Medical practice is shaped by probability. Clinicians are trained to prioritise the most common explanations for symptoms, an approach that supports efficiency and safety in everyday care. For people living with rare diseases, however, this logic can contribute to prolonged diagnostic pathways. Symptoms may initially resemble frequent conditions and are often managed in isolation, across multiple encounters and specialties. Over time, relevant clinical information accumulates, but it may remain fragmented across different systems and organisations. The difficulty lies less in recognising individual signs than in assembling them into a coherent picture. Against this background, artificial intelligence has been proposed as a means of supporting the integration of longitudinal clinical data to help identify rare disease patterns that may otherwise remain unrecognised for many years.
Rare Diseases with Common Presentations
Several rare conditions are characterised by symptoms that overlap with more prevalent diagnoses, contributing to extended diagnostic delays. Acute intermittent porphyria (AIP) is associated with episodes of pain, fatigue and weakness that may initially be attributed to conditions such as fibromyalgia, chronic fatigue or anxiety. The average time from first symptoms to diagnosis is reported as 10–15 years. The underlying cause is a rare disorder of heme metabolism, with attacks occurring when balance is disrupted by factors such as certain medications, stress or hormonal changes. Genetic or biochemical testing may confirm the diagnosis, often long after symptoms first appear.
Fabry disease presents a similar challenge. Patients may spend more than a decade under the care of different specialists, including rheumatologists, neurologists and cardiologists, because symptoms can resemble nerve pain, autoimmune conditions or multiple sclerosis. Reported average diagnostic delays are approximately 14 years in men and 16 years in women. The diagnosis becomes clearer when features such as skin changes, kidney involvement and subtle cardiac thickening are considered together, but these elements may be documented separately and assessed independently.
Must Read:Human–AI Collaboration Lifts Diagnostic Decisions
Transthyretin amyloidosis (ATTR) provides another example of delayed recognition. ATTR can affect the heart over a prolonged period, with early signs that appear unrelated to later cardiac involvement. Carpal tunnel syndrome is described as a frequent early manifestation and is commonly treated as an isolated orthopaedic condition. Research cited indicates an average delay of 6–8 years before ATTR is correctly identified following the onset of initial symptoms. In each case, the clinical indicators are present, but they emerge across time and specialties rather than within a single episode of care.
Structural Barriers to Early Recognition
These prolonged diagnostic journeys are reflecting structural characteristics of healthcare delivery rather than individual clinical oversight. Care is often organised around discrete encounters, with each consultation capturing a limited snapshot of the patient’s condition. Laboratory results, imaging studies and clinical notes may be generated years apart and are rarely reviewed as part of a single, continuous narrative.
Data fragmentation further complicates this process. Laboratory information, imaging reports and genetic results may be stored in separate systems, sometimes across different institutions. As patients move between providers and healthcare organisations, continuity of information can be lost. Without straightforward access to a comprehensive historical record, recognising patterns that evolve slowly over time becomes challenging.
The tendency to prioritise common conditions also plays a role. Because rare diseases are uncommon by definition, the likelihood of considering them at the outset of a clinical assessment is low. This probability-based reasoning is repeatedly applied when each new clinician begins evaluation with limited access to prior information. As a result, the same symptoms may be reinterpreted multiple times without triggering reconsideration of less frequent diagnoses.
Building Longitudinal Views With AI
AI is a tool that could support the integration of fragmented clinical information into a longitudinal, multi-modal patient record. Such a record would bring together data generated across years, including laboratory tests, imaging reports, pathology results, clinical notes and data from wearable devices. By aggregating information into a connected timeline, AI systems may help surface patterns that are difficult to detect when data are reviewed in isolation.
Within this framework, certain combinations of findings could prompt consideration of rare diagnoses earlier in the care pathway. A history of recurrent abdominal pain, dark urine and episodes triggered by stress or medication could align with known patterns associated with AIP when assessed collectively. Similarly, documentation of bilateral carpal tunnel syndrome followed years later by changes in cardiac markers could support reassessment for conditions such as ATTR. The emphasis is on visibility of longitudinal patterns rather than automated diagnosis.
Several functional elements are considered parts of this approach. One involves connecting disparate data sources to form a unified patient timeline. Another focuses on defining characteristic combinations of features associated with specific rare diseases, sometimes described as digital fingerprints. These patterns may be derived from established medical knowledge or identified through analysis of historical patient data. Importantly, clinical decision-making remains the responsibility of healthcare professionals, with AI providing support by highlighting cases that may merit further review.
Rare diseases are often difficult to identify not because evidence is absent, but because it is dispersed across time, settings and information systems. Symptoms that appear routine in isolation can form recognisable patterns when viewed longitudinally. Approaches that support continuity of data and enable integrated review may help address long-standing diagnostic delays. AI-enabled aggregation and pattern recognition can be a way to assist clinicians in identifying patients who may require reassessment for rare conditions. The relevance for healthcare decision-makers lies in improving access to comprehensive patient histories and supporting earlier, more informed diagnostic consideration, while maintaining clinical judgement at the centre of care.
Source: HIT Consultant
Image Credit: iStock
