The American Journal of Managed Care October 2006
Implementation of Evidence-based Alcohol Screening in the Veterans Health Administration
Based on the prevalence of alcohol use disorders among VA populations and on research showing that alcohol misuse is seldom recognized by primary care providers without screening, a 1997 VA performance measure required annual alcohol screening with any validated screening questionnaire. Following baseline performance of 2% in 1996, 40% of patients were screened by the end of 1997. By 2000, the rates of alcohol screening were high (85%), but most sites were only screening for alcohol use disorders using the CAGE questionnaire.40 Although this screening resulted in increased identification of patients with alcohol use disorders,10 only a minority of patients with alcohol misuse reported receiving any alcohol-related advice.41 The VA Office of Quality and Performance became especially concerned by survey data indicating that, among patients who wanted help with their drinking, only 12% of heavy drinkers and 17% of very heavy drinkers reported receiving it.42
Educating Quality Managers
In response to the observed deficits, the VA Office of Quality and Performance consulted with researchers and organized 2 video presentations in early 2003 to educate national and facility quality managers regarding evidence-based alcohol screening and brief alcohol counseling. These presentations noted the following 2 deficits in current VA performance: (1) the CAGE questionnaire alone was being used and (2) brief alcohol counseling was offered predominantly to severely affected patients or to those with medical contraindications to drinking. Patients with milder misuse who benefit from brief alcohol counseling were not being identified or receiving counseling.41 In addition, the presentation introduced managers to a validated 3-item screening questionnaire for alcohol misuse, the AUDIT-Consumption (AUDIT-C) questions (Table 2).43,44 Screening for alcohol misuse was proposed as the essential first step9 toward implementation of evidence-based brief alcohol counseling.
Adopting the New Performance Measure
In June 2003, after reviewing the evidence for screening validity and efficiency, the VA adopted the AUDIT-C as the only brief screen to meet the alcohol misuse screening performance measure. Sites that wanted additional screening information could add questions from the CAGE questionnaire or use the full AUDIT. At that time, 91% of alcohol screening nationwide in the VA used the CAGE questionnaire, and VA facilities had until December 2003 to switch to an acceptable screen for alcohol misuse.
The announcement in July 2003 generated many questions about the rationale for this performance measure and created almost instant, unanticipated interest in education regarding evidence-based alcohol screening and counseling. Informal consultations indicated that local clinical leaders did not understand the "paradigm shift" to the broader spectrum of alcohol misuse. For example, some sites substituted the AUDIT-C for the CAGE questionnaire in electronic clinical reminders and mistakenly initiated referral to specialized alcohol treatment for all patients who screened positive. Given that more than 20% of VA patients screen positive on the AUDIT-C at the recommended thresholds for VA (=4 points for men and =3 points for women), clinicians in addiction treatment programs became concerned about the anticipated high volume of inappropriate referrals.
The VA's Substance Use Disorders Quality Enhancement Research Initiative, Center of Excellence in Substance Abuse Treatment and Education, and VA Office of Quality and Performance collaborated to address these educational needs. Initially, responses were made individually via e-mail correspondence, conference calls, or sharing of the original presentations with quality managers. After distributing a frequently-asked-questions document and revising the performance measure technical manual to explain the change, requests for technical assistance subsided. The frequently-asked-questions document addressed the paradigm shift from screening for alcohol dependence to screening for the spectrum of risky drinking to alcohol use disorders. In addition, a strategy for risk-stratifying patients based on prior alcohol treatment or AUDIT-C scores of 8 or higher45 was included to help providers and managers identify patients most likely to have alcohol use disorders.
CPRS Clinical Reminder for Alcohol Screening
The VA programmers revised a self-scoring clinical reminder for the AUDIT-C in the fall of 2003 and provided it to a national VA clinical reminder developer and opinion leader, who made it available nationwide. Many clinicians believed that screening nondrinkers with the 3-item AUDIT-C would be too burdensome, so the clinical reminder included a "skip out" question for nondrinkers: "In the past 12 months, has the patient had any drinks containing alcohol?"
Implementing the New Performance Measure
Implementation of the performance measure for annual alcohol misuse screening was delayed until December 2003 to allow facilities to install the CPRS self-scoring AUDIT-C clinical reminder. The target screening rates were set at 82% (successful) and 89% (exceptional) of primary care patients within each network. Patients who had consistent documentation of no alcohol use or who had evidence of involvement in substance use disorder treatment were not required to be screened.
Many clinicians and quality managers questioned the recommended screening threshold for the AUDIT-C. Interview studies43,44 among VA and national samples found that 4 points or higher on the AUDIT-C is the optimal screening threshold for alcohol misuse in men and 2 to 3 points or higher in women. However, patients can screen positive for alcohol misuse at these thresholds even when they report drinking below recommended limits. Specifically, patients who report drinking 1 to 2 drinks per day or 3 to 4 drinks 2 to 3 days per week have positive AUDIT-C scores of 4 (Tables 1 and 2). Some of these patients have false-positive screening test results, but interview studies43,44 have shown that many are underreporting their typical alcohol use and actually drink at risky levels or meet criteria for alcohol use disorders.
Ongoing educational efforts have sought to inform clinicians and managers about these issues. Specifically, education has stressed that the AUDIT-C score (range, 0-12) is a valid screening measure, although responses to the AUDIT-C are often underestimations of actual alcohol consumption. Therefore, many patients with AUDIT-C scores of 4 to 5 who report drinking below recommended limits on the AUDIT-C will meet criteria for risky drinking or alcohol use disorders based on in-depth interviews. Concerns about the AUDIT-C screening threshold were addressed in the frequently-asked-questions document, which encouraged providers to assess screen-positive patients' alcohol use and to provide explicit education about recommended drinking limits (Table 1).5 For patients who screened positive on the AUDIT-C but who appeared not to misuse alcohol, providers were advised to explicitly educate patients about recommended drinking limits and to encourage them to stay below those limits.
COMPARISON OF 2 METHODS FOR MONITORING PERFORMANCE
Two complementary systems have been used by the VA Office of Quality and Performance to monitor network performance on alcohol screening and follow-up, namely, medical record reviews and patient surveys. Herein, we describe these systems and report the results in the first 15 months after the new performance measure was implemented.
Nationwide Medical Record Reviews
The VA contracts with an outside agency to conduct standardized medical record reviews of outpatients. The External Peer Review Program (EPRP) selects patients who have been receiving VA care for more than 1 year based on a complex sampling strategy that has been refined over time. In general, patients are sampled from each facility monthly to permit quarterly estimates of network performance. The eligible sample for each EPRP measure is also explicitly defined (footnotes to Table 3).
All 21 VA networks met the new alcohol screening performance measure based on medical record review in 2004, and the self-scoring CPRS clinical reminder for AUDIT-C screening was used more than 1.5 million times. Whereas the screening prevalence of alcohol use disorders had been 4.2% with the CAGE questionnaire,10 as expected the screening prevalence of alcohol misuse with the AUDIT-C or the full AUDIT was much higher. Winter 2005 was the first quarter in which only screening with the AUDIT-C or the full AUDIT satisfied the performance measure, and Table 3 shows high rates of meeting the screening performance measure based on medical record reviews. The national prevalence of positive screens for alcohol misuse based on EPRP was 25% (range, 11%-36% across networks).
EPRP also monitored medical record documentation of follow-up assessment for patients who screened positive for alcohol misuse as a "supporting indicator." Supporting indicators are used to evaluate gaps in the quality of care and often lead to adoption of future performance measures. To satisfy the supporting indicator for follow-up for alcohol misuse, medical records of patients who screened positive for alcohol misuse at least 6 weeks before an EPRP review were required to document assessment for alcohol abuse or dependence. In the first 3 months of 2005, the rates of documented follow-up assessment for alcohol use disorders ranged from 5% to 84% among eligible patients who screened positive for alcohol misuse. However, these estimates were based on a small number of screen-positive patients in each network (27-80 patients per network), some of whom were not eligible for the follow-up measure (footnotes to Table 3). The EPRP medical record reviews are not designed to provide precise estimates for such selective subgroups of patients. Moreover, because alcohol counseling and follow-up assessment might be incorporated into clinical care but not documented for a number of reasons, medical record review may underestimate the actual rates of follow-up assessment.
National Patient Satisfaction Surveys