58. Two Variations of Recommended Alcohol Screening Questionnaires did not Improve Screening Accuracy in a VA General Medical Population
KA Bradley MD MPH, Northwest Health Services Research and Development Center of Excellence; Primary and Specialty Medical Care Service, VA Puget Sound Health Care System; Departments of Medicine and Health Services, University of Washington; DR Kivlahan PhD, Center of Excellence in Substance Abuse Treatment and Education, VA Puget Sound Health Care System; Department of Psychiatry and Behavioral Sciences, University of Washington; KR Bush MPH, Northwest Health Services Research and Development Center of Excellence, VA Puget Sound Health Care System; MB McDonell MS, Northwest Health Services Research and Development Center of Excellence, VA Puget Sound Health Care System; SD Fihn MD MPH, Northwest Health Services Research and Development Center of Excellence; Primary and Specialty Medical Care Service, VA Puget Sound Health Care System; Departments of Medicine and Health Services, University of Washington
Objectives: Guidelines for physicians regarding how to help patients with alcohol problems, developed by the NIAAA, include two approaches to alcohol screening that have not been previously validated. One is a CAGE questionnaire with a past-year timeframe, and the other is a question about the maximum number of drinks patients have on any occasion. The objectives of this study were to validate these two screening approaches, and compare them to the standard CAGE questionnaire and a previously validated screening question for episodic heavy drinking.
Methods: As a pilot study for the Ambulatory Care Quality Improvement Project (ACQUIP), alcohol screening questionnaires were evaluated in male Veterans Affairs (VA) general medicine patients (N=238; mean age 67). Mailed questionnaires included two scoring options for the CAGE questionnaire (standard and past-year timeframes) and two questions about episodic heavy drinking. All participants in this pilot study participated in a diagnostic interview to establish criteria for DSM-III-R active alcohol abuse or dependence, past-year at-risk drinking (>= 14 drinks in a typical week or >= 5 drinks on an occasion in a typical month), and either or both.
Results: Compared to all interview criteria, areas under ROC curves were similar for the past-year version of the CAGE (0.67-0.79) and the standard CAGE (0.70-0.82). However, the sensitivity of the past-year CAGE was lower than the standard CAGE for all comparison standards. For instance, at a cut-point of > 1, the sensitivity of the past-year CAGE was 75% for active alcohol abuse or dependence, 54% for at-risk drinking, and 58% for either or both, compared to 88%, 75%, and 78% respectively for the standard CAGE. The standard CAGE using a cut-point of >= 2, performed similarly to the past-year CAGE using a cut-point of >= 1. The question about the maximum number of drinks per occasion had relatively low sensitivity for detecting active alcohol abuse or dependence (69%), at-risk drinking (58%), and either or both (58%). In contrast, the comparable sensitivities of a question about the frequency of drinking 6 or more drinks on any occasion in the past year were 78%, 82%, and 78%. The question about the maximum number of drinks also had lower specificity than the latter question for all comparisons.
Conclusions: In this study of male VA general medical patients, a CAGE questionnaire with a past-year time frame offered no advantage over the standard CAGE. Furthermore, the past-year CAGE questionnaire failed to identify one fifth of all patients with at-risk drinking or active alcohol abuse and dependence, who were identified by the standard CAGE. A question about the maximum number of drinks per occasion was an inferior screen for at-risk drinking compared with a previously validated question about the frequency of drinking 6 or more drinks on an occasion.
Impact: Clinicians, researchers, and administrators may be tempted to modify validated alcohol-screening questionnaires in efforts to improve their accuracy. This study found that two previously untested approaches consistent with NIAAA recommendations missed 20% of at-risk and problem drinkers identified by validated questionnaires.