4 Classification and assessment in psychiatry4.3 Clinical assessment
4.3.6 Rating scales in psychiatry
18 min read3 figures
Written by Resident Doctor Committee · Published
What scales are for
Purpose | What the scale gives |
|---|---|
Screening | Picks out probable cases in an unselected population for further assessment |
Severity | Quantifies an established or suspected disorder |
Monitoring | Tracks change against a pretreatment baseline |
Adverse-effect surveillance | Records symptoms related to treatment |
Function | Assesses disability or participation |
Diagnosis | Applies diagnostic criteria through a standardised interview |
Research | Standardises study outcomes |
A threshold on a symptom scale does not by itself establish a diagnosis; the score supports clinical assessment.
Choose the measure that fits the question, weighing its validity in that population, language, accessibility, burden and whether the result will change care.
Types of instrument
Rater | Who scores it | Strength | Limitation |
|---|---|---|---|
Self-report | The patient | Direct account of subjective experience | Depends on literacy, understanding and response style |
Informant | A carer or relative | Sees behaviour across settings | Limited by opportunity, perspective and relationship |
Observer-rated | A trained clinician, after interview and observation | Combines interview and observation | Needs time, training and attention to rater bias |
Among depression scales, the Hamilton and Montgomery-Asberg scales are clinician-rated, while most others are patient-completed, including the Beck Depression Inventory, Patient Health Questionnaire-9, Hospital Anxiety and Depression Scale and Geriatric Depression Scale.
A structured instrument uses fixed items with set scoring.
A projective test presents an ambiguous stimulus onto which the person projects, and it allows a free response.
Measures of the same syndrome are not interchangeable, because their items and weighting differ.
When raters disagree, the cause may be different settings, timeframes, symptoms or perspectives, so explore it rather than assume poor insight or exaggeration.
Reliability
Reliability and validity are separate properties. Agreement across raters or occasions is reliability, not a subtype of validity, and reliability alone does not establish validity.
Type | What it captures | How it is expressed |
|---|---|---|
Inter-rater | Agreement between different raters scoring the same subjects on the same occasion with the same instrument, so the rater is the only source of variation | Kappa, which can be weighted |
Test-retest | Correlation between the same subjects' scores on two occasions while their true state is stable | Correlation coefficient |
Internal consistency | How far the items of one scale correlate because they measure the same construct | Cronbach's alpha, Kuder-Richardson 20 or split-half correlation, from a single administration |
A high alpha may reflect redundant items and does not prove a single underlying construct.
Validity
Type | Question it answers | How it is shown |
|---|---|---|
Content | Do the items cover the whole domain of the construct? | Systematic judgement by experts and people with lived experience |
Construct | Do scores behave as theory about the attribute predicts? | Theory-driven hypothesis testing, such as convergent and discriminant correlations |
Concurrent (criterion) | Does it agree with an accepted criterion? | Criterion gathered at the same sitting |
Predictive (criterion) | Does it forecast a later outcome? | Scale scored first, outcome recorded independently later |
Incremental | Does it add anything over the tests already in use? | Comparison with existing tests |
Construct validity is not shown by expert item review, by prediction of a later outcome or by stability on retesting.
Discriminant (divergent) validity is one arm of construct validity, shown by an expected weak correlation with a measure of a different construct.
Concurrent and predictive validity are the two subtypes of criterion validity, and they differ only in when the criterion is measured.
Ecological validity asks whether performance relates to everyday functioning.
Responsiveness is the ability to detect relevant change, and a scale can be valid at one time point yet insensitive to change with treatment.
Screening performance
Sensitivity is the proportion of cases a test detects, and specificity is the proportion of non-cases it classifies correctly.
Where higher scores mean illness, lowering the threshold raises sensitivity at the cost of specificity; a receiver operating characteristic (ROC) curve compares the two across thresholds.
Predictive values depend on prevalence: with 90% sensitivity and specificity, the positive predictive value is 8.3% at 1% prevalence (9 true and 99 false positives per 1,000) but 85.7% at 40%.
Scores, change and error
Identical totals can hide different symptom profiles, so read bands, subscales and items alongside measurement error.
Use a validated version for the population and language, follow its instructions, anchors and missing-data rules, and record any adaptation.
A different rater, setting or timeframe can produce apparent change, so keep conditions consistent on repeat testing.
Floor and ceiling effects limit how much change a scale can register.
A practice effect is better scores on repeated exposure to the same test items through familiarity rather than true change.
A practice effect is told apart from the Hawthorne effect, the halo effect, a ceiling effect and regression to the mean.
Regression to the mean is the statistical tendency for extreme measurements to be followed by values nearer the average, independent of any true change.
Regression to the mean arises when a group is selected for extreme scores on an imperfectly reliable measure.
A statistically significant improvement is not necessarily clinically meaningful.
Source of bias | Effect on the score |
|---|---|
Literacy | Can make self-report inaccessible or incomplete |
Language and culture | Translation alone does not make versions equivalent |
Education and age | Shift cognitive test results and call for suitable norms |
Cognitive or sensory impairment | Limits recall, self-report and performance; collateral history helps |
Response style | Acquiescence, extreme responding or minimisation distort self-report |
Rater expectation | Sways scoring; training and blinding help |
Physical illness | Somatic symptoms inflate mood-scale scores |
Depression scales
Adult scales
Scale | Rater | Items and scoring | Distinctive features |
|---|---|---|---|
Hamilton Depression Rating Scale (HAM-D) | Clinician | 17 items, mixed 0 to 2 and 0 to 4 | Heavy loading on sleep, appetite, weight loss and somatic anxiety |
Montgomery-Asberg Depression Rating Scale (MADRS) | Clinician | 10 items, each 0 to 6, maximum 60 | The depression scale most sensitive to change; light somatic content |
Beck Depression Inventory, second edition (BDI-II) | Patient | 21 items, each 0 to 3, total 0 to 63 | Past two weeks; includes a suicide item; validated from age 13 |
Patient Health Questionnaire-9 (PHQ-9) | Patient | 9 items, each 0 to 3, maximum 27 | Past two weeks; items mirror the diagnostic criteria for major depression |
Hospital Anxiety and Depression Scale (HADS) | Patient | 14 items in two 7-item subscales, each item 0 to 3, each subscale 0 to 21 | Omits bodily symptoms, so physical illness does not inflate scores in medical patients |
The HAM-D has separate items for initial, middle and late insomnia, and for decades it was the standard outcome measure in antidepressant trials.
In frail older people, the HAM-D's somatic items pick up physical illness and inflate the total.
MADRS items were chosen from a larger psychopathology scale because they changed most with antidepressant treatment.
Because the MADRS carries little somatic content, drug side-effects inflate it less than they inflate the HAM-D.
Clinician-rated depression scales measure current depressive severity in a working-age adult and track it over time, unlike scales built for mania, antidepressant discontinuation, late-life depression or broad psychopathology.
The Beck Depression Inventory is built on Beck's cognitive model, so it gives more content to negative views of self and future than the somatic HAM-D or the core-mood MADRS.
Because the BDI-II is self-report and validated from age 13, it can screen adolescents for depression; always specify the edition.
Depression trials commonly define response as a 50% fall in score and remission as a 17-item HAM-D of 7 or less, though definitions vary.
PHQ-9 and HADS bands
Traditional PHQ-9 bands are 0 to 4, 5 to 9, 10 to 14 (moderate), 15 to 19 (moderately severe) and 20 to 27 (severe), with cut-offs at 5, 10, 15 and 20.
A PHQ-9 of 10 or more suggests probable depression, but a self-report screen does not establish the diagnosis.
NICE NG222 (National Institute for Health and Care Excellence) divides depression at a PHQ-9 of 16: below 16 is less severe and 16 or more is more severe.
HADS bands apply to each subscale separately: 0 to 7 is normal, 8 to 10 borderline abnormal and 11 to 21 abnormal (case range).
Depression in particular groups
Scale | Population | Format | Distinctive features |
|---|---|---|---|
Calgary Depression Scale for Schizophrenia | Schizophrenia | 9 items: 8 from a structured interview, 1 observed | Covers mood, hopelessness, guilt and suicidal thinking; omits slowing, blunted affect and drug-induced parkinsonism, which overlap with negative and extrapyramidal features |
Cornell Scale for Depression in Dementia | Dementia, where self-report is unreliable | 19 items, each 0 to 2 | Scored after a carer interview, then observation and interview of the patient |
Geriatric Depression Scale (GDS) | Older people with intact cognition | Self-report yes or no answers; 15-item short form | Few or no somatic items; a short-form score of 5 or more suggests depression |
Brief Assessment Schedule Depression Cards (BASDEC) | Older medical inpatients, including those with hearing loss | 19 printed statements sorted into 'true' or 'false' piles | Scored out of 21 |
Perinatal mood
The Edinburgh Postnatal Depression Scale (EPDS) is a 10-item self-completed screen with a four-response format and a maximum of 30.
It covers the past seven days, not the two weeks of the PHQ-9.
It is validated in pregnancy and beyond the early postnatal weeks, and it screens rather than diagnoses.
Its items cover mood, anhedonia, self-blame, anxiety, panic, feeling overwhelmed, tearfulness and thoughts of self-harm.
It deliberately omits somatic, bonding, infant-harm and psychotic content.
Any endorsement of the self-harm item needs direct assessment.
EPDS total | Meaning |
|---|---|
0 to 9 | Low likelihood of depression |
10 to 12 | Moderate likelihood needing further assessment (10 or more is possible depression) |
13 or more | Probable depression that should prompt clinical assessment |
Anxiety, trauma and obsessive-compulsive scales
Scale | Rater | Structure | Use |
|---|---|---|---|
GAD-7 | Patient | 7 items, maximum 21 | Anxiety screening and monitoring; the threshold depends on setting and purpose |
Impact of Event Scale-Revised | Patient | 22 items on intrusion, avoidance and hyperarousal over the past week, anchored to the event | Brief and sensitive to change, so it follows post-traumatic stress disorder (PTSD) symptoms, including with co-occurring substance misuse |
Yale-Brown Obsessive Compulsive Scale (Y-BOCS) | Clinician | 10 severity items, each 0 to 4, maximum 40 | The severity standard in obsessive-compulsive disorder (OCD), repeated to track response |
Y-BOCS items rate time, interference, distress, resistance and control for obsessions and for compulsions, so it tracks severity whatever the symptom content.
Mania
The Young Mania Rating Scale (YMRS) is a clinician-rated 11-item measure of the severity of current manic symptoms.
It is scored from the patient's report of the preceding two days and the interviewer's observation.
Four items are double weighted: irritability, speech, thought content and disruptive behaviour.
The YMRS is neither a self-report questionnaire nor a diagnostic tool.
Psychosis scales
Positive and Negative Syndrome Scale
The Positive and Negative Syndrome Scale (PANSS) has 30 clinician-rated items, each scored 1 to 7, so totals run from 30 to 210.
Its items fall into seven positive, seven negative and sixteen general psychopathology items.
It covers the past week and draws on informants alongside a semi-structured interview.
It rates the severity of both positive and negative symptoms in schizophrenia and related psychoses, and it is repeated to track change.
The positive subscale reflects an excess or addition to normal function and includes hostility and grandiosity; poor impulse control sits under general psychopathology.
The negative subscale is a specific seven-item set that includes difficulty in abstract thinking and stereotyped thinking.
Look-alike deficit items, such as loss of drive, slowing and social avoidance, sit under general psychopathology rather than the negative subscale.
A fall of at least 20% in total PANSS score from baseline is the usual minimum for calling a patient a responder.
Because the lowest possible total is 30, percentage improvement = (baseline minus follow-up) divided by (baseline minus 30), multiplied by 100.
Brief Psychiatric Rating Scale
The Brief Psychiatric Rating Scale (BPRS) has 18 items in the standard version and 24 in the expanded version, each rated 1 to 7 by a clinician.
The first 14 items are rated from the patient's report and the rest from observation, so the patient needs no reading or writing ability.
The expanded version adds six items, including suicidality, self-neglect and elevated mood; always specify the version.
Its items include somatic concern and uncooperativeness, rated from the patient's behaviour towards the interviewer, but there is no substance item.
The BPRS is a short, broad measure of overall severity across diagnoses, sensitive to change, rating psychotic and non-psychotic items on one form.
It is used mainly in schizophrenia and other psychotic illness.
Comparing psychosis scales
Feature | PANSS | BPRS |
|---|---|---|
Items | 30 | 18 (24 expanded) |
Time to complete | About 30 to 50 minutes | About 15 to 30 minutes |
Rater | Clinician | Clinician |
Subscales | Separate positive, negative and general psychopathology scores | No separation of positive and negative symptoms |
The main advantage of the PANSS over the BPRS is fuller coverage with separate subscales, which gives a separate negative symptom score.
Both scales rate mood items such as anxiety and depression, but neither directly measures real-world functional outcome or diagnoses schizophrenia.
A general symptom-severity scale such as the BPRS measures current psychotic and affective symptoms, not enduring personality functioning, so it has no role in screening for personality disorder.
The Scale for the Assessment of Negative Symptoms (SANS, Andreasen) rates five domains, with items scored 0 to 5: affective flattening, alogia, avolition-apathy, anhedonia-asociality and attention.
The SANS has a companion scale for positive symptoms (SAPS); each covers half the picture, whereas the PANSS covers both.
The Clinical Global Impression, like the BPRS, is a broad scale that does not separate positive from negative symptoms.
Global, outcome and quality-of-life measures
Measure | Structure | Use |
|---|---|---|
Clinical Global Impression (CGI) | Two single-item, seven-point clinician judgements: severity against the rater's experience of similar patients, and improvement against the patient's own baseline | Works across diagnoses, but has no item definitions or individual symptoms, so it needs a clinician familiar with the patient group |
Health of the Nation Outcome Scales (HoNOS) | 12 clinician-rated items, each 0 to 4, over behaviour, impairment, symptoms and social problems | Routine outcome measure developed by the Royal College of Psychiatrists in the 1990s to track change in National Health Service mental health care |
Global Assessment of Functioning (GAF) | A single figure out of 100 | Axis V of the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV); dropped from DSM-5 in 2013, partly for poor reliability |
WHO Disability Assessment Schedule (WHODAS 2.0) | Cognition, mobility, self-care, relationships, life activities and participation, within the International Classification of Functioning, Disability and Health | A WHO measure of functioning and disability, distinct from an ICD-11 diagnostic assessment |
Camberwell Assessment of Need (CAN) | 22 domains, each rated no need, met need or unmet need | Records staff and service user views separately |
SF-36 | Brief, patient-completed, eight scales covering physical and mental health | Generic measure of health-related quality of life |
Flanagan's Quality of Life Scale (QOLS, 16 items) | Normally completed by the patient | Self-report quality of life |
Drug side-effect scales
Scale | What it rates | Detail |
|---|---|---|
Abnormal Involuntary Movement Scale (AIMS) | Tardive dyskinesia and other drug-induced abnormal movements | 12-item clinician examination: seven body areas rated 0 to 4, global judgements and dental items |
Barnes Akathisia Rating Scale | Akathisia | Observed restlessness, the patient's awareness of it, distress about it and global severity |
Simpson-Angus Scale | Drug-induced parkinsonism | Clinician-rated; rigidity (the main emphasis), tremor and bradykinesia |
LUNSERS | Antipsychotic adverse effects | 51 patient-completed items, ten of them sham ('red herring') items to detect over-reporting |
Discontinuation-Emergent Signs and Symptoms scale (DESS) | Symptoms that emerge when an antidepressant is stopped | Each symptom rated as new or worsened |
Arizona Sexual Experiences Scale (ASEX) | Sexual dysfunction | 5 items, each 1 to 6, total 5 to 30, higher meaning more dysfunction; suits both sexes, any orientation and people without a partner |
Extrapyramidal side-effect scales (parkinsonism, akathisia, dyskinesia) are a separate family from scales that rate depression or mania severity.
Akathisia is an inner urge to move that starts within weeks of an antipsychotic.
Because the AIMS is observer-rated, it suits a patient who is unaware of the movements.
The Simpson-Angus Scale scores parkinsonism, whereas the AIMS scores the choreoathetoid and orofacial movements of tardive dyskinesia.
The Lane scale rates minor physical anomalies in a developmental dysmorphology examination and does not measure drug-induced movement disorder.
A scale describes symptoms, but treatment timing, examination and differential diagnosis decide what caused them.
Cognition and older adults
Instrument | Structure | Use |
|---|---|---|
Mini-Mental State Examination (MMSE) | Out of 30: orientation, attention, recall, language and construction | Brief screen with limited executive assessment |
Montreal Cognitive Assessment (MoCA) | Ten minutes, out of 30; more executive, visuospatial, abstraction and delayed-recall tasks; one point added for 12 or fewer years of education | Preferred brief screen, because it picks up mild cognitive impairment |
Addenbrooke's Cognitive Examination III | Out of 100: attention, memory, fluency, language and visuospatial skills; three parallel versions | Tracking cognitive change |
Bristol Activities of Daily Living Scale | 20 carer-rated items on self-care and instrumental tasks such as cooking, finances and telephone use | Built for dementia and sensitive to change |
Neuropsychiatric Inventory | Informant-rated, 12 domains, each scored for frequency, severity and carer distress; a nursing home version exists | Behavioural and psychological symptoms of dementia |
Clifton Assessment Procedures for the Elderly (CAPE) | A short direct cognitive assessment and a behaviour rating scale | Disability and care need in residential and nursing care |
Dementia Questionnaire for People with Learning Disabilities (DLD) | Adapted for a pre-existing learning disability | Dementia in adults with learning disability, such as Down's syndrome |
The DLD suits an adult with Down's syndrome because general cognitive screens and the unadapted CAMDEX assume a typical premorbid baseline.
Interpret cognitive scores against premorbid ability, education, language, sensory needs and collateral history, and remember that a normal result does not exclude impairment.
Alcohol and substance instruments
Instrument | Structure | Use |
|---|---|---|
Alcohol Use Disorders Identification Test (AUDIT) | World Health Organization (WHO), 10 items on consumption, dependence symptoms and alcohol-related harm; 0 to 40, cut-off 8 | Built for primary care to detect hazardous and harmful drinking and possible dependence |
AUDIT-C | The first three AUDIT questions | Consumption screen |
Fast Alcohol Screening Test (FAST) | 4 AUDIT-derived items, including memory loss after drinking; 0 to 16, cut-off 3 or more | Brief screen, also used in the emergency department |
CAGE | 4 items on lifetime problems; 2 or more positive answers is significant | Brief screen, neither diagnostic nor an outcome measure; a positive result only prompts further assessment |
Paddington Alcohol Test | Links drinking to the current attendance | Brief screen built for emergency department attenders |
RAPS4 | 4 items: remorse, amnesia, performance and starter drinking | Screen for alcohol dependence |
Severity of Alcohol Dependence Questionnaire (SADQ) | Questionnaire | Severity of dependence |
Clinical Institute Withdrawal Assessment for Alcohol, revised (CIWA-Ar) | Withdrawal scale | Alcohol withdrawal |
Short Alcohol Withdrawal Scale (SAWS) | 12 or more is moderate to severe withdrawal needing medication; below 12 is mild | Alcohol withdrawal |
WHO ASSIST alcohol involvement score | 0 to 10 is low risk; the moderate band starts at 11 | Low risk calls for feedback, general health advice and routine monitoring |
AUDIT bands are 8 to 15 hazardous use, 16 to 19 harmful use and 20 or more possible dependence needing specialist assessment; the thresholds guide intervention rather than make a diagnosis.
The AUDIT is the best-validated screen across hazardous use, harmful use and dependence.
NICE recommends the AUDIT for identifying alcohol misuse, rather than CAGE or the MAST.
An opioid withdrawal scale does not belong among alcohol instruments such as the AUDIT, CAGE, SADQ and CIWA-Ar.
Other screening tools
The General Health Questionnaire (GHQ, Goldberg) is a brief, self-completed screen that compares recent symptoms with the person's usual state.
The GHQ counts probable cases of non-psychotic (common mental) disorder in community and primary care samples without making a diagnosis, and the 12-item version is usual in population surveys.
The GHQ-28 has 28 items, each with four responses scored 0 to 3 under Likert scoring, giving a maximum of 84.
The SCOFF questionnaire is a brief screen for anorexia nervosa and bulimia nervosa, designed for non-specialist and primary care settings.
The NODS-CLiP is the brief screening instrument for problem gambling.
Diagnostic interviews
These instruments differ by edition and classification. Older ICD-10 or DSM algorithms do not establish an ICD-11 diagnosis; apply the current clinical descriptions and diagnostic requirements separately.
Interview | Format | Interviewer | What it yields |
|---|---|---|---|
Present State Examination (PSE) | Semi-structured; keeps the character of a clinical mental state examination, using a glossary and free cross-examination | Clinician | Standardised rating of mental state symptoms |
SCAN | WHO successor to the PSE, built around the International Classification of Diseases (ICD), with PSE-10 as its core | Clinician | Has replaced the PSE in epidemiological research |
Composite International Diagnostic Interview (CIDI) | Fully structured, read verbatim, no clinical judgement; developed from the Diagnostic Interview Schedule | Lay interviewer | Lifetime ICD and DSM diagnoses; used in the World Mental Health Surveys |
Structured Clinical Interview for DSM (SCID) | Semi-structured; mirrors DSM criteria item by item | Trained interviewer, chiefly in research | Reliable DSM diagnosis |
International Personality Disorder Examination (IPDE) | Semi-structured interview | Interviewer, plus a self-report screen | Older categorical personality disorder diagnoses under ICD-10 or DSM criteria |
CATEGO is the computer program that turns PSE and SCAN ratings into syndromes and diagnostic classes by rule, removing variation in how clinicians combine symptoms.
Culture-bound syndromes
Syndrome | Where described | Features |
|---|---|---|
Dhat | Young South Asian men | Weakness, low energy and sexual difficulty blamed on semen draining away in the urine, semen being held a life-giving essence |
Koro | Mainly southern China and South East Asia | Acute terror that the penis is retracting into the abdomen and that full retraction will be fatal |
Latah | Mainly middle-aged Malay and Indonesian women | Exaggerated startle followed by echolalia, echopraxia, coprolalia, automatic obedience and a dazed state |
Taijin kyofusho | Japan | Fear that one's appearance, gaze, blushing or odour will offend or embarrass others |
Susto ("fright") | Latin American communities | A frightening event is thought to dislodge the soul, followed by sadness, poor sleep and appetite, listlessness and bodily complaints |
Ataque de nervios | Mostly Caribbean Latino people | Sudden uncontrollable shouting, crying, trembling, heat rising to the head and aggression, set off by a family stressor and followed by partial amnesia |
Taijin kyofusho differs from Western social anxiety in its focus: the fear is of offending others, not of one's own humiliation.
Recording a score
Record the instrument and version, date, rater, timeframe, total and maximum, useful subscales, limitations and the action taken, and explain what the result means for the person.
A falling total can coexist with persistent disability or rising risk, so assess these separately.