Classification Systems

Classification Systems: How Diagnoses Are Built and Where They Break

The diagnosis lessons teach you the criteria for each disorder. This lesson teaches the system behind them: why we sort people into categories, how the DSM got its shape, how it differs from the ICD, and which newer models want to replace boxes with dimensions. The EPPP tests the reasoning here, not symptom lists.

Why This Matters for Psychologists

A client leaves your office with "major depressive disorder" in the chart. That label gets her insurance to pay, tells the next clinician what you saw, and points to a treatment. It may also change how she sees herself. Every classification choice has a payoff and a cost. You need to know both.

A classification system does several jobs at once:

• Common language so clinicians and researchers mean the same thing
• Treatment planning and a guess about prognosis
• Research, since studies need defined groups
• Public health statistics on how common disorders are
• Billing and records, through diagnostic codes

What Counts as a Disorder?

DSM-5-TR defines a mental disorder as a syndrome: a clinically significant disturbance in thinking, emotion regulation, or behavior that signals a dysfunction in the psychological, biological, or developmental processes underneath. It usually comes with significant distress or disability. Two exclusions matter. An expectable or culturally approved response to a common stressor or loss, such as grief after a death, is not a disorder. Socially deviant behavior (political, religious, sexual) is not a disorder unless it comes from a dysfunction in the person.

Being unusual, or breaking a cultural norm, is not enough by itself. Being left-handed is unusual. It isn't a disorder.

Wakefield's harmful dysfunction model sits underneath this definition. A disorder needs two things: an internal mechanism that fails at its natural function (dysfunction), and real negative consequences judged by the person's culture (harm). A fear of spiders so strong that you avoid your own basement and quit your job after seeing one at work is both. A mild dislike of spiders that never gets in your way causes no harm, so it isn't a disorder.

At the other pole, Thomas Szasz called mental illness a "myth" and recast symptoms as problems in living.

Two fine points from DSM-5-TR. First, having a diagnosis is not the same as needing treatment. Second, the clinical significance criterion ("causes clinically significant distress or impairment") is how the DSM separates disorder from normal variation when no lab test can.

Labels Help, and Labels Hurt

Rosenhan's Pseudopatients

The classic labeling study is Rosenhan's (1973) "On Being Sane in Insane Places." Eight people with no psychiatric history, Rosenhan among them, sought admission to 12 hospitals. Each reported one symptom: a voice saying "empty," "hollow," or "thud." All were admitted. Once inside, they dropped the symptom and acted normally. Every admission but one ended in a schizophrenia diagnosis (the other was manic-depressive psychosis). Stays ran 7 to 52 days, 19 on average. Those labeled with schizophrenia were discharged with "schizophrenia in remission." The label stayed even after the symptom was gone.

Staff never spotted the fakes. Fellow patients did: on the first three admissions, 35 of 118 patients voiced suspicion. In a follow-up, one hospital was told pseudopatients would try to get in over the next three months. At least one staff member flagged 41 of 193 new patients as likely fakes. Rosenhan had sent none.

Rosenhan argued that the label and the setting, not the person's behavior, drove what staff saw. It's like a teacher who hears a student is a "troublemaker" and then reads every fidget as defiance. The study helped push the APA toward the reforms of DSM-III.

The standard exam answer: Rosenhan shows how labels and context distort diagnostic judgment. Caveat: Spitzer (1976) called it "pseudoscience presented as science," and a 2023 historical review, drawing on records uncovered by journalist Susannah Cahalan, concluded the study was scientific fraud (Scull, 2023); exam items still key on the classic finding.

Modified Labeling Theory and Self-Labels

Modified labeling theory says that once you receive a label, the negative beliefs you already held about that group now apply to you, and they hurt. In 427 youth followed for two years, adopting a mental illness self-label lowered self-esteem, and dropping it raised self-esteem (Harari et al., 2023). Person-first language follows the same logic: a person has schizophrenia; they are not "a schizophrenic."

A Short History of the DSM

The DSM is published by the American Psychiatric Association. Know the turning points.

DSM-III is the big break. It replaced theory about causes with descriptions of observable symptoms, so clinicians of any orientation could use the same criteria. Then it tested whether they actually agreed. Think of it as moving from "describe the cake" to "here's the recipe with exact measurements."

How DSM-5-TR Is Organized

DSM-5-TR has three sections. Section I (DSM-5 Basics) explains how to use the manual. Section II (Diagnostic Criteria and Codes) holds the official disorders. Section III (Emerging Measures and Models) holds tools and ideas that need more research: the assessment measures, the Cultural Formulation Interview, the alternative personality disorder model, and conditions for further study, such as attenuated psychosis syndrome and internet gaming disorder.

The Five Axes and Where They Went

DSM-IV recorded diagnoses on five axes. DSM-5 dropped them.

What DSM-5-TR Added

DSM-5-TR (2022) was mostly a text update, but it also added:

• Prolonged grief disorder (criteria in the Trauma/Stressor-Related, Dissociative, and Somatic Symptom Disorders lesson)
• Unspecified mood disorder (F39), for mood presentations where you can't yet tell bipolar from depressive, such as acute agitation
• Symptom codes for suicidal behavior and nonsuicidal self-injury (NSSI), current or past history

The codes record a behavior. Suicidal behavior disorder and NSSI disorder are still conditions for further study in Section III, not official disorders.

DSM and ICD: Two Manuals, One Code Set

The International Classification of Diseases (ICD) comes from the World Health Organization (WHO) and covers all health conditions, not just mental disorders. The DSM covers mental disorders only and adds detailed criteria and text.

In the U.S., the official codes for billing and records are ICD-10-CM codes, required since October 1, 2015. ICD-10-CM is the U.S. clinical version of the WHO's ICD-10. The DSM doesn't have its own codes. It lists ICD-10-CM codes next to each disorder (for example, F39 for unspecified mood disorder). Psychosocial problems use Z codes.

ICD-11 took effect worldwide on January 1, 2022, but each country chooses when to adopt it. At DSM-5-TR's publication, the U.S. had no timeline, so ICD-10-CM remains the U.S. code set.

The two systems now match more closely than they have since DSM-II and ICD-8. Of 103 shared disorders, 31 were essentially identical and 20 had major differences (First et al., 2021).

ICD-11's Personality Disorder Model

ICD-11 dropped the ten personality disorder types. You decide whether a personality disorder is present and rate it mild, moderate, or severe. Below that line, personality difficulty can be noted, but it isn't a mental disorder. You may add trait domain qualifiers: negative affectivity, detachment, dissociality, disinhibition, and anankastia, plus an optional borderline pattern qualifier (Bach & First, 2018).

Categorical vs. Dimensional Approaches

A categorical approach asks, "Does this person have the disorder, yes or no?" A dimensional approach asks, "How much of this trait or symptom does this person have?" Categorical is a light switch. Dimensional is a dimmer.

Why Categories Strain

• Comorbidity: many people meet criteria for more than one disorder. In the U.S., about 41% of adults with OCD have had major depressive disorder at some point.
• Heterogeneity: one diagnosis covers very different people. A big reason: most DSM criteria sets are polythetic, needing a minimum number of symptoms from a list, not all of them. It's a combo meal where you pick any five of nine items. MDD's criteria can be met 227 different ways (Zimmerman et al., 2015), so two people with MDD may share only one criterion. Among 3,703 depressed outpatients, researchers found 1,030 unique symptom profiles, and the most common was shared by only 1.8% (Fried & Nesse, 2015). The opposite, a monothetic set, requires every criterion (DSM-III-R moved the personality disorders to polythetic sets; Widiger et al., 1988).
• Arbitrary thresholds: when DSM-5 lowered bulimia nervosa's binge-and-purge frequency from twice to once a week, patients who moved from "eating disorder NOS" to bulimia looked much like those already diagnosed (MacDonald et al., 2014).
• No natural boundaries: a meta-analysis of 317 taxometric findings favored dimensional over categorical models five to one. No content area showed overall support for true categories (Haslam et al., 2020).

DSM-5-TR's Answer: Categorical With Dimensional Elements

For clinical usefulness and ICD coding, DSM-5-TR stays primarily categorical with dimensional elements. It also says categories don't have sharp edges and people with the same diagnosis differ. The dimensional elements include:

• Severity specifiers: alcohol use disorder is mild (2-3 symptoms), moderate (4-5), or severe (6 or more). Autism spectrum disorder uses Levels 1 to 3 ("requiring support" up to "requiring very substantial support"). Intellectual developmental disorder severity rests on adaptive functioning, not IQ.
• Spectrum diagnoses: autism spectrum disorder replaced autistic disorder, Asperger's disorder, and PDD-NOS.
• Cross-cutting symptom measures: a "review of mental systems" across diagnoses. The adult Level 1 measure has 23 questions covering 13 domains over the past 2 weeks. Positive screens lead to Level 2 measures.
• Subtypes and specifiers: subtypes are mutually exclusive, so you pick one (anorexia nervosa is restricting or binge-eating/purging type). Specifiers are not, so more than one can apply (MDD with anxious distress and with mixed features).
• WHODAS 2.0: a 36-item WHO measure of disability across six areas (such as getting around, self-care, and getting along with people) over the past 30 days, for adults. It replaced the GAF.

When a Presentation Doesn't Fit: Other Specified, Unspecified, Provisional

DSM-IV called leftover presentations NOS ("not otherwise specified"). DSM-5 offers two options:

• Other specified: the clinician records why criteria aren't met. Example: "other specified schizophrenia spectrum and other psychotic disorder, with persistent auditory hallucinations."
• Unspecified: the clinician doesn't state the reason, often because there isn't enough information yet (for example, in an emergency room).

The difference is only whether you name the reason. "Provisional" is used when you strongly expect the criteria to be confirmed once more information arrives, or when the diagnosis depends on how long symptoms last. Schizophreniform disorder stays provisional while symptoms are ongoing, because you don't yet know whether they will end before 6 months. A provisional diagnosis is a pencil entry on the calendar: you expect it to hold, and you ink it in once it's confirmed.

DSM-5-TR calls its criteria guidelines for clinical judgment, not a cookbook. A diagnosis can be given when a presentation falls just short of full criteria if the symptoms are persistent and severe.

The Alternative DSM-5 Model for Personality Disorders (AMPD)

Section II keeps the ten categorical personality disorders, carried over from DSM-IV-TR. Section III offers the Alternative DSM-5 Model for Personality Disorders, a hybrid dimensional-categorical model. Why? Patients often met criteria for several personality disorders at once, and "other specified" or "unspecified personality disorder" was often the correct but unhelpful diagnosis. The Personality Disorders lesson covers Section II.

Criterion A: Level of Personality Functioning. You rate impairment in four areas, two for the self and two interpersonal. The Level of Personality Functioning Scale (LPFS) runs from Level 0 (little or no impairment) to Level 4 (extreme). A personality disorder requires moderate impairment (Level 2) or greater.

Criterion B: Pathological Personality Traits. Five broad domains contain 25 trait facets. A diagnosis needs at least one pathological trait. The domains are maladaptive versions of the Big Five:

The Personality Inventory for DSM-5 (PID-5) was built to measure these traits (Krueger et al., 2012).

The AMPD keeps six specific disorders: antisocial, avoidant, borderline, narcissistic, obsessive-compulsive, and schizotypal. It drops paranoid, schizoid, histrionic, and dependent. Anyone with a personality disorder who doesn't fit the six gets personality disorder-trait specified (PD-TS), described by their own trait profile.

The AMPD and ICD-11 look alike, and the exam likes to swap their details:

The labels differ, but a published crosswalk matches ICD-11's mild level to the AMPD's Level 2. Dissociality matches antagonism. Anankastia has no AMPD domain; it matches the facets rigid perfectionism and perseveration (Bach & First, 2018).

RDoC: A Research Framework, Not a Diagnostic Manual

The Research Domain Criteria (RDoC) project began at the National Institute of Mental Health (NIMH) in 2009. It flips the DSM's direction. The DSM starts from symptoms and works toward causes. RDoC starts from basic brain-behavior systems and asks how they go wrong, studying the full range from normal to abnormal.

Domains (rows): negative valence (threat, loss), positive valence (reward), cognitive systems, social processes, and arousal/modulatory systems. A sixth, sensorimotor systems, was added in 2019.

Units of analysis (columns): genes, molecules, cells, circuits, physiology, behavior, and self-report.

The DSM is a field guide that sorts birds by how they look. RDoC is a biology lab studying how wings and lungs work in every bird.

The exam point: RDoC is for research. It isn't meant for clinical diagnosis or billing, and it doesn't try to cover every DSM disorder.

HiTOP and the p Factor

The Hierarchical Taxonomy of Psychopathology (HiTOP) is a dimensional system built from statistical patterns in symptom data rather than expert consensus. It stacks symptoms that occur together into levels:

1. Symptom components and maladaptive traits (for example, social interaction anxiety)
2. Syndromes (closest to DSM disorders)
3. Subfactors (for example, a fear subfactor including social phobia, specific phobia, and agoraphobia)
4. Spectra: internalizing, thought disorder, disinhibited externalizing, antagonistic externalizing, detachment, and somatoform
5. Superspectra, topped by a general factor

That top factor is the p factor. In the Dunedin cohort, followed from adolescence to midlife, disorders first sorted into three factors: internalizing, externalizing, and thought disorder. One general psychopathology dimension explained them even better. Caspi et al. (2014) named it p, parallel to g in intelligence. Higher p went with more life impairment and worse developmental histories.

HiTOP handles comorbidity by grouping related syndromes into spectra. It handles heterogeneity by splitting symptoms into narrow components. It's still a work in progress. DSM-5 already nods to this structure: after the neurodevelopmental disorders, its chapters run in internalizing groups, then externalizing groups. For a related model of anxiety and depression, see the tripartite model in the Cognition, Mood, and Temperament lesson.

Reliability of Diagnosis

Diagnostic reliability asks how often two clinicians, working independently, land on the same diagnosis for the same patient. It's usually measured with kappa, which corrects for chance agreement (formula in the Item Analysis and Test Reliability lesson).

Under DSM-I and DSM-II, clinicians often agreed only fairly or poorly on major diagnoses. In the DSM-III field trials, Axis I kappa was .78 when clinicians interviewed together and .66 after separate interviews (Spitzer et al., 1979).

The DSM-5 field trials used a tougher design. Two clinicians interviewed each patient separately, in real clinical settings, with ordinary unstructured interviews. The authors rated 5 diagnoses very good (kappa .60-.79), 9 good (.40-.59), 6 questionable (.20-.39), and 3 unacceptable (below .20) (Regier et al., 2013). PTSD reached .67, while major depressive disorder was .28 and generalized anxiety disorder .20, both in the questionable range (Regier et al., 2013, as cited in Rosellini et al., 2015).

Method matters. In one study of DSM-IV diagnoses, kappa averaged .80 when raters scored the same audio recording but .47 with separate interviews (Chmielewski et al., 2015). Two judges scoring the same video agree more than two judges watching different performances. Remember too: reliability is necessary but not sufficient for validity.

Culture and Classification

Where normal ends and disorder begins shifts across cultural contexts. Direct eye contact reads as honest in the U.S. and as rude in many other cultures.

Two classification changes matter here. Since DSM-5, cultural concepts of distress (cultural idioms, explanations, and syndromes, such as ataque de nervios) have replaced the old idea of "culture-bound syndromes." And DSM-5-TR names racism as a source of misdiagnosis, such as African Americans with mood disorders being misdiagnosed with schizophrenia; a work group reviewed the full text for stigmatizing language.

The Cultural Formulation Interview (CFI) in Section III is meant for any client, not only clients whose background is unfamiliar to you. The Differential Diagnosis lesson covers cultural concepts of distress and the CFI in depth; emic and etic perspectives are in the Cross-Cultural Issues lesson.

EPPP Traps and Common Misconceptions

Misconception 1: "DSM-5-TR diagnoses are recorded on Axes I through V."

• Reality: DSM-5 went nonaxial. Axes I-III are listed together, Axis IV became Z codes, and the GAF was replaced by the WHODAS 2.0.

Misconception 2: "The DSM supplies the diagnostic codes used for billing."

• Reality: U.S. codes are ICD-10-CM codes, the U.S. version of the WHO's ICD-10. The DSM lists them next to each disorder.

Misconception 3: "DSM-5 switched to a dimensional system."

• Reality: It is primarily categorical with dimensional elements (severity specifiers, cross-cutting measures, WHODAS 2.0, the AMPD).

Misconception 4: "The AMPD replaced the ten personality disorders."

• Reality: Section II's ten are unchanged. The AMPD is a Section III alternative with six disorders plus PD-TS.

Misconception 5: "RDoC is NIMH's new diagnostic manual."

• Reality: RDoC is a research framework. Clinicians still diagnose with DSM criteria and ICD codes.

Misconception 6: "Other specified and unspecified mean different levels of severity."

• Reality: The only difference is whether the clinician records the reason criteria aren't met.

Misconception 7: "Low DSM-5 field trial kappas prove DSM-5 is less reliable than DSM-III."

• Reality: The designs differed. Separate interviews in real clinics produce lower kappas than joint interviews or shared recordings.

Misconception 8: "NSSI is an official DSM-5-TR disorder."

• Reality: DSM-5-TR added codes to record the behavior. NSSI disorder is still a condition for further study.

Misconception 9: "The AMPD and ICD-11 use the same five trait domains."

• Reality: They share negative affectivity, detachment, and disinhibition. The AMPD adds antagonism and psychoticism; ICD-11 adds dissociality and anankastia, and codes schizotypal disorder with the psychotic disorders.

Misconception 10: "In Rosenhan's study, hospital staff caught the pseudopatients once they acted normally."

• Reality: Staff never caught them; some fellow patients did. In the follow-up, staff flagged real patients as fakes.

Memory Aids

• DSM-III "CAN": Criteria, Axes, Neutral about causes (plus field trials for reliability)
• Where the axes went: "1-2-3 merge, 4 goes Z, 5 goes WHO"
• LPFS: the self has an ID (Identity, self-Direction); relationships need EI (Empathy, Intimacy)
• AMPD domains: the Big Five gone bad. Four flip familiar traits (stability, extraversion, agreeableness, conscientiousness); the fifth, Psychoticism, is the opposite of lucidity
• AMPD kept at least one from each cluster: schizotypal (A); antisocial, borderline, narcissistic (B); avoidant, obsessive-compulsive (C)
• ICD-11 vs. AMPD: ICD-11 keeps three AMPD domains, has Dissociality where the AMPD has antagonism, drops psychoticism, and adds Anankastia
• Polythetic: pick some from the list. Monothetic: must have them all.
• Other specified: you say the reason. Unspecified: you don't.
• Provisional: waiting on info or waiting on time
• p factor: p is to psychopathology what g is to intelligence

Key Takeaways

• Classification gives a common language, guides treatment and research, supports statistics, and enables billing. Labels can also distort perception (Rosenhan, now disputed) and lower self-esteem (modified labeling theory).
• DSM-5-TR defines a disorder as a clinically significant dysfunction that usually causes distress or disability; Wakefield calls it harmful dysfunction.
• DSM-III (1980) brought explicit criteria, a descriptive approach neutral on cause, the multiaxial system, and field trials. DSM-5 (2013) dropped the axes; the GAF gave way to the WHODAS 2.0.
• DSM-5-TR (2022) added prolonged grief disorder, unspecified mood disorder, and codes for suicidal behavior and NSSI.
• U.S. billing uses ICD-10-CM codes, the U.S. version of the WHO's ICD-10. ICD-11 (in effect worldwide since 2022) rates personality disorder by severity with trait qualifiers.
• Categorical systems are familiar and decisive but struggle with comorbidity, heterogeneity (driven by polythetic criteria), and arbitrary cutoffs. Dimensional systems are more reliable and match the evidence but are harder to use. DSM-5-TR is categorical with dimensional elements.
• Subtypes are mutually exclusive; specifiers can stack.
• Other specified records the reason; unspecified does not; provisional flags expected confirmation.
• AMPD: Criterion A = level of personality functioning (moderate or worse); Criterion B = five trait domains, 25 facets; six disorders plus PD-TS. ICD-11 has dissociality where the AMPD has antagonism, adds anankastia, and has no psychoticism.
• RDoC (NIMH) is a research framework of brain-behavior domains across units of analysis. HiTOP is a statistical hierarchy of spectra topped by the p factor.
• Kappa depends on method: DSM-5 field trials showed MDD at .28 and GAD at .20 with separate interviews.
• Culture shapes thresholds and expression; cultural concepts of distress replaced culture-bound syndromes, and the CFI is for every client.

Now cover the tables and rebuild the five axes, the AMPD domains, and the five systems from memory. That retrieval practice is what makes it stick.