Five years in, the folder of records is two inches thick. It holds notes from nine specialists, three MRI reports, a genetic test and pages of blood results, nearly all normal. Most visits end the same way: nothing serious, come back if it gets worse. It has gotten worse.
Medicine calls this the diagnostic odyssey, and the federal government built a program to end it. That program’s results offer some of the clearest numbers on how often a determined search succeeds. Between 2015 and 2017, the National Institutes of Health’s Undiagnosed Diseases Network received 1,519 applications and accepted 601. Of the 382 people who completed an evaluation, 132 got a diagnosis. That is 35% of those evaluated, and fewer than 9 in 100 of everyone who applied.
Those numbers cut against two opposite beliefs. One holds that the next test or the next specialist will surely produce a name. The other holds that no name means nothing is wrong. Here is what the evidence says about both, and about the clinics that move into the space between them.
Diagnosis goes wrong more often than people assume
In a 2015 report, the National Academies of Sciences, Engineering, and Medicine defined diagnostic error in two parts: failing to arrive at an accurate and timely explanation of a patient’s health problem, or failing to communicate that explanation to the patient. The committee concluded that most people will experience at least one diagnostic error in their lifetime, sometimes with grave consequences.
A widely cited US estimate comes from a 2014 study in BMJ Quality & Safety that combined three earlier US studies. It put the rate at 5.08% of adults who use outpatient care each year, roughly 12 million people, and judged that about half of those errors could cause harm. The National Academies report used that roughly 1-in-20 figure, and cited research suggesting diagnostic errors contribute to about 10% of patient deaths.
The report’s first recommendation includes a point few patients hear: health professionals and organizations should treat patients and their families as members of the diagnostic team. What that partnership sounds like in an exam room is the subject of the guide to doctors who don’t seem to be listening, and the opposite hazard, testing that does more harm than good, gets separate treatment in our overtesting article.
Rare diseases are rare one at a time
Many people on a long diagnostic search turn out to have a rare disease. Under the Rare Diseases Act of 2002, a disease counts as rare in the US if it affects fewer than 200,000 people. NIH’s Genetic and Rare Diseases Information Center (GARD) says experts estimate more than 10,000 rare diseases, and an NIH Common Fund page estimates that about 25 million Americans have one. Each is rare. Together they are common.
How long do they take to name? A large peer-reviewed survey comes from Europe. EURORDIS, a rare-disease patient alliance, surveyed 6,507 people living with 1,675 rare diseases in 41 countries and published the results in the European Journal of Human Genetics in 2024. Diagnosis took an average of 4.7 years from the first symptoms, and 56% were diagnosed more than six months after first seeing a health professional. We know of no comparable peer-reviewed national figure for the US; the numbers that circulate here come from industry-sponsored surveys.
The survey’s links to longer delays are telling. Symptoms that began in childhood or adolescence, being a woman, a previous misdiagnosis and a large number of professionals consulted all went with longer waits. A previous misdiagnosis, for instance, came with about two and a half times the odds of a longer delay. The professionals finding is easy to misread. Seeing many clinicians more likely marks a hard case; the survey can’t show that seeing more of them causes the delay.
Inside the undiagnosed diseases network
The federal effort started small. In 2008, the NIH Clinical Center opened the Undiagnosed Diseases Program. In its first two years it reviewed 1,191 patients’ records, admitted 160 for a full evaluation, and diagnosed 24% of them.
NIH’s Common Fund launched a national version, the Undiagnosed Diseases Network (UDN), in 2013. Its first seven clinical sites were in place in 2014, and a public application portal opened in September 2015. Common Fund support ended in July 2023. The network is now led by the National Institute of Neurological Disorders and Stroke (NINDS), and in August 2024 NIH funded 12 sites as Diagnostic Centers of Excellence, bringing the network to more than 20 clinical sites. NINDS says the network helped diagnose more than 650 people in the decade before that announcement.
Who gets in
The network can’t take everyone, and its rules show what a strong case looks like. According to the 2018 report on its first 1,519 applicants, published in The New England Journal of Medicine, acceptance required:
- a condition still unexplained after a thorough evaluation by a health care provider;
- at least one objective finding, something a clinician can see or measure, not a symptom alone;
- consent to store and share medical data and samples;
- the ability to travel to a site or take part by telemedicine, and willingness to undergo more testing.
Priority went to cases most likely to be solved or to teach researchers something new. Adults fared worse than children: 28% of adult applicants were accepted, against 57% of children. Most applicants had neurologic symptoms, and most were white and lived near a site, limits the authors acknowledged.
Applying takes paperwork. The network’s application page asks for a signed letter on letterhead from a health care provider (a relative can’t write it) covering symptoms, when they began, earlier diagnoses, tests and treatments. It also asks for available records and a one-page account in the patient’s own words. Once the records reach a clinical site, a decision typically takes six to eight weeks.
What the search found
Of the 132 diagnoses, 74% came from exome or genome sequencing, tests that read all of a person’s genes or their entire genetic code. Another 11% came from expert review of the existing records alone. The network also described 31 syndromes no one had defined before. Later data tell a similar story: a 2025 analysis of the national network reported diagnosis rates of 40% for participants with an eye-related finding and 28% for those without one.
A name rarely meant a cure. Of those diagnosed, 21% received a recommendation to change treatment, 37% saw other changes in care, such as narrower testing, and 36% got genetic counseling without any change in tests or treatment. Among the 28 patients with treatment recommendations, 8 showed a benefit and 6 an unclear or negative effect; the rest had not started treatment or had no recorded outcome.
Money is part of the story. In a small sample of 14 patients whose care stayed within one health system, medical care before acceptance averaged $198,651, while the network’s evaluation averaged $15,116. Fourteen people can’t settle a cost question.
Who pays has changed. The 2018 report said: “Diagnostic evaluation is provided at no cost to patients.” Under the funding rules NIH issued in 2023 for the network’s current phase, clinical sites are expected to bill insurance for clinical services, routine tests and sequencing when allowed, and to direct NIH money first to participants who are uninsured or underinsured. Anyone applying now can ask a site what they would owe.
A negative genetic test is not the last word
One of the network’s most useful findings concerns people told their genetic testing was normal. About a third of accepted patients had already had exome sequencing. Among 147 patients sequenced again after earlier testing, 48, or 33%, received a diagnosis, close to the 36% rate among patients who had never had exome sequencing. Eleven of those 48 were diagnosed by reanalyzing data they already had, and some diagnoses in the group relied on case reports published after the original evaluation.
A 2026 report from the network’s Duke University site, published in Genetics in Medicine, pushes the point further. Of 192 participants whose earlier sequencing had found nothing, 77, or 40%, eventually received diagnoses. Research reanalysis of raw sequencing data, old and new, produced more than half of them. The authors judged that 72% of the genetic diagnoses should have been possible from the earlier exome test. The first laboratories had most often passed over variants in genes not yet tied to disease, or set aside a known gene because the symptoms didn’t look typical. It is one site’s experience, drawn from a look back at its own records.
The same logic runs the other way. A long search without a name doesn’t mean nothing is wrong. Functional disorders, for example, are diagnosed by positive signs on examination, not by ruling everything else out, a point our entry on why symptoms persist works through in detail.
Where the search goes wrong: clinics that sell certainty
Years without an answer leave a vacuum, and some clinics fill it with a confident diagnosis delivered fast. The details differ, but the pattern repeats: a test that flags healthy people too, a diagnosis no result can disprove, a treatment that runs for months, and improvement credited to that treatment even though placebo groups improve too.
“Chronic Lyme disease”
Lyme disease is a real tick-borne infection, treated with a standard course of antibiotics. Some people go on to have fatigue, pain or thinking problems that linger after treatment. The Centers for Disease Control and Prevention (CDC) names this post-treatment Lyme disease syndrome and says nobody yet knows the cause. The agency avoids the phrase “chronic Lyme disease” because it suggests the bacteria are still there. CDC doesn’t dispute the symptoms; it disputes that explanation. What those lingering symptoms are, and how common, belongs to the library’s article on illness after infection.
The treatment question has been tested directly. In two NIH-funded trials published in The New England Journal of Medicine in 2001, people with well-documented, previously treated Lyme disease and persistent symptoms got 30 days of intravenous antibiotics followed by 60 days of oral antibiotics, or matching placebos. Both trials stopped early. Among patients with antibodies to the Lyme bacterium, 37% improved on antibiotics and 40% on placebo. The authors concluded that across the two trials, “treatment with intravenous and oral antibiotics for 90 days did not improve symptoms more than placebo”.
A Dutch trial of 281 people, in the same journal in 2016, gave everyone two weeks of intravenous antibiotics, then 12 weeks of one of two antibiotic regimens or a placebo. Quality-of-life scores ended up nearly identical across the three groups. All three improved from where they started, the placebo group included. That detail matters: a patient who feels better during a long course has no way to know that a placebo might have helped as much. Four serious drug-related side effects occurred during the two-week intravenous phase.
The 2020 Lyme guideline from the Infectious Diseases Society of America, the American Academy of Neurology and the American College of Rheumatology (ACR) recommends against more antibiotics for persistent, nonspecific symptoms after recommended treatment when there is no objective evidence of reinfection or treatment failure. It is a strong recommendation based on moderate-quality evidence. Objective evidence means signs such as arthritis, meningitis or nerve damage, not tiredness alone. Months of antibiotics for symptoms that linger after Lyme treatment fall into the category this library labels tested and not supported.
The harms are on record. In 2017, CDC’s Morbidity and Mortality Weekly Report described five patients seriously harmed while being treated for “chronic Lyme disease.” Two developed septic shock, a life-threatening reaction to a bloodstream infection. Others had an infection of the spine, a severe bowel infection called C. difficile colitis, and an abscess beside the spine. One patient died of septic shock tied to the intravenous catheter used for treatment. One had been diagnosed at an alternative medicine clinic, another after tests that had not been validated, and another had received immune-globulin infusions every three weeks for more than 10 years. The report’s authors concluded that these treatments lack proof of effectiveness and can cause serious complications.
Testing feeds the cycle. The ACR’s 2013 Choosing Wisely list noted, citing earlier studies, that Lyme tests are ordered appropriately in only 20% to 40% of cases where the disease is common, and under 10% elsewhere. Between 10% and 79% of inappropriately tested patients received antibiotics whatever the result. The ACR advises against testing people whose only symptoms are chronic fatigue or widespread aches, even where Lyme is common.
Provoked urine metal tests
The second pattern relies on a test built to come back high: urine collected after a dose of a chelating drug, which binds metals so the body flushes them out, and so raises urine metal levels even in healthy people. In their 2013 Choosing Wisely list, the American College of Medical Toxicology and the American Academy of Clinical Toxicology advised against that test and against chelation without documented metal poisoning; our chelation and detox article covers the harms and the trial record.
Questions to ask before you pay for an answer
A legitimate specialist can answer every one of these. A clinic selling certainty tends to stumble on the first two.
- What result does a healthy person get on this test? If nearly everyone tests positive, a positive result tells you little.
- What result would mean I don’t have this condition? A diagnosis that no finding could ever rule out isn’t doing the work of a diagnosis.
- Which objective finding supports it? An examination sign, a lab value with an established normal range, an image. Symptoms alone are what you arrived with.
- When this treatment was compared with a dummy treatment, what happened? For long antibiotic courses after Lyme disease, three placebo-controlled trials found no added benefit.
- How will we know if it isn’t working, and when would it stop?
- Will it need a catheter or port in a vein, and who looks after it? In CDC’s report, the patient who died had a bloodstream infection tied to an intravenous catheter.
- Will the results go to my regular clinicians in a form they can check?
- What will the full course cost, and does insurance cover any of it?
Legitimate routes exist alongside these questions. GARD suggests doctors at larger university or academic medical centers, which tend to have newer technology and may run clinical trials. Its information specialists can point to resources but give no medical advice. A UDN application has to come with a health care provider’s letter, so the referring clinician is part of it from the start. And whether a fresh specialist might read the case differently is the territory of our second-opinion guide.
When to get help now
A long search can make every new symptom feel like more of the same. Some aren’t.
- New, changed or worsening symptoms deserve their own examination, whatever label came before.
- Objective changes such as a measured fever, unexplained weight loss, a new lump or an abnormal examination finding are worth documenting and taking to a clinician, since they are exactly what specialist networks look for.
- During treatment through an intravenous line or port, fever, chills, or redness, swelling or pus where the catheter enters need same-day medical care. In one of the CDC cases, a catheter-related bloodstream infection proved fatal.
- Emergencies mean 911: one side of the face or body suddenly going weak or numb, slurred or garbled speech, an abrupt and severe headache, chest pain, or a struggle to breathe.
- Hopelessness. Years without answers wear people down. If that weariness becomes thoughts of suicide or self-harm, call or text 988, or chat online, to reach the 988 Suicide & Crisis Lifeline; it is free, confidential and staffed around the clock.
The Takeaway
Even the federal network built for medicine’s hardest cases names the problem for about one person in three who completes its evaluation, and a name often changes care without curing anything. That is a reason to keep the search tied to evidence, not a reason to stop. The evidence here points to two things that improve the odds: objective findings, written down and kept, and existing test data reanalyzed as knowledge grows.
Before paying anyone for certainty, ask the question every sound test survives: what result would a healthy person get?
Sources
- The New England Journal of Medicine. “Splinter K, Adams DR, Bacino CA, et al. "Effect of Genetic Diagnosis on Patients with Previously Undiagnosed Disease." N Engl J Med, 2018;379(22):2131-2139. PubMed. DOI: 10.1056/NEJMoa1714458.” pubmed.ncbi.nlm.nih.gov
- NIH Common Fund. “Undiagnosed Diseases Network (last reviewed April 12, 2024).” commonfund.nih.gov
- National Institute of Neurological Disorders and Stroke. “NIH funds 12 Diagnostic Centers of Excellence in Phase III of the Undiagnosed Diseases Network (August 2, 2024).” ninds.nih.gov
- National Institutes of Health. “NIH Guide funding opportunity PAR-23-289, Undiagnosed Diseases Network Diagnostic Centers of Excellence (posted August 31, 2023).” grants.nih.gov
- Undiagnosed Diseases Network (Harvard Medical School coordinating center). “Apply.” undiagnosed.hms.harvard.edu
- Genetic and Rare Diseases Information Center (GARD), National Center for Advancing Translational Sciences. “About GARD.” rarediseases.info.nih.gov
- National Academies of Sciences, Engineering, and Medicine. “Improving Diagnosis in Health Care (2015), Summary.” nationalacademies.org
- BMJ Quality & Safety. “Singh H, Meyer AN, Thomas EJ. "The frequency of diagnostic errors in outpatient care: estimations from three large observational studies involving US adult populations." BMJ Qual Saf, 2014;23(9):727-31. PubMed. DOI: 10.1136/bmjqs-2013-002627.” pubmed.ncbi.nlm.nih.gov
- European Journal of Human Genetics. “Faye F, Crocione C, Anido de Peña R, et al. "Time to diagnosis and determinants of diagnostic delays of people living with a rare disease: results of a Rare Barometer retrospective patient survey." Eur J Hum Genet, 2024;32(9):1116-1126. PubMed. DOI: 10.1038/s41431-024-01604-z.” pubmed.ncbi.nlm.nih.gov
- Genetics in Medicine. “Spillmann RC, Walley N, Schoch K, et al. "Not the end of the road: Achieving diagnoses following non-diagnostic exome sequencing: experiences at a clinical site of the Undiagnosed Diseases Network." Genet Med, 2026:102736 (published online September 28, 2026). PubMed. DOI: 10.1016/j.gim.2026.102736.” pubmed.ncbi.nlm.nih.gov
- American Journal of Ophthalmology. “Tinker RJ, Smith LM, Bastarache LA, et al. "The Undiagnosed Diseases Network (UDN) Solves Ocular Syndromic Diagnostic Dilemmas." Am J Ophthalmol, 2025;280:51-63. PubMed. DOI: 10.1016/j.ajo.2025.07.028.” pubmed.ncbi.nlm.nih.gov
- The New England Journal of Medicine. “Klempner MS, Hu LT, Evans J, et al. "Two controlled trials of antibiotic treatment in patients with persistent symptoms and a history of Lyme disease." N Engl J Med, 2001;345(2):85-92. PubMed. DOI: 10.1056/NEJM200107123450202.” pubmed.ncbi.nlm.nih.gov
- The New England Journal of Medicine. “Berende A, ter Hofstede HJ, Vos FJ, et al. "Randomized Trial of Longer-Term Therapy for Symptoms Attributed to Lyme Disease." N Engl J Med, 2016;374(13):1209-20. PubMed. DOI: 10.1056/NEJMoa1505425.” pubmed.ncbi.nlm.nih.gov
- Infectious Diseases Society of America. “Clinical Practice Guidelines by the IDSA, AAN, and ACR: 2020 Guidelines for the Prevention, Diagnosis and Treatment of Lyme Disease (published November 30, 2020; Lantos PM et al., Clin Infect Dis 2021;72(1):e1-e48, DOI: 10.1093/cid/ciaa1215).” idsociety.org
- Centers for Disease Control and Prevention. “Chronic Symptoms and Lyme Disease (last reviewed June 26, 2026).” cdc.gov
- Centers for Disease Control and Prevention, Morbidity and Mortality Weekly Report. “Marzec NS, Nelson C, Waldron PR, et al. "Serious Bacterial Infections Acquired During Treatment of Patients Given a Diagnosis of Chronic Lyme Disease, United States." MMWR, 2017;66(23):607-609. DOI: 10.15585/mmwr.mm6623a3.” cdc.gov
- American College of Medical Toxicology and American Academy of Clinical Toxicology. “Choosing Wisely: Things Physicians and Patients Should Question (items released September 26, 2013).” clintox.org
- American College of Rheumatology (Arthritis Care & Research). “Yazdany J, Schmajuk G, Robbins M, et al. "Choosing wisely: the American College of Rheumatology's Top 5 list of things physicians and patients should question." Arthritis Care Res (Hoboken), 2013;65(3):329-39. PubMed. DOI: 10.1002/acr.21930.” pubmed.ncbi.nlm.nih.gov
- 988 Suicide & Crisis Lifeline. “988 Suicide & Crisis Lifeline.” 988lifeline.org