The DSM Is a Billing Document Wearing a Lab Coat

There is no blood test for depression. There is no brain scan for anxiety. There is no biomarker for ADHD, no genetic panel for borderline personality disorder, no lab result that distinguishes bipolar II from major depressive disorder with mixed features. After more than a century of modern psychiatry and billions of dollars in neuroscience research, the diagnostic process for mental illness works like this: a clinician sits across from a patient, asks questions, listens to the answers, and checks symptoms against a list in a book. If the patient reports enough symptoms from the list for a long enough period of time, they receive a diagnosis. The diagnosis generates a billing code. The billing code gets sent to insurance. Insurance pays the clinician. The clinician prescribes medication or therapy. The cycle continues.

The book is the Diagnostic and Statistical Manual of Mental Disorders, now in its fifth edition, text revision. It is published by the American Psychiatric Association and it is the foundational document of American mental health care. Every diagnosis a psychiatrist, psychologist, or therapist makes in a clinical setting references the DSM. Every insurance reimbursement for mental health treatment requires a DSM code. Every prescription for a psychiatric medication is justified by a DSM diagnosis. The manual is not advisory. It is infrastructural. It is the operating system on which American mental health care runs.

And it contains no objective diagnostic criteria. Not one.

Five of Nine for Two Weeks

The diagnostic criteria in the DSM are symptom checklists with thresholds. Major depressive disorder requires five of nine symptoms present for at least two weeks, with at least one symptom being either depressed mood or loss of interest. The nine symptoms include depressed mood, diminished interest, weight change, sleep disturbance, psychomotor agitation or retardation, fatigue, feelings of worthlessness, difficulty concentrating, and recurrent thoughts of death.

The threshold is not derived from biology. Nobody discovered that five symptoms constitutes depression while four does not. Nobody identified a neurological event that occurs at the two-week mark. The numbers are consensus decisions made by committees of psychiatrists who had to draw a line somewhere. Four symptoms for ten days? Maybe that should count too. Six symptoms for three weeks? Maybe that is more specific. The committee chose five and fourteen days because it seemed right; because it balanced sensitivity against specificity in a way that clinical experience suggested was workable.

This is not a criticism of the people on the committee. They were working with what they had, which was clinical observation and professional judgment. The problem is the distance between what the DSM claims to be and what it actually is. The DSM presents itself as a diagnostic manual on the model of medical classification systems. It looks like the ICD. It uses numerical codes. It speaks in the language of criteria and specifiers and differential diagnosis. It performs the aesthetics of medical precision. But a diagnostic manual for pneumonia tells you to look for specific pathogens. The DSM tells you to count reported symptoms. The epistemological gap between those two operations is enormous, and the manual does not acknowledge it.

The symptom overlap between diagnoses makes this worse. Fatigue is a criterion for major depressive disorder, generalized anxiety disorder, and PTSD. Sleep disturbance appears in depression, anxiety, PTSD, bipolar disorder, and several substance use disorders. Difficulty concentrating is in the criteria for depression, anxiety, ADHD, and PTSD. A person presenting with fatigue, insomnia, and concentration problems could plausibly receive any of those diagnoses depending on which questions the clinician asks and which narrative the patient tells. The diagnosis often depends less on the patient’s underlying condition than on the clinician’s training, orientation, and the order in which symptoms are elicited.

The Insurance Architecture

The reason the DSM’s checklist structure has become immortal is not scientific. It is financial. The American health insurance system requires a diagnosis before it will pay for treatment. No diagnosis, no billing code. No billing code, no reimbursement. A clinician who wants to get paid for treating a patient’s mental health needs a DSM code on the claim form.

This creates a structural incentive to diagnose. Not to over-diagnose, necessarily; most clinicians are trying to do honest work. But the system makes it impossible to treat without diagnosing, and impossible to diagnose without using the DSM’s categories, and impossible to use the DSM’s categories without committing to a specific label that may or may not correspond to the patient’s actual experience. A therapist who sees a patient struggling with a difficult life transition; a divorce, a job loss, a death in the family; cannot bill for “helping someone through a hard time.” They have to call it something. Adjustment disorder. Major depressive episode. Generalized anxiety. The label is the price of admission to the reimbursement system.

The DSM-III, published in 1980 under Robert Spitzer’s leadership, was explicitly designed to solve a reliability problem. Before the DSM-III, psychiatric diagnosis was a mess. Two clinicians could interview the same patient and come away with entirely different diagnoses. The DSM-III’s innovation was the checklist: specific, countable, observable symptoms with explicit thresholds. This dramatically improved inter-rater reliability. Two clinicians could now look at the same checklist, apply it to the same patient, and arrive at the same code.

But reliability is not validity. Two clinicians can agree that a patient meets five of nine criteria for depression without either of them knowing whether “depression” as defined by those nine criteria is a real, discrete entity or an arbitrary slice of continuous human experience. The checklist solved the billing problem; consistent codes, consistent reimbursement; without solving the scientific problem; what are we actually diagnosing?

The insurance industry did not need the scientific problem solved. Insurance needs categories that are specific enough to generate codes, consistent enough to be auditable, and broad enough to cover the patient population seeking treatment. The DSM delivers exactly this. It is a masterpiece of bureaucratic engineering. It converts the messy, continuous, deeply individual experience of human psychological suffering into a standardized taxonomy that interfaces cleanly with billing software. The fact that the taxonomy may not carve nature at its joints is, from the insurance company’s perspective, irrelevant.

The Categories Shift by Edition

If the DSM’s categories reflected stable, discovered entities; real diseases with real boundaries; they would not change much between editions. Tuberculosis is tuberculosis in every edition of every medical textbook. But DSM categories change constantly. Conditions are added, removed, split, merged, and redefined with every revision.

The DSM-II had 182 diagnoses. The DSM-III had 265. The DSM-IV had 297. The DSM-5 reorganized everything and the count depends on how you count, but the trend is unambiguous: the number of diagnosable mental conditions keeps growing. Human psychology did not become more complex between 1968 and 2013. The manual did.

Some of these changes reflect genuine scientific progress. PTSD was added to the DSM-III in 1980 after extensive research on Vietnam veterans demonstrated that trauma produces a consistent syndrome. That was a real advance. But many changes reflect committee politics, cultural shifts, and stakeholder pressure. Asperger’s disorder was added in the DSM-IV in 1994 and removed in the DSM-5 in 2013, folded into autism spectrum disorder. Bereavement exclusion; the provision that said you should not diagnose depression in someone who is grieving; was removed in the DSM-5, meaning that grief that looks like depression can now be billed as depression two weeks after a loss. The removal was controversial. Critics argued it pathologized normal human grief. Supporters argued it allowed treatment for people who needed it. Both arguments had merit. The decision was made by committee, not by discovery.

Premenstrual dysphoric disorder was promoted from the DSM-IV’s appendix to a full diagnosis in the DSM-5 after lobbying by researchers and pharmaceutical companies. The pharmaceutical interest was transparent: Eli Lilly had rebranded fluoxetine (Prozac) as Sarafem specifically for PMDD. A full diagnostic code made prescribing and billing cleaner. A condition that had been considered “needing further study” became a treatable disorder in the same edition that happened to benefit the pharmaceutical company that had already developed the treatment.

Follow the Money from Symptom to Quarterly Earnings

The financial architecture connecting the DSM to the pharmaceutical industry is not hidden. It is structural and acknowledged by everyone involved, which somehow makes it easier to ignore.

A DSM diagnosis creates a market. Once a condition exists in the manual, pharmaceutical companies can develop drugs to treat it, run clinical trials to demonstrate efficacy, and market the drug to clinicians who now have a diagnosis to match it to. The drug does not need to cure the condition. It does not need to address the underlying cause. It needs to reduce symptoms more effectively than placebo in a controlled trial. If it does that, it gets approved. If there is a DSM code for the condition it treats, it gets prescribed. If it gets prescribed, it gets reimbursed. If it gets reimbursed, it generates revenue.

The financial conflicts of interest on DSM committees have been documented repeatedly. A 2006 study by Cosgrove and Krimsky found that 56 percent of DSM-IV task force members had financial ties to pharmaceutical companies. For some diagnostic panels the number was 100 percent. Every member of the mood disorders panel had pharma ties. Every member of the schizophrenia panel had pharma ties. The DSM-5 implemented conflict-of-interest disclosure requirements, but disclosure is not the same as elimination. Knowing that a committee member consults for Pfizer does not undo the influence of that consulting relationship on their clinical judgment.

The symptom-to-prescription pipeline runs through the DSM like plumbing through a building. The patient reports symptoms. The clinician matches symptoms to a DSM code. The code authorizes a prescription. The prescription generates a pharmacy claim. The pharmacy claim generates revenue for the drug manufacturer. The drug manufacturer funds research on the condition. The research produces findings that support the diagnosis. The diagnosis stays in the next edition of the DSM. The cycle is self-reinforcing. Each component validates the others, and no single component is individually corrupt. The clinician is trying to help. The researcher is studying real phenomena. The drug company is selling a product that passed FDA trials. The insurance company is paying for an authorized treatment. Everyone is operating within the system. The system is the problem.

The Reliability Trap

The DSM’s great achievement; inter-rater reliability; has become its prison. The checklist structure works so well for billing, for research recruitment, for clinical communication, and for institutional coordination that abandoning it would collapse the infrastructure of American mental health care. Every clinical trial recruits subjects based on DSM criteria. Every insurance form requires DSM codes. Every training program teaches DSM categories. Every outcome study measures improvement against DSM baselines.

The National Institute of Mental Health attempted an alternative. In 2013, Thomas Insel, then the director of NIMH, announced the Research Domain Criteria project, which aimed to classify mental disorders based on neuroscience; on brain circuits, genetic markers, and biological processes rather than reported symptoms. Insel explicitly stated that the DSM categories lacked validity and that research should move beyond them. The announcement generated headlines and professional anxiety and then very little changed, because the RDoC framework does not yet produce the kind of discrete, billable categories that the clinical and financial infrastructure requires. You cannot bill for a dysregulated reward circuit. You can bill for major depressive disorder.

The DSM persists not because it is correct but because it is necessary. It is the interface between suffering and payment. It is the language that clinicians, insurance companies, pharmaceutical researchers, and patients have all agreed to speak, not because the language accurately describes the territory but because everyone needs a common language and this is the one that exists. Replacing it would require rebuilding the financial, clinical, and regulatory infrastructure of American mental health care from the ground up.

Nobody is going to do that. The manual will get a sixth edition. The committees will convene. Some categories will be added, some modified, some cautiously removed. The conflicts of interest will be disclosed and not eliminated. The checklist structure will remain because the billing structure requires it. And clinicians will continue sitting across from patients, asking questions, counting symptoms, and assigning labels that carry the weight of medical authority without the foundation of medical certainty.

The DSM is not a scientific fraud. The people who write it are not cynics. They are professionals trying to impose order on genuinely disordered human experience using the tools available to them. But the primary tool available to them is consensus, and the primary constraint on that consensus is the reimbursement architecture of American health insurance. The result is a document that looks like medicine, functions like bureaucracy, and gets used like scripture. It is a billing document wearing a lab coat. It is very good at billing. The lab coat is the part that should make you nervous.