|
CAGE Questionnaire
Please note that this section is an archive (last updated in June 2006). [disclaimer]
Sections:
Overview |
Instrument Reviews |
Construct Overviews |
Book Compendium Reviews |
Internet Site Reviews
|
|
|
|
Practical Information
|
Instrument Name:
|
CAGE Questionnaire
|
Instrument Description:
|
The development of the CAGE is described in a paper presented at an International Conference on Alcoholism; the paper was not published outside of those proceedings and therefore is not readily available. Thus, description of the CAGE is culled from the many publications in which it is cited. (Ref: 2)
The 4 questions of the CAGE (Cutting down, Annoyance by criticism, Guilty feeling, and Eye-openers) were part of a larger item pool used in a clinical study undertaken at North Carolina Memorial Hospital. One hundred thirty medical and surgical patients were randomly selected for the study. They were interviewed using a full set of alcohol detection questions developed by previous authors. Sixteen out of the 130 patients were diagnosed as alcoholic by their physicians; their diagnoses were then confirmed through chart review. Responses to the alcohol use interview questions were compared between patients diagnosed as alcoholic versus nonalcoholic. The authors sought to determine the minimum number of questions that could be used to distinguish the two groups; this resulted in a set of 4 questions, which were placed in a specific order to produce the mnemonic “CAGE.” CAGE items were then incorporated into an interview given to patients in an alcoholism treatment center to determine the percentage of alcoholics answering the CAGE affirmatively at various cut-scores.
A clinician may paraphrase CAGE questions to best suit the occasion, as there is no indication from prior studies that this compromises the interpretability of CAGE scores. (Ref: 2) The CAGE does not discriminate between active and inactive drinking and is meant as a screening test, not as a diagnostic measure. (Ref: 4) Because its time frame is “ever,” the CAGE is sometimes supplemented by other questions to distinguish between active and inactive drinking status. (Ref: 8) It is reported that the most frequent positively answered CAGE item is the first. (Ref: 6)
The author suggests that a score of 1 represents the need for further inquiry into a patient’s drinking behavior. (Ref: 2) Nevertheless, a cut-score of 2 has been more generally used. (Ref: 3,7) Some researchers suggest that in an elderly medical population, a score of 1 or more should be considered a positive screening test, due to the high prevalence of drinking problems or a history of such problems in this population. A cut-score of 1 is associated with an 84% probability of drinking problems in males and 54% probability in females. (Ref: 4)
|
Price:
|
Free; public domain.
|
Administration Time:
|
No information found.
|
Publication Year:
|
1970
|
Item Readability:
|
Flesch-Kincaid grade-level of 5.1; there are only 4 sentences that are very short in length; persons with a 5th grade reading level should easily comprehend them.
|
Scale Format:
|
Dichotomous (Yes/No).
|
Administration Technique:
|
Self-administered or interview.
|
Scoring and Interpretation:
|
Affirmative answers each receive a score of 1, with scores ranging from 1-4. Higher scores indicate higher probability of alcohol abuse or dependence.
|
Forms:
|
No information found.
|
|
|
Research Contacts
|
Instrument Developers:
|
John A. Ewing and Beatrice Rouse
|
Instrument Development Location:
|
University of North Carolina School of Medicine
Bowles Center for Alcohol Studies
2311 Canterwood Dr.
Wilmington, NC 28401
|
Instrument Developer Email:
|
john_ewing@med.unc.edu
|
Instrument Developer Website:
|
www.med.unc.edu/alcohol/welcome.htm
|
|
|
Annotated Bibliography
|
1. Mayfield D, McLeod G, & Hall P. The CAGE questionnaire: Validation of a new alcoholism screening instrument. Am J Psychiatry 1974;131(10):1121-3. [PMID: 4416585]
 |
Purpose: To evaluate the usefulness of the CAGE in terms of its brevity, ease of administration, sensitivity, and validity. |
|
Sample: A total of 366 patients were surveyed. Sample characteristics were as follows: 99% male; 77% white; 63% from 35-55 years old; primarily lower SES classes, with Classes IV and V equaling 73% of the sample; 60% married; 79% alcohol users; 39% later classified as alcoholics by multi-disciplinary team diagnosis. |
|
Methods: All patients entering care over a 1-year period were administered the CAGE 1-7 days after admission to a psychiatric service. Demographic questions were asked during patient interviews along with the 4 CAGE questions. Patients were next categorized as being either alcoholic or non-alcoholic on the basis of a diagnosis made by a multidisciplinary team. Finally, alcoholic category status was correlated with CAGE score. |
|
Implications: The CAGE is not a sensitive detector of alcoholism if a total score of positive 4 is the criterion; however, the instrument is a much more useful screening instrument if a score of positive 2 or 3 is used. The second CAGE question, “Have people annoyed you by criticizing your drinking,” appears to have lower discriminative ability. Development of a weighted scoring system might improve the validity of the questionnaire. |
2. Ewing JA. Detecting alcoholism: The CAGE questionnaire. JAMA Oct 12, 1984;252(14):1905-7. [PMID: 6471323]
|
Purpose: To explain how the 4 original CAGE questions were first identified; to describe the first uses of the CAGE in clinical/research settings. |
|
Sample: The development sample used to derive the original 4-item screening instrument consisted of 130 randomly selected medical and surgical patients from a general hospital population, 16 of whom were diagnosed as alcoholics. Validation Sample A consisted of 166 male patients from an alcoholism rehabilitation center, compared with 68 general hospital male patients who used alcoholic beverages but were clearly not alcoholic; validation Sample B was comprised of 48 alcoholic patients (47 males, 1 female) from London treatment centers who ranged in age from 23-61 years old, came from all social classes, and originated primarily from Ireland or Scotland. |
|
Methods: Development-sample patients were interviewed using a series of alcohol detection questions as written by previous authors. Responses were compared, those of alcoholics versus those of non-alcoholics, with a set of 4 questions found to adequately distinguish between the 2 groups. Validation Sample A patients provided responses to the 4-question CAGE instrument as well as their self-perceptions regarding being (a) alcoholic, (b) a heavy drinker but not alcoholic, or (c) non-alcoholic. Responses of known alcoholics were then contrasted with those of known non-alcoholics. Validation Sample B patients, all alcoholics, also provided responses to the 4-item CAGE. Total score analysis was then conducted with both validation samples to determine 1) response levels most indicative of the presence of alcoholism and 2) the minimal response level at which suspicion of alcoholism should be investigated. |
|
Implications: The CAGE was not developed for the purpose of diagnosing alcoholism, rather, to determine when further inquiry into a patient’s alcoholic drinking should be made. Although a score of 1 should arouse suspicion and initiate an investigation into patient level of alcoholic consumption, scores of 2 and 3 appear to clearly indicate a high likelihood of the presence of alcoholism, even among patients who deny being alcoholic. |
3. Buchsbaum DG, Buchanan RG, Centor RM, Schnoll SH, Lawton MJ. Screening for alcohol abuse using CAGE scores and likelihood ratios. Ann Intern Med 1991;115(10):774-77. [PMID: 1929025]
|
Purpose: To assess the ability of the CAGE to distinguish between alcoholic and non-alcoholic general medicine clinic outpatients; to develop and assess the usefulness of additional risk information as provided by estimating a separate likelihood ratio for each possible CAGE score, with total scores ranging from 0 to 4. |
|
Sample: The study enrolled 821 outpatients who were English-speaking, 18 years or older, and attending a general medicine clinic at an urban university teaching hospital in Virginia for a new or follow-up appointment. Outpatients were screened for lifetime history of alcohol abuse or dependence. Of the 821 study enrollees, 294 were determined by DSM-III criteria to have a lifetime history of alcohol abuse and/or dependence. |
|
Methods: Patients were screened using the alcohol module of the Diagnostic Interview Schedule (DIS), the Composite International Diagnostic Interview – Substance Abuse Module (CIDI-SAM), and the CAGE. Patients were defined as having a drinking problem (a) if they met DSM-III criteria for history of alcohol dependence or abuse, or (b) if they reported 1 or more symptoms on the DIS though did not otherwise meet DSM-III criteria. DSM-III diagnoses were converted to be compatible with DSM-III-R diagnoses. Diagnoses and independently determined CAGE scores were then compared. |
|
Implications: The CAGE performed well in identifying medical outpatients diagnosed by DSM-III criteria with an alcohol disorder (i.e., abuse or dependence). Using likelihood ratios for individual CAGE scores improved the ability to assess patient’s risk for alcohol abuse or dependence. The practical usefulness of this supplemental risk information, however, is not so readily apparent, given the CAGE’s designed function as a first-level screening instrument. |
4. Buchsbaum DG, Buchanan RG, Welsh J, Centor RM, Schnoll SH. Screening for drinking disorders in the elderly using the CAGE questionnaire. J Am Geriatr Soc 1992 Jul;40(7):662-5. [PMID: 1607581]
|
Purpose: To assess the performance of the CAGE questionnaire and its ability to identify elderly medicine outpatients with drinking problems. |
|
Sample: Three hundred twenty-three patients from an outpatient medical practice (>=60 years old) were interviewed at an urban university teaching hospital. One hundred six patients met DSM-III criteria for inclusion in the analyses. |
|
Methods: The alcohol module of the Diagnostic Interview Schedule (DIS) was used as the criterion standard; patients were then given the CAGE for comparative purposes. Patients were defined as having a history of alcohol abuse or dependence if they met DSM-III criteria, or if they had >=1 symptom on the DIS even though not meeting DSM-III criteria. DSM-III/DIS-based diagnoses were then compared to CAGE scores. |
|
Implications: The CAGE was found to effectively distinguish between elderly patients with a history of drinking problems and those without. The authors suggested that the screening cut-off be flexible, considering the context of it use (i.e., the prevalence of drinking disorders in the population being screened). For all patients and using a cut-off score of 1, sensitivity and specificity were reported as 86% and 78%, respectively; using a cut-off score of 2, sensitivity and specificity were found to be 70% and 91%, respectively. Study findings associated with a cut-off score of 2 can be compared to findings previously reported using the same cut-off score of 2 for general population screening use, where sensitivity and specificity have been reported as being 80% and 85%, respectively. |
5. Girela E, Villanueva E, Hernandez-Cueto C, Luna JD. Comparison of the CAGE questionnaire versus some biochemical markers in the diagnosis of alcoholism. Alcohol & Alcoholism 1994;29(3):337-43. [PMID: 7945575]
|
Purpose: To compare the screening characteristics (i.e., sensitivity, specificity, and predictive value) of the CAGE with those of the following objective biochemical markers: plasma levels of ethanol and acetate, mean corpuscular volume, gamma glutamyl transferase, and glycosylated hemoglobin. |
|
Sample: 1) Control group: 50 healthy non-alcoholic patients (24 males and 26 females, aged 17-65 years, with ethanol consumption <=20 g/day); 31 patients with non-alcoholic viral-origin liver disease (20 males and 11 females, aged 17-81, with ethanol consumption <20 g/day). 2) Experimental group: 40 alcoholic patients (32 males and 8 females, aged 24-64, with a minimum ethanol consumption of 80 g/day for men and 60 g/day for women); alcohol consumption had to have taken place over a period of at least 1 year, and patients had to still be drinking or had stopped within <3 weeks of study enrollment). |
|
Methods: Patients were given the CAGE questionnaire and had a medical history taken. Specimens were drawn for biochemical analysis, and measures of the following biomarkers were obtained for each study participant: 1) MCV; 2) plasma GGT; 3) glycosylated hemoglobin; 4) ethanol plasma level; and 5) acetate plasma level. CAGE scores as well as biomarker levels were then compared for the 3 groups, after which patients from the control group were collapsed into a single non-alcoholic group (50 healthy controls and 31 patients with non-alcoholic liver disease) versus the experimental alcoholic group (40 patients). |
|
Implications: Of the multiple measures obtained, including objective laboratory tests, the CAGE questionnaire proved to be the most efficient. When respondents answered at least 1 question positively, CAGE sensitivity was found to be 96%, and its specificity 92%. The most powerful classification model employed each of the 4 CAGE questions as independent variables to predict the presence of alcoholism; this model’s sensitivity was 90%, and its specificity 99%. The single most powerful discriminating item from the CAGE was reported to be #1: “Cutting down on drinking.” |
6. McIntosh MC, Leigh G, Baldwin NJ. Screening for hazardous drinking: using the CAGE and measures of alcohol consumption in family practice. Canadian Family Physician 1994;40:1546-53. [PMID: 7920048]
|
Purpose: 1) To determine the drinking behaviors of a clinical population; 2) to examine the ability of the CAGE and a joined set of specific drinking behavior questions to function as a hazardous drinking screening method; and 3) to examine the relationship between alcohol consumption and individual CAGE items, especially among a group of hazardous drinkers. |
|
Sample: One thousand three hundred thirty-four patients of a family practice clinic, including 1015 females (mean age=40) and 319 males (mean age=42). |
|
Methods: This research was done as part of a larger study to assess interventions for reducing alcohol use. A hazardous drinking event was study-defined as having had 4 or more drinks on any 1 day in the past month, or as having a score of 1 on the CAGE. CAGE items were asked along with a set of demographic questions and questions about alcohol consumption and specific drinking behaviors. |
|
Implications: Although the CAGE, in conjunction with the set of drinking behavior questions, performed well in identifying patients at risk for alcohol problems, the questionnaire appeared to function somewhat differentially for males and females. Males in this study drank more frequently and more heavily than did females. Thirty percent of males reported at least 1 day of hazardous drinking without answering “yes” to any CAGE question; for females it was 9%. Conversely, 21% of males answered “yes” to at least 1 CAGE question without reporting a single day of hazardous drinking, while for females it was 53%. |
7. Morton JL, Jones TV, Manganaro MA. Performance of alcoholism screening questionnaires in elderly veterans. Am J Med. 1996 Aug;101(2):153-9. [PMID: 8757354]
|
Purpose: To validate 3 alcoholism screening questionnaires - the CAGE, the MAST-G, and the AUDIT - in elderly male veterans. |
|
Sample: One hundred twenty male veteran outpatients, >= 65 years of age, from a general medicine practice of the Omaha VAMC. Mean age=72 years, 64% married, 65% living in an urban area, and 36% met DIS-III-R/Alcohol Module criteria for current or past alcohol abuse or dependence. |
|
Methods: Demographic information was collected from hospital computer records, and patients were interviewed in person using the Alcohol Module of the Revised Diagnostic Interview Schedule (DIS-III-R); they were then asked to provide responses to the 3 screening instruments - the CAGE, the MAST-G, and the AUDIT. Performance characteristics of the screening tools (e.g., sensitivity, specificity, positive predictive value) were then compared, using diagnoses based on the DIS-III-R Alcohol Module as the gold standard. |
|
Implications: The MAST-G and the CAGE outperformed the AUDIT in detection of alcohol abuse and dependence in elderly male veterans. The brevity of the CAGE (4 items) offers an advantage over the MAST-G (24 items) in both administration time and patient burden. Reported performance suggests moderate accuracy on the part of each instrument in distinguishing between patients with drinking problems and those without (i.e., sensitivity and specificity for the MAST-G (cut-score >=5): 70% and 81%, respectively; for the CAGE (cut-score >=2): 63% and 82%, respectively). Somewhat lower performance of the CAGE than previously reported was perhaps due to its use in a population in which alcoholism had a lower prevalence (i.e., 23% diagnosed with alcohol abuse or dependence and 42% abstainers, compared with 40% diagnosed with alcoholism and 35% abstainers for males in the McIntosh et al study of a family practice clinical population). (See Annotated Bibliography entry #6). |
8. Bradley KA, Kivlahan DR, Bush KR, McDonell And MB, Fihn SD. Variations on the CAGE alcohol screening questionnaire: strengths and limitations in VA general medical patients. Alcohol Clin Exp Res. 2001 Oct;25(10):1472-8. [PMID: 11696667]
|
Purpose: To evaluate a recommended time-frame modification and the addition of several quantification of drinking behavior questions to the CAGE. |
|
Sample: Two hundred twenty-seven male veteran general medicine patients (mean age=66 years, 64% white). |
|
Methods: As part of the ACQUIP pilot study, patients were mailed the CAGE questionnaire with a two time-frame response option format (standard time frame of “ever” and modified “last-year” time frame); in addition, several new questions concerning quantity and frequency of alcohol consumption as well as episodic drinking events were added. Patient classification derived from screening question responses were then compared to a gold standard diagnosis, based on having either a DSM-III-R diagnosis of alcohol abuse or dependence and/or having had at least 1 hazardous drinking event in the past year (where hazardous drinking was defined as having 14 or more drinks per week, or 5 or more drinks on an occasion). |
|
Implications: Using a past-year time frame for the CAGE was more specific (.82 versus .59), but less sensitive (.57 versus .77), than the original “ever” time frame in detecting the presence of alcoholism (i.e., a DSM-III-R diagnosis of alcohol abuse or dependence and/or reported hazardous drinking). An 8-item CAGE was shown to be most sensitive (.92), but its specificity was only 50%. A single question about the frequency of drinking (AUDIT question #3: >=6 drinks on an occasion) outperformed (sensitivity .77; specificity .83) the standard CAGE, making it an optimal brief screening instrument. However, this single item was administered as part of the 8-item augmented CAGE. Its performance as a stand-alone item/screen was not assessed. The CAGE “past year” time frame instrument was therefore considered insensitive to detecting alcoholism, while the 8-item CAGE was found to be most sensitive. |
9. Moore AA, Seeman T, Morgenstern H, Beck JC, Reuben DB. Are there differences between older persons who screen positive on the CAGE questionnaire and the Short Michigan Alcoholism Screening Test-Geriatric Version? J Am Geriatr Soc. 2002 May;50(5):858-62. [PMID: 12028172]
|
Purpose: 1) To assess the degree of agreement between the CAGE and the SMAST-G in identifying older persons with possible alcohol use disorders, and 2) to determine if persons identified by these instruments as having potential alcohol disorders differ per instrument in demographic, drinking, or health-related characteristics. |
|
Sample: N=1,889 (mean age=69 years old, 51% female, 94% white, 63% married, 70% completed some college, 37% had annual incomes of >=$50,000). Members from three different organizations were surveyed: 1) members of the American Association of Retired Persons, 2) managed care enrollees from a large medical group in southern California, 3) attendees of a community-based senior health center, again, in southern California. |
|
Methods: This study involved the secondary analysis of data obtained from a mailed survey. Participants were sent the Health Risk Appraisal for the Elderly (HRA-E), which also included the CAGE (4 items), SMAST-G (10 items), and additional items about drinking behavior, health status, and general demographic characteristics. A positive CAGE was defined as having 1 or more positive responses; a positive SMAST-G was defined as having 2 or more positive responses. The analysis sample was then comprised of 3 groups used for testing differences: participants positive on the CAGE alone, positive on the SMAST-G alone, and positive on both the CAGE and the SMAST-G. |
|
Implications: Twenty-six percent of all study participants screened positive on the CAGE and/or on the SMAST-G. Less than half of the persons screening positive on either measure screened positive on both, implying that the measures may assess different aspects of unsafe drinking. A combined measure that employs both screening instruments may identify more alcohol disorders among older persons with differing demographic and health characteristics. The two screening instruments were found to be differentially sensitive to: (a) amount of drinking per week (positive CAGEs drank more than positive SMAST-Gs); (b) gender (positive CAGEs were more frequently male than positive SMAST-Gs); (c) marital status (positive CAGEs were more frequently married than positive SMAST-Gs); and (d) income (positive CAGEs had higher incomes than positive SMAST-Gs). No gold standard comparison was available to assess the diagnostic accuracy of the CAGE or SMAST-G. |
|
|
top
|
Factors and Norms
|
Factor Analysis Work:
|
No information found.
|
Normative Information Availability:
|
No information found.
|
|
|
Reliability Evidence
|
Test-retest:
|
No information found.
|
Inter-rater:
|
No information found.
|
Internal Consistency:
|
No information found.
|
Alternate Forms:
|
No information found.
|
|
|
Validity Evidence
|
Construct/ Convergent/ Discriminant:
|
One study compared the CAGE to the SMAST-G and found that fewer than half of all persons who screened positive on at least 1 of the measures (n=393) screened positive on both measures (n=169 or 43%). This suggests that the CAGE may assess different aspects of unsafe drinking than the SMAST-G. (Ref: 9)
|
Criterion-related/ Concurrent/ Predictive:
|
In a sample of outpatients from an urban university teaching hospital, using a diagnostic interview as a gold standard, the CAGE with a cut-score of 1 had a sensitivity of 89% and specificity of 81%; with a cut-score of 2 sensitivity decreased to 74%, but specificity increased to 91%. (Ref: 3) In a sample of elderly medical outpatients, using a diagnostic interview as a gold standard, similar findings were observed: a cut-score of 1 produced a sensitivity of 86% and specificity of 78%, while a cut-score of 2 decreased sensitivity to 70% and increased specificity to 91%. (Ref: 4) In a sample of elderly male veterans, using a diagnostic interview as a gold standard, a cut-score of 2 had a sensitivity of 63% and a specificity of 82%, somewhat lower characteristics than reported elsewhere. This may have been due to its use in a population in which alcoholism was less prevalent. (Ref: 7) Using a “past-year” time frame for the CAGE resulted in its being more specific (0.82 versus 0.59), but less sensitive (0.57 versus 0.77) than the original “ever” time frame in detecting hazardous drinking or DSM-III-R alcohol abuse or dependence diagnosis. (Ref: 8)
Comparing CAGE classification to a gold standard of diagnostic interview, receiver operating characteristic (ROC) curve analysis produced an area under the curve (AUC) of 0.89 in a study of medical outpatients and 0.86 in a study of elderly medical outpatients. (Ref: 3,4) In a group of elderly male veterans, the AUC was observed to be 0.77 (Ref: 7) and 0.71. (Ref: 8)
Likelihood ratios for CAGE cut-scores of 0 to 4 were as follows: 0.14 (7% posterior probability of alcohol abuse or dependence disorder), 1.5 (46% probability of alcohol use disorder), 4.5 (72% probability of disorder), 13 (88% probability of disorder), and 100 (98% probability of disorder). (Ref: 3)
|
Content:
|
No information found.
|
Responsiveness Evidence:
|
No information found.
|
Scale Application in VA Populations:
|
Yes. (Ref: 1,7-8)
|
Scale Application in non-VA Populations:
|
Yes. (Ref: 2-6,9)
|
|
|
Comments
|
There is consistent evidence, for the most part, supporting use of the CAGE as a screening tool for the detection of possible alcohol abuse and/or dependence. There is also consistent evidence suggesting that the CAGE is inappropriately used when it is employed as a diagnostic instrument, when its scores are considered to represent a continuum of increasingly greater risk for alcohol abuse and/or dependence, or when its accuracy in detecting alcohol use disorder is assumed to be equal across populations whose alcoholism prevalence varies.
The CAGE was first conceived of as an “index of suspicion.” As such, a patient’s positive response to at least 1 CAGE item indicates to a care provider that the patient’s drinking behavior should be investigated for possible alcohol use disorder. The CAGE was therefore not intended as a decision-making tool, rather as a tool for determining the appropriateness of a drinking behavior investigation; that is, to screen in order to differentiate among patients for whom more in-depth interviewing is warranted. Evidence presented here indicates that the CAGE is most sensitive when a cut-score of 1 is used, and that its sensitivity falls off considerably when a cut-score of 2 is used. Higher CAGE scores have not consistently or uniformly been indicative of a greater likelihood of alcoholism; therefore, to view CAGE scores as locations on an alcoholism risk continuum is misguided.
There are several limitations to use of the CAGE that must be kept in mind when employing it as a screening tool. First, with its “ever” time frame for drinking issues, the CAGE does not distinguish between active and inactive problems; follow-up questions must identify whether problem-drinking behaviors are in active status. Second, the CAGE may not be an effective screening tool for patients who are incompetent, who deny, underestimate, or minimize their drinking behaviors, who are defensive, or who are simply uncooperative. Such patients may require the use of more objective measures of drinking status. Finally, the CAGE both over- and under-identifies persons with problem drinking behaviors. In one study, 30% of males reported at least 1 day of hazardous drinking without answering “yes” to any CAGE question; for females it was 9%. Conversely, 21% of males answered “yes” to at least 1 CAGE question without reporting a single day of hazardous drinking, while for females it was 53%. (Ref: 6)
Note: The CAGE’s item presentation order was selected for no other reason than to create the mnemonic “CAGE.”
|
|
|
|
