Abstract
Background:
Accurate prevalence estimates are critical to epidemiological research but discordant responses on self-report surveys can lead to over- or underestimation of drug use. We sought to examine the extent and nature of underreported cannabis use (among those later reporting blunt use) from a national drug survey in the US.
Methods:
We used data from the 2015–2019 National Survey on Drug Use and Health (N = 281,650), a nationally representative probability sample of non-institutionalized populations in the US. We compared self-reported prevalence of past-year cannabis use and blunt use and delineated correlates of underreporting cannabis use, defined as reporting blunt use but not overall cannabis use.
Results:
An estimated 4.8% (95% CI: 4.4–5.2) of people reported blunt use but not cannabis use. Although corrected prevalence, cannabis use recoded as use only increased from 15.2% (95% CI: 15.0–15.4) to 15.5% (95% CI: 15.3–15.7), individuals who are aged ≥50 (aOR = 1.81, 95% CI: 1.06–3.08), female (aOR = 1.35, 95% CI: 1.12–1.62), Non-Hispanic Black (aOR = 1.43, 95% CI: 1.16–1.76), or report lower English proficiency (aOR = 3.32, 95% CI: 1.40–7.83) are at increased odds for providing such a discordant response. Individuals with a college degree (aOR = 0.57, 95% CI: 0.39–0.84) and those reporting past-year use of tobacco (aOR = 0.75, 95% CI: 0.62–0.91), alcohol (aOR = 0.42, 95% CI: 0.33–0.54), cocaine (aOR = 0.50, 95% CI: 0.34–0.73), or LSD (aOR = 0.52, 95% CI: 0.31–0.87) were at lower odds of providing a discordant response.
Conclusion:
Although changes in prevalence are small when correcting for discordant responses, results provide insight into subgroups that may be more likely to underreport use on surveys.
Keywords: Marijuana, national surveys, survey reliability
Introduction
Accurately obtaining prevalence estimates of drug use is inherently important to epidemiological research on substance use, particularly when such data inform policy recommendations. The need for accurate estimates is further underscored for certain drugs, such as cannabis, owing to rapidly shifting policies that have swiftly influenced the legality and nature of use. In the United States (US), for example, fifteen states have legalized or voted to legalize recreational recreational cannabis use as of November 2020,1 while another sixteen states have decriminalized its use.2 Furthermore, prevalence of cannabis use has increased significantly over the past decade in states that legalized and among certain age groups, as has the frequency of use and the prevalence of cannabis use disorder.3–7 Common methods of cannabis consumption have also been changing, with vaping of cannabis growing greatly in popularity among adolescents in recent years.8
Self-report via survey methods is among the most common means of estimating drug use prevalence in epidemiological research primarily due to cost-efficiency and practical advantages over biological testing methods such as collecting urine samples.9 However, the potential for inconsistent or discordant reporting exists, which can potentially lead to under- or overestimates of the actual prevalence of drug use as a result. Indeed, data from national drug surveys show discordant responses for some substances.10–12 For example, a study using data from Monitoring the Future—a nationally representative high school sample in the US—found that over a quarter (28.7%) of respondents who reported nonmedical Adderall use reported no nonmedical amphetamine use on the same survey.13 Moreover, study findings have demonstrated that self-reported cannabis use may be underreported,14–17 with one study showing that hair or urine testing detected past-month cannabis use among 11.2% of those who otherwise did not report use in a self-report survey.18 Similarly, demographic factors such as race or level of education have also been shown to influence discordant reporting.10,19
While research into potential reasons underlying discordant responses would be welcome, one can speculate that a contributing factor may be respondents’ lack of familiarity with varying drug nomenclature. For example, in the aforementioned study on nonmedical amphetamine use, people who used Adderall may not have been aware that Adderall is one of several commercial names for amphetamine salts.13 Cannabis itself is a drug with a myriad of names and associations, such as (but not limited to) pot, grass, weed, hash, joints, and blunts.20 To this end, the annual National Survey on Drug Use and Health (NSDUH) provides a unique opportunity to examine potential discordant responses regarding cannabis use because it queries blunt use in a section that is independent of responses to the cannabis use assessment earlier on the survey (i.e., not completely based on gate questions and skip-logic).21 Correctly estimating prevalence of blunt use is important given that it increases risk for adverse health outcomes such as cannabis use disorder, nicotine dependence, elevated heart rate, and acute increases in carbon monoxide levels.22–24 Accordingly, this study seeks to determine the extent to which possible underreporting of cannabis use on the NSDUH takes place, as well as potential correlates of underreporting based on factors known to affect discordant reporting. Findings may be used to inform future study designs.
Method
Procedure and participants
We conducted a secondary data analysis using NSDUH, a cross-sectional nationally representative survey of non-institutionalized individuals aged ≥12 in the US. Each year, a sample is obtained using a multistage design. Census tracts are first selected in each of the 50 US states and within the District of Columbia, and then segments within each tract are randomly assigned for data collection. Dwellings within segments and then participants within dwellings are selected. Surveys are administered via computer-assisted interviewing, which are conducted by an interviewer using audio-computer-assisted interviewing. We utilized data collected from the most recent five cohorts (2015–2019). The aggregate sample size was 281,650 and the weighted interview response rates from these years ranged from 64.9% to 68.4%. This secondary analysis was exempt from review by the New York University Langone Medical Center institutional review board.
Measures
For cannabis use, participants were asked whether they had used marijuana or hashish (cannabis) in the past 12 months. It was explained that marijuana is also called pot or grass, and that it is usually smoked (in cigarettes called joints or in a pipe) and that it is sometimes cooked in food.25 It was further explained that hashish is a form of cannabis also called hash, which is typically smoked in a pipe. They were also reminded that another form of hashish is hash oil. This question did not, however, ask specifically about blunt use. Later in the survey, participants were asked about past-year use of blunts, which was defined as when someone takes some tobacco out of a cigar and replaces it with marijuana. Since questions about blunts did not stem directly from the general cannabis use questions, some responses could be discordant or contradictory (i.e., self-reported blunt use without earlier reported use of cannabis). We created a variable indicating whether a discordant response was provided. Specifically, we created a binary variable indicating who reported blunt use but did not report cannabis use compared to those who reported cannabis use and did not provide a discordant response. We also created a variable indicating reported cannabis use or blunt use in order for us to provide a corrected estimate of use.
In addition to questions about cannabis/blunt use, participants were also asked about past-year use of other drugs including tobacco, alcohol, cocaine, and LSD, as well as about past-year prescription opioid misuse. Misuse was defined as using in any way not directed by a doctor, including use without a prescription, more often, in greater amounts, or longer than directed to use them, or use in any other way not directed to use.26 Participants were also asked about their age, sex, race/ethnicity, educational attainment, annual family income, and marital status. Participants were also asked how well they speak English. At the end of the survey, participants were debriefed, and the interviewer recorded whether the participant mentioned having had trouble understanding questions asked during the interview and whether the participant made any comments about the interview being too long. These two items were provided by NSDUH as binary responses of “yes” and “no”.
Analysis
First, we compared past-year self-report of cannabis use to past-year self-report of blunt use by estimating the percentage of individuals providing a discordant response with respect to cannabis/blunt use. We then compared whether a discordant response was provided according to participant characteristics in a bivariable manner using Rao-Scott chi-square.27 These independent variables were chosen based on previous research examining discordant reporting in national samples.10,13 Next, all of these variables were fit into a multivariable logistic regression model with our cannabis discord variable as the dependent variable. This allowed us to examine associations with all else being equal and this model generated adjusted odds ratios (aORs) for each independent variable. Finally, we estimated “corrected” prevalence of cannabis use including the original discordant responses coded affirmatively as reported cannabis use. We also estimated corrected prevalence of use stratified by each level of each independent variable to determine the absolute and relative differences in use by each characteristic.
We used sample weights to account for the complex survey design, selection probability, population distribution, and non-response, and weights were divided by five to account for aggregation of data from five cohorts. Stata 13 SE was used to conduct these analyses and we used Taylor series estimation methods to provide accurate standard errors.27
Results
An estimated 15.2% (95% confidence interval [CI]: 15.0–15.4) of individuals ageed ≥12 in the US used cannabis in the past year (uncorrected estimate) and an estimated 6.7% (95% CI: 6.6–6.8) of individuals in the US used blunts in the past year. We estimated that 4.8% (95% CI: 4.4–5.2) of people who reported blunt use did not report cannabis use, suggesting that some cannabis use is underreported. When accounting for these discordant responses, our corrected prevalence estimate for past-year cannabis use increased to 15.5% (95% CI: 15.3–15.7). Among those reporting past-year cannabis use, 42.1% (95% CI: 41.4–42.7) reported past-year blunt use, but this estimate increased to 43.3% (95% CI: 42.6–43.9)—a 1.2% absolute increase—when examining the percentage of those reporting blunt use using the corrected cannabis use variable.
With respect to correlates of providing a discordant response, as is shown in Table 1, bivariable tests indicated significant differences regarding age (p < .001), sex (p = .002), race/ethnicity (p < .001), education (p < .001), income (p = .001), past-year use of tobacco (p < .001), alcohol (p < .001), cocaine (p < .001), and LSD (p < .001), and prescription opioid misuse (p < .001), and ability to speak English (p < .001). In the multivariable model, with all else being equal, compared to adolescents, those aged ≥50 were at higher odds of providing a discordant response (adjusted odds ratio [aOR] = 1.81, 95% CI: 1.06–3.08), and compared to males, females were at higher odds of providing a discordant response (aOR = 1.35, 95% CI: 1.12–1.62). Compared to Non-Hispanic White individuals, Non-Hispanic Black individuals were at higher odds for providing a discordant response (aOR = 1.43, 95% CI: 1.16–1.76), and compared to those with a high school education or less, those with some college (aOR = 0.69, 95% 0.55–0.85) or a college degree (aOR = 0.57, 95% CI: 0.39–0.84) were at lower odds for providing a discordant response. With regard to other drug use, those reporting past-year use of tobacco (aOR = 0.75, 95% CI: 0.62–0.91), alcohol (aOR = 0.42, 95% CI: 0.33–0.54), cocaine (aOR = 0.50, 95% CI: 0.34–0.73), or LSD (aOR = 0.52, 95% CI: 0.31–0.87) were at lower odds for providing a discordant response. In addition, compared to those who speak English very well, those reporting that they only speak English well (aOR = 1.50, 95% CI: 1.10–2.06) or not well or not at all (aOR = 3.32, 95% CI: 1.40–7.83) were at increased odds for providing a discordant response.
Table 1.
Correlates of reporting past-year blunt use but not past-year cannabis use.
| Full sample % (95% CI) | No discord % (95% CI) | Discord % (95% CI) | aOR (95% CI) | |
|---|---|---|---|---|
| Age | ||||
| 12–17 | 9.2 (9.0–9.3) | 93.1 (92.1–93.9) | 6.9 (6.1–7.9)c | 1.00 |
| 18–25 | 12.6 (12.5–12.8) | 96.1 (95.6–96.5) | 3.9 (3.5–4.4) | 1.05 (0.81–1.36) |
| 26–34 | 14.5 (14.3–14.7) | 95.9 (95.2–96.5) | 4.1 (3.5–4.8) | 1.15 (0.86–1.54) |
| 35–49 | 22.4 (22.1–22.7) | 94.9 (93.8–95.8) | 5.1 (4.2–6.2) | 1.15 (0.83–1.61) |
| ≥50 | 41.3 (40.8–41.8) | 91.7 (87.5–94.6) | 8.3 (5.4–12.5) | 1.81 (1.06–3.08)a |
| Sex | ||||
| Male | 48.5 (48.2–48.8) | 95.8 (95.2–96.2) | 4.2 (3.8–4.8)b | 1.00 |
| Female | 51.5 (51.2–51.8) | 94.5 (93.9–95.1) | 5.5 (4.9–6.1) | 1.35 (1.12–1.62)b |
| Race/ethnicity | ||||
| Non-Hispanic White | 62.8 (62.3–63.3) | 96.1 (95.6–96.5) | 3.9 (3.5–4.4)c | 1.00 |
| Non-Hispanic Black | 12.0 (11.7–12.4) | 93.1 (92.0–94.0) | 6.9 (6.0–8.0) | 1.43 (1.16–1.76)b |
| Hispanic | 16.8 (16.4–17.2) | 94.9 (93.9–95.7) | 5.1 (4.3–6.1) | 1.13 (0.89–1.43) |
| Other/mixed | 8.4 (8.2–8.6) | 96.3 (94.9–97.3) | 3.7 (2.7–5.1) | 0.90 (0.65–1.24) |
| Education | ||||
| High School or Less | 34.1 (33.7–34.5) | 94.0 (93.3–94.7) | 6.0 (5.3–6.7)c | 1.00 |
| Some College | 28.0 (27.7–28.3) | 96.5 (95.8–97.1) | 3.5 (2.9–4.3) | 0.69 (0.55–0.85)b |
| College Degree or Higher | 28.7 (28.3–29.1) | 97.2 (96.2–97.9) | 2.8 (2.1–3.8) | 0.57 (0.39–0.84)b |
| Annual Family Income | ||||
| <$20,000 | 16.2 (15.9–16.5) | 94.2 (93.3–95.0) | 5.8 (5.0–6.7)c | 1.00 |
| $20,000–$49,999 | 29.3 (28.9–29.7) | 95.2 (94.7–95.7) | 4.8 (4.3–5.3) | 0.88 (0.70–1.10) |
| $50,000–$74,999 | 15.8 (15.6–16.1) | 95.1 (93.8–96.2) | 4.9 (3.8–6.2) | 0.94 (0.67–1.32) |
| ≥$75,000 | 38.7 (38.1–39.2) | 96.4 (95.8–97.0) | 3.6 (3.1–4.2) | 0.80 (0.62–1.03) |
| Married | ||||
| No | 50.8 (50.4–51.2) | 95.5 (95.1–95.9) | 4.5 (4.1–4.9) | 1.00 |
| Yes | 49.2 (48.8–49.6) | 94.8 (93.3–95.9) | 5.2 (4.1–6.7) | 1.03 (0.76–1.40) |
| Past-Year Other Drug Use | ||||
| Tobacco | 27.7 (27.3–28.0) | 95.9 (95.5–96.3) | 4.1 (3.7–4.5)c | 0.75 (0.62–0.91)b |
| Alcohol | 65.4 (65.1–65.7) | 96.0 (95.6–96.3) | 4.0 (3.7–4.4)c | 0.42 (0.33–0.54)c |
| Cocaine | 2.0 (1.9–2.1) | 98.1 (97.2–98.7) | 1.9 (1.3–2.8)c | 0.50 (0.34–0.73)b |
| LSD | 0.8 (0.7–0.8) | 98.5 (97.6–99.1) | 1.5 (0.9–2.4)c | 0.52 (0.31–0.87)a |
| Prescription Opioid Misuse | 4.0 (3.9–4.1) | 97.0 (96.2–97.7) | 3.0 (2.3–3.8)c | 0.74 (0.54–1.00) |
| Felt Survey Took Too Long | ||||
| No | 90.5 (90.2–90.7) | 95.3 (94.9–95.7) | 4.7 (4.3–5.1) | 1.00 |
| Yes | 9.5 (9.3–9.8) | 94.3 (92.3–95.7) | 5.8 (4.3–7.7) | 1.10 (0.79–1.55) |
| Trouble Understanding | ||||
| No | 94.2 (94.0–94.4) | 95.3 (94.9–95.6) | 4.7 (4.4–5.1) | 1.00 |
| Yes | 5.8 (5.6–6.0) | 93.9 (89.2–96.6) | 6.1 (3.4–10.8) | 0.94 (0.48–1.81) |
| English Proficiency | ||||
| Very Well | 86.7 (86.4–87.0) | 95.5 (95.1–95.8) | 4.5 (4.2–4.9)c | 1.00 |
| Well | 8.7 (8.5–8.9) | 92.4 (90.0–94.3) | 7.6 (5.7–10.0) | 1.50 (1.10–2.06)a |
| Not Well or Not at All | 4.6 (4.4–4.8) | 84.0 (69.5–92.4) | 16.0 (7.6–30.5) | 3.32 (1.40–7.83)b |
aOR: adjusted odds ratio; CI: confidence interval.
p < .05,
p < .01,
p < .001.
Finally, changes in prevalence of cannabis use between uncorrected and corrected estimates are presented in Table 2. Those reporting not speaking English well or at all had the highest discrepancy between the corrected and uncorrected estimate (16.0%), followed by those aged ≥50 (8.3%), and those who only speak English well (7.6%). The largest absolute increases in prevalence upon considering discordant responses were among those reporting past-year use of cocaine (1.1%) or LSD (1.0%), and among those identifying as Non-Hispanic Black (0.9%).
Table 2.
Prevalence estimates of cannabis use corrected for underreporting.
| Estimated prevalence % (95% CI) | Corrected prevalence % (95% CI) | Discord % | Absolute difference % | Relative difference % | |
|---|---|---|---|---|---|
| Full population | 15.2 (15.0–15.4) | 15.5 (15.3–15.7) | 4.8 | 0.3 | 2.0 |
| Age | |||||
| 12–17 | 12.5 (12.2–12.9) | 13.1 (12.8–13.5) | 6.9 | 0.6 | 4.8 |
| 18–25 | 34.1 (33.6–34.6) | 34.9 (34.5–35.4) | 3.9 | 0.8 | 2.3 |
| 26–34 | 24.3 (23.8–24.8) | 24.8 (24.3–25.3) | 4.1 | 0.5 | 2.1 |
| 35–49 | 14.0 (13.6–14.3) | 14.2 (13.8–14.6) | 5.1 | 0.2 | 1.4 |
| ≥50 | 7.5 (7.2–7.8) | 7.6 (7.3–7.9) | 8.3 | 0.1 | 1.3 |
| Sex | |||||
| Male | 17.9 (17.6–18.2) | 18.3 (18.0–18.6) | 4.2 | 0.4 | 2.2 |
| Female | 12.6 (12.4–12.9) | 12.9 (12.7–13.2) | 5.5 | 0.3 | 2.4 |
| Race/ethnicity | |||||
| Non-Hispanic White | 15.7 (15.4–15.9) | 15.9 (15.7–16.2) | 3.9 | 0.2 | 1.3 |
| Non-Hispanic Black | 17.5 (17.0–18.1) | 18.4 (17.8–19.0) | 6.9 | 0.9 | 5.1 |
| Hispanic | 13.2 (12.7–13.6) | 13.5 (13.1–14.0) | 5.1 | 0.3 | 2.3 |
| Other/mixed | 12.2 (11.7–12.8) | 12.5 (11.9–13.0) | 3.7 | 0.3 | 2.5 |
| Education | |||||
| High School or Less | 14.8 (14.4–15.1) | 15.2 (14.9–15.6) | 6.0 | 0.4 | 2.7 |
| Some College | 18.6 (18.2–19.0) | 18.9 (18.5–19.2) | 3.5 | 0.3 | 1.6 |
| College Degree or Higher | 13.3 (12.8–13.8) | 13.4 (12.9–13.9) | 2.8 | 0.1 | 0.8 |
| Annual Family Income | |||||
| <$20,000 | 19.7 (19.2–20.3) | 20.4 (19.9–20.9) | 5.8 | 0.7 | 3.6 |
| $20,000–$49,999 | 15.5 (15.1–15.8) | 15.8 (15.5–16.2) | 4.8 | 0.3 | 1.9 |
| $50,000–$74,999 | 14.2 (13.8–14.6) | 14.5 (14.1–14.9) | 4.9 | 0.3 | 2.1 |
| ≥$75,000 | 13.5 (13.2–13.8) | 13.7 (13.4–14.0) | 3.6 | 0.2 | 1.5 |
| Married | |||||
| No | 22.6 (22.3–22.9) | 23.1 (22.8–23.4) | 4.5 | 0.5 | 2.2 |
| Yes | 8.7 (8.4–8.9) | 8.8 (8.5–9.0) | 5.2 | 0.1 | 1.1 |
| Past-year other drug use | |||||
| Tobacco | 33.4 (32.9–33.8) | 34.1 (33.7–34.5) | 4.1 | 0.7 | 2.1 |
| Alcohol | 21.0 (20.7–21.2) | 21.4 (21.1–21.6) | 4.0 | 0.4 | 1.9 |
| Cocaine | 82.8 (81.4–84.1) | 83.9 (82.5–85.1) | 1.9 | 1.1 | 1.3 |
| LSD | 92.2 (90.6–93.6) | 93.2 (91.7–94.5) | 1.5 | 1.0 | 1.1 |
| Prescription Opioid Misuse | 48.6 (47.4–49.9) | 49.4 (48.2–50.7) | 3.0 | 0.8 | 1.6 |
| Felt survey took too long | |||||
| No | 15.3 (15.1–15.6) | 15.7 (15.4–15.9) | 4.7 | 0.4 | 2.6 |
| Yes | 13.8 (13.2–14.4) | 14.1 (13.5–14.7) | 5.8 | 0.3 | 2.2 |
| Trouble understanding | |||||
| No | 15.6 (15.4–15.7) | 15.9 (15.7–16.1) | 4.7 | 0.3 | 1.9 |
| Yes | 9.2 (8.5–10.0) | 9.4 (8.7–10.2) | 6.1 | 0.2 | 2.2 |
| English proficiency | |||||
| Very Well | 16.6 (16.4–16.8) | 16.9 (16.7–17.1) | 4.5 | 0.3 | 1.8 |
| Well | 8.6 (8.1–9.1) | 8.9 (8.3–9.4) | 7.6 | 0.3 | 3.5 |
| Not Well or Not at All | 1.7 (1.4–2.1) | 1.9 (1.5–2.3) | 16.0 | 0.2 | 11.8 |
Note. Absolute difference refers to the direct difference in prevalence and relative difference refers to the absolute difference as a percentage of the other prevalence estimate.
Discussion
Accurately estimating the prevalence of substance use is critical for assessing public health risks. Our study estimates that approximately 5% of respondents reported blunt use but did not report cannabis use. The impact of this discordant responding was minimal, with estimated prevalence of cannabis use only rising from 15.2% to 15.5% after correction. While this represents a relatively small portion of people who use cannabis, we determined that the corrected percentage of people using blunts in the past year increased by 1.2%.
Underreporting of blunt use may be of concern since blunt smoking, in some respects, is associated with more problematic cannabis use that can be detrimental to health. For example, in one study, a greater proportion of cannabis users who smoked blunts in the past year, compared to those that did not, reported symptoms associated with classic definitions of cannabis abuse and dependence such as “spending a great deal of time getting, using, or getting over the effects of cannabis,” “reduced or gave up important activities,” and “used in hazardous situations”.24 Furthermore, smoking cannabis mixed with tobacco can lead to acute increases in carbon monoxide levels and elevated heart rate when compared to smoking cannabis without tobacco.22 Moreover, people who smoke blunts are more likely to be dependent on nicotine when compared to people who smoke cannabis but do not smoke blunts.23 Further still, blunt use is associated with more chronic cannabis use as blunts usually contain more cannabis than joints.28 Consequently, blunt use often leads to greater levels of both intoxication and withdrawal than cannabis use by other means, which can place people at higher risk for cannabis use disorder.23,28,29 Indeed, blunt use is known to be associated with a greater likelihood of cannabis use disorders, as well as more psychosocial problems and poorer cannabis cessation outcomes.24,30
To some extent, it is possible that some of the discordance identified in the present study arises from respondents not considering blunt use to be cannabis use. Perhaps these individuals may perceive blunts as a distinct class of drug modality, or perhaps they align blunts more with tobacco use than cannabis use. In any case, other studies have also reported misclassification of drug use whereby participants responded negatively to tobacco use but reported rolling tobacco with cannabis when consuming joints or “spliffs”.31 Discordance may also be related to unfamiliarity with “blunt” nomenclature, particularly if respondents do not adequately read the definitions that are included within survey questions, or if such definitions are lacking. Indeed, studies have previously found that ecstasy use appears to be underreported when the term “Molly” is not included in the definition of ecstasy.32
Although our corrected estimates only suggest small increases in prevalence of use when considering underreporting, exploring subpopulations and demographic groups that underreported use provides important insight regarding who is more likely to underreport use. For example, NSDUH data suggest that older adults (ages 50 and over) are more likely to provide a discordant response as it pertains to blunt use and cannabis use questions. Other studies have also noted that adults in general (ages 30 and over) are more likely to underreport lifetime cannabis use when compared to younger indivduals.16 Moreover, older adults are more likely to satisfice on surveys,33 which is defined as respondents investing minimal attention toward survey questions and/or taking shortcuts with the intention of completing the survey sooner.34
Similarly, our findings demonstrate that female respondents were at significantly greater odds of providing a discordant response with respect to queries on blunt use and cannabis use. This also is an important consideration not only because blunt use has been increasing among women in recent years,35 but because several others have noted higher rates of underreporting among females. For example, female high school seniors were found to be at greater odds than males for reporting nonmedical Vicodin use after not reporting overall nonmedical opioid use on the same survey,10 while others have found that that women were more likely than men to test positive for drug use after not reporting use.36 In a recent study of pregnant women, the estimated prevalence of prenatal cannabis use (based on self-reported survey responses) increased from 4.2% to 7.1% when correcting for positive detection through urine testing.37 While potential reasons underlying these observations remain unclear, females are generally less likely to report drug use than males in national samples or elsewhere,38,39 a discrepancy that we believe may be, in part, due to either intentional or unintentional underreporting.
Our findings also suggest that Non-Hispanic Black respondents were at greater odds of underreporting cannabis use. This aligns with previous studies reporting that Black users were more likely to provide discordant responses when queried about nonmedical amphetamine use and opioid use.10,13 Others have found that Black survey respondents are more likely to satisfice,33 as well as to refer to blunts as “joints”.40 Such a demographic association may be particularly relevant here because Black individuals are more likely to smoke blunts than their white counterparts.41,42 Together, these findings on older, female, and Non-Hispanic Black respondents being more likely to underreport cannabis use are of concern because underestimates of use may result in inadequate prevention and/or treatment efforts targeted at these groups, as well as less funding allocated for research with these vulnerable populations.
In contrast, respondents who achieved higher levels of education were at significantly lower odds of providing a discordant response on the surveys. We speculate that this may reflect, at least in part, a higher degree of knowledge on cannabis’s nomenclature among those with higher educational attainment. Our education results corroborate established findings from other studies.19 In fact, it has further been shown that the odds of recanting previously reported cannabis use on future surveys declined with increasing educational attainment.43 Moreover, studies have shown that individuals with lower levels of education are more likely than those of higher education to satisfice on surveys.44 As such, people who have achieved a college degree or higher may simply pay closer attention to survey questions or read them more carefully. Similar correlations between level of educational attainment and discordant reporting of use of other drugs have also been established.10, 13, 16, 39
Further still, our findings show that respondents who also used other drugs in the past year (e.g., cocaine, LSD, opioids) were at significantly lower odds of providing a discordant response. We believe that this follows the same line of thought, namely, that those with a more experienced drug use history are more versed and familiar with drug nomenclature and classifications. This may also explain why older adults, who have the lowest prevalence of cannabis and blunt use in our study, may tend to have less knowledge of nomenclature and therefore higher odds of responding discordantly.
To the same point, our data suggest that having low to no proficiency with the English language was associated with over triple the odds of providing a discordant response, which may reflect to a potential unfamiliarity with or inability to comprehend different drug names and categories. The NSDUH survey can be completed only in English or Spanish, and certified bilingual interviewer is available for respondents who prefer to complete the interview in Spanish.25 If the respondent does not speak either English or Spanish, the interview is not conducted. However, no standardized protocol is noted in the NSDUH field interviewer manual for assessing English proficiency, and if the interviewer is not bilingual, they rely on a member of the household to serve as a translator for screening questions.45 Language and cultural differences, especially among immigrant or ethnic minority populations, have important implications for the quality of data collected through national surveys.46,47 Therefore, our results of higher discordant responses among respondents with lower English proficiency emphasizes the need to ensure respondents are able to fully understand survey questions as it relates to substance use.
Study design can be a major reason for discrepant reporting, and an argument can be put forth that survey designs should better facilitate respondent comprehension, especially among the aforementioned groups at higher risk for discordant reporting, such as those of lower English proficiency. The NSDUH, for example, currently queries blunt use in a standalone section as opposed to either the cannabis or tobacco sections.25 As such, it is not only assessed separately from both of these drugs, but may serve to psychologically prime respondents—especially those already less familiar with drug nomenclature—to regard blunts as a distinct drug. Furthermore, unlike most other items on the NSDUH, skip-logic did not appear to be applied to this question, which can increase discordant responding as evidenced by other studies based on nationally representative Monitoring the Future data.10, 13
Limitations
Despite being a nationally representative survey, some populations, such as homeless individuals who do not use shelters, are underrepresented. This may affect the overall prevalence of drug use within this study. NSDUH does not ask specifically about vaping or dabbing of cannabis so it is possible that some people who only use in one of these specific forms underreported use. As noted above, satisficing and lack of attention to questions could also be factors related to providing a discordant response, but we had no means to examine whether these factors influenced such responses in these analyses.
Conclusion
While our study found that approximately 5% of respondents reported blunt use but not cannabis use, changes in overall prevalence are small when correcting for discordant responses. However, results indicated that adults aged 50 or older, females, those identifying as Non-Hispanic Black, and those with lower levels of English proficiency were at significantly higher odds of providing discordant responses, while the opposite was true among individuals with higher educational attainment and individuals who indicated past year tobacco, alcohol, cocaine, or LSD use. We speculate that this may be related to knowledge and familiarity (or lack thereof) on various drug names and classifications used on surveys.
Results should provide insight into subgroups that may be more likely to underreport use on surveys and guide future research investigating or confirming speculated reasons for underreporting among said subgroups. Granted, some participants will simply misreport and submit inconsistent responses, but survey designs should attempt to better address this discordance. Blunt use, for example, should likely be assessed in cannabis use sections and classified as both cannabis and tobacco use, with proper skip-logic implemented as needed and adequate definitions included in the questions. Moreover, future research on how to design survey questions to improve respondent comprehension seems warranted, while current research utilizing these data should consider their limitations, including the capacity for underreporting among the foregoing subgroups.
Funding
This project was funded by the National Institutes of Health [R01DA044207 and K23DA043651]. The funding organization had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Footnotes
Disclosure statement
One author has consulted for Alkermes. The authors have no other potential conflicts to declare.
References
- [1].Gomez SA. The States That Legalized Marijuana. https://www.addictioncenter.com/news/2020/11/states-legalized-marijuana/. Published 2020. Accessed February 26, 2021.
- [2].National Conference of State Legislatures. Cannabis Overview. Washington, DC. https://www.ncsl.org/research/civil-and-criminal-justice/marijuana. 2020. [Google Scholar]
- [3].Han BH, Sherman S, Mauro PM, Martins SS, Rotenberg J, Palamar JJ. Demographic trends among older cannabis users in the United States, 2006–13. Addiction. 2017;112(3):516–525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Han BH, Palamar JJ. Trends in cannabis use among older adults in the United States, 2015–2018. JAMA Intern Med. 2020;180(4):609–611. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Hasin DS, Sarvet AL, Cerdá M, et al. US adult illicit cannabis use, cannabis use disorder, and medical marijuana laws: 1991–1992 to 2012–2013. JAMA Psychiatry. 2017;74(6):579–588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Cerda M, Mauro C, Hamilton A, et al. Association between recreational marijuana legalization in the United States and changes in marijuana use and cannabis use disorder From 2008 to 2016. JAMA Psychiatry. 2020;77(2):165. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Cerdá M, Wall M, Keyes KM, Galea S, Hasin D. Medical marijuana laws in 50 states: investigating the relationship between state legalization of medical marijuana and marijuana use, abuse and dependence. Drug Alcohol Depend. 2012;120(1–3):22–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Miech RA, Johnston LD, O’Malley PM, Bachman JG, Schulenberg JE, Patrick ME. Monitoring the Future National Survey Results on Drug Use, 1975–2019: Volume I, Secondary School Students. Ann Arbor: Institute for Social Research, The University of Michigan; 2020. [Google Scholar]
- [9].Safdar N, Abbo LM, Knobloch MJ, Seo SK. Research methods in healthcare epidemiology: survey and qualitative research. Infect Control Hosp Epidemiol. 2016;37(11):1272–1277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Palamar JJ, Shearston JA, Cleland CM. Discordant reporting of nonmedical opioid use in a nationally representative sample of US high school seniors. Am J Drug Alcohol Abuse. 2016;42(5):530–538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Lessler J, Caspar R, Penne M, Barker P. Developing Computer Assisted Interviewing (CAI) for the national household survey on drug abuse. Journal of Drug Issues. 2000;30(1):9–34. [Google Scholar]
- [12].Mensch BS, Kandel DB. Underreporting of substance use in a national longitudinal youth cohort: individual and interviewer effects. Publ Opin Quart. 1988;52(1):100–124. [Google Scholar]
- [13].Palamar JJ, Le A. Discordant reporting of nonmedical amphetamine use among adderall-using high school seniors in the US. Drug Alcohol Depend. 2017;181:208–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Dembo R, Robinson RB, Barrett K, et al. The validity of truant youths’ marijuana use and its impact on alcohol use and sexual risk taking. J Child Adolesc Subst Abuse. 2015;26(4):355–365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Fendrich M, Vaughn CM. Diminished lifetime substance use over time: an inquiry into differential underreporting. Publ Opin Quart. 1994;58(1):96–123. [Google Scholar]
- [16].Palamar J, Le A. Self-correction of unreported marijuana use by participants taking a street intercept survey. Am J Drug Alcohol Abuse. 2020;46(6):708–717. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Bessa MA, Mitsuhiro SS, Chalem E, Barros MM, Guinsburg R, Laranjeira R. Underreporting of use of cocaine and marijuana during the third trimester of gestation among pregnant adolescents. Addict Behav. 2010;35(3):266–269. [DOI] [PubMed] [Google Scholar]
- [18].Palamar JJ, Le A, Guarino H, Mateu-Gelabert P. A comparison of the utility of urine- and hair testing in detecting self-reported drug use among young adult opioid users. Drug Alcohol Depend. 2019;200:161–167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Jensen HAR, Karjalainen K, Juel K, Ekholm O. Consistency in adults’ self-reported lifetime use of illicit drugs: a follow-up study over 13 years. J Stud Alcohol Drugs. 2018;79(3):490–494. [PubMed] [Google Scholar]
- [20].Small E. Evolution and classification of Cannabis sativa (Marijuana, Hemp) in relation to human utilization. Bot Rev. 2015;81(3):189–294. [Google Scholar]
- [21].Palamar JJ, Acosta P, Calderón FF, Sherman S, Cleland CM. Assessing self-reported use of new psychoactive substances: the impact of gate questions. Am J Drug Alcohol Abuse. 2017;43(5): 609–617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Cooper ZD, Haney M. Comparison of subjective, pharmacokinetic, and physiological effects of marijuana smoked as joints and blunts. Drug Alcohol Depend. 2009;103(3):107–113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Timberlake DS. A comparison of drug use and dependence between blunt smokers and other cannabis users. Subst Use Misuse. 2009;44(3):401–415. [DOI] [PubMed] [Google Scholar]
- [24].Fairman BJ. Cannabis problem experiences among users of the tobacco-cannabis combination known as blunts. Drug Alcohol Depend. 2015;150:77–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Center for Behavioral Health Statistics and Quality. 2019 National Survey on Drug Use and Health Public Use File Codebook. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2020. [Google Scholar]
- [26].Center for Behavioral Health Statistics and Quality. Results from the 2018 National Survey on Drug Use and Health: Detailed Tables. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2019. [Google Scholar]
- [27].Heeringa SG, West BT, Berglund PA. Applied Survey Data Analysis. London: Chapman and Hall: CRC Press; 2010. [Google Scholar]
- [28].Mariani JJ, Brooks D, Haney M, Levin FR. Quantification and comparison of marijuana smoking practices: blunts, joints, and pipes. Drug Alcohol Depend. 2011;113(2–3):249–251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Ream GL, Benoit E, Johnson BD, Dunlap E. Smoking tobacco along with marijuana increases symptoms of cannabis dependence. Drug Alcohol Depend. 2008;95(3):199–208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Peters EN, Budney AJ, Carroll KM. Clinical correlates of co-occurring cannabis and tobacco use: a systematic review. Addiction. 2012;107(8):1404–1417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Fix BV, Smith D, O’Connor R, et al. Cannabis use among a nationally representative cross-sectional sample of smokers and non-smokers in the Netherlands: results from the 2015 ITC Netherlands Gold Magic Survey. BMJ Open. 2019;9(3):E024497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Palamar JJ, Keyes K, Cleland CM. Underreporting of ecstasy use among high school seniors in the US. Drug Alcohol Depend. 2016;165:279–282. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Di J, Li Y, Friedman MR, et al. Determining survey satisficing of online longitudinal survey data in the multicenter AIDS cohort study: a group-based trajectory analysis. JMIR Public Health Surveill. 2016;2(2):e150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Tourangeau R. The Psychology of Survey Response. 1st ed. Cambridge University Press; 2000. [Google Scholar]
- [35].Coleman-Cowger VH, Pickworth WB, Lordo RA, Peters EN. Cigar and marijuana blunt use among pregnant and nonpregnant women of reproductive age in the United States, 2006–2016. Am J Public Health. 2018;108(8):1073–1075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Fendrich M, Johnson TP, Wislar JS, Hubbell A, Spiehler V. The utility of drug testing in epidemiological research: results from a general population survey. Addiction. 2004;99(2):197–208. [DOI] [PubMed] [Google Scholar]
- [37].Young-Wolff KC, Tucker LY, Alexeeff S, et al. Trends in self-reported and biochemically tested marijuana use among pregnant females in California from 2009–2016. JAMA. 2017; 318(24):2490–2491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [38].Johnston LD, Miech RA, O’Malley PM, Bachman JG, Schulenberg JE, Patrick ME. Demographic Subgroup Trends among Adolescents in the Use of Various Licit and Illicit Drugs. Ann Arbor, MI: Institute for Social Research, University of Michigan; 2019:1975–2018. [Google Scholar]
- [39].Palamar JJ, Le A. Underreporting of drug use on a survey of electronic dance music party attendees. Addict Res Theory. 2020;28(4):321–327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].Montgomery L, Plano Clark VL, Twitty D, Budney AJ, Prochaska JJ, Winhusen T. Is it “loud” enough?: A qualitative investigation of blunt use among African American young adults. J Ethn Subst Abuse. 2020;1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [41].Montgomery L, Mantey DS, Peters EN, Herrmann ES, Winhusen T. Blunt use and menthol cigarette smoking: an examination of adult marijuana users. Addict Behav. 2020;102: 106153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [42].Montgomery L, Mantey D. Racial/ethnic differences in prevalence and correlates of blunt smoking among adolescents. J Psychoactive Drugs. 2018;50(3):195–205. [DOI] [PubMed] [Google Scholar]
- [43].Percy A, McAlister S, Higgins K, McCrystal P, Thornton M. Response consistency in young adolescents’ drug use self-reports: a recanting rate analysis. Addiction. 2005;100(2):189–196. [DOI] [PubMed] [Google Scholar]
- [44].Hamby T, Taylor W. Survey satisficing inflates reliability and validity measures: an experimental comparison of college and Amazon mechanical turk samples. Educ Psychol Meas. 2016; 76(6):912–932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [45].Center for Behavioral Health Statistics and Quality. 2019 National Survey on Drug Use and Health: Field Interviewer Manual. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2020. [Google Scholar]
- [46].Johnson TP, Cho YI, Holbrook AL, O’Rourke D, Warnecke RB, Chavez N. Cultural variability in the effects of question design features on respondent comprehension of health surveys. Ann Epidemiol. 2006;16(9):661–668. [DOI] [PubMed] [Google Scholar]
- [47].Wenz A, Al Baghal T, Gaia A. Language proficiency among respondents: implications for data quality in a longitudinal face-to-face survey. J Surv Stat Methodol. 2021;9(1):73–93. [Google Scholar]