Reading a law against its own paperwork: a four-part series on the Universal Health Care Act
A case study in AI-assisted policy research, where every claim has to be traceable to a document
At a glance
| Situation | A national health law with a long list of promises, deadlines and money rules, and public claims about it that nobody had lined up against the source documents |
| Approach | Collect the primary documents, read them in full, check public claims against them, write the findings as a four-part series, and publish dated corrections when later reading proved a statement wrong |
| Window | 9 to 24 October 2026: research began on the 9th, the four parts went live on 11, 12 and 13 October, and corrections followed on the 24th |
| Effort | About 40 commits touching the series (the exact count depends on what is counted), covering research, drafts, revisions and publication |
| Output | Four published posts (about 9,200 words), a source library of 203 official documents, and four follow-up research notes (about 32,700 words) ready for later posts |
The situation
The Universal Health Care Act (RA 11223) promised a lot: integrated local health systems, a comprehensive outpatient benefit within two years, rules on how much PhilHealth may hold in reserve. Years later the public conversation ran on press releases and headline numbers. The author, writing for a collaborative essay blog, wanted to answer a plain question for each promise: what does the law say, what do the implementing documents say, and what do PhilHealth's own records show?
That kind of work is slow for one person. The documents are scattered across agency sites, many are scanned PDFs, some sites block automated downloads, and every figure has a definition that matters (billed versus received, registered versus empaneled, current versus total health spending).
How the work was run
- Build the source library first. 203 PDFs from PhilHealth, the Department of Health, the Commission on Audit and the Supreme Court were converted to Markdown (text layer where present, OCR where not), each with its source link in the front matter. That included the Court's December 2025 decision and its 12 separate opinions, 17 national health accounts workbooks for 2014 to 2025, and 14 implementing issuances retrieved from the UP Law Center's ONAR database. Note: the original PDFs are kept outside the repo, because they are large and can be re-downloaded from the links.
- Write a single primer, then split it. The first draft was one long primer. On 10 October it was split into three posts, and a fourth was added the same day for the outpatient benefit, giving a series of four with a shared closing section listing every part.
- Mark what is missing. Gaps in a draft were tagged with numbered
[NEED]markers, and no post could go out with one left. Each marker became a research task, and several became findings (see below). - Check claims, not just facts. For the outpatient benefit, 49 public claims from PhilHealth releases and news reports were lined up against the circulars and PhilHealth's own statistics. 18 were supported, 12 partly supported, 4 not supported, and 15 could not be checked.
- Screen everything, not just titles. A first pass over circular titles was replaced on 12 October by a full-text screen of all 197 PhilHealth circulars from 2019 to 2026.
- State positions and limits separately. Each post has a TLDR, a section on where the author stands, and a section on how the author could be wrong.
- Correct in the open. When newer documents contradicted a statement, the post kept its original text and gained a dated "Corrections and updates" section.
What the series found
- Part 1, what the Act promises and who delivers it. Sets out the two kinds of health services and which agency owns each.
- Part 2, the study that could not be found. Section 41(d) ties nationwide integration of local health systems to a recommendation from an independent study. The author searched for it, for the oversight committee's reports and for the DOH's promised evaluation, and found none. The post says what was searched and what could not be reached, and asks readers to send the study if they have it. It also reports that the implementing rules are about 2.4 times as long as the Act and restate four deadlines as running from their own effectivity.
- Part 3, whose money. After the Supreme Court voided the transfer of PhilHealth funds, the post compares the Finance department's method for the reserve ceiling with the actuarial one the Act requires (P280.6 billion against P560.55 billion), and checks the 2023 to 2025 numbers. The author also states plainly that an earlier position, agreeing that "idle" money should return to the treasury, did not survive the Court's reading of the law.
- Part 4, the outpatient benefit. Maps each item on the Act's list to the circulars that created it and the dates they appeared. Several headline numbers did not match the circulars, "empaneled" most of all: 40.0 million registrations, but only 14.7 million with a first patient encounter.
Where it was wrong, and how that was caught
The most useful discipline was going back to published work and checking it against newer sources. On 24 October, five dated corrections were added:
- Billed is not received. Part 3 gave government premiums for indirect members as billed amounts. The audited 2025 financial statements showed they were receivables, not cash, so the cash shortfall is larger than first written.
- "Collected" and "paid" overstated it. Part 3 repeated the Court's word "collected" and compared claims with premiums "paid", when part of the government premium was billed and never paid.
- A vaccine benefit did exist. Part 4 said no PhilHealth-paid outpatient vaccine benefit was found. A 2012 circular has paid a fixed amount per case for animal-bite treatment, including rabies vaccine, and it is the third-largest paid procedure by amount.
- A test was worded against the wrong list. One of part 4's three tests for "done" named palliative care as if it were on the IRR list. It comes from a different section of the Act.
- PhilHealth does use the word "comprehensive". Part 4 said it found no PhilHealth statement calling the benefit comprehensive. PhilHealth's 2024 statistics say Konsulta is one. The finding that no document says the Act's requirement is fully met still stands.
The same day, a marker left in the local copy of part 2 was removed, and a backlog topic about the independent study was dropped from the list of follow-ups.
What was left open
- The Department of Health and statistics websites blocked the author's searches, so some "I could not find it" statements describe the limit of reach, not proof of absence. Each says so.
- OCR text can contain errors in numbers and document numbers. The source library warns that each figure should be checked against the PDF before quoting.
- Four follow-up topics are researched but not yet written: local health-system integration, a comparison with other countries' systems, reserve and subsidy by source of funds, and what it would take to call the outpatient benefit complete.
- The original source PDFs are not backed up anywhere except by their download links.
Takeaways
- Build the library before the argument. Having every document converted, linked and searchable made later checking cheap, and it is why the corrections could be made in a single day.
- Name what you could not find. "I cannot find the independent study" was both the strongest finding of the series and an honest invitation to be proved wrong.
- Check the claim, not only the number. Most of the useful discoveries came from asking what a word like "empaneled" or "collected" actually means in the source.
- Corrections are part of publishing. Dated corrections under unchanged text let readers see both what was said and what changed.
- Make gaps impossible to ignore. A marker that blocks publication turned every doubt into a task.
About these numbers
Commit counts, dates and word counts come from the repository history and the post files. The claim-check tally comes from the published part 4. Figures about PhilHealth, the Court and national accounts are those given in the series and its cited sources, and are not independently verified here. The series is the author's analysis and is not legal or financial advice.