- Email: [email protected]
Tobacco-reporting validity in an epidemiological drug-use survey
Addictive Behaviors 30 (2005) 175 – 181
Short communication
Tobacco-reporting validity in an epidemiological drug-use survey Michael Fendricha,*, Mary Ellen Mackesy-Amitia, Timothy P. Johnsonb, Amy Hubbella, Joseph S. Wislara a
Department of Psychiatry, Institute for Juvenile Research, University of Illinois at Chicago, 1747 Roosevelt Road (M/C 747), Chicago, IL 60608, USA b Survey Research Laboratory, University of Illinois at Chicago, Chicago, IL, USA
Abstract We compared responses to questions about tobacco use and passive exposure to smoking with biological tests for cotinine in order to estimate tobacco-reporting validity in an epidemiological survey on drug use. Respondents identified via multistage sampling (n = 627) completed household surveys that were administered using an Audio Computer-Assisted Self-Interview (ACASI) procedure. Following the survey, respondents were asked to participate in drug testing. Saliva (oral fluid) was used to screen for the presence of cotinine, a major metabolite of tobacco. Hair, urine, and oral fluid testing were used to detect the presence of illicit drugs such as amphetamines, cocaine, heroin, and marijuana. While underreporting of tobacco was relatively rare, estimates from this recent study suggest that it may be increasing over time. Even after adjusting for passive exposure, self-report sensitivity estimates were still well below the 90% level suggested in prior reviews. Underreporting of marijuana and race/ethnicity showed a strong association with underreporting of tobacco use, suggesting that factors associated with the underreporting of illicit substance use parallel those associated with the underreporting of tobacco use. D 2004 Elsevier Ltd. All rights reserved. Keywords: Drug testing; Epidemiology; Survey research; Tobacco; Validity
* Corresponding author. Tel.: +1-312-413-1084; fax: +1-312-413-1036. E-mail address: [email protected] (M. Fendrich). 0306-4603/$ – see front matter D 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.addbeh.2004.04.009
176
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
1. Introduction U.S. studies conducted over the past 15 years suggest overall validity of smoking selfreports is relatively high, with average sensitivity and specificity estimates typically approaching 90% (Patrick et al., 1994). Nevertheless, it is not clear that previous population-based nonintervention studies provide an accurate barometer of the validity of more recently collected epidemiological surveys. With rare exceptions (Luepker et al., 1981; Ogden, Morgan, Heavner, Davis, & Kinne, 1997; Pierce, Dwyer, DiGiusto, et al., 1987), most comparative studies investigating survey validity have informed subjects about the use of biological testing prior to survey administration, possibly inflating overall estimates of reporting accuracy. General population surveys assessing substance use and other sensitive health information increasingly employ Audio Computer-Assisted Self-Interview (ACASI) procedures (Des Jarlais et al., 1999; Metzger et al., 2000). To our knowledge there are no studies comparing the results of biological testing for smoking to reports provided under these state-of-the-art assessment conditions. Several studies have investigated relative differences between levels of cotinine associated with reports of current smoking and environmental tobacco smoke (ETS; Etter, Duc, & Perneger, 2000; Ogden et al., 1997; Wall, Johnson, Jacob, & Benowitz, 1987). Despite this, we are aware of no previous analyses that have examined the impact of adjustments for respondent passive smoking exposure on conclusions about the overall validity of tobacco-use reports. In recent years, public health concerns about smoking and the adverse impact of ETS have led to increasingly restrictive policies in the United States regarding public smoking behavior; surveys suggest that restrictions on workplace smoking, for example, increased dramatically from the mid to late 1990s (Centers for Disease Control, 2000). Since socially undesirable behavior is often underreported, increasingly undesirable perceptions regarding smoking may result in shifts towards less valid reporting of this behavior in more recent surveys. Finally, while smoking validity studies have measured illicit substance use (Wills & Cleary, 1997), we are aware of no studies examining the links between the underreporting of tobacco and underreporting of other illicit substances. This report summarizes results from a community ACASI study where subjects were asked survey questions without first being informed that responses would be confirmed by drug testing. We provide estimates of current tobacco-prevalence rates and evaluate the level of agreement between indicators of smoking-like exposure derived from surveys and testing. We explore the correlates of smoking exposure underreporting, with a particular emphasis on examining its association with the underreporting of marijuana.
2. Methods From June 2001 through January 2002, 627 residents of the city of Chicago between the ages of 18 and 40 years were randomly selected to participate in a household drug-use survey. This survey used a multistage area probability design (Kish, 1965). Specific details about survey administration and response rates have been reported elsewhere (Fendrich, Johnson, Wislar, & Hubbell, 2004; Fendrich, Johnson, Wislar, Hubbell, & Spiehler, 2004).
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
177
2.1. Tobacco use Respondents were told that questions were pertaining to ‘‘all tobacco products including cigarettes, cigars, pipes, chewing tobacco, and snuff.’’ Those who admitted to ‘‘lifetime’’ tobacco use were then asked: ‘‘How long has it been since you last used tobacco’’ and asked to select the appropriate specified time period category. Categorical responses ranged from last use ‘‘within the past 24 hours’’ to ‘‘more than 3 years ago. ‘‘ In the present study, we define ‘‘current’’ tobacco users as those who reported use within the last 3 days. 2.2. Tobacco passive exposure Subjects were first asked about the last time they were in the same room as someone who was smoking tobacco. All subjects who indicated exposure in the past month were then asked ‘‘on about how many different days in the past month were you in the same room as someone while they were smoking.’’ For the purposes of this study, all subjects who indicated more than 25 days in response to this question were counted as current passiveexposure cases. 2.3. Drug testing Immediately following the drug assessment portion of the survey, subjects were asked to consent to participate in hair, saliva (hereafter referred to as ‘‘oral fluid’’), and urine testing procedures. Note that subjects were invited to participate in the drug testing after substance use questions had been completed. Specimens screened positive for illicit drugs (including marijuana) were confirmed using gas chromatography/mass spectrometry (GC/MS). Cotinine tests were not confirmed. 2.4. Drug test classification In the present analyses, all subjects who took at least one drug test and who were confirmed positive for cannabinoids by any of the three tests were classified as marijuana positive cases. A review of the tobacco screening literature suggests that a cotinine level of 20 is a reasonable cutoff for ‘‘smoking-like’’ exposure (Patrick et al., 1994). All subjects taking the oral fluid cotinine test who screened positive for cotinine at the 20-ng/ml level were counted as cotinine positive for smoking. In analyses of underreporting, marijuana underreporters were those who tested positive for marijuana but did not disclose marijuana use within the past 30 days. Tobacco underreporters were those who tested positive for cotinine but did not report tobacco use within the past 3 days (current use) or who were not classified as passively exposed to tobacco smoke. All prevalence estimates were adjusted using poststratification weights (Lohr, 1999) and confidence intervals were adjusted for design effects using Stata survey (svy) procedures (StataCorp, 2001). Correlates of tobacco underreporting were examined via Stata cluster-adjusted logistic regression procedures (StataCorp, 2001), with standard errors adjusted for clustering within sampling units (blocks within census tracts).
178
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
3. Results Valid cotinine tests, obtained for 536 subjects, ranged from 0 ng/ml to over 500 ng/ml, with a median value of 7 ng/ml. Just over one third of the sample with cotinine tests (n = 207; 38%) met the 20-ng/ml threshold to be classified as cotinine positive for smoking. Table 1 summarizes the survey prevalence estimates of smoking and smoking exposure along with validity statistics comparing the survey to drug test data. Weighted prevalence estimates indicated that about 37% of the sample self-reported current smoking (95% confidence interval [CI]: 31.4, 42.7). When reports of passive exposure are considered, the revised weighted survey estimate of current tobacco use/exposure increases to 44.7% (95% CI 38.1, 51.4). Comparisons between the two self-report tobacco-use indicators and the cotinine test results suggest excellent levels of agreement. When self-report is limited to tobacco use, the sensitivity of self-reports is 78% and the specificity is 92%. The chance-corrected agreement statistic, kappa (Fliess, 1981), approaches an excellent value of .72. When self-report is expanded to include reports of passive exposure to smoking, sensitivity increases to 84% and specificity drops to 82%. In addition, kappa drops to .64. Decreases in the validity statistics reflect the fact that only 11 of the 46 new cases identified through responses to questions about passive exposure actually met the smoking classification threshold of 20 ng/ml. Analyses focusing on the subsample of subjects (n = 72) who tested positive for both marijuana (by any method) and cotinine (by oral fluid) indicated that 49 (68%) disclosed use of marijuana within the past month; the remaining 23 subjects (32%) denied use of this substance within this time frame. Only 6% of those disclosing past-month marijuana use underreported their tobacco use/exposure, compared to 26% of those denying past month marijuana use (Fisher’s exact test P < .05). In the logistic regression model examining correlates of cotinine underreporting among those testing positive for cotinine (Table 2), each square root unit increase in cotinine level resulted in significantly reduced odds (.78) of underreporting of tobacco use/exposure among respondents (95% CI: .69, .88). In addition, compared with White/other subjects, African Table 1 Prevalence of tobacco use and exposure and validity of self-report measures n Current user Exposed and/or user Nonexposed nonusers Total N
188 234 302 536
Prevalencea %
95% CI
36.9 44.7 55.3
31.4 38.1 48.6
42.7 51.4 61.9
Sensitivityb (%)
Specificityc (%)
Kappad
78.3 83.6 –
92.1 81.5 –
.72 .64
Current user: self-reported use of tobacco within past 3 days. Exposed: self-reported exposure to secondhand tobacco smoke on more than 25 days in the past month. a Weighted and cluster adjusted. b Sensitivity: % of cotinine-positive (z 20 ng/ml) cases not reporting tobacco use/exposure. c Specificity: % cotinine-negative cases not reporting tobacco use/exposure. d Agreement between self-report and cotinine test.
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
179
Table 2 Logistic regression predicting underreporting of tobacco use/exposure oral test for cotinine z 20 ng/ml (N = 189)a Variable OR S.E. 95%CI Square root cotinine ng/ml Male vs. female African American vs. White/other Hispanic vs. White/other Age 26 – 30 years vs. 18 – 25 years 31 – 40 years vs. 18 – 25 years SES Low vs. middle Low vs. high Denial vs. disclosure/positive MJ testb Negative vs. positive test/MJ disclosureb a b
0.78 1.95 5.93 1.63
0.05 1.13 4.36 0.90
0.69 0.63 1.40 0.55
0.88 6.06 25.05 4.81
0.46 0.65
0.29 0.28
0.14 0.29
1.55 1.49
0.49 2.44 12.38 1.45
0.27 1.99 10.09 1.74
0.16 0.49 2.51 0.14
1.45 12.10 61.17 15.29
Cluster-adjusted for sampling unit (block). Based on contrasts estimated from a 22 interaction of marijuana test results and marijuana self-report.
American subjects had 5.9 times the odds of underreporting recent tobacco use/exposure. Finally, a 22 interaction of marijuana test results and marijuana self-report was coded into a variable having four levels. The interaction was significant and contrasts showed that among those who tested positive for marijuana, those who failed to report past-month marijuana use had 12.4 times the odds of underreporting tobacco use/exposure compared to those who disclosed past-month marijuana use (95% CI: 2.5, 61.2).
4. Discussion When passive exposure to smoking was accounted for, sensitivity approached, but still fell short of the 90% benchmark described in Patrick et al.’s (1994) review. In light of workforce policy shifts reinforcing the socially undesirable nature of tobacco consumption, our findings may reflect a trend toward diminished tobacco-reporting validity in surveys; tobaccoreporting behavior may increasingly parallel the reporting of illicit substances. The bivariate association between tobacco and marijuana underreporting, along with the results of the logistic regression, suggest that even though tobacco underreporting is rare, when it occurs, it tends to occur at much higher rates among people who also underreport other illicit substances. Furthermore, a key variable that has been associated with drug abuse underreporting in other contexts, race/ethnicity (Fendrich et al., 2004; Johnson & Bowman, 2003), also turned out to be a significant predictor of tobacco-use underreporting. Consistent with prior research suggesting that African Americans are particularly reluctant to report stigmatized drug use behavior, African Americans were more reluctant than other subjects to report tobacco use when they tested positive for cotinine in this study. This is also consistent with prior research on tobacco use focused mainly on adolescents and young adults (Bauman & Ennett, 1994; Wagenknecht, Burke, Perkins, Haley, & Friedman, 1992; Wills & Cleary, 1997). Thus, for a relatively small subset of tobacco users, reluctance to disclose this behavior
180
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
parallels the reluctance to disclose other drug use in other contexts. For this small subgroup of respondents, reporting of tobacco use, like the reporting of drug use, may be perceived of as socially undesirable or stigmatizing behavior. In light of these findings, this study underscores the value of community studies that use a combination of survey questions about use and exposure as well as biological tests to gauge tobacco use/exposure.
Acknowledgements This research was supported by National Institute on Drug Abuse Grant R01DA12425. The data collection for this study was carried out by the University of Illinois Survey Research Laboratory (SRL). We gratefully acknowledge the project support provided for this study by Christine Orland of SRL. An earlier version of this paper was presented at the Society for Research on Nicotine and Tobacco 9th Annual Meeting in New Orleans, Louisiana, in February 2003.
References Bauman, K. E., & Ennett, S. E. (1994). Tobacco use by black and white adolescents: The validity of self-reports. American Journal of Public Health, 84, 394 – 398. Centers for Disease Control (2000). State-specific prevalence of current cigarette smoking among adults and the proportion of adults who work in a smoke-free environment—United States, 1999. Morbidity and Mortality Weekly Report, 49, 978 – 982. Des Jarlais, D. C., Paone, D., Milliken, J., Turner, C. F., Miller, H., Gribble, J., Shi, Q., Hagan, H., & Friedman, S. R. (1999). Audio-computer interviewing to measure risk behavior for HIV among injecting drug users: A quasi-randomized trial. Lancet, 353, 1661. Etter, J. F., Duc, T. V., & Perneger, T. V. (2000). Saliva cotinine levels in smokers and nonsmokers. American Journal of Epidemiology, 151(3), 251 – 258. Fendrich, M., Johnson, T., Wislar, J. S., & Hubbell, A. (2004). Drug test feasibility in a general population household survey. Drug and Alcohol Dependence, 73, 237 – 250. Fendrich, M., Johnson, T., Wislar, J. S., Hubbell, A., & Spiehler, V. (2004). The utility of drug testing in epidemiological research: Results from a general population survey. Addiction, 99(2), 197 – 208. Fliess, J. (1981). Statistical methods for rates and proportions. New York: Wiley. Johnson, T. P., & Bowman, P. J. (2003). Cross-cultural sources of measurement error in substance use surveys. Substance Use and Misuse, 38, 1447 – 1490. Kish, L. (1965). Survey sampling. New York: Wiley. Lohr, S. L. (1999). Sampling: Design and analysis. Pacific Grove, CA: Duxbury Press. Luepker, R. V., Pechacek, T. F., Murray, D. M., Johnson, C. A., Hund, F., & Jacobs, D. R. (1981). Saliva thiocyanate: A chemical indicator of cigarette smoking in adolescents. American Journal of Public Health, 71, 1320 – 1324. Metzger, D. S., Koblin, B., Turner, C., Navaline, H., Francesca, V., Holte, S., Gross, M., Sheon, A., Miller, H., Cooley, P., & Seage, G. R. (2000). Randomized controlled trial of audio computer-assisted self interviewing: Utility and acceptability in longitudinal studies. American Journal of Epidemiology, 152, 99 – 106. Ogden, M. W., Morgan, W. T., Heavner, D. L., Davis, R. A., & Steichen, T. J. (1997). National incidence of smoking and misclassification among the U.S. married female population. Journal of Clinical Epidemiology, 50, 253 – 263.
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
181
Patrick, D., Cheadle, A., Thompson, D. C., Diehr, P., Koepsell, T., & Kinne, S. (1994). The validity of selfreported smoking: A review and meta-analysis. American Journal of Public Health, 84, 1086 – 1093. Pierce, J. P., Dwyer, T., DiGiusto, E., Carpenter, T., Hannam, C., Amin, A., Yong, C., Sarfaty, G., Shaw, J., & Burke, N. (1987). Cotinine validation of self-reported smoking in commercially run community surveys. Journal of Chronic Diseases, 40, 689 – 695. StataCorp (2001). Stata statistical software: Release 7.0. College Station, TX: Stata. Wagenknecht, L. E., Burke, G. L., Perkins, L. L., Haley, N. J., & Friedman, G. D. (1992). Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. American Journal of Public Health, 82, 33 – 36. Wall, M. A., Johnson, J., Jacob, P., & Benowitz, N. L. (1987). Cotinine in the serum, saliva, and urine of nonsmokers, passive smokers and active smokers. American Journal of Public Health, 78, 699 – 701. Wills, T. A., & Cleary, S. D. (1997). The validity of self-reports of smoking: Analyses by race/ethnicity in a school sample of urban adolescents. American Journal of Public Health, 87, 56 – 61.
Short communication
Tobacco-reporting validity in an epidemiological drug-use survey Michael Fendricha,*, Mary Ellen Mackesy-Amitia, Timothy P. Johnsonb, Amy Hubbella, Joseph S. Wislara a
Department of Psychiatry, Institute for Juvenile Research, University of Illinois at Chicago, 1747 Roosevelt Road (M/C 747), Chicago, IL 60608, USA b Survey Research Laboratory, University of Illinois at Chicago, Chicago, IL, USA
Abstract We compared responses to questions about tobacco use and passive exposure to smoking with biological tests for cotinine in order to estimate tobacco-reporting validity in an epidemiological survey on drug use. Respondents identified via multistage sampling (n = 627) completed household surveys that were administered using an Audio Computer-Assisted Self-Interview (ACASI) procedure. Following the survey, respondents were asked to participate in drug testing. Saliva (oral fluid) was used to screen for the presence of cotinine, a major metabolite of tobacco. Hair, urine, and oral fluid testing were used to detect the presence of illicit drugs such as amphetamines, cocaine, heroin, and marijuana. While underreporting of tobacco was relatively rare, estimates from this recent study suggest that it may be increasing over time. Even after adjusting for passive exposure, self-report sensitivity estimates were still well below the 90% level suggested in prior reviews. Underreporting of marijuana and race/ethnicity showed a strong association with underreporting of tobacco use, suggesting that factors associated with the underreporting of illicit substance use parallel those associated with the underreporting of tobacco use. D 2004 Elsevier Ltd. All rights reserved. Keywords: Drug testing; Epidemiology; Survey research; Tobacco; Validity
* Corresponding author. Tel.: +1-312-413-1084; fax: +1-312-413-1036. E-mail address: [email protected] (M. Fendrich). 0306-4603/$ – see front matter D 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.addbeh.2004.04.009
176
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
1. Introduction U.S. studies conducted over the past 15 years suggest overall validity of smoking selfreports is relatively high, with average sensitivity and specificity estimates typically approaching 90% (Patrick et al., 1994). Nevertheless, it is not clear that previous population-based nonintervention studies provide an accurate barometer of the validity of more recently collected epidemiological surveys. With rare exceptions (Luepker et al., 1981; Ogden, Morgan, Heavner, Davis, & Kinne, 1997; Pierce, Dwyer, DiGiusto, et al., 1987), most comparative studies investigating survey validity have informed subjects about the use of biological testing prior to survey administration, possibly inflating overall estimates of reporting accuracy. General population surveys assessing substance use and other sensitive health information increasingly employ Audio Computer-Assisted Self-Interview (ACASI) procedures (Des Jarlais et al., 1999; Metzger et al., 2000). To our knowledge there are no studies comparing the results of biological testing for smoking to reports provided under these state-of-the-art assessment conditions. Several studies have investigated relative differences between levels of cotinine associated with reports of current smoking and environmental tobacco smoke (ETS; Etter, Duc, & Perneger, 2000; Ogden et al., 1997; Wall, Johnson, Jacob, & Benowitz, 1987). Despite this, we are aware of no previous analyses that have examined the impact of adjustments for respondent passive smoking exposure on conclusions about the overall validity of tobacco-use reports. In recent years, public health concerns about smoking and the adverse impact of ETS have led to increasingly restrictive policies in the United States regarding public smoking behavior; surveys suggest that restrictions on workplace smoking, for example, increased dramatically from the mid to late 1990s (Centers for Disease Control, 2000). Since socially undesirable behavior is often underreported, increasingly undesirable perceptions regarding smoking may result in shifts towards less valid reporting of this behavior in more recent surveys. Finally, while smoking validity studies have measured illicit substance use (Wills & Cleary, 1997), we are aware of no studies examining the links between the underreporting of tobacco and underreporting of other illicit substances. This report summarizes results from a community ACASI study where subjects were asked survey questions without first being informed that responses would be confirmed by drug testing. We provide estimates of current tobacco-prevalence rates and evaluate the level of agreement between indicators of smoking-like exposure derived from surveys and testing. We explore the correlates of smoking exposure underreporting, with a particular emphasis on examining its association with the underreporting of marijuana.
2. Methods From June 2001 through January 2002, 627 residents of the city of Chicago between the ages of 18 and 40 years were randomly selected to participate in a household drug-use survey. This survey used a multistage area probability design (Kish, 1965). Specific details about survey administration and response rates have been reported elsewhere (Fendrich, Johnson, Wislar, & Hubbell, 2004; Fendrich, Johnson, Wislar, Hubbell, & Spiehler, 2004).
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
177
2.1. Tobacco use Respondents were told that questions were pertaining to ‘‘all tobacco products including cigarettes, cigars, pipes, chewing tobacco, and snuff.’’ Those who admitted to ‘‘lifetime’’ tobacco use were then asked: ‘‘How long has it been since you last used tobacco’’ and asked to select the appropriate specified time period category. Categorical responses ranged from last use ‘‘within the past 24 hours’’ to ‘‘more than 3 years ago. ‘‘ In the present study, we define ‘‘current’’ tobacco users as those who reported use within the last 3 days. 2.2. Tobacco passive exposure Subjects were first asked about the last time they were in the same room as someone who was smoking tobacco. All subjects who indicated exposure in the past month were then asked ‘‘on about how many different days in the past month were you in the same room as someone while they were smoking.’’ For the purposes of this study, all subjects who indicated more than 25 days in response to this question were counted as current passiveexposure cases. 2.3. Drug testing Immediately following the drug assessment portion of the survey, subjects were asked to consent to participate in hair, saliva (hereafter referred to as ‘‘oral fluid’’), and urine testing procedures. Note that subjects were invited to participate in the drug testing after substance use questions had been completed. Specimens screened positive for illicit drugs (including marijuana) were confirmed using gas chromatography/mass spectrometry (GC/MS). Cotinine tests were not confirmed. 2.4. Drug test classification In the present analyses, all subjects who took at least one drug test and who were confirmed positive for cannabinoids by any of the three tests were classified as marijuana positive cases. A review of the tobacco screening literature suggests that a cotinine level of 20 is a reasonable cutoff for ‘‘smoking-like’’ exposure (Patrick et al., 1994). All subjects taking the oral fluid cotinine test who screened positive for cotinine at the 20-ng/ml level were counted as cotinine positive for smoking. In analyses of underreporting, marijuana underreporters were those who tested positive for marijuana but did not disclose marijuana use within the past 30 days. Tobacco underreporters were those who tested positive for cotinine but did not report tobacco use within the past 3 days (current use) or who were not classified as passively exposed to tobacco smoke. All prevalence estimates were adjusted using poststratification weights (Lohr, 1999) and confidence intervals were adjusted for design effects using Stata survey (svy) procedures (StataCorp, 2001). Correlates of tobacco underreporting were examined via Stata cluster-adjusted logistic regression procedures (StataCorp, 2001), with standard errors adjusted for clustering within sampling units (blocks within census tracts).
178
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
3. Results Valid cotinine tests, obtained for 536 subjects, ranged from 0 ng/ml to over 500 ng/ml, with a median value of 7 ng/ml. Just over one third of the sample with cotinine tests (n = 207; 38%) met the 20-ng/ml threshold to be classified as cotinine positive for smoking. Table 1 summarizes the survey prevalence estimates of smoking and smoking exposure along with validity statistics comparing the survey to drug test data. Weighted prevalence estimates indicated that about 37% of the sample self-reported current smoking (95% confidence interval [CI]: 31.4, 42.7). When reports of passive exposure are considered, the revised weighted survey estimate of current tobacco use/exposure increases to 44.7% (95% CI 38.1, 51.4). Comparisons between the two self-report tobacco-use indicators and the cotinine test results suggest excellent levels of agreement. When self-report is limited to tobacco use, the sensitivity of self-reports is 78% and the specificity is 92%. The chance-corrected agreement statistic, kappa (Fliess, 1981), approaches an excellent value of .72. When self-report is expanded to include reports of passive exposure to smoking, sensitivity increases to 84% and specificity drops to 82%. In addition, kappa drops to .64. Decreases in the validity statistics reflect the fact that only 11 of the 46 new cases identified through responses to questions about passive exposure actually met the smoking classification threshold of 20 ng/ml. Analyses focusing on the subsample of subjects (n = 72) who tested positive for both marijuana (by any method) and cotinine (by oral fluid) indicated that 49 (68%) disclosed use of marijuana within the past month; the remaining 23 subjects (32%) denied use of this substance within this time frame. Only 6% of those disclosing past-month marijuana use underreported their tobacco use/exposure, compared to 26% of those denying past month marijuana use (Fisher’s exact test P < .05). In the logistic regression model examining correlates of cotinine underreporting among those testing positive for cotinine (Table 2), each square root unit increase in cotinine level resulted in significantly reduced odds (.78) of underreporting of tobacco use/exposure among respondents (95% CI: .69, .88). In addition, compared with White/other subjects, African Table 1 Prevalence of tobacco use and exposure and validity of self-report measures n Current user Exposed and/or user Nonexposed nonusers Total N
188 234 302 536
Prevalencea %
95% CI
36.9 44.7 55.3
31.4 38.1 48.6
42.7 51.4 61.9
Sensitivityb (%)
Specificityc (%)
Kappad
78.3 83.6 –
92.1 81.5 –
.72 .64
Current user: self-reported use of tobacco within past 3 days. Exposed: self-reported exposure to secondhand tobacco smoke on more than 25 days in the past month. a Weighted and cluster adjusted. b Sensitivity: % of cotinine-positive (z 20 ng/ml) cases not reporting tobacco use/exposure. c Specificity: % cotinine-negative cases not reporting tobacco use/exposure. d Agreement between self-report and cotinine test.
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
179
Table 2 Logistic regression predicting underreporting of tobacco use/exposure oral test for cotinine z 20 ng/ml (N = 189)a Variable OR S.E. 95%CI Square root cotinine ng/ml Male vs. female African American vs. White/other Hispanic vs. White/other Age 26 – 30 years vs. 18 – 25 years 31 – 40 years vs. 18 – 25 years SES Low vs. middle Low vs. high Denial vs. disclosure/positive MJ testb Negative vs. positive test/MJ disclosureb a b
0.78 1.95 5.93 1.63
0.05 1.13 4.36 0.90
0.69 0.63 1.40 0.55
0.88 6.06 25.05 4.81
0.46 0.65
0.29 0.28
0.14 0.29
1.55 1.49
0.49 2.44 12.38 1.45
0.27 1.99 10.09 1.74
0.16 0.49 2.51 0.14
1.45 12.10 61.17 15.29
Cluster-adjusted for sampling unit (block). Based on contrasts estimated from a 22 interaction of marijuana test results and marijuana self-report.
American subjects had 5.9 times the odds of underreporting recent tobacco use/exposure. Finally, a 22 interaction of marijuana test results and marijuana self-report was coded into a variable having four levels. The interaction was significant and contrasts showed that among those who tested positive for marijuana, those who failed to report past-month marijuana use had 12.4 times the odds of underreporting tobacco use/exposure compared to those who disclosed past-month marijuana use (95% CI: 2.5, 61.2).
4. Discussion When passive exposure to smoking was accounted for, sensitivity approached, but still fell short of the 90% benchmark described in Patrick et al.’s (1994) review. In light of workforce policy shifts reinforcing the socially undesirable nature of tobacco consumption, our findings may reflect a trend toward diminished tobacco-reporting validity in surveys; tobaccoreporting behavior may increasingly parallel the reporting of illicit substances. The bivariate association between tobacco and marijuana underreporting, along with the results of the logistic regression, suggest that even though tobacco underreporting is rare, when it occurs, it tends to occur at much higher rates among people who also underreport other illicit substances. Furthermore, a key variable that has been associated with drug abuse underreporting in other contexts, race/ethnicity (Fendrich et al., 2004; Johnson & Bowman, 2003), also turned out to be a significant predictor of tobacco-use underreporting. Consistent with prior research suggesting that African Americans are particularly reluctant to report stigmatized drug use behavior, African Americans were more reluctant than other subjects to report tobacco use when they tested positive for cotinine in this study. This is also consistent with prior research on tobacco use focused mainly on adolescents and young adults (Bauman & Ennett, 1994; Wagenknecht, Burke, Perkins, Haley, & Friedman, 1992; Wills & Cleary, 1997). Thus, for a relatively small subset of tobacco users, reluctance to disclose this behavior
180
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
parallels the reluctance to disclose other drug use in other contexts. For this small subgroup of respondents, reporting of tobacco use, like the reporting of drug use, may be perceived of as socially undesirable or stigmatizing behavior. In light of these findings, this study underscores the value of community studies that use a combination of survey questions about use and exposure as well as biological tests to gauge tobacco use/exposure.
Acknowledgements This research was supported by National Institute on Drug Abuse Grant R01DA12425. The data collection for this study was carried out by the University of Illinois Survey Research Laboratory (SRL). We gratefully acknowledge the project support provided for this study by Christine Orland of SRL. An earlier version of this paper was presented at the Society for Research on Nicotine and Tobacco 9th Annual Meeting in New Orleans, Louisiana, in February 2003.
References Bauman, K. E., & Ennett, S. E. (1994). Tobacco use by black and white adolescents: The validity of self-reports. American Journal of Public Health, 84, 394 – 398. Centers for Disease Control (2000). State-specific prevalence of current cigarette smoking among adults and the proportion of adults who work in a smoke-free environment—United States, 1999. Morbidity and Mortality Weekly Report, 49, 978 – 982. Des Jarlais, D. C., Paone, D., Milliken, J., Turner, C. F., Miller, H., Gribble, J., Shi, Q., Hagan, H., & Friedman, S. R. (1999). Audio-computer interviewing to measure risk behavior for HIV among injecting drug users: A quasi-randomized trial. Lancet, 353, 1661. Etter, J. F., Duc, T. V., & Perneger, T. V. (2000). Saliva cotinine levels in smokers and nonsmokers. American Journal of Epidemiology, 151(3), 251 – 258. Fendrich, M., Johnson, T., Wislar, J. S., & Hubbell, A. (2004). Drug test feasibility in a general population household survey. Drug and Alcohol Dependence, 73, 237 – 250. Fendrich, M., Johnson, T., Wislar, J. S., Hubbell, A., & Spiehler, V. (2004). The utility of drug testing in epidemiological research: Results from a general population survey. Addiction, 99(2), 197 – 208. Fliess, J. (1981). Statistical methods for rates and proportions. New York: Wiley. Johnson, T. P., & Bowman, P. J. (2003). Cross-cultural sources of measurement error in substance use surveys. Substance Use and Misuse, 38, 1447 – 1490. Kish, L. (1965). Survey sampling. New York: Wiley. Lohr, S. L. (1999). Sampling: Design and analysis. Pacific Grove, CA: Duxbury Press. Luepker, R. V., Pechacek, T. F., Murray, D. M., Johnson, C. A., Hund, F., & Jacobs, D. R. (1981). Saliva thiocyanate: A chemical indicator of cigarette smoking in adolescents. American Journal of Public Health, 71, 1320 – 1324. Metzger, D. S., Koblin, B., Turner, C., Navaline, H., Francesca, V., Holte, S., Gross, M., Sheon, A., Miller, H., Cooley, P., & Seage, G. R. (2000). Randomized controlled trial of audio computer-assisted self interviewing: Utility and acceptability in longitudinal studies. American Journal of Epidemiology, 152, 99 – 106. Ogden, M. W., Morgan, W. T., Heavner, D. L., Davis, R. A., & Steichen, T. J. (1997). National incidence of smoking and misclassification among the U.S. married female population. Journal of Clinical Epidemiology, 50, 253 – 263.
M. Fendrich et al. / Addictive Behaviors 30 (2005) 175–181
181
Patrick, D., Cheadle, A., Thompson, D. C., Diehr, P., Koepsell, T., & Kinne, S. (1994). The validity of selfreported smoking: A review and meta-analysis. American Journal of Public Health, 84, 1086 – 1093. Pierce, J. P., Dwyer, T., DiGiusto, E., Carpenter, T., Hannam, C., Amin, A., Yong, C., Sarfaty, G., Shaw, J., & Burke, N. (1987). Cotinine validation of self-reported smoking in commercially run community surveys. Journal of Chronic Diseases, 40, 689 – 695. StataCorp (2001). Stata statistical software: Release 7.0. College Station, TX: Stata. Wagenknecht, L. E., Burke, G. L., Perkins, L. L., Haley, N. J., & Friedman, G. D. (1992). Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. American Journal of Public Health, 82, 33 – 36. Wall, M. A., Johnson, J., Jacob, P., & Benowitz, N. L. (1987). Cotinine in the serum, saliva, and urine of nonsmokers, passive smokers and active smokers. American Journal of Public Health, 78, 699 – 701. Wills, T. A., & Cleary, S. D. (1997). The validity of self-reports of smoking: Analyses by race/ethnicity in a school sample of urban adolescents. American Journal of Public Health, 87, 56 – 61.