florio HAEMO

Florio GmbH, MunichiOS and Android, freev4.2.3, 22 Jun 2026Returns 404 on the Singapore App Store

What it does

Estimates a plasma factor level between doses and displays it as a single number on the home screen. Logs injections, bleeds, pain, activities and wellbeing, optionally pulling activity data from a wearable through Apple Health. A caregiver mode lets one parent log for several children under 13, and the whole record can be shared with the treatment centre. The store copy carries its own qualifier under the factor number: availability dependent on treatment plan.²

The evidence

A two-part longitudinal survey in Central Europe, funded by Sobi, followed 66 users and then 50 of them about six months later, a 75.8% retention rate. 92% stayed very or rather satisfied, 88% found it intuitive, and 78% had entered all their infusion data for the preceding fortnight.¹ Those are good numbers for a logging tool. In the same paper the authors state plainly that the app does not support on-demand therapy, which is the clinical name for treating a bleed that is happening now.

What users say

One rating on the UK App Store, five stars. That is not a sample. Apple's public review feed returns no written reviews for the storefront at all, so there is nothing to quote and nothing is invented here.

Where it stops: a caregiver with a swollen knee at 2am opens florio HAEMO and is shown a modelled percentage. The app is honest about its purpose in its own evidence base: it exists to produce a trend for the next clinic visit, not an instruction for the next ten minutes. Its own caregiver mode is built around a parent logging for a child under 13, so the wife who infuses her husband and the daughter caring for a father with joint damage are not users it has.

myWAPPS and WAPPS-Hemo

McMaster University Health Information Research UnitiOS and Android, freev4.3, 24 Jul 2026On the Singapore App Store, too few local ratings to display

What it does

The most clinically sophisticated tool in this category. WAPPS-Hemo is a population pharmacokinetic service that takes two or three sparse blood samples and builds an individualised factor VIII or IX curve; myWAPPS is the patient-facing app that renders that curve on a phone. It shows a live estimated concentration, flags when the level enters a warning zone, and holds the regimen for offline reading.³

The gate

It cannot be self-started. A treating physician has to complete the pharmacokinetic report on wapps-hemo.org before the app activates at all.⁴ The Coalition for Hemophilia B, reviewing four tracking apps side by side, lists the same constraint as its leading con and adds that not every treatment centre supports it. That review's conclusion about the whole category is one line: emergency guidance is not discussed.⁶

What users say

One rating on the US App Store, five stars, and no written reviews in Apple's public feed. Again, no sample, and nothing invented.

Where it stops: myWAPPS answers "what is my factor level right now" with real modelling rigour. It has no vocabulary for "and the knee in front of me is swollen", so it cannot connect the number it holds to the symptom the person is looking at. It also fails the 2am test for anyone whose centre never ran the sampling.

Microhealth Bleeding Disorders

MicroHealth Inc, USiOS and Android, free, English and Spanishv3.34.0, 10 Jul 2026On the Singapore App Store, too few local ratings to display

What it does

The best-reviewed tool in the category by a wide margin, started by a man with severe haemophilia A working with a Columbia University researcher, and broadened from haemophilia to bleeding disorders generally. Tracks infusions and bleeds, personalises a treatment plan with reminders, scans product lot numbers, monitors home factor inventory and triggers refill requests, supports several family profiles, and notifies the treatment centre nurse when a bleed is logged.⁵ Its own store copy states it is designed for patient use only and is not a replacement for professional medical advice.

What users say

4.8 out of 5 from 407 ratings on the US App Store: a real sample and a genuinely liked product. Read as a group, the reviews carry a quieter finding. Across nearly a decade of feedback from the most engaged users in this community, nobody is asking the app to tell them what to do about a bleed. They ask for more bleed categories, better scanning and independent access for teenagers. The category has trained its own users not to expect a decision from it.

Where it stops: Microhealth routes information to a nurse or a pharmacy after the event. It does not evaluate a bleed while it is happening, and the nurse it notifies is not awake at 2am either.

HemMobile

PfizeriOS and Android, freev6.0.1, 2 Jul 2025Returns 404 on the Singapore App Store

What it does

Logs the date, time, site and reason of every infusion for any factor product, captures lot number and expiry through the camera, photographs a bleed and sends it to the treatment centre, sets reorder and appointment reminders, and locates nearby treatment centres. It has no category for von Willebrand disease or haemophilia C.⁶

Its own disclaimer, printed inside its own screenshot

The App Store gallery image for the infusion log carries the text in full: this application is not intended for curing, treating, seeking treatment for, managing, or diagnosing a specific disease, disorder, or any specific health condition.⁷ The single most direct statement anywhere in these files that the category is not built for the 2am question comes from a manufacturer's own marketing asset.

What users say

2.9 out of 5 from 7 ratings, a decade after launch, for a tool backed by one of the largest pharmaceutical companies in the world. The negative reviews cluster on two failure modes: data loss on update, and crashes. The basic logging function is not fully trusted, well before any question of guidance arises.

Where it stops: HemMobile disclaims the entire job in its own store copy, and its feature set assumes an open treatment centre relationship during business hours. Photograph the bleed and send it, and nobody reads it until Monday.

Haemtrack

UK Haemophilia Centre Doctors' Organisation, National Haemophilia DatabaseiPhone app and web, freeLaunched 2008UK only, tied to NHS centre registration

What it does

The national electronic home therapy diary for the United Kingdom, with more than 2.5 million treatment records entered since launch.⁹ Registration is compulsory for anyone on home delivery and recommended for anyone on prophylaxis, and the data flows from the patient's phone to their centre and on to the national database. This is the deepest institutional penetration any haemophilia logging tool has achieved anywhere.

The evidence

Its own stewards published a design and implementation paper in Haemophilia titled, in part, "strengths and weaknesses", which is a rare and creditable thing for a national body to do about its own system.⁸ Even at 90% centre coverage, adherence to the logging itself has never been complete.

Where it stops: Haemtrack exists to feed a clinical registry. It records what treatment was given, after it was given. It is compliance and outcomes infrastructure, and it makes no attempt to be anything else.

HIRT, the Hemophilia Injury Recognition Tool

Canadian Hemophilia SocietyMobile self-management guideThe closest existing thing to a bleed decision tool

What it does

The one tool found anywhere in this survey that was built to help someone decide about an injury rather than record one. It is scoped tightly: young men aged 18 to 35 with mild haemophilia, facing a minor injury.¹²

The evidence

A qualitative feasibility study of 12 men. Perceived confidence in injury self-management rose significantly after use (p=0.004), and five themes came out of the interviews: accessibility, credibility, the benefit of alarms, confidence and usefulness. The study's own caveats are the important part: twelve people, self-reported confidence rather than a measured clinical outcome, and a population deliberately restricted to mild disease and minor injury.¹²

Where it stops: HIRT is proof that the shape works and evidence of how little has been built. The population it excludes by design, severe and moderate haemophilia facing an ambiguous or serious bleed, is exactly the population in these files. It is also Canadian, unavailable here, and has not been developed further in the decade since the study.

Eyes on Haemo

Hong Kong Haemophilia Society with The Chinese University of Hong KongFunded by the Hong Kong Jockey Club Charities TrustFree, iOS, v1.0.2

Why this one matters more than its feature list

It did not appear in any web search. It surfaced only through a direct query to Apple's own search index, and it is the closest regional peer to what this project proposes that exists anywhere.⁵⁷ Hong Kong is a comparable small, high-income Asian market with its own patient society, and its society partnered with a university and a local charitable trust to build a haemophilia app. Singapore has a society, two academic medical centres and the same problem, and has not.

What it does

In its own words, tailor-made for patients with haemophilia and their carers, to record, review and share bleeding and injection events. Injection reminders, monthly calendar and body-chart summaries of bleeds and injections, one-click report sharing with medical professionals, and the society's own event listings. The body chart is its one distinctive feature and it is a visualisation of the past, not a way in to a decision about tonight.

Where it stops: no triage logic, no red flags, no urgency output. It is the same structured logging tool as the rest of the category, built for the clinic relationship. Its value to these files is as a delivery model rather than a product: a patient society, a university and a local funder is exactly the shape a Singapore equivalent would take, and it is evidence that shape is fundable in this region.

Ada Health

Ada Health GmbH, BerliniOS, Android and web4.8 out of 5 from about 11,000 US App Store ratings

What it does

A branching question sequence that returns a ranked list of possible conditions with an urgency recommendation. Ada states 10 million users and 20 million completed assessments, self-reported and not independently audited.

The evidence

The best performer in the largest head-to-head study of the category. In Gilbert and colleagues' 2020 BMJ Open comparison of eight apps against a panel of seven general practitioners on roughly 200 vignettes, Ada placed the correct condition in its top three 71% of the time, against a 38% average for the other apps and 82% for the doctors, and gave advice at least as cautious as the panel in 97.0% of cases.¹⁹

What users say

The second review is the category's self-description in a user's own words. The product these apps deliver, and the reason people rate them highly, is reassurance. For haemophilia, reassurance is the failure mode.

Where it stops: Ada is the best of them and it still has no way to know it is speaking to someone with a factor VIII level under 1%. Its priors are built on what the general population presents with, and a general population's tingling ankle is not an intra-articular bleed.

Babylon Health, and what happened to it

Babylon Holdings Ltd, UKValued near two billion dollars at peakChapter 7 bankruptcy, 9 August 2023

Why it belongs in these files

Babylon is in this landscape not as a competitor, since it no longer exists, but as the category's clearest documented safety record. A UK consultant oncologist, David Watkins, spent roughly three years from 2017 testing its chatbot and reporting the results to the Medicines and Healthcare products Regulatory Agency and the Care Quality Commission, going public on BBC Newsnight in February 2020.²⁰

What the testing found

Identical chest-pain symptoms entered for a man and a woman of the same age produced different advice: the man was sent to an emergency department with heart attack information, the woman was told to see her general practitioner in six hours with panic attack information. An older male smoker with chest pain was offered gastritis or a sickle cell chest crisis and told to see a doctor in six hours. A simulated 66-year-old woman entering a painful breast lump as her only symptom was told osteoporosis was among the likely diagnoses.²⁰

What the regulators said

The MHRA's clinical director for devices wrote to Watkins in 2020 stating that his concerns were all valid and ones the agency shared, and in 2021 the agency publicly flagged a legal gap in how triage chatbots are classified.²⁰ Separately, the Care Quality Commission's 2017 inspection of Babylon's GP at Hand service found it was not providing safe care in accordance with relevant regulations in some areas; Babylon took the regulator to the High Court to block publication of the report, lost, and paid eleven thousand pounds in costs.²¹ Babylon's public response to Watkins was to call him a troll.²⁰ It filed for bankruptcy in 2023 and was sold for parts.²²

Where it stops: the best-funded consumer triage chatbot ever built produced gender-biased cardiac advice, fought its own regulator in court, and is gone. Any argument that a generic engine can be pointed at a rare bleeding disorder has to get past this record first.

WebMD Symptom Checker

Internet BrandsiOS, Android and web4.7 out of 5 from about 135,000 US App Store ratings

What it does and how well

Free-text and checkbox symptom entry returning a list of possible conditions. It is the most-tested tool in the field, appearing in eight of the ten studies in the 2022 npj Digital Medicine systematic review, where its accuracy on the primary diagnosis ranged from 3% to 53% depending on which conditions were tested.¹⁸ The spread matters more than the midpoint: the same tool is near-useless on some condition sets and adequate on others, and the user cannot tell which case they are in.

What users say

Where it stops: a reference library with a symptom form attached. Enormously popular, and the reason it is popular is that it is a place to read rather than a place to decide.

ChatGPT and other large language models

Not built for triageUsed for it anywayThe realistic default at 2am in 2026

What the evidence actually shows

The headline numbers are better than they look. A 2025 JMIR study reporting 100% triage accuracy for GPT-4 Turbo was scoring against the Modified Early Warning Score across 1,854 simulated scenarios: a structured single-score classification from vital-sign inputs, which is a materially easier task than open-ended symptom interpretation.²⁶ A small 2025 comparison against NHS 111 online on ten vignettes had all three systems scoring nine out of ten; the language models over-triaged a kidney infection to emergency care while NHS 111 under-triaged an acute angle-closure glaucoma.²⁵

The rare-disease problem

Work on rare-disease symptom checking states the mechanism plainly: these tools are built primarily on published literature, which often contains incomplete data on rare diseases, and diagnostic accuracy is compromised as a result.²⁷ The purpose-built rare-disease tool that does exist, FindZebra, states explicitly that it is meant for physicians and other medical specialists, not for patients. No source found reports language-model performance specifically on inherited bleeding disorders.

Where it stops: a general model will answer a haemophilia question at 2am, fluently and instantly, which is precisely the risk. It does not know this patient's factor level, it has no signed clinical source behind the answer, it leaves no record for the treatment centre, and neither the family nor the receiving doctor can audit where the answer came from.

The evidence on the category as a whole

Four studies2015 to 2022Vignette benchmarks and systematic review

2015, the Harvard benchmark

Semigran and colleagues audited 23 free symptom checkers against 45 standardised vignettes split evenly across emergency, non-emergency and self-care tiers. The correct diagnosis came first in 34% of cases and appeared in the top ten in 58%. Triage advice was correct in 49% of vignettes.¹⁶

2022, the five-year follow-up

The same 45 vignettes were re-run against the surviving tools. Median triage accuracy had moved from 59.1% to 55.8%: flat, within noise, after five years of development. Performance decreased on two specific tasks, and one of them is the only one that matters here: advising when emergency care is required. The median app correctly identified 51.9% of emergencies, and few if any outperformed an ordinary person making the same judgement with no tool at all.¹⁷

2022, the systematic review

Wallace and colleagues screened 177 studies, included 10, covering 48 distinct checkers. Primary diagnosis accuracy ran 19% to 37.9%; triage accuracy ran 48.8% to 90.1%. One included study found ophthalmic emergencies triaged correctly 39% of the time against 88% for non-urgent ophthalmic cases in the same tool, which is the shape of the problem in miniature: headline accuracy conceals far worse performance on the urgent end. The review concluded that reliance on symptom checkers could pose significant patient safety hazards.¹⁸

The pattern: the category is measurably weakest at exactly the judgement a bleeding disorder needs it to be strongest at, and it has not improved. Where a tool is safe, it is safe by escalating, which recreates the over-referral this project is trying to reduce.

Why the logic inverts for haemophilia

The team's own synthesisNo cited source frames it this wayBuilt from the evidence above and the clinical guidance

Stated plainly

No paper found in this survey describes bleeding disorders as a case where symptom-checker logic runs backwards. That framing is this team's, assembled from four separate findings, and it is set out here as analysis rather than as anyone else's claim.

One. These engines are tuned on population base rates, and haemophilia affects about 10.31 per 100,000 males in Singapore. Asked about a warm, sore joint, any well-built model weights strain, arthritis and minor injury far above an internal bleed, because for almost everyone asking, it is.

Two. Absence of symptoms is not reassurance in this condition. In one head-trauma series 72.9% of episodes presented with no symptoms,²⁹ and the Children's Hospital of Philadelphia pathway starts factor correction concurrently with imaging rather than waiting for the picture to clarify.²⁸ A checker reading "knocked head, feels fine" returns self-care, which is the correct answer for the general population and the wrong one here.

Three. The clinical doctrine is the opposite workflow. MASAC Document 257 states that pain and swelling in a joint or muscle directly warrant replacement therapy regardless of investigation results, and that treatment should not be delayed for test results.³⁰ A symptom checker's entire structure, ask more questions, narrow the differential, then recommend a care level, is the inverse of treat first and investigate alongside.

Four. The signal that matters most has no place in a generic question set. The aura, a bubbling or tingling warmth in a joint before any swelling, is the treat-now moment. "Unusual tingling, nothing visible, mild sensation" is the exact input pattern that pushes a general-purpose tool towards reassurance.

What follows: a haemophilia tool cannot be a thinned-down general symptom checker. It has to be a pre-committed decision tree carrying this condition's own inverted risk logic, which is what the precedents in Category G were built to do for other conditions.

KKH U-PAL, the Urgent Paediatric Advice Line

KK Women's and Children's Hospital with SynapxeFree web chatbot, children under 18Now operating around the clock

What it does

The closest thing in Singapore to the tool this project proposes. A chat service, reached from a launcher on the hospital's own page, that performs, in the developer's words, basic medical triaging and provides responses for the majority of common acute concerns, escalating to the care team when it cannot answer. Its stated purpose is to reduce unnecessary emergency department visits among Priority 3 cases.³¹ Originally an office-hours advice line, it has been rebuilt with generative artificial intelligence and breastfeeding support and now, in KKH's own words, operates around the clock at no cost to users, with only standard local data charges applicable.

Who it is for, in its own words

Common paediatric conditions in children under 18: fever, cough, diarrhoea, vomiting, minor injuries. And who it is not for, quoted in full: "For certain pre-existing or chronic medical conditions, KKH U-PAL may not provide the specialised care your child requires. If your child has any of the following medical history that still requires ongoing follow-up and care, please consult directly with your child's attending doctor or designated care team rather than using this service." The list runs: bowel or abdominal diseases, liver conditions, kidney disease, autoimmune disorders, heart conditions, brain or neurological diseases, endocrine disorders, cancer, complications from prematurity, or long-term immunosuppressive medication.³¹

The ambiguity that matters

A bleeding disorder is not named in that list. It is also plainly a chronic, specialist-managed, haematology-led condition with ongoing follow-up, which is the description the clause is written to capture. A haemophilia parent reading it at 2am gets no clear answer about whether the service is meant for them, and the direction they are pointed instead, consult your child's attending doctor or care team, is the door that is closed at that hour. The page also carries its own accuracy caveat: responses may not always be completely accurate, and users should exercise appropriate judgement.

Where it stops: U-PAL is the strongest possible evidence that this is buildable and that Singapore's health system is willing to build it. It is also, on its own published scope, generic paediatrics that stops at the door of specialist-managed chronic disease, covers nobody over 18, and has no bleeding-disorder logic in it. It proves the model and marks the gap in the same paragraph.

NurseFirst, 6262 6262

Woodlands Health for Ministry of HealthFree, nurse-staffed, generic protocols8am to 11pm daily

What it does

Emergency-trained nurses working from protocolised questions with emergency-physician oversight, directing callers to general practitioners, urgent care centres, mobile medical teams or teleconsults. From 1 June 2025 a nationwide trial began warm-transferring non-life-threatening 995 callers to it, with patient details passed across and an ambulance dispatched immediately if the caller deteriorates.³²

The overnight gap, in the Ministry's own words

"Outside of these hours, if a 995 call is assessed to be non-life-threatening, SCDF will deal with it according to its current protocols."³² Since April 2019 those protocols mean the Singapore Civil Defence Force does not convey assessed non-emergency cases to hospital at all; the caller is advised to see a general practitioner or arrange their own transport. This is the clearest single confirmation of the overnight coverage gap available, and it comes from the Ministry rather than from inference. No outcome data, extension decision or published evaluation of the nationwide trial could be found up to September 2026.

Where it stops: the right model, on the wrong clock, with the wrong protocols. A generic nurse line reading generic questions to a haemophilia caller will either over-refer defensively or miss an inverted presentation, and between 11pm and 8am it does neither, because it is closed.

HealthHub

Synapxe, Health Promotion Board and Ministry of Health4.7 out of 5 from 28,573 App Store ratings, 4.1 from 47,909 on Google Play1M+ installs, v6.0.13

What it actually contains, checked rather than assumed

Biometric login, a customisable dashboard, appointment booking and mobile pre-registration, caregiver access allowing a parent to view records and transact for a dependent, bill payment, prescription viewing and refill management, and the National Electronic Health Record.³⁵ There is no symptom checker, no triage tool and no care-navigation feature anywhere in the feature list, the marketing description or the national programme page. The team re-verified this against the live listing rather than repeating the claim.

What users say

That review matters more than its star rating. Caregiver access to a dependent's medication record is the exact function a haemophilia caregiver depends on, and it is the function the reviewer lost.

Where it stops: the national platform holds the record and books the appointment. It has never been asked to answer a question, and none of its 28,600 reviewers mentions after-hours guidance, because there is no occasion to.

sgcovidcheck.gov.sg, the precedent that was switched off

MOH Office for Healthcare Transformation, NUHS and NCIDLaunched March 2020Domain no longer resolves, checked 3 Sep 2026

What it was

Singapore's government took age, travel and exposure history and symptoms and returned an immediate care recommendation: monitor at home, or seek care in a named setting. It was written up in BMJ Health and Care Informatics.³⁷ It is the closest thing the state has built to a public-facing "should I seek care and where" tool for the general population.

Where it stops: it has been retired, and the retirement is confirmed at the level of the domain rather than assumed: sgcovidcheck.gov.sg no longer resolves at all. The precedent is that Singapore builds this kind of tool when a condition is urgent enough, and decommissions it when the emergency passes. Haemophilia's emergency does not pass.

Doctor Anywhere

24/7 video consults, about S$20 to S$254.9 out of 5 from 26,969 Singapore App Store ratings4.9 out of 5 from 50,220 Google Play ratings, 1M+ installs

What its own terms say

In capitals, in its Terms and Conditions: do not use the video consultation functionality on the application for a medical emergency. The terms then list what video consultation is not appropriate for, including urgent care or emergency conditions, severe pain, breathlessness, undiagnosed chronic conditions, and anything requiring physical examination or laboratory tests.³⁹ The wording is precise: it excludes undiagnosed chronic conditions, so a diagnosed haemophilia patient is not automatically shut out. The emergency exclusion still applies, and no bleeding-disorder pathway exists.

What users say

The first is the single most on-point review found anywhere in this survey. A real user, at 4am, in pain, using the best-rated telehealth app in Singapore, with no visibility into when help would arrive. The 4.9 rating is real and so is that night.

Where it stops: a generalist on a ten-minute video call, with no factor level, no inhibitor status and no bleed history, facing a rare condition where the safe-looking answer can be the wrong one. The defensible thing for that doctor to do is send the family to the emergency department, which is the over-referral this project exists to reduce.

WhiteCoat

24/7 teleconsults with medication delivery4.7 out of 5 from 4,442 App Store ratings, 4.9 from 2,880 on Google PlayFirst telemedicine provider in MOH's regulatory sandbox

What users say, including the case against these files' own argument

The first review is included because it is the honest counter-example and these files are worse without it. A competent generalist recognised something outside their scope and escalated correctly, at a weekend, and a child kept his sight. That is the system working.

Where it stops: the escalation worked because the doctor could see the eye. Haemophilia's dangerous bleeds are the ones that cannot be seen: intracranial, retroperitoneal, iliopsoas, a compartment filling in a forearm. Recognition is the whole task, and it is the part a video call cannot do.

MaNaDr, and the regulatory record

Teleconsults with an AI assistant4.8 out of 5 from 5,736 App Store ratings, 4.8 from 5,062 on Google PlayLicence revocation notice, 24 October 2024

What the Ministry of Health found

On 16 August 2024 the Ministry directed MaNaDr Clinic to stop providing outpatient medical services by teleconsultation, having found more than 100,000 teleconsultations involving video calls of one minute or less, the shortest lasting one second; more than 1,500 patients issued medical certificates on five or more occasions in a single sampled month, one patient receiving 19; and case notes either implausibly detailed for the length of the call or extremely sparse.⁴² On 24 October it issued a notice of intended licence revocation, citing an entrenched culture of disregard for the applicable clinical and ethical standards, and referred 41 doctors to the Singapore Medical Council.⁴³

How far this can fairly be taken

The Ministry framed this as a failure of teleconsultation ethics and medical certificate issuance, not as an emergency mis-triage incident, and it should not be presented as a haemophilia case. What it does establish, on the regulator's own record, is that Singapore's generalist telehealth sector has a documented quality-control problem in exactly the format a rare-disease patient would meet at 2am: the brief, generic consultation.

Where it stops: the app is still rated 4.8 by its users. Consumer satisfaction and clinical adequacy are measuring different things, and for a bleeding disorder only one of them matters.

Speedoc

Video consults 7am to midnight, house calls and IV care advertised as 24/74.7 out of 5 from 3,944 Singapore App Store ratings4.5 out of 5 from 3,374 Google Play ratings

What it says it is for

The one provider in this category that names chronic-condition follow-up as in scope, alongside mild and common symptoms, urgent care and specialist referrals. Anything life-threatening is routed to 995 or the emergency department.⁴¹

What users say

Three reviews, one failure mode: the advertised hours and the delivered hours are different things, and the gap opens after dark. The first is the sharpest piece of evidence in this category, because it is a Singapore family in a real emergency being told by the product that they should have called 995. Fairness requires two caveats: these are 2022 reviews, which is as recent as Apple's public feed reaches for this app, and the 4.7 average is real.

Where it stops: the policy is correct on paper and this is what it looked like on one night. A haemophilia family cannot plan around a service whose overnight availability is the part users dispute, and Speedoc's own answer for an emergency is the ambulance number the family was trying to avoid needing.

The doctrine, and the evidence that it does not survive contact with 2am

WFH 3rd edition, MASAC 257, SingHealth, HSSAll free, all published, all correct

What exists

The World Federation of Hemophilia's third-edition guidelines set the system-level expectation of round-the-clock access to emergency medicines and clotting factor.⁴⁵ MASAC Document 257 codifies the rules a family needs: treat before diagnostic studies, treat on suspicion rather than documentation, assume a factor level of zero when it is unknown, and get factor into the patient within an hour of arrival.³⁰ SingHealth's own haemophilia page gives Singapore families the operative instruction, that during office hours they call the Children's Day Therapy Centre or the haematology resource nurse, that after hours they attend the Children's Emergency at KKH Basement 1 and notify the nurse the next working day, and that treatment should happen within one to two hours of onset. Its summary line is three words: if in doubt, treat.⁴⁶

What happens anyway

A 2025 audit in Pediatric Blood and Cancer measured the gap between doctrine and delivery: a median 2.9 hours from triage to ordering haemostatic therapy, 4.2 hours to administration, against MASAC's one-hour standard, and fewer than a third of emergency visits by haemophilia patients ending in haemostatic therapy at all.⁴⁸

The emergency card schemes

The field has already tried the obvious fix. The World Federation of Hemophilia issues an international identity card designed to be unobtrusive but available. The UK's National Haemophilia Database issues a bleeding disorder card to every registered patient, carrying condition details, an emergency contact and treatment advice, to be shown to emergency staff on arrival. The US programme is called FactorFirst, and its purpose is to shortcut the diagnostic workup entirely.⁴⁷

Where it stops: this is the most important finding in these files and the one that most sharpens what the build has to be. Haemophilia already has the right doctrine, and a physical delivery mechanism for it, and it still fails at the point of care. The unmet need is not another statement of the rule. It is getting a specific family to the right decision at 2am, and giving them something the receiving clinician recognises fast enough to act on. A 300-page guideline is the right content at the wrong moment, unreadable in the state of brain switch-off a family hits at 2am.

What families reach for at 2am

From the team's own HSS interviews and the 30 Aug 2026 patient journey summary⁵⁵

Spreadsheets

Work well for routine prophylaxis, and break on travel, on a missed dose, and whenever the recorded stock drifts from what is actually in the fridge.

Group chats and forums

Disorganised, unsearchable and unvalidated. Patients themselves say they disregard them, which is the significant part: the informal layer is not trusted even by the people inside it.

Paper discharge notes and saved screenshots

The reference layer families scour once the clinic lines are closed. Undated, uncontrolled, and specific to whichever visit produced them.

Memory and trial and error

What is left when the paper and the screenshots run out. The team's interviews found that off-hours guidance seeking largely does not happen at all: families repeat the usual dose and monitor on self-invented windows, an hour for a hip bleed, up to a day elsewhere, before deciding anything more.

Where it stops: this is the real incumbent, and the thing any new tool has to be better than on the first night. None of it is signed, current or personal, and none of it leaves a record anyone can hand to a doctor.

The ASCIA anaphylaxis action plan

Australasian Society of Clinical Immunology and AllergyOne page, colour-coded, device-specificRedesigned 2023, first major change in 20 years

Given to every patient prescribed an adrenaline autoinjector, in a general version plus four device-specific versions. The 2023 revision added colour coding by device and QR codes linking to short injection videos on the plan itself.⁴⁹ The structure is what matters: recognise a defined symptom set, use the device now, call emergency services. There is no differential diagnosis step anywhere in it.

The lesson: at the moment of crisis, a pre-agreed instruction beats an assessment. This is the benchmark row in the scorecard below, and it is the only artefact in these files that passes all four tests.

Asthma action plans, green, yellow and red

The best-evidenced action plan format in medicineCochrane review of 18 randomised trials

A written, individualised traffic-light plan: green for baseline control, yellow for loss of control with a defined step-up, red for urgent action. The Cochrane review of 18 randomised trials found that a written plan combined with education and regular review significantly reduces hospitalisations, emergency visits, unscheduled doctor visits, days lost from work and school, and nocturnal symptoms.⁵⁰ The yellow zone is specified to the dose: trials show a four to five-fold increase in inhaled corticosteroid for 7 to 14 days works, and simply doubling the dose does not.

The lesson, and the warning: specificity is the active ingredient, and a good artefact does not implement itself. The same literature describes written asthma action plans as a niche practice and an example of ineffective guideline implementation despite the evidence behind them.⁵⁰ Being correct is necessary and nowhere near sufficient.

Seizure action plans, and a methodology worth copying

American Academy of Pediatrics, Seizure Action Plan CoalitionIndividualised plan plus rescue medication

An individualised plan capturing seizure type and history with step-by-step instructions, paired with a rescue benzodiazepine to abort a prolonged seizure before it becomes status epilepticus. An expert panel published formal recommendations to standardise the format across settings.⁵¹

The lesson: the epilepsy field published a paper explicitly distilling insights from other chronic conditions' emergency action plans to design its own. That is the same exercise these files are performing for haemophilia, and it is worth reading before the format is fixed.