A practical guide for pharma and patient support programmes, registries and patient organisations, and clinics and laboratories on the data decisions that decide whether a rare disease programme can help families and generate evidence that holds up.
Each rare disease affects relatively few people, but together they are anything but rare. More than 6,000 rare diseases have been described, and between them they affect an estimated 3.5% to 5.9% of the world's population, somewhere between 263 and 446 million people.1 Around 72% are genetic, and around 70% of those with a known age of onset begin exclusively in childhood.1
For the people running programmes in this space, the data challenge looks different from most other areas of healthcare. There are too few patients in any one place to rely on volume, so every record matters. The story of each patient is usually spread across years, specialists, hospitals and sometimes countries. And the evidence that regulators, payers and clinicians need often has to be built from records that were never designed to be combined.
We wrote this guide for three groups of readers. If you lead a pharma-supported or patient support programme, it sets out what your data needs in order to support patients well and to show outcomes. If you run a registry or a patient organisation, it covers what it takes to build a dataset that can be trusted, shared and reused. If you work in a clinic or laboratory, it explains how to keep a patient's diagnostic and genetic record coherent as it grows. The twelve points apply to all three, and where the emphasis differs, we say so.
In a survey of more than 6,500 people living with a rare disease across 41 countries, the average time from first symptoms to a confirmed diagnosis was 4.7 years.10 In Australia, 30% of adults living with a rare disease have waited more than five years for a diagnosis, and almost half have received at least one misdiagnosis.11
Every step of that journey leaves a trace: a referral letter, a set of blood results, an imaging report, a clinic note from a specialist in another city or country, a genetic test report. By the time a diagnosis arrives, the most useful record of how the condition began is often held by the family, in a folder at home or across several email inboxes.
An illustrative journey. Each step holds part of the story in a different system, format or language, and by the time of diagnosis the family is often the only one holding all of it.
The twelve points below are grouped into four themes: getting the clinical picture right, making the data usable, connecting the longer story, and building for trust and scale.
There is no single global definition of a rare disease. Most definitions come from orphan medicine regulation or from funding rules, and they measure different things: some use a share of the population, others an absolute number of patients in one country.
This matters as soon as a programme works across more than one market. A condition may qualify as rare in one country and not in another, prevalence estimates stop being comparable, and eligibility rules for funding or early access can differ for the same patient depending on where they live.
What good looks like: the definition and jurisdiction behind every eligibility decision, cohort and prevalence figure recorded explicitly, so that data combined across markets stays comparable.
Matters most to: Pharma programmes, Registries and patient groups
In rare disease, the pattern of clinical features (the phenotype) is often what leads to a diagnosis and what defines who belongs in a cohort. When those features are recorded as free text, one clinician's "low muscle tone" becomes another's "floppy infant" and a third's "hypotonia", and the patients who share a pattern become very hard to find.
The Human Phenotype Ontology (HPO) was developed for exactly this purpose. It is a standardised vocabulary of phenotypic abnormalities, now with more than 17,000 terms and annotations for more than 8,100 rare diseases, and it is widely used in genetic diagnosis and research.7 Recording features against a vocabulary like this makes it possible to compare patients, support genetic analysis and match cases across sites.
What good looks like: clinical features coded to a standard vocabulary, with the age at which each was observed, and important absent features recorded as well as present ones.
Matters most to: Clinics and laboratories, Registries and patient groups
Many rare diseases have no specific code of their own in the classification systems most hospitals use. An analysis of the Orphanet nomenclature found that only about 6% of rare disorders map exactly to a code in the International Classification of Diseases, 10th revision (ICD-10), rising to about 15% in the 11th revision (ICD-11).8 The rest are grouped under broader codes, which means patients with a specific rare condition can be invisible in hospital data.
ORPHAcodes, the codes of the Orphanet rare disease nomenclature, were created to close this gap. In 2025 the World Health Assembly encouraged countries to consider interoperable coding systems for rare diseases, such as the Orphanet nomenclature, alongside ICD-11.9
What good looks like: a rare disease specific code carried alongside the local or national codes a hospital already uses, with local code sets mapped rather than replaced.
Matters most to: Registries and patient groups, Pharma programmes, Clinics and laboratories
Most systems record the diagnosis a patient has today. For rare disease programmes, the journey that led there carries just as much value: when symptoms first appeared, which specialists were seen, which tests were done and when, and which diagnoses were considered and later ruled out. This is the information that shows where diagnostic delay happens, supports earlier identification of future patients and gives context to every later result.
Rebuilding that journey usually means working from referral letters, discharge summaries, laboratory results, imaging reports and records from other countries, many of them scanned, handwritten or in another language. Earlier diagnoses should not simply be overwritten when a new one arrives, because the misdiagnosis is itself part of the evidence.
What good looks like: a dated, chronological record assembled from every source document, with each entry linked back to its original, and superseded diagnoses kept and clearly marked as revised.
Matters most to: Pharma programmes, Registries and patient groups, Clinics and laboratories
Genetic results are central to rare disease care, yet they usually arrive as PDF reports written for a clinician to read. The details that matter for a programme or registry sit inside the narrative: the gene, the variant described in standard nomenclature, whether one or both copies are affected, the classification the laboratory gave, the type of test (a gene panel, an exome or a genome), the laboratory and the date.
When those details stay locked in the PDF, it becomes very hard to find every patient with a particular variant, to check whether a family member has been tested, or to see which patients might benefit from a new trial or treatment. Laboratories also format these reports differently, which is why extraction on its own is not enough. As we explain in Reading a Lab Report with AI Is Not the Same as Making It Usable, the findings need to be structured, checked and linked to the right person and date.
What good looks like: each genetic result stored as structured fields (gene, variant, zygosity, classification, test type, laboratory and date), with clinical review, and the original report kept alongside.
Matters most to: Clinics and laboratories, Pharma programmes
A genetic result is not always final. Laboratories classify variants using the standards published by the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP), which set out five categories: pathogenic, likely pathogenic, uncertain significance, likely benign and benign.12 As evidence builds, classifications change. In a study of more than three million people who had hereditary disease genetic testing over eight years, 4.6% of variants were reclassified, and most reclassified variants of uncertain significance moved towards benign.13
Among reclassified variants of uncertain significance in a study of more than three million tested people, 78.8% moved to likely benign or benign.13
An ACMG statement on re-evaluating genomic results recommends that laboratories have clear policies for reanalysis and respond promptly to requests, and it notes that who is responsible for starting a re-evaluation is not yet settled.14 For a programme, the practical question is simpler: if a classification changes, can you find every patient it affects?
What good looks like: every classification stored with its date, the laboratory that made it and the standard used, so updates can be recorded without losing history and affected patients can be identified quickly.
Matters most to: Clinics and laboratories, Registries and patient groups, Pharma programmes
Because most rare diseases are genetic,1 one diagnosis often has consequences for parents, siblings and children. Family members may be offered testing, carriers may be identified, and a second affected child may be diagnosed far sooner than the first. Records that treat each person in isolation lose this picture, while records that link people too casually create real privacy risks for relatives who never agreed to take part.
What good looks like: family relationships recorded explicitly and only where each person has consented, with each relative's data, consent and access rights managed separately.
Matters most to: Registries and patient groups, Clinics and laboratories
Understanding how a rare disease progresses without treatment is essential for designing trials, choosing endpoints and showing that a treatment makes a difference. The United States Food and Drug Administration (FDA) notes in its draft guidance on natural history studies that this information is usually not available, or is incomplete, for most rare diseases.15
Because most rare diseases begin in childhood, natural history also has to survive the move from paediatric to adult care, which is where records are often lost. In a European survey of centres caring for people with rare metabolic diseases, about two thirds had no transition coordinator.16
What good looks like: the same measures and scales used consistently over time, the patient's age recorded at every observation, and a record designed to carry on through the transition from children's to adult services.
Matters most to: Pharma programmes, Registries and patient groups
Patients with rare diseases may receive treatment through routine prescribing, clinical trials, early access programmes, imported medicines or dedicated funding schemes such as Singapore's Rare Disease Fund.6 Doses are often adjusted by weight as children grow, infusions may be given at different sites, and treatment can pause when access or funding changes.
How a patient reached treatment matters for evidence as much as what they received. Gaps caused by funding or supply look very different from stopping because of side effects or lack of benefit, and a programme that cannot tell them apart will struggle to show outcomes.
What good looks like: each treatment recorded with product, dose, route, site, start and stop dates and the access route, with reasons for pauses and stops captured in consistent categories.
Matters most to: Pharma programmes, Clinics and laboratories
Removing names and identity numbers is not enough in rare disease. When only a handful of people in a country share a diagnosis, a combination of age, location and condition can be enough to identify someone. The European Medicines Agency (EMA) states that clinical trials in rare diseases and small populations carry a higher risk of re-identification.17
This does not mean rare disease data cannot be shared. It means privacy protections need to be designed for small numbers from the start, as we set out in our guide to de-identifying health data on the Jonda Health resources page.
What good looks like: data pseudonymised for any use beyond direct care, a re-identification risk assessment before each release, small counts suppressed or grouped, and controlled access with audit trails.
Matters most to: Registries and patient groups, Pharma programmes, Clinics and laboratories
With so few patients in any one place, rare disease evidence nearly always depends on combining data across sites, programmes and countries. The FAIR principles (data that is Findable, Accessible, Interoperable and Reusable) were set out for this kind of reuse,18 and European rare disease initiatives now publish practical guidance on applying them to patient registries.19 The EMA has also published a guideline on the methodological, regulatory and operational aspects of using registry-based studies to support regulatory decisions.20
Designing for this early is far cheaper than retrofitting it. A programme that captures data in its own format, without a data dictionary or consent for reuse, may find its most valuable dataset cannot be used for the research it was meant to support.
What good looks like: a documented data dictionary, standard codes for diagnoses, features and results, data structured to interoperability standards such as Fast Healthcare Interoperability Resources (FHIR), clear provenance, and consent that covers the reuse you intend.
Matters most to: Registries and patient groups, Pharma programmes
In rare disease, patients and families are often the most complete source of their own history, and patient organisations are frequently the ones who start registries in the first place. They know when symptoms began, which treatments were tried and what daily life with the condition is really like. Programmes that treat families only as a source of consent miss much of what they can contribute.
What good looks like: families able to contribute documents and patient-reported outcomes, see what has been recorded about them, correct errors and take a copy of their record with them when they move between services or countries.
Matters most to: Registries and patient groups, Clinics and laboratories, Pharma programmes
You do not need to solve all twelve at once. The right starting point depends on what your programme needs its data to do.
Teams tend to get the most from these twelve points when they use them as a working tool rather than a reading list.
If you would like support with this, the Health Data Foundations work within Jonda Health Services takes a programme through exactly these questions, from what data you hold and where it sits to what it needs before it can support analytics, AI or evidence generation. Where the goal is to show outcomes, our insight and analytics work helps turn a harmonised dataset into measures and benchmarks your stakeholders can use.
Much of what sits between a well-intentioned rare disease programme and a dataset that can support patients and evidence is data work. It means gathering years of letters, results, imaging reports and genetic reports, making them usable across formats and languages, cleaning them, resolving the many ways the same thing gets captured, reviewing them clinically and mapping them to the standards and code sets your programme uses. This is the work we do every day.
JondaX, our health data transformation engine, turns laboratory reports, clinical documents, medical device readings and wearable data into structured, harmonised data through a single API, with other clinical data types, such as genetic reports, configured to your needs. Every dataset goes through human-in-the-loop review by clinically trained reviewers, so that the burden of checking does not fall on your team. JondaX can be deployed and processed within your country where required, supports pseudonymisation of data for secondary use, and works with your own ontology and code sets. Jonda Health Services supports teams earlier in the journey, from programme and registry design to deciding what to build yourselves and what to bring in. Jonda Health is ISO 27001 certified, and our platform is designed to comply with HIPAA, GDPR and PDPA.
This guide is for teams designing and running rare disease programmes, registries and services. It is not clinical advice for individuals.
