Problem Drinking Scale (CAGE)

1. Abstract

The Problem Drinking Scale, universally designated by the acronym CAGE (Ewing, 1984), represents one of the most widely implemented, parsimonious, and clinically validated screening instruments for identifying alcohol use disorders (DSM-5 Alcohol Use Disorder), harmful drinking, and physiological alcohol dependence. Originally developed by Dr. John A. Ewing at the University of North Carolina at Chapel Hill, the CAGE questionnaire condenses complex behavioral, affective, social, and physiological manifestations of pathological alcohol consumption into four straightforward binary (Yes/No) questions. The instrument’s acronym operationalizes its core diagnostic dimensions: the subjective perception of the need to Cut down on drinking, social friction and emotional reactivity manifested as being Annoyed by external criticism, internal psychological distress characterized by feeling Guilty about drinking behaviors, and the physiological imperative of using alcohol as an Eye-opener to alleviate morning withdrawal symptoms or tremors.

Extensive psychometric investigations over five decades demonstrate that the CAGE exhibits exceptional sensitivity (routinely ranging from 74% to 93%) and specificity (typically exceeding 70% to 90%) across inpatient psychiatric, general medical, and emergency department settings when using the conventional diagnostic cut-off score of ≥2. While psychometric indices such as internal consistency coefficients (Cronbach’s α and Kuder-Richardson Formula 20) vary across diverse racial and gender demographic groups—ranging from modest values (α ≈ .42–.55) in low-prevalence non-clinical male cohorts to robust levels (α ≈ .64–.75) among clinical, female, and racially diverse populations—the scale’s clinical power derives from its high negative predictive value and Guttman-like hierarchical scaling structure. This comprehensive academic review delineates the historical origins, theoretical underpinnings, psychometric validation metrics, factor structure, and clinical implementation parameters of the CAGE Problem Drinking Scale.

2. Keywords

CAGE questionnaire, problem drinking, alcohol use disorder, screening instrument, alcohol dependence, psychometrics, John A. Ewing, diagnostic validity, alcoholism screening, internal consistency, Guttman scaling

3. Authors

The CAGE questionnaire was conceptualized and empirically developed by John A. Ewing, M.D. (1924–2006), an internationally renowned physician, clinical psychiatrist, and pioneering addiction researcher. Dr. Ewing served as the founding director of the Bowles Center for Alcohol Studies and was a distinguished Professor of Psychiatry at the University of North Carolina at Chapel Hill School of Medicine.

Dr. Ewing’s clinical investigations into the phenomenology of alcoholism commenced in the late 1960s at North Carolina Memorial Hospital in Chapel Hill. Disturbed by the routine failure of general internal medicine services and surgical wards to detect chronic, life-threatening alcohol dependence in hospitalized patients, Ewing set out to construct an ultra-brief, non-stigmatizing, and easily memorized clinical interview tool. In collaboration with colleagues including Beatrice A. Rouse, M.S.W., Ewing evaluated extensive diagnostic interview protocols before distilling the four emblematic inquiries in 1968. Although Dr. Ewing presented preliminary validation findings at international symposia as early as 1969 and 1970, his definitive, seminal paper titled “Detecting Alcoholism: The CAGE Questionnaire” was published in the Journal of the American Medical Association (JAMA) in 1984.

4. Purpose

The overarching clinical and scientific objective of the Problem Drinking Scale (CAGE) is to serve as a rapid, primary-level screening instrument capable of distinguishing individuals who exhibit severe alcohol-related problems, hazardous drinking patterns, or physical dependence from non-problematic social drinkers. The tool was not formulated to function as an exhaustive diagnostic manual or a quantitative measure of total ethanol volume consumed; rather, its primary purpose is triage, identification, and risk stratification across diverse clinical, forensic, and epidemiological domains.

Clinical Applications

In routine medical and psychiatric practice, the CAGE serves as a critical first-line triage instrument. Standard medical examinations historically omitted targeted inquiries regarding ethanol consumption due to time constraints, clinician discomfort, or patient evasiveness. The four CAGE items bypass quantitative volumetric estimations—which patients often consciously underreport or unconsciously distort—focusing instead on experiential, emotional, and neurobiological sequelae. In ambulatory primary care, hospital inpatient admissions, trauma centers, and emergency departments, the administration of the CAGE permits clinicians to identify individuals who warrant comprehensive diagnostic workups, brief motivational interventions (such as SBIRT: Screening, Brief Intervention, and Referral to Treatment), or medically supervised alcohol withdrawal management.

Research and Epidemiological Applications

In population-based public health investigations, psychiatric epidemiology, and clinical trials, the CAGE provides a standardized, low-burden metric for categorizing alcohol involvement across large sample sizes. It has been incorporated into multinational surveys, epidemiological catchment area studies, and health risk appraisals. Researchers frequently deploy the scale to control for problematic alcohol use in studies evaluating medical comorbidities (e.g., hepatic dysfunction, cardiovascular disease, traumatic brain injury) or psychiatric dual diagnoses (e.g., major depressive disorder, post-traumatic stress disorder, bipolar disorder).

Theoretical and Diagnostic Rationale

The conceptual rationale driving the CAGE is rooted in the recognition that problem drinking is characterized by behavioral dysfunction, loss of volitional control, interpersonal alienation, internal moral distress, and neurochemical adaptation. Rather than cataloging exact quantities of drinks consumed per week—metrics characterized by significant measurement error and varying tolerance thresholds—the CAGE captures qualitative markers across four discrete psychological and behavioral domains. By evaluating these key markers, the instrument successfully identifies individuals whose drinking has surpassed recreational thresholds and entered the realm of clinically significant pathology.

5. Psychological Construct

The CAGE measures the multi-faceted psychological construct of Problem Drinking, conceptualized as a continuum ranging from risky behavioral patterns to severe alcohol use disorder as defined by contemporary nosological systems (World Health Organization ICD-11; American Psychiatric Association DSM-5). The construct is synthesized through four distinct behavioral, interpersonal, intrapsychic, and neurophysiological dimensions.

1. Volitional Control Impairment: The ‘Cut Down’ Dimension

Item 1 (“Have you ever felt you should cut down on your drinking?”) evaluates the intra-individual awareness of behavioral dysregulation. From a cognitive-behavioral perspective, this dimension taps into the respondent’s recognition that their ethanol intake exceeds their personal desires, intentions, or physiological limits. This item reflects executive control failure, where the individual attempts self-regulation, experiences behavioral lapses, and consciously perceives an incongruence between their desired consumption goals and their actual drinking behavior. It assesses the early cognitive awareness of loss of control—a cardinal diagnostic criterion of substance dependence.

2. Social Friction and Interpersonal Reactivity: The ‘Annoyed’ Dimension

Item 2 (“Have people annoyed you by criticizing your drinking?”) measures the relational and interpersonal friction generated by the individual’s drinking habits, coupled with the individual’s psychological defense mechanisms. Rather than inquiring neutrally whether family members, peers, or colleagues have expressed concern, the item specifically interrogates the affective response of annoyance. This captures psychological projection, defensiveness, and cognitive denial. When external observers identify hazardous behavioral patterns, the individual frequently responds with irritation, resentment, or hostility, reflecting interpersonal conflict and the external social consequences of alcohol-related dysfunction.

3. Intrapsychic Affective Distress: The ‘Guilty’ Dimension

Item 3 (“Have you ever felt bad or guilty about your drinking?”) taps into affective distress, remorse, and internal moral conflict. Pathological alcohol use consistently disrupts an individual’s value systems, occupational commitments, and domestic obligations. Following episodes of intoxication, behavioral disinhibition, or memory blackouts, individuals frequently experience intense feelings of guilt, shame, and self-reproach. Psychometrically, this dimension serves as a highly sensitive affective barometer of internal conflict, signaling that the respondent’s alcohol use is causing subjective suffering and moral distress.

4. Neurochemical Adaptation and Withdrawal Relief: The ‘Eye-Opener’ Dimension

Item 4 (“Have you ever had a drink first thing in the morning to steady your nerves or to get rid of a hangover (eye opener)?”) taps directly into neurobiological dependence, physiological tolerance, and the manifestation of physical alcohol withdrawal. The subjective experience of morning tremulousness, autonomic hyperactivity (“the shakes”), anxiety, or nausea signifies neuroadaptation within gamma-aminobutyric acid (GABA) and N-methyl-D-aspartate (NMDA) receptor pathways. The pharmacological administration of morning ethanol to terminate these withdrawal phenomena indicates severe physiological dependence. In psychometric terms, this item possesses the highest discrimination parameter and highest specificity within the instrument.

6. Theoretical Framework

The architecture of the CAGE questionnaire is grounded in several foundational paradigms of addiction psychology, including E.M. Jellinek’s classical disease model, the neuroadaptive framework of physiological dependence, self-regulation theory, and cognitive dissonance theory.

Jellinek’s Classical Disease Model of Alcoholism

During the mid-20th century, E.M. Jellinek (1960) formalized the conceptualization of alcoholism as a progressive, chronic disease entity characterized by identifiable symptom clusters and physiological progression. Jellinek’s phase model posited that individuals transition through pre-alcoholic, prodromal, crucial, and chronic phases. The CAGE items mirror this progressive architecture:

• The awareness of needing to reduce intake (Cut down) and internal remorse (Guilty) emerge during the prodromal and early crucial phases.
• Social friction and external confrontation (Annoyed) mark the escalation of social impairment during the crucial phase.
• Morning drinking (Eye-opener) exemplifies the chronic phase characterized by physiological adaptation, prolonged tremors, and the necessity of exogenous ethanol to restore autonomic baseline.

The Alcohol Dependence Syndrome (Edwards & Gross)

The CAGE strongly aligns with the theoretical framework of the Alcohol Dependence Syndrome (ADS) formulated by Griffith Edwards and Milton M. Gross (1976). Edwards and Gross conceptualized dependence not as an all-or-none categorical state, but as a biological-behavioral syndrome characterized by:

1. Narrowing of the drinking repertoire.
2. Salience of drink-seeking behavior over competing obligations.
3. Increased tolerance to ethanol.
4. Repeated withdrawal symptoms.
5. Relief or avoidance of withdrawal symptoms by further drinking (captured directly by the CAGE “Eye-opener” item).
6. Subjective awareness of the compulsion to drink (reflected in the CAGE “Cut down” item).

Cognitive Dissonance and Psychological Defense

Leon Festinger’s (1957) theory of Cognitive Dissonance provides the psychological explanation for the “Annoyed” and “Guilty” items. When an individual’s deeply held self-concept (as a competent, responsible, and controlled adult) conflicts with the behavioral realities of uncontrolled drinking and social reprimand, profound psychological tension ensues. The individual manages this dissonance either through negative affect directed inward (guilt and shame) or defensive externalization directed outward (annoyance, hostility toward observers, and cognitive rationalization). Ewing deliberately designed the CAGE to capture both trajectories of dissonance resolution.

7. Validity

The CAGE questionnaire has undergone exhaustive empirical evaluation regarding its construct, criterion, convergent, and discriminant validity across hundreds of published studies encompassing diverse medical, community, and clinical settings.

Criterion-Related and Predictive Validity

The criterion validity of the CAGE has been established against gold-standard psychiatric interviews, including the Structured Clinical Interview for DSM (SCID), the Diagnostic Interview Schedule (DIS), and standardized clinical diagnoses of alcohol dependence and alcohol use disorder. In Dr. Ewing’s original 1984 report evaluating 130 medical patients, a cut-off score of 2 or more affirmative responses yielded a sensitivity of 100% and a specificity of 89% in detecting alcoholism. Subsequent large-scale clinical investigations have provided nuanced benchmarks:

• Sensitivity: In inpatient psychiatric and substance abuse treatment populations, the sensitivity of a CAGE score ≥2 consistently ranges between 85% and 93%. In general medical outpatient settings, sensitivity typically falls between 70% and 85%.
• Specificity: Specificity across clinical cohorts is remarkably stable, ranging from 75% to 96%. A score of 0 or 1 reliably identifies non-problem drinkers.
• Threshold Modifications: Lowering the threshold to ≥1 affirmative answer dramatically increases sensitivity (approaching 95%–98%) while decreasing specificity (dropping to 60%–70%), an approach frequently recommended in perinatal and prenatal screening where missing hazardous exposure carries severe teratogenic risks.

Convergent Validity

The convergent validity of the CAGE has been established through strong correlations with other validated psychometric scales and physiological biomarkers:

• Correlation with the MAST: Studies evaluating the CAGE against the Michigan Alcoholism Screening Test (MAST; Selzer, 1971) report substantial Pearson correlation coefficients typically ranging between r = .70 and .86.
• Correlation with the AUDIT: Comparative analyses with the Alcohol Use Disorders Identification Test (AUDIT; Babor et al., 2001) demonstrate strong convergence (correlations ranging from r = .65 to .78), although the AUDIT is superior for detecting hazardous drinking before dependence develops, while the CAGE is optimized for established problem drinking and dependence.
• Biomarker Convergence: CAGE positive scores demonstrate statistically significant positive correlations with objective biological markers of heavy ethanol intake, including elevated Serum Gamma-Glutamyl Transferase (γ-GGT; odds ratios of 2.5 to 4.2), Carbohydrate-Deficient Transferrin (%CDT), and Mean Corpuscular Volume (MCV).

Discriminant Validity

The CAGE demonstrates adequate discriminant validity by differentiating alcohol use disorders from primary depressive episodes, generalized anxiety disorder, and non-substance psychiatric conditions. While depressive and anxious patients may endorse subjective guilt, they do not endorse the “Eye-opener” or “Cut down” items unless co-occurring alcohol pathology is present.

8. Reliability

The psychometric evaluation of reliability for brief binary scales presents unique methodological characteristics. Because the CAGE consists of only four dichotomous items designed to capture distinct hierarchical phases of a clinical construct, classical internal consistency estimates are inherently constrained by test length.

Internal Consistency Across Diverse Populations

Internal consistency metrics calculated via Cronbach’s coefficient alpha (α) and the Kuder-Richardson Formula 20 (KR-20) demonstrate variability contingent on the underlying base rate and severity of problem drinking within the sampled demographic:

• Caucasian Samples: Typical internal consistency coefficients in validation studies demonstrate α values around .67.
• African American Samples: Studies examining African American cohorts report internal consistency coefficients of .64.
• Female Cohorts: In diverse female patient cohorts, reliability indices average approximately .64.
• Hispanic Populations: In general community samples of Hispanic individuals, the coefficient has been documented at approximately .43.
• Male Community Samples: In non-clinical, general population male samples, internal consistency often manifests around .42.

These lower alpha coefficients in general community samples do not indicate fundamental measurement failure; rather, they reflect the brevity of a 4-item scale and the extreme skewness of item distributions in cohorts where severe dependence criteria (such as morning eye-openers) exhibit low base rates. When evaluated in high-prevalence clinical samples (e.g., inpatient detoxification units), Cronbach’s alpha routinely exceeds .75 to .82.

Test-Retest and Inter-Rater Reliability

Test-retest stability has been demonstrated across clinical and research environments. Investigations administering the CAGE at 24-hour, 7-day, and 30-day intervals report test-retest reliability coefficients ranging from r = .80 to .95, indicating high temporal stability of patient self-report when measured longitudinally. Furthermore, inter-rater reliability between independent clinical interviewers (e.g., physician vs. triage nurse) and between self-administered paper/digital formats versus clinician-administered interviews demonstrates high concordance, with Cohen’s kappa (κ) values regularly exceeding .85.

9. Factor Analysis

The structural dimensionality of the CAGE has been scrutinized through Exploratory Factor Analysis (EFA), Confirmatory Factor Analysis (CFA), and modern Item Response Theory (IRT) models.

Unidimensionality vs. Two-Factor Models

The prevailing consensus in psychometric literature confirms that the CAGE functions predominantly as an essentially unidimensional construct representing problem drinking severity. In principal component analyses, the first unrotated eigenvalue typically accounts for 48% to 62% of the total variance across clinical samples, meeting standard criteria for unidimensionality.

However, several extensive factor analytic studies (e.g., Heck & Williams, 1995) have identified a secondary two-factor structure in non-clinical or community cohorts:

• Factor 1: Subjective/Affective Reactions: Comprising the Annoyed and Guilty items, this factor captures emotional, intrapsychic, and social responses to drinking. Factor loadings for these items on the subjective component typically range from .70 to .84.
• Factor 2: Behavioral and Physiological Control Impairment: Comprising the Cut down and Eye-opener items, this factor represents the behavioral attempt at self-regulation and physiological withdrawal relief. Factor loadings for these items on the behavioral component typically range from .68 to .82.

Confirmatory Factor Analysis Fit Indices

When evaluated via CFA, the single-factor model yields robust fit indices when accounting for the binary nature of the data using diagonally weighted least squares (DWLS) or robust maximum likelihood estimation (WLSMV):

• Comparative Fit Index (CFI): Routinely exceeds .96 to .99.
• Tucker-Lewis Index (TLI): Typically ranges between .95 and .98.
• Root Mean Square Error of Approximation (RMSEA): Consistently ≤ .05 (90% CI [.02, .07]).
• Standardized Root Mean Square Residual (SRMR): Typically ≤ .04.

Item Response Theory (IRT) and Guttman Scalogram Properties

Item Response Theory analyses demonstrate that the CAGE operates akin to a classic Guttman scale, displaying hierarchical difficulty parameters (β) and strong item discrimination (α):

1. Item 1 (Cut down): Lowest difficulty (β ≈ −0.50 to 0.00); high sensitivity; serves as the broad entry threshold into problem drinking.
2. Item 3 (Guilty): Moderate difficulty (β ≈ 0.20 to 0.60); reflects escalating intrapsychic conflict.
3. Item 2 (Annoyed): Moderate-to-high difficulty (β ≈ 0.50 to 0.90); reflects interpersonal confrontation.
4. Item 4 (Eye-opener): Highest difficulty (β ≈ 1.50 to 2.20) and highest discrimination parameter (α > 2.0); represents severe, established neurochemical dependence. Individuals endorsing Item 4 almost universally endorse Items 1, 2, and 3.

10. Instrument / Measurement Tool

The CAGE Problem Drinking Scale is structured as an ultra-brief, four-item clinical interview or self-report screening tool. Its operational specifications are outlined below:

• Instrument Designation: Problem Drinking Scale (CAGE Questionnaire).
• Instrument Class: Clinical Screening Instrument / Psychological Assessment Tool.
• Item Count: 4 items.
• Response Format: Dichotomous binary scale (Yes = 1, No = 0).
• Administration Modality: Clinician-administered oral interview, self-administered paper-and-pencil questionnaire, or computerized/digital assessment.
• Estimated Administration Time: Under 60 seconds.
• Scoring Rules:
  • Each affirmative answer (“Yes”) receives a score of 1 point.
  • Each negative answer (“No”) receives a score of 0 points.
  • Total possible score ranges from 0 to 4.
• Diagnostic Cut-off Interpretations:
  • Score of 0: Negative screen; indicates non-problematic drinking patterns.
  • Score of 1: Equivocal or borderline; warrants further clinical inquiry, exploration of drinking history, or re-screening in high-risk populations (e.g., obstetric patients, patients with hepatic dysfunction, psychiatric comorbidities).
  • Score of 2 or 3: Clinically positive screen; indicates clinically significant problem drinking or high index of suspicion for Alcohol Use Disorder. Indicates the necessity of formal diagnostic evaluation.
  • Score of 4: Highly positive screen; virtually pathognomonic for severe physiological alcohol dependence. Immediate diagnostic assessment, medical withdrawal evaluation, and structured addiction treatment referral indicated.

11. Permissions & Fee and Test Year

The CAGE questionnaire was formally published in its standardized format in 1984 by Dr. John A. Ewing in the Journal of the American Medical Association (JAMA), following earlier academic presentations in 1970. The instrument resides firmly in the public domain. Dr. Ewing intentionally released the four-question metric without copyright restrictions or licensing fees to facilitate unhindered global implementation across clinical medicine, public health research, and community care.

Consequently, clinicians, hospitals, universities, and independent researchers are free to utilize, reprint, digitize, and integrate the CAGE questionnaire into electronic health record (EHR) systems without purchasing permissions, paying royalties, or securing written licenses. Professional academic attribution citing Dr. Ewing’s foundational 1984 publication remains the standard ethical and scholarly convention.

12. References

• Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B., & Monteiro, M. G. (2001). AUDIT: The Alcohol Use Disorders Identification Test: Guidelines for use in primary care (2nd ed.). World Health Organization. https://www.who.int/publications/i/item/audit-the-alcohol-use-disorders-identification-test
• Bush, K., Kivlahan, D. R., McDonell, M. B., Fihn, S. D., & Bradley, K. A. (1998). The AUDIT alcohol consumption questions (AUDIT-C): An effective brief screening test for problem drinking. Archives of Internal Medicine, 158(16), 1789–1795. https://doi.org/10.1001/archinte.158.16.1789
• Edwards, G., & Gross, M. M. (1976). Alcohol dependence: Provisional description of a clinical syndrome. British Medical Journal, 1(6017), 1058–1061. https://doi.org/10.1136/bmj.1.6017.1058
• Ewing, J. A. (1984). Detecting alcoholism: The CAGE questionnaire. Journal of the American Medical Association, 252(14), 1905–1907. https://doi.org/10.1001/jama.1984.03350140051025
• Ewing, J. A., & Rouse, B. A. (1970). Identifying the hidden alcoholic. In Proceedings of the 29th International Congress on Alcohol and Addictions (pp. 52–55). Sydney, Australia.
• Festinger, L. (1957). A theory of cognitive dissonance. Stanford University Press. https://doi.org/10.1515/9781503620766
• Heck, E. J., & Williams, M. D. (1995). Using the CAGE to screen for drinking problems in college students. Journal of College Student Development, 36(3), 278–283.
• Jellinek, E. M. (1960). The disease concept of alcoholism. Hillhouse Press.
• Mayfield, D., McLeod, G., & Hall, P. (1974). The CAGE questionnaire: Validation of a new alcoholism screening instrument. The American Journal of Psychiatry, 131(10), 1121–1123. https://doi.org/10.1176/ajp.131.10.1121
• O’Brien, C. P. (2008). The CAGE questionnaire for detection of alcoholism. Journal of the American Medical Association, 300(17), 2054–2056. https://doi.org/10.1001/jama.2008.570
• Selzer, M. L. (1971). The Michigan Alcoholism Screening Test: The quest for a new diagnostic instrument. The American Journal of Psychiatry, 127(12), 1653–1658. https://doi.org/10.1176/ajp.127.12.1653

13. Items of the Scale (Questionnaire)