Executive Summary
The Age of Diagnosis by Dr. Suzanne O’Sullivan presents a compelling critique of overdiagnosis and overmedicalisation in modern medicine. The book challenges the conventional wisdom that “earlier diagnosis is always better,” demonstrating through extensive case studies that diagnostic expansion is pathologizing normal human variation and converting healthy people into patients. O’Sullivan’s distinctive contribution lies in showing how diagnostic labels themselves function as powerful interventions that can create real physical symptoms through the nocebo effect and predictive coding—often causing more harm than benefit for mild cases. The author argues that while diagnosis provides substantial benefits for severe disease, the harm-to-benefit ratio becomes unfavorable for mild presentations where treatment risks remain identical but potential benefits diminish dramatically.
Overview and Core Thesis
Dr. Suzanne O’Sullivan examines the troubling phenomenon of overdiagnosis and overmedicalisation in modern medicine. The book argues that while diagnosis provides substantial benefits for those with severe disease, diagnostic expansion—driven by technology, financial incentives, and social pressure—is converting healthy people into patients by pathologizing normal human variation.
The central challenge to conventional wisdom is that “earlier diagnosis is always better.” O’Sullivan demonstrates how diagnostic labels themselves can be powerful interventions that create real physical symptoms through the nocebo effect and predictive coding, often causing more harm than benefit for mild cases.
The Problem of Diagnostic Expansion
Overdiagnosis refers to correct diagnoses that harm patients because treatment is neither needed nor beneficial—detecting medical problems before they cause clinical harm. Overmedicalisation turns ordinary human differences, behaviors, and life stages into medical problems requiring doctor intervention. Both arise through overdetection (using new technologies to find earlier/milder disease) and expanded disease definitions (moving diagnostic thresholds for what counts as abnormal, also called “diagnostic creep”).
The statistics reveal alarming trends: ADHD diagnoses increased 48%, depression rates rose 25% in a single year, autism prevalence exploded from 4 in 10,000 fifty years ago to 1 in 100 today, cancer diagnoses projected to exceed 2 million in the US in 2024, and diabetes affects 537 million people globally.
Case Studies in Diagnostic Expansion
Huntington’s Disease: The Burden of Predictive Diagnosis
Huntington’s disease testing exemplifies the profound dilemma of predictive genetic diagnosis—knowing decades in advance that you will develop an incurable neurodegenerative condition. HD is caused by a single gene variant on chromosome 4 (autosomal dominant), meaning 50% chance of inheritance, absolute certainty of developing the disease if the gene is present, mid-life emergence causing progressive dementia, movement disorders, and psychiatric problems, with no effective treatment.
The testing paradox: Early surveys found 90% of at-risk people said they would take the test; however, when actually offered testing, only 5-18% globally accept it. This gap suggests many people find that hope sustains them better than certainty.
Psychological consequences include depression, suicidal ideation, loss of healthy years to worry, impacts on driving and employment, potential medical discrimination, and interpreting ordinary bodily changes as disease signs.
Stephanie’s experience illustrates the protective power of delayed diagnosis. She lived without an HD diagnosis for 30 years, allowing her to pursue career, family, and ambitions freely. When diagnosed at age 52, this “blissful ignorance” was revealed as protective—she accomplished more during those decades without knowing her genetic status than she might have with early knowledge.
Lyme Disease: Test Subjectivity and Diagnostic Controversy
Lyme disease’s discovery history reveals diagnosis as fundamentally subjective despite appearing objective. The disease is caused by bacterium Borrelia burgdorferi transmitted by tick bite, typically causing bullseye rash, flu-like symptoms, and potentially multi-system disease. Testing involves two stages: ELISA (enzyme-linked immunosorbent assay) followed by Western blot. Crucially, these are not diagnostic tests—they are supportive evidence requiring interpretation within clinical context.
The problem of test interpretation: a positive test does not mean Lyme disease but indicates exposure or past infection; false positives occur from cross-reactivity with other infections or autoimmune disorders; false negatives happen if tests target wrong bacterial strains, are done too early, or are poorly calibrated. In the New Forest (UK), 25% of forestry workers tested positive for Borrelia burgdorferi without any symptoms—showing exposure does not equal disease.
The CDC found that 1,016 of 1,261 people referred with a Lyme disease diagnosis had no evidence of active or recent infection (80% misdiagnosis rate). In 2022, while 63,000 cases met CDC standards, electronic health records showed 476,000 people treated for Lyme disease—over 400,000 without official diagnosis confirmation. Chronic Lyme disease (CLD) represents the extreme of this expansion; the US National Institute of Allergies and Infectious Diseases defines CLD as “symptoms in people who have no clinical or diagnostic evidence of a current or past infection”—essentially a misdiagnosis by definition.
Long Covid: Patient-Driven Diagnosis Without Definition
Long COVID was coined by Elisa Perego via Twitter hashtag on May 20, 2020, to describe persistent symptoms after mild COVID-19 infection. From inception, it lacked a disease definition, specific symptoms, or diagnostic test requirements—making it a self-diagnosis with no proof of infection needed. A negative COVID test didn’t count against it; 70% of one UK support group had tested negative for the virus.
Most infectious illnesses cause more severe long-term effects in those most severely ill acutely, but long COVID was more common after mild infection than in hospitalized severe cases. A substantial proportion of long COVID—particularly in those with mild or self-diagnosed infection—is best explained as psychosomatic illness. Multiple studies show anxiety, depression, perceived stress, loneliness, and negative life events predict long COVID better than positive viral tests.
Rather than a single illness, long COVID represents consequences of pandemic isolation with no institutional support, combined with nocebo effects, excess attention to the body, predictive coding, and fear spread through social media.
Autism Diagnosis: Expansion Beyond Recognition
Autism has undergone dramatic diagnostic expansion since Leo Kanner’s 1943 description of 11 severely impaired children with “extreme autistic aloneness” and complete inability to relate to others. Diagnostic prevalence exploded from 4 in 10,000 fifty years ago to 1 in 100 today. In 2023, California reported 1 in 22 eight-year-olds autistic; Northern Ireland 1 in 20; Texas 1 in 64; France 1 in 144. The male-to-female ratio shifted from 4:1 (1980s) to 3:1 currently to approaching 2:1. Adult diagnoses increased 150% between 2008-2016, and between 1998-2018, autism diagnoses rose 787% in the UK.
DSM-5 changes (2013) dramatically reorganized autism diagnosis by reducing essential symptoms needed for diagnosis, combining social and communication problems into one symptom list, abolishing PDD-NOS and Asperger’s as separate diagnoses, removing requirement for symptom appearance before age 3 (now just “early developmental period”), and allowing diagnosis based on “masking” even if no autistic traits are visible.
Diagnostic inconsistency is severe: a 2022 US study reassessing community-diagnosed autistic children by research standards found 47% didn’t actually meet research criteria, and a recent UK study found diagnostic rates varying from 35-85% across assessment centers. Official assessment ideally involves hours of semi-structured interviews (ADOS and ADI-R), observations across environments, teacher corroboration, and multidisciplinary consensus.
Systemic pressures undermine proper assessment: England has 1.2 million people waiting for autism assessment, lengthy multidisciplinary assessments are difficult to maintain, some US autism centers train education professionals in ADOS, and Oregon allows school diagnosis without medical evaluation. “Impairment”—essential for diagnosis—is undefined at the mild end. Consequences of expansion are severe: as discourse is dominated by mild, articulate, self-identified voices, severe autism becomes invisible. Uta Frith, a pioneering autism researcher, warns that “the diagnosis of autism has been stretched to breaking point.”
ADHD: Diagnostic Inflation and Biologizing
ADHD diagnosis has exploded globally—from 7% global prevalence in children to 22% in Iran, 14% in Tunisia. In the US, ADHD rose from 6% (1990s) to 10% (2016) in children, and UK teenagers doubled 2000–2018. Most striking: adult ADHD diagnoses went from rare to 1 in 20 in some places, almost all mild, with the UK seeing a 400% increase in adults seeking ADHD diagnosis 2020–2023.
Diagnostic inconsistency is significant: Norway (with free healthcare) varies from <1% to >8% diagnosis rates between regions, and US rates range 5% (California) to 14% (Mississippi). The word “often” in DSM-5 criteria is subjective, and immaturity may be confused with neurodevelopment as youngest children in school cohorts are diagnosed more frequently than older classmates.
Biological evidence has major limitations: brain differences exist (slightly smaller volumes in some studies) but are not abnormalities—only group comparisons. Radiologists cannot diagnose ADHD on brain scans; people with ADHD have normal scans. Twin studies show 76–88% heritability, but genome-wide association studies show only 22%, suggesting non-genetic factors dominate.
“Neurodiversity” was coined in 1998 by sociologist Judy Singer and is not a medical term but sounds like one, lending false authority. Singer explained: “Neuroscientists were the new priesthood, so I thought, let’s put them together. Neurodiversity sounds really important.”
ADHD is now framed as a neurobiological disorder despite weak biomedical evidence. This obscures social and environmental factors: childhood abuse, neglect, trauma, and witnessing violence increase ADHD risk. The DSM-5 notes ADHD signs may be “completely absent when the individual is receiving frequent rewards, is under close supervision, in a novel setting, or engaged in interesting activities.”
Medication concerns are substantial: among DSM-5 ADHD working group advisers, 78% disclosed financial ties to drug companies. Stimulant prescriptions increased sevenfold in the UK (last 10 years), tenfold for adults in New Zealand (2006–2022), and 250% in the US (2006–2016). A 2022 Cochrane review of 24 trials (5,066 people) found no good evidence methylphenidate (Ritalin) was superior to placebo in adults.
BRCA Variants and Risk-Reducing Surgery
The discovery of BRCA1 (1994) and BRCA2 (1995) genes revolutionized cancer risk assessment. However, the predictive value differs critically from Huntington’s disease: a BRCA variant is a risk factor, not a certainty. High-risk BRCA1 variants carry 60–85% lifetime breast cancer risk and 40–60% ovarian cancer risk; BRCA2 variants confer 40–65% and 10–20% respectively. Unlike Huntington’s disease (incurable), women with BRCA variants can choose preventative surgery or intensive surveillance.
Risk-reducing surgery reduces respective cancer risks by 95%. However, 10–15% of women undergoing mastectomy and 40% undergoing oophorectomy would never have developed cancer if untreated—making them overdiagnosed by definition. These women received unnecessary surgery with permanent consequences.
Risk prediction problems are significant: models accurate for high-risk familial populations may not apply to those without cancer family histories, and the UK Biobank shows many healthy people in their 60s–70s carry high-risk BRCA variants yet never developed cancer. Direct-to-consumer genetic testing dangers are substantial: over 26 million people have used commercial genetic tests costing £129–148, bypassing counseling requirements and using non-clinical-grade sequencing with up to 96% false positive rates.
Cancer Screening: Saving Lives While Creating Overdiagnosis
Cancer screening saves lives but also creates substantial overdiagnosis. NHS England estimates screening saves 10,000 lives yearly, but among those lives are people treated for early cancers that would never have progressed. A 2023 US study estimated 31% of breast cancers diagnosed in women over 70 were overdiagnosed, and a French study found €100+ million spent on overdiagnosed thyroid cancer in four years.
For prostate cancer screening, while some lives are saved, as many as 20 men per 1,000 screened are diagnosed with and treated for cancer that would never have caused problems. A 2023 meta-analysis of 2 million screened people found that for large bowel cancer, screening extended lifespan by only 110 days. For other cancers, there was no evidence screening allowed people to live longer (all-cause mortality). The core problem: scientists cannot distinguish slow-growing, indolent cancers from aggressive ones. Detroit autopsies found early prostate cancer in 45% of men in their 50s and 70% in their 60s, yet only 13% develop clinically significant disease.
The Power of Diagnostic Labels
The Nocebo Effect and Predictive Coding
Diagnostic labels are not inert facts—they are powerful interventions that can create real physical symptoms through the nocebo effect (opposite of placebo). Through predictive coding, the brain uses past experience and expectations to predict bodily responses. When someone learns they have a disease or are at risk of one, this knowledge fundamentally alters how they perceive their body.
This expectation-driven process causes people to notice and worry about ordinary bodily changes they would normally filter out, interpret normal sensations as symptoms, and behave in ways consistent with their diagnosis.
Valentina’s case exemplifies this mechanism. After learning she had a 50% chance of inheriting Huntington’s disease from her mother, she began experiencing clumsy movements, dizziness, and mood swings. The more attention she paid to these sensations, the worse they became. When genetic testing revealed she did not carry the Huntington’s gene, these symptoms either disappeared or became manageable—demonstrating they resulted from expectation and fear, not pathological process.
Research shows this labeling effect operates across multiple conditions. A 2018 sham MRI study demonstrated children’s powerful responsiveness to expectation—children told they received a placebo that would relax and focus them showed strong symptom reduction simply from being told to expect it. Multiple studies demonstrate that anxiety, depression, perceived stress, loneliness, and negative life events predict long COVID symptoms better than positive viral tests.
The Severity-Based Harm-Benefit Analysis
The Critical Importance of Disease Severity
For severe disease, diagnosis provides undeniable benefits: a child with severe autism supported by a one-to-one tutor has “all to gain and little to lose” through labeling because the severity speaks for itself. Severe depression, severe ADHD, and severe autism all benefit from pathways to treatment, expert support, and accommodations.
However, the harm-to-benefit calculation reverses for mild cases. The same label carries identical treatment risks but dramatically less potential benefit, making the harm-to-benefit ratio unfavorable. A person with subtle, masked autism faces high vulnerability to labeling effects yet gains “substantially less from medication, school accommodations and other types of support.”
The author uses cancer treatment as an analogy: chemotherapy’s horrific side effects are justified for someone dying of cancer but not for someone with a few cells that may never grow. Similarly, a person with subtle autism faces high vulnerability to labeling effects yet gains substantially less from treatment.
Pathologizing Normal Human Variation
Society has developed unrealistic expectations of health, success, and aging, leading to pathologization of ordinary human experience. Sadness, even when understandable (grief, disappointment), has become medicalized as depression. Failure to achieve desired goals is increasingly explained through medical diagnosis rather than accepted as normal limitation.
The author argues these are cultural problems, not medical ones: “An expectation of constant good health, graceful ageing and an obedient body and mind has left people unprepared for those ordinary bodily declines that affect us all.”
Practical Strategies for Navigating Diagnosis
Severity-Based Diagnostic Decision-Making
When considering whether a diagnosis is appropriate, assess whether the individual’s symptoms/traits cause genuine impairment across multiple life domains (functioning, relationships, work, education). Mild traits affecting only specific contexts may not warrant diagnosis.
Ask: “What is the specific harm this causes to this specific person?” If harm is mild and inconsistent, diagnosis may cause more harm through labeling than benefit through intervention. Before accepting a diagnosis, seek a second opinion from a generalist physician (not a specialist) who knows you as a whole person. Verify that symptoms actually impair multiple areas of life, not just appear unusual. Understand what treatments/accommodations the diagnosis unlocks—and whether they’re worth the identity shift and potential limitations.
Clinical Context Interpretation of Medical Tests
Remember that tests alone cannot diagnose. All tests have false positive and false negative rates dependent on disease prevalence in the tested population, calibration, timing, and confounding variables. Ask your doctor about your pretest probability (how likely am I to have this disease based on my symptoms and risk factors), the false positive and false negative rates for this specific test, what confounding variables might affect this result, whether a negative result would change the clinical impression, and what the next step is if this result is positive.
Assessing Harm-to-Benefit Ratios
Before accepting a diagnosis or beginning treatment, explicitly weigh potential benefits against potential harms. Ask what specific evidence shows this treatment/diagnosis improves your life, what the documented harms are, whether there’s evidence showing benefit outweighs harm for your specific severity level, and what alternatives exist.
Recognizing Nocebo Effects and Protecting Against Diagnosis-Induced Harm
Understand that diagnostic labels can create symptoms through expectation and predictive coding. Once someone learns they have a disease, their expectations change, often causing them to notice ordinary bodily changes they would normally filter out.
Protective strategies include limiting exposure to symptom lists and support groups focused entirely on illness before establishing your own baseline experience, noticing whether you develop new symptoms after learning about them, considering whether your symptoms existed before diagnosis or emerged after, and seeking balance between medical support and activities/identities unrelated to your diagnosis.
Seeking Generalist Medical Perspective
Specialists have financial stakes in expanding their disease definitions and identifying more patients within their specialty. Generalists (primary care physicians), by contrast, have no stake in increasing diagnoses and are better positioned to notice when too much diagnosis has made a patient worse. If you’ve accumulated multiple diagnoses from different specialists, ask your primary care physician to review your entire medical file holistically. Ask whether these diagnoses explain your overall condition, whether there’s a simpler unifying explanation, whether any treatments are contradicting each other, and whether you’ve become over-medicalized.
Key Insights and Counterintuitive Perspectives
The Protective Power of Diagnostic Ignorance
Common belief holds that earlier diagnosis is always better and prevents harm. The book reveals that for conditions with no effective treatment or those decades away from symptom onset, the knowledge of disease can create more harm than benefit through decades of anxiety and life limitation.
Tests Are Subjective, Not Objective
Medical tests provide objective truth about disease presence—this is the common belief. The book reveals that all tests have false positive and false negative rates depending on calibration, timing, population, and interpretation. The same test result means different things in different contexts—making diagnosis fundamentally subjective despite appearing technical.
Severe Autism Becomes Invisible When Mild Cases Expand
More diagnosis of autism is good because it identifies previously missed cases—this is the conventional view. The book reveals that as diagnostic criteria broaden to include milder presentations, discourse shifts toward mild cases, resources concentrate on high-functioning individuals, and severe autism becomes invisible.
Psychosomatic Illness Is Not Imaginary or Malingering
Many believe “psychosomatic” means “not real” or “all in your head.” The book reveals that psychosomatic illness is entirely real—the brain genuinely creates symptoms through expectation, fear, and predictive coding. The distinction matters: recognizing an illness as psychosomatic allows different treatment approaches.
Risk Factors Are Not Diseases, Yet Society Treats Them As Such
Having a BRCA variant means you will get cancer—this is how many people understand it. The book reveals that BRCA variants confer risk, not certainty. Unlike Huntington’s disease (certain if gene present), many BRCA carriers never develop cancer. Yet risk-reducing surgery decisions are made as though risk equals certainty.
Medication Efficacy Is Often Overstated
ADHD medication significantly improves functioning in adults—this is the common assumption. The book reveals that a 2022 Cochrane review found methylphenidate no better than placebo in adults. Depression’s “serotonin theory” was disproven by a 2023 Nature meta-analysis.
Critical Warnings and Considerations
Mental Health Warnings
This book challenges psychiatric diagnosis and medication efficacy, particularly for mild-to-moderate presentations. For people with severe mental illness, psychiatric diagnosis and medication can be life-saving. However, for those with mild depression, anxiety, or ADHD, non-medical interventions may be equally or more effective.
Do not discontinue psychiatric medication based on this book’s arguments without consulting your prescriber; abrupt cessation can cause harm.
When to Seek Professional Help
This book is not an argument against medical care—it’s an argument against unnecessary and potentially harmful medicalization. Seek professional help if your symptoms genuinely impair multiple areas of your life, you’ve suffered significant trauma or abuse, you’re experiencing suicidal thoughts, medical investigation has revealed an actual disease, or you’re isolated and struggling without support.