and amenable to automated processing, but instead, Philhealth persisted with human processors who, like all of us, can get tired and make mistakes. The human process became a chokepoint because even as thousands of electronic claims were arriving at Philhealth, humans can only adjudicate a certain number per day before the quality of their evaluation suffers. Non-standardized data require human interpretation while standardized data can be processed automatically.

In addition, standardizing the terminology will offer efficiencies not just to Philhealth but also to hospitals and research institutions. It opens up huge potential for multi-institutional research, population health, and outcomes research.

3) Smart contracts and blockchain

People may be surprised to learn that even if the claim is electronic, a human adjudicator is still required to evaluate data quality (#2).

If the eClaims files are simplified (#1) and standardized (#2), then it gets more interesting. Philhealth can now design algorithms with business rules embedded within them. Then the eClaims can be processed "automatically" by computers who don't get tired or make mistakes. Philhealth can even commission several adjudication algorithms and have every claim pass thru each one, and reimburse based on a quality score. Claims can be paid on a more granular level (such as completeness, accuracy, timeliness for each data element) rather than all-or-none. If the claim obtains a 90% quality score, then they get paid 90% of the case rate rather than an outright return-to-hospital (RTH) which pushes them into financial difficulties.  On top of these, artificial intelligence can be applied to assess the risk of fraud. The AI can assign each claim a risk score and those that get high numbers will warrant the more detailed review by a human adjudicator.

But wait there's more. 

Philhealth can define the business rules to reimburse good claims automatically on the 60th day as provided by law. And if that claim turns out to be bad, Philhealth can use blockchain technology to deduct penalties from future claims. Blockchain will ensure there is transparency and will enable stakeholders to track payments simultaneously.

4) Strict AI-guided post-audit

Let's go back to #1. The history behind requiring scanned documents is that the Commission on Audit asks for proper documentation to support payments made to hospitals. While that is true, COA didn't actually say hospitals have to submit the whole scanned paper medical chart (caveat: this is my interpretation of that ruling). Rather (and I hope COA can correct me if I am wrong), COA is simply stating that there should be a clear auditable process by which Philhealth can verify the transaction and for hospitals to provide proof to support their claims. I interpret this as allowing hospitals to maintain the medical chart as evidence within their records departments and surrendering them to Philhealth for post-audit when requested ("post-audit" here refers to the conduct of a review after the claim has been filed). Which records should be audited? Refer to #3 -- the claims that breach the parameters of "good" claims based on the consensus algorithm or the artificial intelligence or both.

Therefore, there is no need to submit any scanned medical chart. The only thing needed is for Philhealth and hospitals to agree to collaborate and make the national health insurance program work for the benefit of the people.

Can we reform Philhealth? Yes. Is Philhealth ready for reform? As long as the leadership decides to do so, they can. The changes needed are obviously more than the 4S mentioned here but these four are doable interventions that can be applied now or in the first 100 days of any administration.

(* Disclosure: the author was a senior vice-president and first chief information officer of Philhealth from 2011 to 2013 and was responsible for scaling up the eClaims system from pilot to nationwide implementation.)

For comments, pls email admarcelo AT up DOT edu DOT ph

IT Governance and Enterprise Architecture: a Call to Action to the Incoming 2022 Philippine Administration

Much has been said about the tepid growth of the Philippine information and communications technology (ICT) sector (Rappler) but rather than point fingers, I would like to use this blog to recommend some clear actions the incoming 2022 administration can take to help our country catch up with our neighbors. Don't get me wrong, the Philippines has a deep bench of highly skilled, internationally recognized technology professionals who can stand shoulder-to-shoulder with their counterparts in first-world countries. Many have left the country out of frustration, but given the right environment, these technocrats, working together, can help catapult the country to a leadership position in Asia. 

But how? 

 1) Governance - IT is merely a tool albeit a very effective one. Done well, ICT has the power to transform organizations and in turn, societies. And while "digital transformation" has been bandied about as the be-all-and-end-all solution, the operative word here is not "transform" but something more primordial and implicit - and that is "govern". ICT, like any tool, when governed well, can do mighty things. And any tool, used haphazardly, can cause harm. 

What then is "governance"? 

Many definitions exist but one that resonated with me very well is "ethical decision-making to achieve goals". This is a high-level definition but it shows the key elements such as ethics (which is rooted in character), decision-making (which includes selecting leaders who have the power, mandate, and will to invest major resources), and achieving goals (which includes strategic planning, project execution, and optimizing performance). 

I prefer not to provide any more details because it might scare off the future president who will have the mandate to govern. To keep it simple, Mr (or Ms) President 2022, you are accountable for the development of ICT in the Philippines. Yes, you can delegate this to your Secretary of DICT who takes responsibility, but the buck stops with you. Choose your secretary very well and be brave enough to replace him/her if s/he fails to deliver. 

 2) Architecture - If your house contractor goes to you and says give me 2 million pesos and I will build you a house, would you shell it out? Probably not. For starters, is the building a bungalow, a two-story house, or one with a basement? Does that price include the lot? Will there be monthly expenses in excess of the two million? Is the house sturdy enough so you can add more floors above it later on? Does the price include after-sales support and warranty? Where is the blueprint? While the intent is good (help you get a "house" at the end), the process is very dubious, to say the least, and it is because of a major flaw in the proposition: there is no transparency. 

This is why we have certified architects who know the technology and the ethics of architecture. They come up with a blueprint that serves as a reference point for all stakeholders to be involved in the project. After an interview with the owner to understand the vision and desires (including the budget), the architect goes to the drafting board and returns with a drawing (in blue paper with blue ink thus the name blueprint) with the intent of showing stakeholders a plan they can both see and understand. And that is exactly what a blueprint is: a plan of how the house might look like after it is finished. It is a very powerful tool because it gives a single transparent unified view of what is to be built. Since it is on paper, it is easy to make revisions like moving stairs to different parts of the house or adding or removing rooms, toilets, garages. You can even put your future dreams drawn on the blueprint and just have the higher floors constructed later confident that the foundations to hold them are already factored in. If at all possible, these aspirations are best drawn as early as feasible on the blueprint. 

But ICT systems are even more complex than houses and buildings thus making blueprints more important in this domain. They are more likened to cities. And if there is no clear city plan (that is, no ICT blueprint), just wait a few years and you won't even know how to start solving the problems of organizing all the crisscrossing electrical lines, leaking water pipes, and seeping sewage that come with a rapidly growing population. An ICT blueprint, or an enterprise architecture, is a key enabler for a well-governed, justly financed, transparent ICT system. Without it, any ICT project will be shrouded in doubt - suffering from reputational loss even before it can meet its objectives. The enterprise architecture is so important that it must be formalized through a sign-off by the owner, in the case of national ICT systems - the President. The leader of the country's ICT affixes his/her signature on the plan and this commences its construction. 

Governance and architecture are not expensive activities. For governance, it can be as simple as coming up with just laws and policies. It can be as easy as meeting the right principals regularly to gain consensus and support prior to making major decisions. It is just what good leaders are supposed to do. For architecture (which is more technical), we need to spend for the services of professional and certified enterprise architects, and doing so helps prevent the wastage that comes from a poorly governed and non-architected ICT solution. After one group of architects completes the design, another group can help monitor the construction ensuring that all components of the blueprint are being built as planned. And yet another group of architects can certify completion. Corruption is eliminated if the blueprint is published and accessible to all. I do not have statistics of how much has been wasted in ICT projects. If we can just get 5% of these funds, that should be able to fund the government EA practice. In fact, if we did spend the 5% upfront at the start of every administration, we may not have wasted the 95% anymore. 

In summary, whoever becomes the president, all s/he needs to do is take ownership and accountability for ICT in the country. S/he sets up a governance structure, asks them to publish a transparent enterprise architecture visible to all stakeholders, and assign agency chief information officers (CIOs) responsible for every ICT project of the government. Governance and architecture come together like two sides of the same coin. Governance alone without architecture is hot air. Architecture without governance is just a drawing.

 The ICT stakeholders are already waiting and ready to participate in the national ICT construction.

Resurrection 2

 I realized it is hard to keep and maintain a blog and appreciate Blogger.com for just being there (and protecting our blogs from bots).

Resurrecting again -- maybe next post is after 8 years again :)

At any rate, stay safe -- 

Resurrection

I completely forgot all about this blog. With recent events, it seems a good time to bring it back to life.

What makes us tick? Service to God it is for me. And it is in anything that I do. And today was no different no matter how painful it had to be.

We may not know the future but we can plan for it by living the present as truthful as we can.