Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Clinical Psychological Assessment and Differential Diagnosis

Abstract. Clinical psychological assessment is an evidence-guided process of integrating symptom reports, observed behavior, developmental and temporal histories, functional consequences, medical and substance information, and contextual factors for clinical decisions. Differential diagnosis compares plausible explanations through discriminating observations and revisable hypotheses. This fundamental research monograph reconstructs the development of operational diagnosis and structured interviews, examines screening versus diagnostic use, reference standards, test agreement, Bayesian updating, longitudinal inference, and common clinical differential problems. It critically evaluates cultural and developmental validity, safety, functioning, clinical ethics, structural bias, and artificial decision support. Drawing on original methodological studies, diagnostic-accuracy research, professional standards, and WHO, APA, and NICE clinical guidance, the synthesis identifies the conditions under which diagnostic information is reliable, clinically useful, and responsibly interpreted.

Keywords: clinical psychological assessment; differential diagnosis; diagnostic interviews; evidence-based assessment; SCID; MINI; sensitivity and specificity; clinical decision-making; diagnostic validity; screening; ICD-11; DSM-5-TR; medical differential; cultural formulation; diagnostic uncertainty

Clinical Assessment as an Integrative Scientific Inference

Clinical psychological assessment is the systematic acquisition, interpretation, and integration of information about a person's symptoms, experience, behavior, functioning, development, and environmental context for a defined clinical purpose. Differential diagnosis is the comparative evaluation of plausible explanations for a presentation, including distinct mental disorders, co-occurring conditions, substance- or medication-associated states, physical illnesses, developmental phenomena, and responses to circumstances. The assessment process produces a reasoned clinical conclusion at a specified point in time, supported by sources whose reliability and relevance are made explicit. Clinical impressions, classifications, dimensional descriptions, functional formulations, and decisions about care are related outputs, yet each answers a different question. An effective assessment explains how the clinician moved from presenting concerns to a finite set of hypotheses, what evidence was sought to separate them, and which conclusions remain provisional. This object of inquiry connects the evidence-based assessment program articulated by Hunsley & Mash (2007) to the clinical purposes of the WHO (2024) guidance and the American Psychiatric Association (2022) diagnostic manual.

Assessment is fundamentally sequential. The clinician first establishes the referral question and immediate safety or medical priorities, then gathers a developmental and temporal account, examines present mental status, seeks corroborating records or informant observations when appropriate, and uses selected measures or structured interviews to refine hypotheses. As information is accumulated, the potential interpretation of earlier observations can change. Insomnia accompanied by fatigue may initially support several alternatives; a history of distinct periods of increased energy and reduced need for sleep changes the diagnostic comparison. A symptom score represents information about a construct within a reference population, whereas diagnosis requires interpretation of course, duration, impairment, exclusion criteria, and sometimes contextual meanings. This is why a positive brief screen can justify further investigation without establishing that the screened disorder is present. Kroenke et al. (2001), Levis et al. (2019), and NICE (2022) describe distinct roles of symptom measurement, diagnostic accuracy, and contextual clinical evaluation.

The present monograph examines the science of clinical assessment as a methodological system: the design and reliability of interviews, the information value of screening and observation, hypothesis testing in differential diagnosis, the interpretation of reference standards, comparative reasoning about common clinical presentations, contextual validity, diagnostic communication, and prospective reassessment. Its scientific boundary is explicit. General psychometric theory is treated where it changes decisions, while comprehensive scale construction belongs to Psychometric Measurement of Mental Processes. The taxonomy of disorders is used as a decision framework, while its theoretical critique belongs to Psychiatric Classification and Diagnostic Boundaries. Etiological explanation and treatment choice are addressed only as consequences of assessment quality. The aim is to characterize what can be known, by which procedures, and with what uncertainty in a clinical encounter.

From Descriptive Psychiatry to Operational Diagnostic Research

Modern diagnostic assessment developed through a succession of attempts to make clinical descriptions reproducible across observers and institutions. Earlier psychiatric traditions recorded patterns of behavior, mood, cognition, and longitudinal course but often used categories whose interpretation varied by school and clinician. In an influential validity framework, Robins & Guze (1970) proposed evaluating a diagnostic construct using clinical description, laboratory investigations where applicable, exclusion or delimitation from other disorders, follow-up, and family studies. These validators were not a mechanical checklist by which every condition became certain; they formed a research program for asking whether a proposed category cohered across independent forms of evidence. Their historical importance is the explicit demand to test diagnostic distinctions rather than simply reproduce them. Clinical assessment was thus linked to longitudinal observation and external validity, not just to agreement about symptom vocabulary.

The diagnostic criteria reported by Feighner et al. (1972) continued this research trajectory through explicit operational rules intended to improve comparability across studies. Subsequently, the Schedule for Affective Disorders and Schizophrenia, described by Endicott & Spitzer (1978), provided a structured way to obtain relevant symptom and history information. The Structured Clinical Interview for DSM-III-R later formalized an interviewer-led route through specific diagnostic criteria Spitzer et al. (1992). These tools demonstrate that standardization can reduce variability in the questions clinicians ask and in how evidence is recorded. Their existence does not imply that all relevant phenomena can be captured by a single protocol. The interview's reference period, respondent recall, interviewer expertise, cultural fit, and clinical setting remain substantive parts of the measurement process.

The contemporary diagnostic landscape includes DSM-5-TR, published by the American Psychiatric Association in 2022, and the ICD-11 Clinical Descriptions and Diagnostic Requirements issued by the World Health Organization in 2024 American Psychiatric Association (2022) WHO (2024). These sources are designed for clinical diagnosis while operating within their respective frameworks and revision histories. The WHO reports that ICD-11 entered into effect for health reporting in January 2022 and that the 2024 CDDR brings clinically oriented guidance into its mental, behavioral, and neurodevelopmental chapter. Any report must specify which classification edition, criterion set, language version, and jurisdictional requirements were used; a code copied from one system can fail to convey the same threshold or clinical meaning in another. The historical movement toward operationalization therefore creates an obligation for precise version control in research and practice.

The Epistemic Architecture of a Diagnostic Conclusion

A diagnostic conclusion is an inference from observations to a model of the individual's condition. The observations may include reported experiences, witnessed behavior, test responses, historical records, physiological findings, and longitudinal change. The model may be categorical, dimensional, functional, etiological, or several of these at once. The inferential problem is that many signs are nonspecific: reduced concentration may accompany depression, anxiety, sleep deprivation, medication effects, neurological conditions, and ordinary overload. A single sign therefore has limited discriminating value until its time course, accompanying features, base rate, and context are examined. A responsible conclusion makes the link between evidence and interpretation traceable. It names the discriminating observations, explains why alternatives have become less plausible, and identifies the information still needed to resolve the remaining uncertainty.

The formal distinction between reliability and validity structures this inquiry. Reliability concerns the consistency of a judgment or measurement under specified conditions. Validity concerns whether the interpretation and use of information are justified for the intended purpose. Two clinicians may agree on an inaccurate category, and an unstable classification may contain clinically important information when its instability reflects real change in the patient's state. The tradition of construct validity and contemporary testing standards insist that interpretations be evaluated against convergent and discriminant evidence, response processes, and consequential uses AERA, APA, & NCME (2014). Field trials of DSM-5 evaluated test–retest reliability in clinical settings Regier et al. (2013), but agreement alone cannot establish the causal nature of a condition or the appropriateness of every treatment inference. Reports must preserve these separate evidence questions.

A practical assessment can be represented as a structured chain: referral question, source selection, observations, temporal reconstruction, candidate hypotheses, discriminating tests, classification and functional interpretation, communication, and review. This chain creates opportunities to audit failures. An interview can omit a critical historical domain; a questionnaire can be applied outside its validated population; a collateral report can reflect a different observation context; a medical explanation can be insufficiently investigated; an interpretation can overstate certainty. Importantly, the chain is revisable. Later follow-up may clarify episodic patterns that were invisible at an initial appointment. Reassessment should be treated as an ordinary feature of diagnosis, especially when longitudinal course is itself part of a condition's definition.

Clinical Utility, Decision Consequences, and the Meaning of Accuracy

Assessment quality has at least four interacting dimensions: diagnostic accuracy, reliability, clinical utility, and fairness. Accuracy concerns correspondence between a decision and a defensible target or reference process; reliability concerns reproducibility; clinical utility asks whether the information improves an action that matters to patients and services; fairness examines systematic differences in error or access across groups. These dimensions may diverge. A lengthy interview can increase completeness but become impractical in emergency or low-resource settings. A quick screen can identify more people for further evaluation yet produce additional false positives. A categorical label may facilitate communication while underdescribing disability, strengths, culture, and personal priorities. Evidence-based assessment therefore evaluates instruments as elements within decision pathways rather than as autonomous diagnostic authorities Hunsley & Mash (2007) AERA, APA, & NCME (2014).

The consequences of error are asymmetric and context-specific. Missing an acute medical cause of altered consciousness can carry immediate danger, while an unwarranted enduring psychiatric label may affect self-concept, eligibility, treatment exposure, and records over years. In other settings, delaying recognition of a treatable mental health condition can cause prolonged impairment. The appropriate threshold for seeking more information depends on prevalence, consequence of false reassurance, consequence of false alarm, available follow-up, and whether a positive result leads to effective services. Screening trials and diagnostic-accuracy studies answer different questions. The US Preventive Services Task Force (2023) recommendation for depression screening in adults assumes systems for diagnosis, treatment, and follow-up; it cannot be generalized into the claim that administering a questionnaire is itself a complete care pathway.

Clinical utility also depends on matching assessment to the actual question. A consultation about current safety requires immediate and potentially repeated evaluation of risk and protective factors. A differential diagnosis involving persistent attention problems requires developmental chronology and evidence across settings. A question about treatment response requires repeated comparable observations and assessment of adherence, stressors, medical change, and adverse experiences. These projects share tools, but they demand different comparison standards and time horizons. The scientific challenge is to select a sufficient, proportionate assessment process and to document why its evidence is adequate for its particular use.

Scope of Evidence and Method of This Research Synthesis

This monograph is a critical narrative scientific synthesis grounded in original interview-development studies, validity and reliability research, diagnostic-accuracy meta-analyses, methodological guidance, and clinical recommendations. It brings together historical work such as Robins & Guze (1970), Feighner et al. (1972), and Endicott & Spitzer (1978); structured-interview research such as Spitzer et al. (1992), Sheehan et al. (1998), and Kessler & Üstün (2004); studies of brief screening including Kroenke et al. (2001), Spitzer et al. (2006), and Levis et al. (2019); and formal clinical frameworks including WHO (2024) and American Psychiatric Association (2022). The sources represent different designs and answer distinct questions. This is neither a preregistered systematic review nor a meta-analysis; no pooled effect estimates or comprehensive search coverage are asserted. Quantitative claims are reserved for directly identifiable studies with a traceable reference.

The literature is interpreted through an evidence matrix organized by study question: agreement, criterion or construct validity, clinical decision impact, predictive value, and equity or transportability. The review asks whether a study's sample and setting resemble the decision context; whether the reference standard is independent of the instrument under test; whether interviewers are blinded when necessary; whether missing results are accounted for; and whether test characteristics were obtained in the relevant clinical spectrum. Diagnostic findings cannot be made more secure through numerical precision alone when the target construct or reference comparison is unstable. Findings from primary care, specialist outpatient services, epidemiological surveys, and emergency practice are considered separately where their populations differ.

The synthesis maintains four attribution levels. A classification rule states the operational definition adopted by DSM or ICD. An empirical result describes observations under a particular design. A clinical recommendation expresses a practice judgment that may combine evidence, feasibility, and values. A conceptual claim explains how these levels relate and remains open to critical testing. The chapters follow the entire assessment pathway, then examine contrasting diagnostic problems, vulnerable populations, and the implications of digital and artificial systems. This organization keeps the subject of assessment and differential reasoning central even where the clinical examples vary.

The Referral Question and the Architecture of a Clinical Encounter

A scientifically useful assessment begins by specifying who seeks the evaluation, which decision depends on it, and what time horizon the decision concerns. A referral framed as “rule out depression” may conceal a wider problem involving occupational impairment, a change in sleep, an unexplained medical symptom, or disagreement between a patient and a referring institution. Restating the question in observable terms creates an explicit scope: current symptom pattern, historical course, functioning, possible underlying conditions, and the disposition to be reached. It also prevents a favored hypothesis from determining the questions asked. Evidence-based assessment favors a staged process that applies more intensive measures to unresolved hypotheses rather than distributing a large test battery indiscriminately Hunsley & Mash (2007).

Early clarification includes consent, confidentiality and its limitations, language needs, the availability of interpreters, the role of collateral sources, and the patient's own account of what would constitute a useful outcome. The assessment context may be routine outpatient care, crisis services, primary care, consultation-liaison medicine, telehealth, or disability evaluation; each creates different selection effects. A high-prevalence specialist setting and a low-prevalence general population differ in the predictive value of the same screen. The presence of referral pressure from a school, employer, court, or family further requires documentation of whose questions are being answered and how the report may be used. Professional ethics concerning assessment competence, consent, appropriate instruments, and interpretation apply to these decisions American Psychological Association (2017).

A competent intake does not postpone a time-sensitive concern until the end of a formal protocol. Acute changes in consciousness, confusion, severe agitation, possible intoxication or withdrawal, urgent threats to safety, and rapidly evolving neurological signs call for appropriate same-day clinical or medical evaluation in their setting. These priorities influence the ordering of questions and the degree to which a conventional symptom interview can be trusted. In unstable conditions, the immediate objective can be stabilization and a working differential rather than a final psychiatric classification. Once the patient can participate adequately, the broader evaluation may resume with explicit documentation of why some data were deferred. The NICE (2010/2023) guidance emphasizes detection of acute changes and appropriate assessment when delirium is possible.

Clinical Interviewing: Narrative, Structure, and Diagnostic Depth

The clinical interview generates information through a relationship between the interviewer, the respondent, the questions, and the setting. Open-ended inquiry allows patients to describe experiences in their own sequence and language; targeted follow-up then clarifies phenomenology, frequency, severity, duration, triggers, consequences, and explanatory context. These stages are complementary. A highly constrained question can impose assumptions that distort the respondent's account, whereas a wholly unstructured conversation can omit essential exclusions or criterion requirements. The scientific design of the interview therefore concerns both elicitation quality and standardization. The history from the Schedule for Affective Disorders and Schizophrenia Endicott & Spitzer (1978) to the Structured Clinical Interview Spitzer et al. (1992) illustrates the effort to make diagnostic domains explicit while preserving opportunities to probe ambiguous answers.

Structured interviews prescribe items and decision rules; semistructured interviews guide the clinician through required content while allowing elaboration and clinical judgment. The SCID family is closely tied to DSM criteria, while the Mini-International Neuropsychiatric Interview was designed as a comparatively brief structured interview for DSM-IV and ICD-10 diagnostic research Sheehan et al. (1998). The World Mental Health Composite International Diagnostic Interview was developed for epidemiological application with standardized procedures Kessler & Üstün (2004). None of these instruments should be treated as interchangeable merely because each yields diagnostic information. They differ in populations, training expectations, branching logic, time burden, reference periods, and intended uses. Clinicians should select the version and administration method supported for their particular task and document deviations that may alter interpretation.

Good interviewing involves probing for examples rather than accepting criterion words at face value. “Racing thoughts” can describe anxiety, emotional arousal, fast associative thinking, or manic activation; interpretation requires chronology, sleep–energy relationship, observed speech, and degree of behavioral change. “Hearing a voice” can refer to intrusive thought, ordinary inner speech, culturally interpreted experience, hypnagogic phenomena, or psychotic symptoms. “Panic attack” may be used colloquially for many episodes of fear. Clinicians can improve the fidelity of assessment by asking what happened, how long it lasted, what the person believed at the time, and what occurred before and after. The aim is an adequately specified phenomenon from which a classification decision can be reasoned, rather than a vocabulary match between patient and criterion list.

Temporal Reconstruction and Longitudinal Diagnostic Reasoning

Symptoms acquire diagnostic significance within a trajectory. Onset, persistence, recurrence, remission, and relation to developmental transitions or external events often discriminate between otherwise similar descriptions. A temporal reconstruction begins with earliest changes, includes a baseline before the current presentation, and distinguishes persistent traits from episodes. It places symptoms alongside sleep, illnesses, medication exposure, substance use, reproductive or endocrine changes when relevant, accidents, stressors, and changes in relationships or functioning. The timeline should be detailed enough to evaluate whether two variables merely coincided or whether their sequence is compatible with a causal hypothesis. Temporal proximity increases the plausibility of a medication-associated condition but does not automatically establish it.

Retrospective histories are vulnerable to incomplete recall, telescoping, current-mood effects, and reinterpretation of earlier behavior. Follow-up records and contemporaneous notes can improve chronology but are themselves selective observations. Informants may remember highly visible behavior better than private distress and may observe the person in only one environment. Cross-setting discrepancies therefore constitute data to interpret rather than errors to suppress. A young person's attention difficulties that emerge only after a recent sleep disruption differ in evidentiary meaning from a long developmental history of similar difficulties observed at school and home. The most useful timeline describes episodes, exposure changes, and functioning independently before applying diagnostic labels.

Longitudinal observation is especially important for bipolar-spectrum questions, psychotic episodes, recurrent depression, and emerging neurocognitive disorders. The NICE (2014/2025) guideline directs attention to past episodes of elevated or irritable mood, overactivity, course, impairment, and collateral information when appropriate. The validity tradition described by Robins & Guze (1970) similarly recognizes follow-up as an independent route for evaluating proposed diagnostic distinctions. A working formulation may therefore retain multiple hypotheses at an early encounter. When evidence is incomplete, the clinician can specify what future observations would change the conclusion: persistence after removal of an exposure, recurrence of a distinctive episode, reliable records of prior function, or improvement of an acute medical state.

Mental Status Examination and Behavioral Observation

The mental status examination provides a structured description of the person's observable and reported current state. Domains typically include appearance and behavior, level of alertness, speech, affect, reported mood, form and content of thought, perception, orientation, attention, memory, insight, and judgment, with variations determined by setting and population. Its special contribution is the distinction between a report about a period of time and observation during an encounter. A patient can report severe insomnia while appearing alert; another can report no cognitive concern yet display fluctuating attention. Such discrepancies may guide further inquiry. Observation should be recorded in concrete language that another clinician can interpret: response latency, continuity of speech, ability to maintain a question, or change in attention over the interview.

Current mental status is also an imperfect sampling frame. A single encounter can miss intermittent symptoms, performative masking, situational anxiety, and impairments that emerge only under particular cognitive or social demands. Interpersonal behavior differs across unfamiliar clinical settings, culturally specific communication norms, and sensory environments. Recording observed facts separately from interpretive labels reduces the risk of turning a brief impression into a global trait. For example, limited eye contact alone is a weak foundation for conclusions about social motivation or psychotic illness. Interpretation depends on developmental, cultural, sensory, and situational evidence and on the complete differential question Kirmayer (2012) NICE (2012).

When altered consciousness or rapid fluctuation is suspected, observational findings need timely medical integration. Delirium may involve acute changes in attention, awareness, cognition, and sleep–wake patterns, sometimes with variability across hours. The NICE (2010/2023) guidance supports formal evaluation of possible delirium with relevant tools in the appropriate care setting. A psychological interview may document abnormal attention but cannot replace medical investigation of its cause. Likewise, severe disorganization or agitation may impair the validity of self-report measures administered at that moment. The state of the respondent during testing is a substantive condition of any interpretation.

Multiple Informants, Records, and the Problem of Disagreement

Multisource assessment combines patient report, clinician observation, prior clinical records, developmental history, direct measurement, and collateral information when lawful and appropriate. These sources have different access to events. Patients have privileged access to many subjective experiences; observers may detect behaviors or functional consequences that patients have not noticed; medical records can establish chronology of diagnoses, injuries, treatments, or laboratory findings; repeated measures may reveal change outside the interview. Their convergence increases confidence when the sources are partly independent. Their disagreement can also expose context dependence, differences in recall periods, or contrasting operational definitions. Assessment gains depth by analyzing such differences instead of forcing them into a single score.

Collateral information requires particular care with consent, confidentiality, independence, and potential conflicts of interest. A family member may be a reliable observer of sleep, expenditure, activity, or everyday functioning while having limited knowledge of internal anxiety or trauma. An employer's concern may be shaped by workplace requirements, and school reports may depend on instructional conditions. The clinician should specify what each source observed, when, and under what conditions. A report that “the patient is impulsive” is less useful than a description of concrete behavior, its frequency, and changes over time. A missing or unavailable informant should be documented as an evidence limitation, not treated as proof that a suspected condition is absent.

The practice of triangulation is strongest when an observation from one source makes a testable prediction about another. If a patient reports abrupt episodes of extreme activity, the chronology predicts possible changes in sleep, communication, and observable functioning that an independent record might corroborate. If cognitive complaints are prominent, contextualized tests may help distinguish objective deficits, fatigue, and subjective distress. A source should be weighted by the question it can answer, its sampling limitations, and the likelihood that it has been influenced by the same prior label. In this way, a multisource assessment remains an inferential process rather than an exercise in counting agreeing witnesses Hunsley & Mash (2007).

Physical Health, Medication, and Substance-Associated Explanations

A differential diagnostic process considers medical and substance-related explanations whenever their features or chronology make them plausible. Endocrine disease, neurological conditions, inflammatory illness, sleep disorders, pain, adverse medication effects, intoxication, withdrawal, and physiological stress can generate mood, perceptual, cognitive, or behavioral complaints. The appropriate medical workup is selected by the qualified health professional based on history, examination, acuity, and local standards; a psychological evaluator should identify the need for such evaluation without turning every presentation into an indiscriminate laboratory panel. The NICE (2014/2025) recommendations explicitly include drug and alcohol misuse and physical conditions in bipolar differential assessment. The WHO (2024) diagnostic framework similarly places mental disorders within a broader clinical evaluation.

Medication review must go beyond a static list of prescriptions. The clinically meaningful questions concern changes in dose or adherence, recent initiation or discontinuation, interactions, timing relative to symptoms, over-the-counter and nonprescribed agents, and whether use was regular or episodic. Substance histories similarly require attention to route, frequency, pattern, tolerance, withdrawal, and the temporal relation between exposure and symptom onset. Confidential, nonjudgmental interviewing improves the chance that relevant information will be volunteered. For a patient whose anxiety intensifies after a new medication and resolves after a clinically supervised change, the trajectory has a different inferential meaning from long-standing panic occurring independently of that exposure. Even then, coexisting disorders remain possible.

Differential diagnosis must also distinguish psychiatric presentations from medical emergencies. Abrupt confusion and fluctuating attention deserve a different pathway from chronic worry; acute changes in speech or neurological function require appropriate urgent medical attention. Testing ordered or interpreted outside professional scope creates clinical and ethical risk. The assessment record should therefore identify which medical questions have been resolved, which specialist findings were reviewed, and which possibilities remain under investigation. An assessment can be high quality while explicitly deferring an etiological conclusion that depends on unavailable medical evidence.

Screening, Diagnostic Assessment, and Symptom Severity as Separate Tasks

Screening attempts to identify people who may benefit from further assessment. Diagnostic assessment evaluates whether a specified condition is present under a given classification framework, while symptom measurement characterizes severity or change on a continuous scale. Prognostic assessment estimates a future outcome conditional on currently observed variables. These purposes call for different reference standards, threshold choices, and validation studies. The Patient Health Questionnaire family illustrates the distinction. Spitzer et al. (1999) evaluated a self-report approach to common mental disorders in primary care; Kroenke et al. (2001) evaluated the PHQ-9 as a depression severity measure; and Levis et al. (2019) used individual participant data to examine the accuracy of PHQ-9 thresholds for detecting major depression against interview reference standards. The studies' different objectives explain why one instrument can be useful in several ways without its score becoming equivalent to a diagnosis.

A brief screen may be administered to patients with little prior indication of the target condition, or to patients in a high-symptom clinical service. The probabilities that follow a positive result differ considerably across those groups. A threshold that performs acceptably when the clinical aim is broad case finding may produce too many false positives for a definitive classification decision. Conversely, a high threshold may miss clinically important subthreshold distress or presentations that require specialized questioning. The US Preventive Services Task Force (2023) adult depression-screening recommendation is a recommendation about a care process that includes diagnostic evaluation, treatment capability, and follow-up. Its separate assessment of evidence for suicide-risk screening in asymptomatic adults should not be misread as a directive to ignore disclosed risk or other clinical indications for a safety evaluation.

Severity scales remain valuable after a clinical diagnosis because they provide a repeatable record of how symptoms change. The Spitzer et al. (2006) GAD-7 study described a brief scale for generalized anxiety symptoms in primary care. Yet serial measurements must be interpreted in the light of measurement error, changing context, treatment exposure, patient priorities, and functional outcomes. A falling anxiety score could coexist with avoidance that leaves a person's life constricted; a stable score could mask changes in sleep, substance use, or interpersonal functioning. Scores therefore enter the assessment as fallible indicators of selected domains. Their clinical meaning depends on the purpose, the population, and the hypotheses being tested.

Diagnostic Accuracy, Base Rates, and Bayesian Updating

The interpretation of a diagnostic test starts with its intended target. Sensitivity is the proportion of people meeting the reference definition who have a positive result; specificity is the proportion who do not meet the reference definition and test negative. Positive and negative predictive values describe the proportions of positive and negative test results that correspond to the reference outcome within the tested population. Predictive values depend on prevalence. For example, at a hypothetical sensitivity of 0.80, specificity of 0.90, and prevalence of 0.05, a cohort of 1,000 would contain approximately 40 true positives, 10 false negatives, 95 false positives, and 855 true negatives. The positive predictive value would be about 30%; a positive result would therefore justify follow-up rather than amount to near-certainty. These values are an illustration derived from the stated assumptions, not a finding attributed to a particular questionnaire.

Likelihood ratios provide a way to summarize how a result changes odds. The positive likelihood ratio is sensitivity divided by one minus specificity; the negative ratio is one minus sensitivity divided by specificity. Prior odds multiplied by the relevant likelihood ratio yields updated odds when assumptions are met. This formalism clarifies that the diagnostic information of a score depends on the credible probability before testing. The prior estimate cannot be taken automatically from a convenience sample, and the test's operating characteristics may change across populations and severity spectra. Evidence reviews such as Levis et al. (2019) are valuable partly because they examine accuracy in relation to clinical reference interviews, thresholds, and sample variation rather than presenting a single score as universally decisive.

There are important limitations to simple sequential updating. Symptoms measured by two questionnaires may be highly correlated; multiplying both likelihood ratios as though they were conditionally independent overstates their information. Interview questions may have been used to establish the reference diagnosis, creating incorporation bias. Referral and verification procedures can leave some test-negative participants without full diagnostic evaluation. Age, language, medical comorbidity, and setting alter the range of presentations to which an instrument is exposed. Bayesian reasoning is therefore an organizing discipline, not a license to invent precise pretest probabilities. When relevant base rates or accuracy estimates are unavailable, qualitative updating with explicit uncertainty is preferable to pseudoquantitative certainty.

Reference Standards and the Risk of Circular Validation

Diagnostic-accuracy studies require a comparator that is sufficiently independent, reproducible, and appropriate for the intended classification. In mental health research that comparator is frequently a trained clinician's structured or semistructured diagnostic interview. Such an interview is a clinical reference standard rather than a biological gold standard. It relies on symptom reports, judgment about threshold and impairment, classification rules, and the time frame of assessment. Some conditions lack an independently verifiable single marker; agreement with the interview therefore establishes agreement with an operational criterion set, not proof that the instrument isolates an underlying disease entity. This distinction applies particularly when evaluating brief questionnaires against SCID-derived judgments or comparable interviews Sheehan et al. (1998) Spitzer et al. (1992).

A reference interview should be administered by appropriately trained examiners who are unaware of the index test result where blinding is needed to limit expectation bias. Ideally the study evaluates an appropriately sampled spectrum of cases, ensures comparable assessment timing, reports exclusions and indeterminate results, and includes follow-up when the clinical question concerns course or prognosis. If the same symptom items determine both the index result and the reference classification, independence is limited. If only positive screens are interviewed, sensitivity and specificity may be distorted by verification bias. A research report must disclose these design features before an accuracy estimate is transported into practice.

Agreement statistics are likewise conditional on their design. Simple percentage agreement does not adjust for agreement expected under a specified chance model. Cohen's kappa is sensitive to category prevalence and marginal distributions, so a low value can arise in studies where a rare category is nearly always absent even when raw agreement appears high. Weighted kappa, intraclass correlation, and other indices answer different questions, especially when severity is ordinal or continuous Koo & Li (2016). The DSM-5 field-trial investigation reported by Regier et al. (2013) illustrates the difficulty of evaluating test–retest diagnostic agreement across heterogeneous real-world clinical settings. A scientific appraisal should show the estimate, confidence interval where reported, case mix, assessment interval, and plausible reasons for disagreement rather than merely labeling reliability “good” or “poor.”

Empirical Studies of Diagnostic Practice and Their Limitations

A meta-analysis by Mitchell et al. (2009) examined how accurately primary-care practitioners identified depression relative to research-based assessments. The authors' synthesis illustrated the substantial tension between recognition of true cases and false-positive identification in routine practice. Its practical lesson concerns the setting and reference procedures, not an accusation that individual clinicians are inherently incapable of diagnosis. Studies of routine practice often combine clinicians with different amounts of time, case information, access to longitudinal follow-up, and criteria for recognizing depression. Each of these can change performance. Estimates from one health system or era should not be used as fixed properties of clinical judgment in all services.

The individual participant data meta-analysis by Levis et al. (2019) examined PHQ-9 accuracy and the consequences of threshold selection when a diagnostic interview served as reference. The analysis helps distinguish screening performance from clinical diagnosis and also demonstrates why pooled estimates need attention to interview type and cutoffs. Accuracy depends on who enters a study and on when the symptom scale and the reference interview are administered. A report comparing different screening procedures should examine whether all participants received the same adequate reference assessment and whether the cutoff was prespecified. Post hoc selection of the threshold that maximizes accuracy in one sample is vulnerable to optimistic performance estimates.

The research tradition of evidence-based assessment gives stronger weight to well-described clinical populations, independent validation, comparisons with plausible alternatives, and tests of actual decision benefit Hunsley & Mash (2007). Even high diagnostic accuracy may yield limited improvement if the information never reaches a clinician, if positive results cannot obtain confirmatory assessment, or if treatment pathways are unavailable. Conversely, careful assessment can be valuable by resolving a medical alternative, reducing an inappropriate label, documenting functioning, or improving a shared decision despite a result that lacks a perfect categorical match. The scientifically meaningful endpoint is the quality of decisions and outcomes under specified conditions, rather than the prestige of an instrument or the amount of data collected.

Clinical Judgment, Statistical Prediction, and Cognitive Bias

Diagnostic work requires both flexible interpretation of complex phenomena and discipline about predictable errors. Clinical judgment permits inquiry into unusual events, context, and exceptions to standardized patterns, but unstructured judgment can be influenced by anchoring, availability, confirmation bias, halo effects, diagnostic momentum, and premature closure. Croskerry (2003) examined cognitive sources of diagnostic error in medicine and strategies for limiting them; the conceptual mechanisms are relevant to clinical psychological assessment even though diagnostic evidence is discipline-specific. A clinician who receives a prior label of panic disorder may interpret all later chest sensations through that frame, overlooking a new medical problem or a medication effect. Reviewing alternative explanations at defined points in the encounter helps preserve a broader differential.

The tradition established by Meehl (1954) and revisited in the meta-analysis of Grove et al. (2000) compares informal clinical prediction with standardized mechanical combinations of information. These works concern prediction under particular evidence sets and outcomes. Their implications should be applied carefully: a validated algorithm may integrate known predictors consistently, but an algorithm trained on incomplete records or a mismatched population can be systematically wrong. A clinician may supply crucial observations missing from a predictor; a statistical model may expose a pattern that intuition underweights. The correct comparison therefore evaluates performance for a given task in an independent population, while retaining professional responsibility for the meaning, fairness, and safety of the decision.

Concrete debiasing procedures include recording the initial problem without a diagnosis, generating more than one plausible hypothesis, identifying a disconfirming observation for the leading alternative, checking criteria and exclusions, and conducting a brief review before final sign-off. A diagnostic time-out is especially appropriate when a severe condition is proposed with limited history or when the observed presentation conflicts with the existing record. Calibration is the long-term relation between expressed confidence and actual correctness, and requires outcome feedback or follow-up cases rather than a feeling of certainty. These practices operationalize the broader goal of auditable inference: clinicians can explain what they believed, what they observed, what changed their judgment, and what remains to be checked.

Longitudinal Assessment, Treatment Response, and Reassessment

Assessment continues after the first conclusion. Clinical conditions change; diagnostic definitions themselves may require evidence of duration or recurrence; and intervention can reveal patterns that were previously obscured. Follow-up makes it possible to examine the stability of a syndrome, emergent alternatives, functional recovery, treatment adverse effects, and evolving social circumstances. Repeating a symptom scale in a consistent way can reveal direction of change, but interpretation is strengthened by repeated interviews, functioning measures, and a record of new exposures. The WHO (2010) framework provides one example of a standardized approach to describing activity limitations and participation across areas of life. Assessment of functioning complements categorical diagnosis because people with the same category can have very different patterns of disability and adaptation.

A change after treatment cannot automatically confirm the original diagnosis. Improvement with a nonspecific intervention may occur across multiple disorders or through changes in sleep, expectation, interpersonal support, or regression to the mean. Similarly, the absence of improvement may reflect implementation problems, competing conditions, or a mistaken causal account. Interpretation of treatment response requires an explicit comparison between competing explanations and review of outcomes that mattered at baseline. Care plans should state what evidence would trigger reassessment of the diagnosis, the formulation, or the chosen treatment pathway. Relevant practice recommendations for depression NICE (2022) make assessment of course, treatment context, and ongoing monitoring part of care rather than a one-time administrative event.

A high-quality longitudinal record contains a reproducible baseline description, dates of change, clearly described outcomes, and the reason for each updated judgment. This supports transfer of care and reduces the tendency for a historical label to become self-validating. Where the underlying evidence remains mixed, the record can state that the presentation is best understood provisionally and name the most important rival hypotheses. Diagnostic revision is an expected consequence of new information, especially when a disorder's defining pattern unfolds across time. The ethical and scientific task is to make revision explainable to patients and other professionals.

Depressive Presentations and the Bipolar Differential

The initial complaint of depressed mood, loss of interest, fatigue, sleep change, or impaired concentration supports a broad differential. A clinically adequate inquiry evaluates the temporal clustering of symptoms, associated impairment, prior episodes, bereavement or other stressors, medications, substance exposure, physical health, and safety. The distinction between a depressive syndrome and a current major depressive episode depends on criteria and course within the chosen classification system American Psychiatric Association (2022) WHO (2024). Symptom severity scales can inform this inquiry, but somatic overlap is consequential: chronic pain, sleep apnea, neurological disease, endocrine disturbance, and medication effects can influence many items used in depression measures. Evaluators should record whether criteria are met through an interview that addresses the full picture rather than assuming that a high questionnaire score resolves the differential.

The clinically decisive question in many recurrent depressive presentations is whether the lifetime history includes an episode compatible with mania or hypomania. The NICE (2014/2025) guideline recommends examining earlier periods of overactivity or disinhibited behavior and considering the broader differential, including substance use, ADHD, personality patterns, and thyroid disorders. Asking only “Have you ever been manic?” is poorly suited to discovery because patients may not recognize the term or may regard an elevated period as productive rather than impairing. Concrete inquiry addresses distinctly increased energy, reduced need for sleep, mood change, speech, goal-directed activity, risk taking, duration, and observable change from baseline. Collateral accounts can supply missing behavioral chronology, especially for episodes accompanied by impaired insight.

Classification must also accommodate co-occurrence. An individual may meet criteria for a mood disorder and simultaneously experience anxiety, trauma-related symptoms, problematic substance use, or a physical condition. A single dominant label can obscure treatable complications, while indiscriminate accumulation of diagnoses may obscure shared symptom sources. The diagnostic record should distinguish confirmed episodes from suspected past episodes and identify whether an apparent activation was temporally associated with a medication, an intoxication state, or an independently recurring mood pattern. The recommendation is a disciplined historical differential, not an assumption that every depressive patient belongs to a bipolar spectrum.

Anxiety, Panic, Somatic Symptoms, and Medical Mimics

Anxiety complaints may encompass persistent worry, sudden surges of fear, avoidance, bodily alarm, intrusive thoughts, trauma responses, or episodes produced by medical and substance-related conditions. The differential begins by describing the event before assigning an anxiety-disorder label. For panic-like episodes, clinicians ask whether fear rose abruptly, which physical sensations occurred, what circumstances were present, and whether subsequent behavioral changes or anticipatory worry developed. Similar sensations can occur in cardiac arrhythmias, respiratory illness, endocrine disorders, medication effects, and substance intoxication or withdrawal. The implication is a careful medical history and appropriate clinical investigation when the presentation warrants it. Psychological diagnosis and medical evaluation can proceed cooperatively rather than in competition WHO (2024).

The temporal relationship between fear and bodily sensation can aid formulation but rarely supplies a conclusive physiological test. Some patients become frightened after noticing palpitations; others report fear before somatic arousal, and retrospective accounts can compress these sequences. Anxiety measures such as the GAD-7 Spitzer et al. (2006) describe frequency of selected symptoms during a recent period. They do not independently distinguish generalized anxiety disorder from depressive comorbidity, stress reactions, panic disorder, or medical illness. Assessment therefore adds the distribution of worry across domains, its perceived controllability, duration, impairment, avoidance, and the individual's explanations. Culture and health literacy influence whether distress is described primarily through emotion, social concern, or bodily symptoms Kirmayer (2012).

A strong differential also tests whether repetitive thinking represents worry, obsession, depressive rumination, delusional conviction, or a problem-solving attempt under chronic uncertainty. The question is phenomenological: how intrusive the thought is, what the person believes about it, how they respond, and whether rituals or avoidance follow. A report of recurrent physical symptoms should not be dismissed as psychological solely because ordinary tests are unrevealing; symptom meaning, prior investigations, health anxiety, and genuine disease may coexist. The clinical aim is to identify the best-supported set of explanations and the next observations needed, with careful attention to urgent physical red flags and functional consequences.

Obsessions, Compulsions, Psychosis, and Belief Conviction

Intrusive thoughts can be terrifying, repetitive, and difficult to control across obsessive-compulsive, trauma-related, depressive, psychotic, and other clinical presentations. The meaning of the experience depends on whether it is recognized as unwanted, how much conviction attaches to its content, whether neutralizing behaviors occur, and how much time and functioning the pattern consumes. Obsessions and compulsions may be mental as well as overt, and insight is variable. A rigid assumption that all people with obsessive-compulsive disorder always recognize their beliefs as unreasonable would mischaracterize the range of presentations addressed in contemporary classifications American Psychiatric Association (2022) WHO (2024). A useful assessment obtains concrete examples of the sequence from trigger to thought, affect, response, temporary consequence, and return of the symptom.

Differentiating obsessional concerns from delusional beliefs requires attention to phenomenology and longitudinal context. A belief can be strongly held because of cultural background, personal history, or immediate threat rather than psychiatric illness. Conversely, an idiosyncratic belief associated with impaired reality testing, perceptual disturbance, and disorganization warrants assessment within a possible psychotic syndrome. Clinicians should investigate broader functioning, duration, mood episodes, sleep, substances, neurological or medical causes, and relevant family or personal history. The NICE (2014) guidance emphasizes comprehensive assessment and engagement rather than a conclusion drawn from one unusual statement.

Apparent diagnostic boundaries are further complicated by comorbidity and changing insight. A person may experience both obsessions and psychotic symptoms, or a mood episode with psychotic features. Treatment implications often make the distinction clinically meaningful, yet the appropriate inference depends on the total pattern and should remain revisable. Evaluators can document what was directly reported, which aspects are corroborated, and whether the experience occurs only in specific states. In high-stakes settings, attention to immediate safety and appropriate specialist involvement takes priority over resolving every nosological detail in a single interview.

Trauma, Dissociation, Acute Stress, and the Meaning of Avoidance

Trauma-related presentations require an assessment that preserves symptom chronology, context, exposure history, functional consequences, and a patient's control over disclosure. Intrusive recollections, avoidance, arousal, negative mood, emotional numbing, and dissociative experiences can overlap with depressive, anxiety, sleep, psychotic, and neurological phenomena. The WHO (2024) CDDR includes clinical guidance for post-traumatic stress disorder and complex post-traumatic stress disorder; the DSM-5-TR American Psychiatric Association (2022) has its own diagnostic categories and criteria. A clinician must specify the framework in use and avoid treating superficially similar labels as interchangeable without checking their requirements. The assessment aim is to elicit sufficient information for the question while minimizing unnecessary repetition of potentially distressing detail.

Dissociation describes a family of experiences involving disruptions or alterations in normally integrated functions of consciousness, memory, identity, perception, or self-experience. Depersonalization and derealization can arise in several conditions and may also be reported during intense stress, sleep disruption, or substance exposure. An isolated report of feeling unreal should be characterized by duration, recurrence, associated fear, intactness of reality testing, and relevant exposures. Memory discontinuity requires particular caution because routine forgetting, intoxication, neurological causes, and contextual avoidance can resemble dissociative explanations. The nature of the reported memory gap and available independent corroboration are evidentiary issues, not rhetorical proof of any one cause.

Trauma-informed interviewing emphasizes predictability, consent, collaboration, and the person's capacity to pause or redirect an interview. These procedures have a clinical rationale and also improve information quality by reducing coercive conditions that interfere with disclosure. The evaluator should separate observed symptoms from assumptions about an event's exact mechanism or historical truth; a clinical symptom report is not, by itself, forensic proof of disputed facts. When trauma-focused histories are obtained in legal contexts, methods appropriate to evidentiary interviewing must be considered separately from treatment-oriented inquiry. Follow-up may clarify the development of avoidance and persistent functional impact.

Attention, Neurodevelopment, and the Lifespan Differential

Attention problems can indicate neurodevelopmental vulnerability, current emotional distress, sleep deprivation, substance use, chronic pain, acquired brain injury, medication effects, or environmental overload. A neurodevelopmental hypothesis generally requires attention to early developmental patterns, longstanding features, and manifestations across relevant settings, interpreted through current classification guidance American Psychiatric Association (2022) WHO (2024). Adult symptom endorsement alone does not reproduce the evidence that would have been observable earlier in life. Equally, imperfect childhood records do not automatically rule out a developmental condition. The assessment can work with documented episodes, educational or occupational history, retrospective informants, and the stability of complaints across contexts while transparently describing the limits of retrospective reconstruction.

Autism-related assessment illustrates the importance of developmental and cultural context. Social communication, sensory differences, restricted interests, adaptations to unfamiliar settings, and masking can affect what is observed in a brief interview. NICE (2012) emphasizes comprehensive assessment rather than reliance on a screening score, with consideration of childhood history, functioning, other mental health conditions, and alternative explanations. A behavior such as limited eye contact may be consistent with several interpretations and varies with culture, anxiety, sensory conditions, and personal choice. The evaluator should examine patterns and their developmental continuity, rather than making a categorical judgment from a small set of stereotyped signs.

In older adults, new problems with attention, memory, or behavior raise a different set of possibilities: delirium, depressive syndromes, neurocognitive disorders, medication effects, sensory loss, and acute physical illness. The distinction between gradual change and acute or fluctuating change is central, as described in NICE (2010/2023). A brief cognitive screen has to be interpreted in relation to education, language, sensory accessibility, baseline function, and practice effects. Neuropsychological testing may contribute a detailed profile, but inference about etiology often requires neurological and medical investigation. Across the lifespan, the assessment method must fit the developmental and physiological conditions of the person being evaluated.

Substance Use, Intoxication, Withdrawal, and Independent Disorders

Substances and medications can affect sleep, attention, mood, anxiety, perception, movement, and behavior through intoxication, withdrawal, long-term exposure, or interaction with health conditions. The diagnostic problem is chronological: whether symptoms arose during a plausible exposure window, preceded substance use, persisted during abstinence, recurred independently, or developed in parallel with another disorder. Reliable inquiry covers multiple substance classes, patterns of use and cessation, self-medication motives, past episodes, and collateral medical data when obtainable. A person who uses alcohol to manage anxiety may have an independent anxiety disorder, an alcohol-related condition, or both; chronology and careful follow-up help adjudicate the alternatives. The classifications and the WHO (2016) mhGAP guidance provide frameworks for clinically attending to such presentations.

The interpretation of symptoms during intoxication or withdrawal may be constrained by fluctuating awareness, inconsistent recall, and acute medical risk. In these circumstances the evaluator may record a syndrome or immediate concern without assigning a durable psychiatric diagnosis until the person can participate reliably and a relevant course has been observed. A second assessment after stabilization can reveal which symptoms persist. This does not make the first assessment useless: it can establish urgent needs, exposure details, and a plan for reassessment. It is important to distinguish a symptom's temporal association with substance use from proof of a specific pharmacological mechanism, particularly when multiple agents or medical comorbidities are present.

Substance-related explanations should be investigated without moral assumptions. Stigmatizing questions can reduce disclosure, and the act of obtaining information must respect confidentiality and the person's safety. A standardized form may facilitate completeness, but the quality of dialogue, the interpretation of local terms, and the possibility of inaccurate self-report remain relevant. Under conditions of uncertainty, the clinical record can state which diagnoses are being considered and what duration of observation or specialist examination would clarify them.

Personality Patterns, Acute States, and Transdiagnostic Comorbidity

Personality-related assessment concerns persistent patterns in self-functioning, affect regulation, interpersonal functioning, and behavior across contexts. Acute episodes of depression, mania, psychosis, intoxication, trauma response, or intense stress may temporarily change how these features appear. A diagnostic inference about enduring patterns is therefore stronger when it incorporates developmental history, longitudinal observations, and functioning across relationships and roles. The WHO (2024) CDDR includes a dimensional approach to personality disorder, reflecting the relevance of severity and trait domains. DSM-5-TR uses its own structures and criteria American Psychiatric Association (2022). Comparing these frameworks requires attention to which aspects are descriptive, which are diagnostic thresholds, and which relate to impairment or risk.

A crisis encounter is a particularly poor setting for assuming that current interpersonal behavior reveals a stable personality structure. Agitation, distress, mistrust, and fear can affect cooperation with evaluators; the institutional environment itself may amplify certain behaviors. Whenever possible, the assessment uses information outside the acute state and explains when longitudinal evidence is missing. The clinical record can identify features that require further observation without presenting them as confirmed traits. A differential should also ask whether symptom patterns are better accounted for by neurodevelopmental conditions, trauma-related responses, mood episodes, or combinations of conditions.

Comorbidity poses both conceptual and practical difficulties. Diagnoses can co-occur because their processes overlap, because risk factors are shared, because one condition influences another, or because thresholds divide correlated dimensions. Dimensional research such as Kotov et al. (2017) offers ways to analyze broad patterns of co-occurring symptoms, although a statistical hierarchy does not itself settle causal mechanisms. In an individual assessment, multiple classifications may be justified when each contributes meaningful information; alternatively, an overextended list can obscure the most important current processes and needs. The purpose of differential work is to record the evidence supporting each conclusion and its implications for a coherent care plan.

Cultural Formulation, Language, and Diagnostic Interpretation

Every clinical interview is conducted through a vocabulary for distress. Different languages and social traditions offer different distinctions among fear, sadness, bodily discomfort, dissociation, spiritual experience, and interpersonal obligation. Translation can preserve the apparent topic of a question while changing its intensity or assumptions about ordinary behavior. A respondent's explanation may employ concepts unfamiliar to the assessor without thereby constituting a psychiatric symptom. Cultural formulation requires the clinician to ask about the person's own terms for the problem, perceived causes, available support, experiences of discrimination, social constraints, and expectations for care. These questions supplement direct symptom assessment and can change how a complaint is interpreted. Kirmayer (2012) analyzes the intellectual and institutional challenges of cultural competence in health settings.

The Cultural Formulation Interview introduced with DSM-5 provides a structured route for exploring the social and cultural framing of clinical problems. The international field-trial research reported by Lewis-Fernández et al. (2017) investigated its feasibility, acceptability, and perceived clinical utility across participating settings. Such studies support analysis of implementation and contextual value, while their design and populations should not be inflated into proof that every diagnostic decision will improve in every culture. Practical use requires interpreters who understand clinical confidentiality and are trained for the task, attention to dialect and literacy, and recognition that the interpreter mediates a three-way conversation. Clinicians should document whether an instrument's validated language version was used and whether modification of wording occurred.

Cross-cultural validity is a question for evidence, not a courtesy added after diagnosis. The same reported symptom may have different normative prevalence, thresholds of disclosure, or social consequences across settings. Conversely, genuine clinical distress may be minimized when professionals regard unfamiliar expressions as simply customary. The WHO (2024) CDDR incorporates culture-related guidance in diagnostic descriptions; its global orientation is one response to the need for contextual interpretation. Assessment should compare an individual's experience against relevant local and developmental context and be transparent when local norms or adequate instruments are unavailable.

Developmental, Disability, and Accessibility Requirements

Children, adolescents, older adults, and people with communication or sensory disabilities require adjustments to assessment procedures while preserving the clinical question. Development influences attention span, memory, insight, vocabulary, dependency on caregivers, and how symptoms are expressed. A child's difficulty concentrating can have a different range of explanations from an older adult's abrupt attentional fluctuation. Multiple informants and context-specific observations may be particularly important when the person cannot reliably describe the relevant developmental period or when behaviors occur only in one setting. However, dependence on informants can conceal the person's own experiences, especially for internalizing symptoms. The interview should preserve opportunities for direct expression using accessible communication methods.

Assessment accommodations may involve adapted pacing, assistive communication technology, interpretation or captioning, visual aids, sensory modifications, and accessible forms. The choice should be guided by the individual's needs and by evidence about how modifications affect instrument comparability. Enlarging text or allowing additional breaks can improve access, but changing the construct-relevant content or response conditions may invalidate comparison with published norms. The test interpreter must document accommodations and any consequent limits on standard score interpretation AERA, APA, & NCME (2014) American Psychological Association (2017). A clinical interview often remains useful even when a standardized score cannot be validly computed.

An assessment that focuses only on symptoms can overlook environmental barriers that are central to functioning. Disability is shaped by the interaction of health conditions and physical, social, institutional, and technological contexts. The WHO (2010) framework covers domains of cognition, mobility, self-care, getting along, life activities, and participation; the instrument's content illustrates a broader conception of functioning than diagnosis alone. This perspective assists evaluation of accommodations, support needs, and outcomes. It also helps separate a person's capacity under supportive conditions from performance in an inaccessible or adverse environment.

Functioning, Distress, Impairment, and Quality of Life

Many clinical classifications require clinically significant distress, impairment, or changes in functioning, although the exact requirements vary by disorder and classification system. These judgments require explicit evidence. A person may report substantial symptoms while maintaining employment through extraordinary effort; another may have comparatively modest symptom frequency but severe difficulty in a key life domain. Occupational and educational performance are important sources of information but cannot serve as universal proxies for mental health, particularly when social opportunity and accommodations differ. Assessment should examine daily activity, self-care, relationships, housing, finances, safety, and valued participation according to the referral question and patient's own priorities WHO (2024) WHO (2010).

The relationship between symptom burden and functioning is moderated by resources, role expectations, work conditions, illness severity, coping strategies, and social support. Functional changes can clarify onset and course, but causal interpretation remains difficult. A deterioration after relocation may be connected to anxiety, economic insecurity, loss of support, or all of these processes. Likewise, a successful occupational outcome does not prove absence of illness. Clinically meaningful assessment specifies what is observable, how it changed, and which alternative explanations are plausible. When an instrument measures participation or disability, the evaluator must understand whether its score represents self-report, observed performance, or a composite and what reference period it covers.

Quality of life is multidimensional and person-centered; it encompasses outcomes that may extend beyond symptom reduction. A diagnosis that is accurate but disconnected from actual needs can have limited utility. Longitudinal assessment should therefore revisit functioning and participation alongside symptom change, especially when treatment or service interventions target everyday recovery. Interpreting these outcomes also requires attention to adverse effects, social barriers, and whether care goals were defined collaboratively. The scientific argument is that diagnosis forms one part of the evidence required for an adequate account of a person's situation.

Risk and Safety Assessment Within Diagnostic Evaluation

Safety evaluation is a clinical decision process that may intersect with, but should not be subordinated to, formal diagnostic classification. A person can experience suicidal thinking, severe self-neglect, vulnerability to exploitation, or acute behavioral risk across many diagnostic categories. Assessment requires direct, context-sensitive inquiry when indicated; evaluation of current and past behaviors, intent where relevant, access to means, precipitants, protective supports, and opportunities for urgent intervention; and documentation of decisions made. A brief screening instrument can be one source of information but cannot guarantee prediction of an individual future event. The NICE (2014/2025) guidance places risk evaluation within the assessment and care planning of bipolar disorder. Emergency pathways and legal duties depend on location and clinical setting.

It is important to separate population-level prediction from the immediate assessment of a specific clinical presentation. A method can have moderate group discrimination and still be poorly calibrated for an individual with unusual circumstances or changing risk. Risk also changes over time with intoxication, social stress, agitation, medication changes, and access to support. Clinicians should document the time of assessment, the sources consulted, foreseeable limitations, and the follow-up plan. In uncertainty, a proportionate safety response may be justified without waiting for a definitive DSM or ICD label. The question is what action is supported by the available information and immediate clinical need.

The US Preventive Services Task Force (2023) evidence statement on suicide risk in asymptomatic adult screening populations concerns the evidence for universal screening in that defined preventive setting. It must be distinguished from evaluation of people who present with symptoms or disclosed risk. This difference illustrates the danger of transporting a screening recommendation beyond its population. The present research addresses the methods and limits of clinical inference; individualized triage, emergency intervention, and treatment require qualified clinicians and locally appropriate services.

Primary Care, Specialty Services, Emergency Settings, and Teleassessment

Clinical context changes both the prevalence of conditions encountered and the information available. Primary care commonly begins with nonspecific complaints and multiple medical conditions. Specialist outpatient services receive selected populations whose prior screening and referral history alter base rates. Emergency settings must prioritize acute medical and safety questions under time pressure. A research diagnostic interview administered in a controlled setting may have greater procedural consistency than an interview performed amid interruptions or crisis. Generalizing performance across these environments requires evidence of transportability. A positive depression questionnaire can have different predictive value in an unselected clinic than in a specialty cohort enriched for mood disorders Levis et al. (2019) Mitchell et al. (2009).

Teleassessment can expand access across geography and disability barriers while introducing additional conditions of observation. Connection quality, privacy at the remote location, access to records, cultural and language mediation, and opportunities to observe behavior all affect data collection. The clinician should establish the identity and location of the participant as required by the setting, confirm suitable privacy, explain contingencies for disrupted communication and emergencies, and consider whether a particular instrument or procedure has appropriate evidence for remote administration. It is methodologically unsound to assume that an instrument validated in-person is necessarily equivalent when administered through an altered mode. Accessibility gains and potential measurement effects require separate evaluation AERA, APA, & NCME (2014) American Psychological Association (2017).

Resource-limited services may require stepped assessment pathways. A first encounter can establish presenting concerns, immediate safety, medical questions, and likely diagnostic domains; focused additional appointments can address unresolved conditions or gather collateral data. The WHO (2016) mhGAP guide provides an example of structured decision support for nonspecialist settings. Its application should follow the actual service capacity and scope of competence, not the assumption that a manual removes the need for professional training. Staged assessment is defensible when the clinical record states what has and has not been evaluated and ensures that patients with urgent needs reach appropriate care.

Forensic and Administrative Uses: A Distinct Evidentiary Threshold

Questions about legal responsibility, parenting capacity, disability eligibility, risk, or workplace fitness may require psychological assessment, but the purposes and standards differ from treatment-oriented diagnosis. In clinical care, information is often collected to guide support and shared decisions. In forensic or administrative contexts, the evaluator may be asked to address a legally defined capacity or functional standard and produce a report for a third party. A psychiatric diagnosis alone rarely answers that legal or administrative question. The relevant causal and functional links must be examined using methods appropriate to the question, the jurisdiction, and the evidence available. The assessment should state its scope and avoid implying expertise in legal standards beyond the evaluator's competence.

Response context matters. Some individuals may have incentives to minimize symptoms, while others may be motivated to emphasize them; fear, misunderstanding, shame, or illness may also influence responding without deliberate distortion. Clinical observations and selected validity measures can inform an assessment, but such indicators should not be interpreted as direct proof of dishonesty. Multiple explanations must be compared, including cognitive limitations, cultural or linguistic mismatch, unusual symptom patterns, and adversarial institutional conditions. The appropriate evaluation requires specialized training and, where relevant, formal forensic methods. Routine clinical screens should not be repurposed as definitive measures of credibility.

The ethical significance of purpose-specific validation is especially high when a result may determine housing, employment, custody, liberty, or access to care. A test's validity for identifying symptoms in a treatment clinic does not establish validity for a forensic claim, because the reference outcome, stakes, case selection, and error costs have changed. Reporting should separate diagnostic observations from opinions about the external question, with transparent limitations and the evidence supporting each inference. General principles in AERA, APA, & NCME (2014) and American Psychological Association (2017) support a disciplined account of test use, competence, and consequences.

Scientific Equity: Selection, Measurement Invariance, and Structural Bias

Differences in diagnostic rates can reflect genuine epidemiological variation, selection into care, differences in symptom expression, unequal access to assessment, variable thresholds, or systematic measurement error. A single prevalence comparison cannot separate these possibilities. Fair assessment investigates who enters the sample, whose symptoms are recognized, which languages and norms the instrument supports, and whether reference procedures operate similarly across groups. Measurement invariance concerns whether the same latent construct is reflected comparably across groups; its absence may make unadjusted score comparisons misleading AERA, APA, & NCME (2014). In clinical practice, the principle becomes concrete: one cutoff or question set should not be presumed equivalent across populations without relevant validation.

Structural bias may also enter through the target used to train predictive systems. In a prominent health services investigation, Obermeyer et al. (2019) showed that an algorithm using healthcare costs as a proxy for need could embed racial inequities in the allocation of services. The setting was population health management rather than psychiatric differential diagnosis, but the methodological lesson is directly relevant: predictors and outcomes can encode unequal access, and calibration on an administrative proxy does not guarantee fair representation of underlying need. An algorithm predicting past service use may reproduce service disparities rather than identify untreated clinical difficulty.

Equity evaluation therefore combines psychometric scrutiny, clinical context, and institutional analysis. Audits may compare sensitivity, specificity, false-positive rates, calibration, and missingness across groups, but such metrics can conflict when underlying base rates differ. No single metric resolves the ethical problem independently of the use case. Patient involvement, interpreters, accommodation, qualified assessment, and accessible pathways to confirmatory care address parts of the problem that cannot be solved through test-score adjustment alone. The broader objective is to ensure that the inferential pathway is transparent and proportionate for the population to which it is applied.

Reconciling Categorical and Dimensional Clinical Evidence

Categorical diagnostic systems provide a common vocabulary for communicating recognized disorders, identifying service pathways, and defining research samples. Dimensional approaches capture severity, trait variation, and patterns of symptoms that may cross categorical boundaries. A coherent assessment can use both forms of information when their functions are clearly distinguished. The Kotov et al. (2017) literature describes a hierarchical framework for organizing covariation among psychopathology features, while WHO (2024) and American Psychiatric Association (2022) supply current clinical classification rules. A latent dimension observed across populations does not automatically define a clinical syndrome in an individual, just as meeting categorical criteria does not fully describe symptom intensity and functional need. The choice of model should follow the particular inference being attempted.

A problem arises when a threshold is interpreted as a natural border rather than as a rule that serves a defined classificatory purpose. Patients close to a symptom-duration or severity cutoff may be similar on many outcomes while receiving different codes. Assessment should record both the categorical conclusion and the observations that justify it, including subthreshold or transdiagnostic features relevant to care. Dimensional information may also reveal residual symptoms that matter for functioning after a categorical episode no longer meets full criteria. Conversely, a high score on a broad distress dimension can obscure an acute medical or substance-associated state that requires a different line of investigation.

The scientific test of integrative models is not simply whether they accommodate more observations. Competing models should be compared for reproducibility, calibration, explanatory value, longitudinal predictive performance, and usefulness in actual decisions. An elaborate network of symptom relations may describe covariance in a dataset without demonstrating causal links. Clinical assessment must therefore maintain the distinction between statistical structure and individual causal inference. A good report can acknowledge the usefulness of both categorical and dimensional descriptions while remaining precise about what the available evidence actually establishes.

Reliability, Replication, and Diagnostic Change Over Time

Diagnostic reliability is a property of a procedure under specified conditions, not an immutable property of the disorder being assessed. Agreement can vary with interviewer training, criterion complexity, patient characteristics, reference period, mode of administration, and the interval between assessments. The DSM-5 field trials discussed by Regier et al. (2013) and methodological interpretation by Kraemer et al. (2012) made these complexities visible in practical clinical settings. A test–retest design may detect unstable judgment, real changes in a person's condition, or both. Interrater agreement within the same interview and agreement between independent interviews at different times answer different questions. Research reports need to name the design before comparing numerical values.

Replication is complicated by differences in clinical samples. A structured interview developed among research volunteers may behave differently in patients with multiple conditions, acute illness, linguistic diversity, or unstable housing. A screening tool's diagnostic accuracy may change when the prevalence and severity spectrum shift, even when the wording stays identical. The evidence required for broad applicability therefore includes external validation and transparent reporting of recruitment, missing data, interviewer qualifications, and threshold rules. When a study provides an accuracy estimate without describing verification of screen-negative cases, the result may overstate what can be inferred.

Diagnostic change is not always evidence of error. New episodes, remission, developmental maturation, treatment exposure, and more complete history can legitimately alter a conclusion. An audit can distinguish process errors from warranted revision by examining what was knowable at the time and what additional observations later became available. The scientific objective is a learning assessment system capable of recording uncertainty, checking its decisions, and improving through case follow-up. This is especially important for conditions in which longitudinal course is diagnostically informative Robins & Guze (1970).

Evidence Quality, Contradictions, and Limits of Generalization

The strongest empirical designs depend on the question under consideration. Interview reliability studies examine agreement under repeated or independent evaluations. Diagnostic-accuracy studies compare an index test with a reference assessment, ideally with appropriate blinding and complete verification. Longitudinal cohorts establish temporal patterns and prognostic relationships, although confounding limits causal interpretation. Randomized trials can test whether adding an assessment procedure changes subsequent decisions or outcomes, but their findings depend on what intervention follows a positive result. Meta-analyses of diagnostic accuracy, such as Levis et al. (2019), can improve precision while remaining subject to differences in instruments, reference interviews, and settings. A critical narrative synthesis cannot collapse these distinct outcomes into one ranking without explaining the assumptions.

Apparent contradictions often disappear when the research questions are aligned. A study may show that a questionnaire correlates strongly with symptom severity while another finds that its positive scores have modest diagnostic predictive value in a low-prevalence population. Both results can be accurate. Agreement with a structured interview can be substantial for common syndromes in a specialist clinic yet modest for poorly defined presentations or conditions with low base rates. Similarly, a broad clinician impression can be helpful for referral while being insufficient for a final classification. The task is to identify the unit of analysis, the denominator, and the target decision before treating two estimates as incompatible.

Gaps remain important. Many instrument validation studies underrepresent complex comorbidity, severe illness, cultural minorities, low-literacy groups, or patients with disabilities. High-quality interview validation may require resources unavailable in ordinary practice. Publication and selective reporting can favor favorable thresholds and settings. Changes to diagnostic manuals complicate longitudinal comparison of instruments whose target definitions have shifted. A conservative scientific appraisal should therefore describe the populations for which evidence is strongest, those for which extrapolation is likely, and the additional studies required to justify wider use AERA, APA, & NCME (2014) Hunsley & Mash (2007).

Artificial Intelligence, Digital Phenotyping, and Decision Support

The Artificial Era introduces computational systems capable of processing large volumes of narrative, behavioral, and clinical data. Possible applications to assessment include automated transcription, information extraction from records, structured intake assistance, support for differential hypothesis generation, longitudinal symptom visualization, and prediction of service needs. These functions have different evidence requirements. Accurate extraction of symptoms from text does not establish accurate diagnosis; good prediction of an administrative outcome does not establish understanding of a patient's clinical condition. Any automated component must be evaluated against an explicitly defined task and independent outcome, including subgroup performance, sensitivity to missing data, and handling of contradictory evidence. The conceptual object is the relationship between human clinical judgment, recorded information, and artificial decision support.

Diagnostic systems face distinctive failure modes when trained on electronic health records. Historical diagnoses may be wrong or inconsistent. Clinical notes may contain negations, hypothetical possibilities, copied text, and information about family members rather than the patient. Treatment selection can determine which symptoms are observed and coded, creating selective labels. Training data may reflect inequitable access to care. The demonstration by Obermeyer et al. (2019) of racial bias introduced through an administrative healthcare cost proxy illustrates the broader risk of substituting measurable service use for underlying health need. A model that recreates recorded diagnoses may reproduce the original system's errors while reporting strong apparent agreement.

The proper standard for clinical deployment is prospective, purpose-specific evaluation with clear human accountability. Evidence should include an independent reference process, comparison against usual assessment, calibration, false-positive and false-negative consequences, drift monitoring, privacy, data retention and consent rules, and review of changes in clinician behavior. A system that produces fluent explanations can still cite information incorrectly or conceal uncertainty. Its role should be explicitly delimited in the clinical workflow, with escalation when urgent medical or safety issues appear. Artificial Cognition and Human–Artificial Relations are relevant research horizons, yet claims about artificial selfhood or reasoning status are philosophical and theoretical propositions distinct from demonstrated medical diagnostic validity.

Clinical Documentation as an Auditable Research Product

A strong assessment report offers more than a list of criteria. It identifies the referral context and sources, specifies the interview and classification versions, presents a concise symptom and developmental history, records meaningful mental status observations, describes the medical and substance questions investigated, and states the evidence supporting each important differential conclusion. It distinguishes confirmed findings, reported experiences, reasonable inference, and open questions. The interpretation of test scores includes their purpose, administration conditions, relevant validity limits, and the relationship between score and diagnosis. The structure enables another qualified professional to see why the conclusion follows from the evidence and where further evaluation may change it American Psychological Association (2017) AERA, APA, & NCME (2014).

A differential section is most useful when it links alternatives to discriminating observations rather than stating that a diagnosis was “ruled out” without explanation. For example, an assessment of episodic activation may identify the lack of a sufficiently established distinct episode, the presence of a temporally associated substance exposure, and the absence of corroborating course evidence. The report could reasonably retain bipolar disorder as a possibility requiring further longitudinal information, while describing current symptoms and functioning directly. By contrast, a definitive exclusion statement would exceed the observed data. The same principle applies to medical explanations when relevant investigations have not yet been completed.

Documentation should preserve provenance. The date and version of a classification or instrument, the identity and role of each source, the context of informant reports, and the reason for amendments all matter when a conclusion is revisited. Technical systems can help maintain these records, but they can also propagate a copied label into future reports. A reliable workflow makes amendments visible and prevents tentative formulations from being converted automatically into established diagnoses. Assessment quality is thus partly an information-governance problem, linking clinical science to data architecture and professional responsibility.

A Research Agenda for Better Differential Assessment

Several research programs follow from the synthesis. First, prospective multicenter studies should compare clinically feasible structured pathways with usual assessment across primary care, specialty, emergency, and community settings, reporting not only agreement with a reference interview but also changes in patient-relevant decisions. Second, diagnostic-accuracy studies should include complete verification or defensible corrections for partial verification, prespecified thresholds, and transparent reporting of indeterminate results. Third, repeated within-person assessment should examine how sleep, stress, substance exposure, developmental transition, and treatment alter symptom expression and criterion stability. These programs would improve the interpretability of results in patients who present with multiple plausible conditions rather than textbook examples.

Fourth, research should examine the incremental value of collateral histories, medical records, functional measures, and digital monitoring after accounting for a careful clinical interview. More information is not automatically better if it increases bias, creates privacy harms, or consumes resources without improving decisions. Fifth, cross-language and disability-inclusive validation should test response processes and measurement invariance in the specific populations where a tool is used. Sixth, clinical decision-support systems need prospective studies with independent outcome adjudication and audits of error distribution, including groups underrepresented in training data. These methods can assess whether the system adds useful information or merely automates preexisting documentation patterns.

Finally, conceptual work should distinguish the validity of diagnostic categories from the utility of a particular assessment procedure. Better measurement can improve agreement without resolving whether the category captures a discrete causal entity. Conversely, even imperfect categorical models may serve communication and access purposes when their limitations are recognized. The strongest interdisciplinary program combines clinical phenomenology, psychometrics, epidemiology, causal reasoning, public health, ethics, and human–artificial system evaluation. The objective is an assessment science that is accountable to evidence and to the lives affected by its conclusions.

Conclusion

Clinical psychological assessment is best understood as a structured, revisable scientific inference that integrates symptoms, observable behavior, life history, functional consequences, and contextual and medical evidence. Differential diagnosis is its comparative engine: the clinician identifies plausible explanations, seeks findings that distinguish them, evaluates the reliability of each source, and documents the resulting level of support. The historical development of diagnostic interviews and operational criteria made assessment more reproducible, while evidence from accuracy studies and field trials demonstrates why reproducibility, validity, clinical utility, and equity must be examined separately Robins & Guze (1970) Spitzer et al. (1992) Regier et al. (2013) Hunsley & Mash (2007).

The strongest practical conclusion is that a screen, an interview, a categorical code, and a formulation answer different questions. Screening identifies possible need for further assessment; a diagnostic interview tests defined criteria; functional and dimensional measures describe burden and change; a formulation relates the individual pattern to context and potential mechanisms. Their integration requires attention to reference standards, prevalence, comorbidity, medical and substance-associated explanations, developmental and cultural conditions, safety, and longitudinal course. No instrument eliminates the need to know what is being inferred, from which evidence, for which purpose, and under which limitations WHO (2024) AERA, APA, & NCME (2014).

Within the Artificial Era, computational assistance can improve organization and comparison of evidence when its performance is independently validated and its decision role remains explicit. Its introduction sharpens rather than removes the demand for traceable clinical reasoning. A mature science of assessment measures the accuracy and consequences of its judgments, distinguishes provisional hypotheses from confirmed findings, and learns from follow-up. The enduring purpose of differential diagnosis is to produce an intelligible and appropriately cautious explanation of a clinical presentation that supports proportionate, competent, and humane action.

References

  • American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for educational and psychological testing. American Educational Research Association. https://ncme.org/resources/books/testing-standards/
  • American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.; DSM-5-TR). American Psychiatric Association Publishing. https://doi.org/10.1176/appi.books.9780890425787
  • American Psychological Association. (2017). Ethical principles of psychologists and code of conduct (2002, amended effective 2010 and 2017). https://www.apa.org/ethics/code
  • Croskerry, P. (2003). The importance of cognitive errors in diagnosis and strategies to minimize them. Academic Medicine, 78(8), 775–780. https://doi.org/10.1097/00001888-200308000-00003
  • Endicott, J., & Spitzer, R. L. (1978). A diagnostic interview: The schedule for affective disorders and schizophrenia. Archives of General Psychiatry, 35(7), 837–844. https://doi.org/10.1001/archpsyc.1978.01770310043002
  • Feighner, J. P., Robins, E., Guze, S. B., Woodruff, R. A., Jr., Winokur, G., & Muñoz, R. (1972). Diagnostic criteria for use in psychiatric research. Archives of General Psychiatry, 26(1), 57–63. https://doi.org/10.1001/archpsyc.1972.01750190059011
  • Grove, W. M., Zald, D. H., Lebow, B. S., Snitz, B. E., & Nelson, C. (2000). Clinical versus mechanical prediction: A meta-analysis. Psychological Assessment, 12(1), 19–30. https://pubmed.ncbi.nlm.nih.gov/10752360/
  • Hunsley, J., & Mash, E. J. (2007). Evidence-based assessment. Annual Review of Clinical Psychology, 3, 29–51. https://doi.org/10.1146/annurev.clinpsy.3.022806.091419
  • Kessler, R. C., & Üstün, T. B. (2004). The World Mental Health (WMH) Survey Initiative version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI). International Journal of Methods in Psychiatric Research, 13(2), 93–121. https://doi.org/10.1002/mpr.168
  • Kirmayer, L. J. (2012). Rethinking cultural competence. Transcultural Psychiatry, 49(2), 149–164. https://doi.org/10.1177/1363461512444673
  • Koo, T. K., & Li, M. Y. (2016). A guideline of selecting and reporting intraclass correlation coefficients for reliability research. Journal of Chiropractic Medicine, 15(2), 155–163. https://doi.org/10.1016/j.jcm.2016.02.012
  • Kotov, R., Krueger, R. F., Watson, D., Achenbach, T. M., Althoff, R. R., Bagby, R. M., Brown, T. A., Carpenter, W. T., Caspi, A., Clark, L. A., Eaton, N. R., Forbes, M. K., Forbush, K. T., Goldberg, D., Hasin, D., Hyman, S. E., Ivanova, M. Y., Lynam, D. R., Markon, K. E., ... Zimmerman, M. (2017). The Hierarchical Taxonomy of Psychopathology (HiTOP): A dimensional alternative to traditional nosologies. Journal of Abnormal Psychology, 126(4), 454–477. https://doi.org/10.1037/abn0000258
  • Kraemer, H. C., Kupfer, D. J., Clarke, D. E., Narrow, W. E., & Regier, D. A. (2012). DSM-5: How reliable is reliable enough? American Journal of Psychiatry, 169(1), 13–15. https://doi.org/10.1176/appi.ajp.2011.11010050
  • Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ-9: Validity of a brief depression severity measure. Journal of General Internal Medicine, 16(9), 606–613. https://doi.org/10.1046/j.1525-1497.2001.016009606.x
  • Levis, B., Benedetti, A., & Thombs, B. D., on behalf of the DEPRESsion Screening Data (DEPRESSD) Collaboration. (2019). Accuracy of Patient Health Questionnaire-9 (PHQ-9) for screening to detect major depression: Individual participant data meta-analysis. BMJ, 365, l1476. https://doi.org/10.1136/bmj.l1476
  • Lewis-Fernández, R., Aggarwal, N. K., Lam, P. C., Galfalvy, H., Weiss, M. G., Kirmayer, L. J., Paralikar, V., Deshpande, S. N., Díaz, E., Nicasio, A. V., Boiler, M., Alarcón, R. D., Rohlof, H., Groen, S., van Dijk, R. C. J., Jadhav, S., Sarmukaddam, S., Ndetei, D., Scalco, M. Z., ... Vega-Dienstmaier, J. M. (2017). Feasibility, acceptability and clinical utility of the Cultural Formulation Interview: Mixed-methods results from the DSM-5 international field trial. British Journal of Psychiatry, 210(4), 290–297. https://doi.org/10.1192/bjp.bp.116.193862
  • Meehl, P. E. (1954). Clinical versus statistical prediction: A theoretical analysis and a review of the evidence. University of Minnesota Press. https://meehl.umn.edu/lane-practice
  • Mitchell, A. J., Vaze, A., & Rao, S. (2009). Clinical diagnosis of depression in primary care: A meta-analysis. The Lancet, 374(9690), 609–619. https://doi.org/10.1016/S0140-6736(09)60879-5
  • National Institute for Health and Care Excellence. (2010, updated 2023). Delirium: Prevention, diagnosis and management in hospital and long-term care (Clinical guideline CG103). https://www.nice.org.uk/guidance/cg103/chapter/Recommendations
  • National Institute for Health and Care Excellence. (2012). Autism spectrum disorder in adults: Diagnosis and management (Clinical guideline CG142). https://www.nice.org.uk/guidance/cg142/chapter/Recommendations
  • National Institute for Health and Care Excellence. (2014, updated 2025). Bipolar disorder: Assessment and management (Clinical guideline CG185). https://www.nice.org.uk/guidance/cg185/chapter/Recommendations
  • National Institute for Health and Care Excellence. (2014). Psychosis and schizophrenia in adults: Prevention and management (Clinical guideline CG178). https://www.nice.org.uk/guidance/cg178/chapter/Recommendations
  • National Institute for Health and Care Excellence. (2022). Depression in adults: Treatment and management (NICE guideline NG222). https://www.nice.org.uk/guidance/ng222/chapter/Recommendations
  • Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. https://doi.org/10.1126/science.aax2342
  • Regier, D. A., Narrow, W. E., Clarke, D. E., Kraemer, H. C., Kuramoto, S. J., Kuhl, E. A., & Kupfer, D. J. (2013). DSM-5 field trials in the United States and Canada, Part II: Test-retest reliability of selected categorical diagnoses. American Journal of Psychiatry, 170(1), 59–70. https://doi.org/10.1176/appi.ajp.2012.12070999
  • Robins, E., & Guze, S. B. (1970). Establishment of diagnostic validity in psychiatric illness: Its application to schizophrenia. American Journal of Psychiatry, 126(7), 983–987. https://doi.org/10.1176/ajp.126.7.983
  • Sheehan, D. V., Lecrubier, Y., Sheehan, K. H., Amorim, P., Janavs, J., Weiller, E., Hergueta, T., Baker, R., & Dunbar, G. C. (1998). The Mini-International Neuropsychiatric Interview (M.I.N.I.): The development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. Journal of Clinical Psychiatry, 59(Suppl. 20), 22–33. https://pubmed.ncbi.nlm.nih.gov/9881538/
  • Spitzer, R. L., Kroenke, K., & Williams, J. B. W. (1999). Validation and utility of a self-report version of PRIME-MD: The PHQ Primary Care Study. JAMA, 282(18), 1737–1744. https://doi.org/10.1001/jama.282.18.1737
  • Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006). A brief measure for assessing generalized anxiety disorder: The GAD-7. Archives of Internal Medicine, 166(10), 1092–1097. https://doi.org/10.1001/archinte.166.10.1092
  • Spitzer, R. L., Williams, J. B. W., Gibbon, M., & First, M. B. (1992). The Structured Clinical Interview for DSM-III-R (SCID). I: History, rationale, and description. Archives of General Psychiatry, 49(8), 624–629. https://doi.org/10.1001/archpsyc.1992.01820080032005
  • US Preventive Services Task Force. (2023). Screening for depression and suicide risk in adults: US Preventive Services Task Force recommendation statement. JAMA, 329(23), 2057–2067. https://doi.org/10.1001/jama.2023.9297
  • World Health Organization. (2010). Measuring health and disability: Manual for WHO Disability Assessment Schedule (WHODAS 2.0). World Health Organization. https://www.who.int/publications/i/item/measuring-health-and-disability-manual-for-who-disability-assessment-schedule-%28-whodas-2.0%29/
  • World Health Organization. (2016). mhGAP Intervention Guide for mental, neurological and substance use disorders in non-specialized health settings: Mental Health Gap Action Programme (version 2.0). https://www.who.int/europe/publications/i/item/9789241549790
  • World Health Organization. (2024). Clinical descriptions and diagnostic requirements for ICD-11 mental, behavioural and neurodevelopmental disorders (CDDR). World Health Organization. https://www.who.int/publications/i/item/9789240077263