BMJ  2003;327:E263-E264 (20 December), doi:10.1136/bmj.327.7429.E263

BMJ USA: Editorial

Alcohol screening in primary care

Behind the numbers

What should primary care doctors do about screening for alcohol problems? Beich and colleagues (p 590) seem to have provided the answer: Nothing. Some will consider their results authoritative and conclude these activities are ineffective in general practice. But as with all systematic reviews and meta-analyses, there can be devils in the details.

Beich et al have addressed a clinically important question: "How effective is screening in general practice for identifying drinkers who can and do benefit from brief intervention?" The most trustworthy answer would come from a randomized controlled trial of screening and brief intervention compared with no systematic screening or intervention, but no such studies exist.¹ The results of such a trial would provide a direct measurement of the number needed to screen (NNS) to benefit one patient.² In the absence of studies evaluating such direct evidence, reviewers are forced to examine and link separate bodies of evidence (Is the screening test accurate? How effective is the intervention?). With the right studies, quantitative screening yield and intervention effectiveness can be combined to estimate the NNS indirectly, but with caution, since such results are subject to propagation of errors.² Perhaps to overcome this potential limitation, Beich et al used a single set of 7 trials testing intervention efficacy in primary care settings to provide both screening and intervention data for their calculated "screening effect" (a corollary of NNS). But this method, too, has potential limitations.

The authors found that the intervention itself can work, based on their number needed to treat (NNT) to benefit one patient (NNT 10, 95% confidence interval 7 to 14). However, they dispute whether a screening program to identify patients for intervention is worth the effort. Their argument largely rests on two questionable foundations: calculations of a low "screening effect" (3 patients changing drinking habits per 1000 screened) and their interpretation of how practitioners might "feel" as a result.

How did they get their numbers? To estimate primary care screening yield, Beich et al assumed that alcohol use screening to identify patients for intervention trials would yield the same proportion of patients as would screening patients for brief intervention in usual care. This assumption is unlikely to be valid. Exclusion criteria for efficacy trials eliminate people from receiving the intervention who would be eligible after usual care screening. In 6 of 7 studies in their meta-analysis, for example, participants receiving recent physician advice or alcohol treatment were excluded.¹ Research trials also carry additional participant demands, such as informed consent and increased time or extra visits for research-required activities. In 5 of 7 studies, screen-positive participants were required to attend a separate 20-30 minute visit to qualify for the study and to collect research-quality measures.¹

It is impossible to say by what amount the research process has inflated the attrition rate from screening in these trials, resulting in the low screening yield that Beich and colleagues report (25 per 1000 screened). However, two researchers involved in 3 of the 7 trials have disputed Beich's interpretation as underrepresenting their screening yield for intervention.^3,4 And in one US study excluded from Beich's meta-analysis only for not reporting the selected outcome, the screening yield was 7.7% after a one-step screening process conducted in HMO primary care waiting rooms.⁵

Any underestimate of the screening yield would critically affect the accuracy of this study's main result. While Beich and colleagues have argued that the research requirements have analogs in the primary care process (eg, informed consent is the same as obtaining permission to address lifestyle issues), this is not really credible. By equating these research-burdened screening yields precisely with those expected from usual care screening, the authors likely understate the number of patients who would benefit from screening followed by brief intervention in primary care.

In addition, their conclusions rely on their interpretation of how practitioners might "feel," given these screening numbers. The authors state that "a practitioner who experiences such a low ratio of success to workload is bound to be disappointed and reluctant to engage any further." This assertion disregards the NNS or NNT for other prevention screening efforts and how practitioners may or may not "feel" about those. Even if their quantitative results for screening yields are correct, nothing in the article supports the conclusion that practitioners will disengage.

What is the bottom line here? Beich et al estimate a positive intervention effect generally consistent with other findings for this outcome. Other concerns about the review, such as its comprehensiveness, data accuracy, and focus on a single alcohol use measure, probably do not result in a misleading conclusion about the overall intervention effect. However, the screening assumptions beneath their calculated "screening effect" raise concerns, as does their opinion of practitioners' feelings. Readers are left to decide whether the questions raised here and elsewhere (p 596) alter their confidence in Beich and colleagues' results and conclusions. I, for one, cannot give them my vote.

Kaiser Permanente Center for Health Research (a partner in the Oregon Evidence-based Practice Center) Portland, OR evelyn.whitlock{at}kpchr.org

---

Competing interests: None declared.

Papers p 590

References

1. Whitlock EP, Green CA, Polen MR. Behavioral counseling interventions in primary care to reduce risky/harmful alcohol use. Systematic evidence review. Rockville, MD, Agency for Healthcare Research and Quality. In press.
2. Rembold C. Number needed to screen: development of a statistic for disease screening. BMJ 1998;317: 307-312.[Abstract/Free Full Text]
3. Anderson P. Transcription errors and erroneous assumptions. [Rapid response, bmj.com] September 13, 2003. http://bmj.bmjjournals.com/cgi/eletters/327/7414/536#36537.
4. Heather N, Richmond R. Screening effect ten times greater than calculated by Beich et al. [Rapid response, bmj.com] October 1, 2003. http://bmj.bmjjournals.com/cgi/eletters/327/7414/536#37261.
5. Senft RA, Polen MR, Freeborn DK, Hollis JF. Brief intervention in a primary care setting for hazardous drinkers. Am J Prev Med 1997;13: 464-470.[Web of Science][Medline]

CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    StumbleUpon    Technorati    What's this?

This article has been cited by other articles:

• Whitlock, E. P., Lin, J. S., Chou, R., Shekelle, P., Robinson, K. A. (2008). Using Existing Systematic Reviews in Complex Systematic Reviews. ANN INTERN MED 148: 776-782 [Abstract] [Full text]  
• Beich, A., Gannik, D., Saelan, H., Thorsen, T. (2007). Screening and brief intervention targeting risky drinkers in danish general practice A pragmatic controlled trial. Alcohol Alcohol 42: 593-603 [Abstract] [Full text]  
• Whitlock, E. P., Polen, M. R., Green, C. A., Orleans, T., Klein, J. (2004). Behavioral Counseling Interventions in Primary Care To Reduce Risky/Harmful Alcohol Use by Adults: A Summary of the Evidence for the U.S. Preventive Services Task Force. ANN INTERN MED 140: 557-568 [Abstract] [Full text]  

Online poll

Find out more>>