Measuring tobacco use in a prison population

Background: Widespread tobacco use and high interest in quitting make prisons an ideal environment for smoking cessation interventions, however little has been done to assist prisoners in their efforts to quit. Measuring tobacco use is central to tobacco research, yet there has been only one published examination of tobacco use measures among prisoners. Methods: Personal interviews were conducted with 200 male prisoners. Four measures of tobacco use were collected: self-report, carbon monoxide (CO) breath test, salivary cotinine measured by enzyme immunoassay (EIA), and salivary cotinine measured by liquid chromatography tandem mass spectrometry (LC/MS/MS). Calculated sensitivity and specificity of the tests were compared using first self-reports, then LC/MS/MS as the gold standard. Optimum cutpoints for continuous measures were selected to maximize the Youden index. Results: With self-reported tobacco use as the standard, all tests showed excellent discrimination (CO: cut point ≥ 4 ppm, sensitivity = 88.3%, specificity = 94.9%; EIA: cut point ≥ 10 ng/ml, sensitivity = 92.2%, specificity = 94.3%; LC/MS/MS: cut point ≥ 9 ng/ml, sensitivity = 98.6%, specificity = 97.8%). Using LC/MS/MS as the gold standard produced similar, but not identical, results (CO: cut point ≥ 4 ppm, sensitivity = 85.4%, specificity = 91.7%; EIA: cut point ≥ 10 ng/ml, sensitivity = 94.2%, specificity = 100.0%;self-report: sensitivity = 98.5%, specificity = 88.9%). Conclusions: In the absence of a perfect predictor of the outcome of interest, investigators must consider how use of an imperfect "gold standard" may impact estimates of test performance.