Screening tests IV

doi:10.1136/bmj.b4365

Published 28 October 2009, doi:10.1136/bmj.b4365
Cite this as: BMJ 2009;339:b4365

Endgames

Statistical question

Screening tests IV

Philip Sedgwick, senior lecturer in medical statistics

¹ Centre for Medical and Healthcare Education, St George’s, University of London, Tooting, London SW17 0RE

p.sedgwick{at}sgul.ac.uk

A study investigated the performance of the CAGE questionnaireas a screening tool for alcohol abuse or alcohol dependencyin primary care. The CAGE questionnaire consists of four questionsabout alcohol behaviour, each with a "yes" or "no" response.Clinical interview was used as the diagnostic test ("gold standard")for alcohol abuse or alcohol dependency. A total of 1970 individualswere investigated and the results obtained are shown in table1. ¹

View this table:
[in this window]
[in a new window]

Table 1 Cross tabulation of the number of "yes" responses on CAGE questionnaire against clinical diagnosis

A minimum number of "yes" responses is required for the screeningtest to identify an individual as screen "positive" and subsequentlyrecommend clinical interview. If the threshold for "positive"on the screening test was increased from one or more ( $≥$ 1) totwo or more ( $≥$ 2) "yes" responses, which of the following, ifany, would be observed?

a) The sensitivity would decrease

b) The specificity woulddecrease

c) The frequency of "true positives" would increase

d)The frequency of "false positives" would increase

Show Answers

Answers

a—Answer a is true. Sensitivity describes the accuracyof the CAGE questionnaire in correctly identifying individualswho have a clinical diagnosis of alcohol abuse or alcohol dependency.Sensitivity is the percentage of the 175 individuals with diagnosedalcohol abuse or alcohol dependency correctly identified bythe CAGE questionnaire as screen "positive."

The previous table has been re-organised to show screen "positive"thresholds of either one or more ( $≥$ 1) "yes" responses and twoor more ( $≥$ 2) "yes" responses.

View this table:
[in this window]
[in a new window]

Table 2 Cross tabulation of the CAGE screening questionnaire results against clinical diagnosis, with a screen "positive" threshold of one or more "yes" responses

View this table:
[in this window]
[in a new window]

Table 3 Cross tabulation of the CAGE screening questionnaire results against clinical diagnosis, with a screen "positive" threshold of one or more "yes" responses

If the threshold for screen "positive" were one or more ( $≥$ 1)"yes" responses, the sensitivity would be (107÷175)x100=61.1%.If the threshold for screen "positive" were increased to twoor more ( $≥$ 2) "yes" responses, the sensitivity would be (80÷175)x100=45.7%.By increasing the threshold of the number of "yes" responses,the number of individuals with diagnosed alcohol abuse or alcoholdependency correctly identified by screening reduced, therebyreducing sensitivity.

Answer b is false. Specificity describes the accuracy of theCAGE questionnaire in correctly identifying individuals whohave not been diagnosed with alcohol abuse or alcohol dependency.Specificity is the percentage of the 1795 individuals withoutdiagnosed alcohol abuse or alcohol dependency correctly identifiedby the CAGE questionnaire as screen "negative." If the thresholdfor screen "positive" were one or more "yes" responses, thespecificity would be (1560÷1795)x100=86.9%. If the thresholdfor screen "positive" were increased to two or more ( $≥$ 2) "yes"responses, the specificity would be (1700÷1795)x100=94.7%.By increasing the threshold of the number of "yes" responses,the number of individuals without diagnosed alcohol abuse oralcohol dependency correctly identified by screening increased,thereby increasing specificity.

Answer c is false. A "true positive" is someone with diagnosedalcohol abuse or alcohol dependency identified correctly bythe screening questionnaire as screen "positive." If the thresholdfor screen "positive" increased from one or more "yes" responsesto two or more, the frequency of "true positives" would decrease.

Answer d is false. A "false positive" is someone without diagnosedalcohol abuse or alcohol dependency incorrectly identified bythe screening questionnaire with a screen "positive" result.If the threshold for screen "positive" increased from one ormore "yes" responses to two or more, the frequency of "falsepositives" would decrease.

Cite this as: BMJ 2009;339:b4365

Competing interests: None

References

Aertgeertsa B, Buntinxa F, Kester A. The value of the CAGE in screening for alcohol abuse and alcohol dependence in general clinical populations: a diagnostic meta-analysis. J Clin Epidemiol 2004;57:30-9.[CrossRef][Web of Science][Medline]