Amyloidosis behavioral signals: late diagnosis and the data gap
The median delay in cardiac amyloidosis diagnosis is 22 months, and in some subtypes it stretches past 34 months. Behavioral data from patient search patterns reveals a pre-diagnostic window filled with fragmented symptom queries, specialist-hopping signals, and misdiagnosis frustration. This gap is not just a clinical failure. It is a data infrastructure failure.
Cardiac amyloidosis has a median diagnostic delay of 22 months. For certain subtypes, that number climbs to 34 months. During that window, patients cycle through an average of five or more physicians, accumulate misdiagnoses of heart failure or hypertensive cardiomyopathy, and generate a trail of behavioral signals that the healthcare system currently ignores.
The existing literature focuses on clinical predictors of delay. Seven factors, including age, race, and presentation pattern, have been shown to predict late diagnosis. But what the clinical literature misses is the behavioral layer: what patients and caregivers search, when they search it, and how those patterns map to the pre-diagnostic window that precedes a confirmed amyloidosis diagnosis.
This post examines the amyloidosis data gap from the behavioral intelligence side. Not the clinical markers that should trigger referral, but the digital exhaust that proves patients were looking for answers long before anyone ordered the right test.
What is the delay in diagnosis of amyloidosis?
Published research from multiple registries puts the median diagnostic delay for cardiac amyloidosis at 22 months from first symptom presentation to confirmed diagnosis. That number masks significant variation. Patients with AL amyloidosis (light chain amyloidosis, driven by a plasma cell disorder) tend to receive diagnosis somewhat faster because hematologic abnormalities surface in routine blood work. Patients with ATTR amyloidosis (transthyretin amyloidosis, either hereditary or wild-type) often wait longer because their symptoms overlap almost entirely with common cardiac conditions.
A study of 1,432 patients found that seven factors predicted delayed diagnosis, including presentation with heart failure symptoms alone, absence of neuropathy, and care delivered outside academic medical centers. For patients who presented with isolated cardiac symptoms and no neuropathic complaints, the median delay stretched to 34 months.
These numbers describe the clinical timeline. They do not capture the behavioral timeline. Patients with unexplained heart failure who search "why does my heart failure treatment not work" or "heart failure getting worse despite medication" are generating signals months or years before the amyloidosis label appears in their chart. That behavioral window is where the data gap lives.
What is the 5-5-5 rule for amyloidosis?
The 5-5-5 rule is a clinical heuristic used to identify patients who should be evaluated for cardiac amyloidosis. It states that clinicians should suspect amyloidosis when a patient has:
The rule exists because amyloidosis mimics more common cardiac conditions. It is not a diagnostic test. It is a screening trigger designed to prompt nuclear scintigraphy (for ATTR) or tissue biopsy. The problem is that the 5-5-5 rule requires a clinician who is already thinking about amyloidosis. In community cardiology settings, where most heart failure is managed, awareness of amyloidosis as a differential diagnosis remains low.
Behavioral data shows this awareness gap from the patient side. Search queries for "thickened heart wall cause" and "heart failure no improvement" spike in populations that epidemiologically overlap with ATTR prevalence: men over 65, African American patients with the V122I variant, and patients with bilateral carpal tunnel syndrome preceding cardiac symptoms. Those searches happen in the same 22-to-34-month window the clinical literature identifies as the delay period.
What famous person has amyloidosis?
Several public figures have brought attention to amyloidosis. Former NFL player Vin Scully's wife, Sandra Scully, died from complications related to amyloidosis. More recently, the condition gained visibility through media coverage of athletes and public figures diagnosed with ATTR cardiomyopathy.
The most significant awareness driver has been the growing recognition that wild-type ATTR amyloidosis (ATTRwt) may affect up to 13% of elderly patients hospitalized with heart failure with preserved ejection fraction. This is not a rare curiosity. It is a systematically underdiagnosed condition hiding inside the most common cardiac diagnosis in older adults.
Celebrity diagnoses tend to generate brief search spikes. Behavioral data consistently shows that disease-name searches surge for 7 to 14 days following a public disclosure, then return to baseline. The structural search patterns, the ones that map to actual patients in the diagnostic pipeline, are the persistent low-volume queries: "carpal tunnel and heart failure connection," "why is my heart wall thick," "amyloid protein heart." These are the signals that matter for identifying undiagnosed populations.
How fast does ATTR amyloidosis progress?
ATTR amyloidosis progression varies significantly by subtype. Wild-type ATTR (ATTRwt), formerly called senile cardiac amyloidosis, tends to progress more slowly. Median survival from diagnosis is approximately 3.5 to 5 years, though some patients live longer with treatment.
Hereditary ATTR (ATTRv) varies by mutation. The V30M mutation, common in Portuguese, Swedish, and Japanese populations, can progress over 7 to 15 years when onset occurs before age 50. The V122I mutation, carried by approximately 3.4% of African Americans, tends to present later in life with cardiac-dominant disease and a median survival of 2.5 to 3.5 years from diagnosis.
The key issue with progression data is that it is measured from diagnosis, not from disease onset. If the median diagnostic delay is 22 months, and median survival from diagnosis is 3.5 years, a significant fraction of the patient's treatable window has already passed before treatment begins. Tafamidis, the first FDA-approved therapy for ATTR cardiomyopathy, showed the greatest benefit in patients with NYHA Class I and II heart failure. By the time most patients are diagnosed, they have progressed to Class III or beyond.
This is why earlier identification matters. And this is why the behavioral data layer matters. The clinical system is not catching these patients early enough. The question is whether the data infrastructure can.
The behavioral signal map in amyloidosis
Patients with undiagnosed cardiac amyloidosis generate a distinctive search pattern that differs from typical heart failure search behavior. The pattern has three phases.
Phase one is symptom normalization. Patients search for common explanations: "shortness of breath causes," "swollen ankles heart," "fatigue older adults." These searches are indistinguishable from general cardiac symptom queries. They generate no signal.
Phase two is treatment failure frustration. This is where the amyloidosis behavioral signature begins to separate. Patients or caregivers search for "heart failure medication not working," "why is my heart failure getting worse," "heart failure progression despite treatment." This phase often coincides with the clinical reality of refractory heart failure, where standard therapies (ACE inhibitors, beta-blockers, diuretics) fail to produce the expected improvement because the underlying pathology is infiltrative, not ischemic or dilated.
Phase three is differential diagnosis seeking. Patients search for "rare causes of heart failure," "cardiac amyloidosis symptoms," "amyloid heart disease," or, critically, "carpal tunnel syndrome and heart failure." Bilateral carpal tunnel syndrome precedes cardiac ATTR diagnosis by a median of 5 to 9 years. Patients who search for the connection between carpal tunnel and cardiac symptoms are, in many cases, patients whose clinicians have not yet made the connection.
Key statistics
The data gap is structural, not incidental
The amyloidosis diagnostic delay is not primarily a knowledge gap among cardiologists at academic centers. Most tertiary centers now include amyloidosis in the differential for unexplained left ventricular hypertrophy. The gap is structural. It exists because the data systems that should connect a patient's carpal tunnel surgery in 2018 to their heart failure hospitalization in 2024 do not talk to each other.
A patient's orthopedic surgeon documents bilateral carpal tunnel release. Their primary care physician manages their blood pressure. A community cardiologist diagnoses heart failure with preserved ejection fraction. An emergency department treats them for fluid overload. None of these records are scored, linked, or evaluated as a longitudinal signal. Each encounter lives in its own EHR silo.
This is a provenance and concordance problem. The data exists. It is scattered across systems that have no trust layer, no concordance scoring, and no mechanism to flag when a pattern of records, spanning multiple providers and years, matches an amyloidosis risk profile.
Recent work in biomarker-based early detection, including the use of plasma biomarkers for neurodegenerative diseases as covered by MedPAGE Today, demonstrates that the field is moving toward earlier identification across multiple disease categories. Cardiac amyloidosis is ripe for the same approach, but the data infrastructure has to be there first.
Why current AI models miss amyloidosis
AI models trained on heart failure cohorts will not find amyloidosis patients unless the training data includes scored, longitudinal records that span the full pre-diagnostic window. Most heart failure prediction models are trained on single-institution datasets. They predict 30-day readmission or mortality. They do not predict etiology.
A model trained on EHR data that lacks concordance scoring will treat a carpal tunnel diagnosis and a heart failure diagnosis as unrelated events. A model trained on data without recency weighting will not distinguish between a patient whose symptoms started 6 months ago and one whose symptoms started 5 years ago. A model trained on data without provenance tracking cannot tell the difference between a confirmed echocardiographic finding and a billing code entered for reimbursement purposes.
This is the core problem with deploying AI for rare disease identification without a trust layer. The signal is there. The data is there. But the data has never been scored for the dimensions that matter: provenance, recency, concordance, and breadth.
The V122I equity gap
The V122I ATTR variant deserves specific attention. Carried by approximately 3.4% of African Americans, this mutation causes cardiac amyloidosis that is systematically underdiagnosed. Studies have shown that African American patients with heart failure and the V122I variant wait longer for diagnosis, are less likely to be referred for nuclear scintigraphy or genetic testing, and are less likely to receive tafamidis.
Behavioral data reflects this disparity. Search volume for "amyloidosis" in geographic regions with high African American populations is disproportionately low compared to the expected prevalence. This is not because patients are not symptomatic. It is because the clinical system has not introduced amyloidosis as a possibility, so patients do not know to search for it.
The behavioral gap here is the absence of signal. Patients who should be searching for amyloidosis are instead searching for "heart failure getting worse" or "why does heart failure treatment not work" because no clinician has raised amyloidosis as a differential. Mapping these negative behavioral signals, the searches that should be happening but are not, requires population-level behavioral intelligence layered on top of geographic and demographic data.
What behavioral intelligence can actually do here
Behavioral intelligence does not diagnose amyloidosis. It identifies populations and individuals who are exhibiting the pre-diagnostic search patterns that precede a confirmed diagnosis. When those patterns are mapped at scale, they can:
This is not speculative. The same approach has been validated across other conditions with long diagnostic delays, including endometriosis (7-year average delay), fibromyalgia, and neuroendocrine tumors. The pattern is consistent: patients leave a behavioral trail that the clinical system does not read.
The connection between data trust and diagnostic delay
Every month of diagnostic delay in cardiac amyloidosis represents disease progression that narrows the treatment window. Tafamidis works best in earlier-stage disease. The difference between NYHA Class II and Class III at treatment initiation is not academic. It is measured in survival.
Closing the diagnostic delay requires two things the current system lacks. First, longitudinal data concordance: the ability to link a carpal tunnel surgery, a spinal stenosis diagnosis, and a heart failure admission across providers and years. Second, data trust scoring that ensures the records being evaluated are current, validated, and properly attributed.
Without these foundations, no AI model, no screening algorithm, and no clinical decision support tool will reliably identify amyloidosis patients earlier. The data gap is not about the absence of data. It is about the absence of scored, trusted, linked data.
VIOLET maps behavioral signals across 750+ oncology and disease-specific search terms before patients reach a clinic. For conditions like cardiac amyloidosis, where the pre-diagnostic behavioral window spans years and the data infrastructure has not caught up to the clinical need, behavioral intelligence is the layer that connects population-level search patterns to actionable identification. If your team is working on amyloidosis cohort identification, trial recruitment for ATTR therapies, or market intelligence for underdiagnosed cardiac populations, schedule a conversation with the SuperTruth commercial team or (215) 918-4140.
Further reading:
Jason Alan Snyder
Co-founder of SuperTruth and Artists & Robots, and an inventor on the Data Trust Index patents. Twenty-plus years building technology inside Interpublic Group. He writes here nearly every day on data trust, provenance, and what AI should be allowed to act on, and publishes essays on his Substack.
About SuperTruth · LinkedIn · Substack · jasonalansnyder.com
See it in practice
DTI scores the record, not the patient.
8 dimensions. 0 to 100. Travels with every record permanently.