This study analyzed daily assessments from a smartphone-based study conducted in California during 2016 and 2017. The study procedure was described in detail elsewhere [10]. Initially, participants completed a baseline survey on their demographics and substance use history. They were then trained on use of the study app to collect data every day for a 30-day period. Each day, participants were prompted between 10 and 11 AM to complete a daily assessment reporting their use of tobacco and cannabis on the entire previous day, including substance use occurrences late at night. To increase participant study compliance and retention, incentives were contingent on level of data-collection completion.

Electronic informed consent was obtained from all participants. The study was approved by the University of California San Francisco Institutional Review Board (15-18033).

Participants were recruited through social media and online advertisements (eg, Facebook and Craigslist). To conduct a nested qualitative substudy, participants were also recruited via the websites of sexual-minority youth organizations, and we oversampled women identifying as a sexual minority. Eligible participants were aged 18 to 26 years, had smoked at least 100 cigarettes in their lifetime, and currently smoked at least one cigarette per day at least 3 days per week. Since the parent study focused on cigarette smoking, cannabis use was not part of the inclusion criteria. Of 184 participants who completed the baseline assessments, 147 who completed at least one daily assessment were included in the analytic sample. There was no statistical difference between the analytic sample (n=147) and those who were excluded from the analysis (n=37) in baseline characteristics (ie, age, sex, educational attainment, race, ethnicity, and past-30-day use of tobacco and cannabis). Baseline characteristics of the study sample are presented in Table 1. The sample had a mean age of 22.7 (SD 2.4) years, 51.7% (76/147) of participants were female, 40.8% (60/147) of participants were non-Hispanic White, and 76.9% (113/147) of participants were currently in college or had a college degree or higher. At baseline, a majority of participants reported past-30-day use of cannabis (96/147, 65.3%) and alcohol (136/147, 92.5%). We included 51 participants who did not report past 30-day use of cannabis at baseline in our sample, since these participants could report co-use during the daily-diary period (and n=7 did), allowing for comparison between co-use and single-substance use within a day.

For each day, participants were asked, “How many times did you use marijuana or hash?” Answer options ranged continuously from 0 to 7 or more times. Those who reported any cannabis use were then asked, “How did you use marijuana or hash?” with answer options including smoking, vaping, and edibles. While we asked about intensity of use of cannabis in general, we did not ask about intensity of use of each cannabis product separately, to avoid overburdening participants. As such, depending on a participant’s interpretation, a smoking occasion of combustible cannabis, a hit of a cannabis vaporizer, or consuming 1 edible may have been considered as a single occasion of cannabis use in our study.

Statistical analyses were performed using Stata (version 15; Stata Corp). Descriptive statistics of sample characteristics at baseline and substance use in daily assessments were summarized. First, to examine associations of type of tobacco and cannabis co-use on the same day, we fitted multilevel logistic regression models examining associations of use of cannabis products (combustible, vaporized, and edible) with each of the binary outcomes (ie, any use of cigarettes, cigarillos, or e-cigarettes on a given day). Second, to examine associations between intensities of tobacco and cannabis use on the same day, we fitted multilevel mixed-effects ordered logistic regression models examining intensity of cannabis use (ie, number of times using cannabis on a given day) with each of the ordinal outcomes (ie, numbers of cigarettes or cigarillos smoked and number of times using e-cigarettes on a given day) [19]. The models also included random intercepts for participants to control for variation in tobacco use intensity attributable to individual participants.

The variable of intensity of cannabis use was decomposed into 2 elements: personal mean (ie, average intensity of cannabis use for each participant, indicating comparisons between participants, in other words, between-person effects), and deviation (ie, the difference between intensity in a particular daily observation and the personal mean, indicating comparisons across study days within a certain participant, that is, within-person effects) [20]. For each ordinal outcome, the proportional odds assumption was checked by fitting a generalized multinomial logit model and comparing its likelihood ratio to that of the ordinal model [19]; this assumption was satisfied for all the models. All models controlled for demographic covariates, day-level alcohol use [21], and time effects (ie, study day and weekend vs weekday). All tests were 2-tailed with a significance level of α<.05. The analyses were not preregistered and thus the results should be considered exploratory.

During the 30-day study period, 147 participants completed an average of 19.7 (SD 9.6) daily assessments with a completion rate of 65.6% (2891 completed assessments of 4410 prompted assessments). Table 2 describes reports of tobacco and cannabis use among the total of 2891 daily assessments. Co-use was reported in 989 daily assessments (34.2%), while use of tobacco without cannabis was reported in 1501 daily assessments (51.9%). The most common intensity of cigarette use reported in the daily assessments was smoking 2 to 5 cigarettes per day. Not using at all was reported the most in the daily assessments for cigarillos, e-cigarettes, and cannabis. On the days when participants used these products, the common intensities of use were smoking 2 to 5 cigarillos per day, vaping e-cigarettes 2 to 5 times per day and using cannabis once a day.

Table 3 presents same-day co-use in terms of combinations of specific products. The most commonly used tobacco product was cigarettes (2407 of 2891 assessments, 83.3%), while combustible cannabis was the most common type of cannabis use (1040 of 2891 assessments, 36%). The 3 most common product combinations on the same day were cigarettes and combustible cannabis (885 of 2891 assessments, 30.6%), cigarillos and combustible cannabis (197 of 2891 assessments, 6.8%), and cigarettes and vaporized cannabis (147 of 2891 assessments, 5.1%).

Results from the mixed-effects models are shown in Table 4. Participants had higher odds of reporting using cigarettes (adjusted odds ratio [AOR] 1.92, 95% CI 1.31-2.81) and cigarillos (AOR 244.29, 95% CI 35.51-1680.62) on days when they used combustible cannabis. Notably, participants had higher odds of using e-cigarettes on days when they used vaporized cannabis (AOR 23.21, 95% CI 8.66-62.24). It should be noted that the CIs for cigarillos and e-cigarettes were quite wide due to the small number of daily assessments with use of these products. In addition, participants had higher odds of smoking cigarettes on days with alcohol use (AOR 2.73, 95% CI 1.99-3.76). The study day was negatively associated with the odds of smoking cigarettes (AOR 0.96, 95% CI 0.95-0.98) and cigarillos (AOR 0.96, 95% CI 0.92-0.99). Older participants (vs younger peers) had higher odds of reporting cigarette smoking (AOR 1.31, 95% CI 1.10-1.57), while Hispanic participants (vs non-Hispanic White peers) had higher odds of reporting cigarillo smoking (AOR 22.93, 95% CI 3.30-159.49); however, this estimate was very wide due to a small number of cigarillo-use reports.

Results from the multilevel ordinal models are shown in Table 5. Participants had higher odds of reporting a greater intensity of using cigarettes (AOR 1.35, 95% CI 1.23-1.48), cigarillos (AOR 2.04, 95% CI 1.70-2.46), and e-cigarettes (AOR 1.48, 95% CI 1.16-1.88) on days when they used more cannabis. In addition, alcohol use on a given day was positively associated with intensity of cigarette use (AOR 1.41, 95% CI 1.35-1.49). Participants with higher average intensity of cannabis use had higher average intensity of cigarillo use (AOR 3.72, 95% CI 2.41-5.73). The study day was negatively associated with intensity of smoking cigarettes (AOR 0.97, 95% CI 0.96-0.98) and cigarillos (AOR 0.94, 95% CI 0.91-0.97), meaning that intensity of smoking cigarettes and cigarillos decreased slightly over the study period. Those with education attainment of college or higher reported lower intensity of cigarette smoking (AOR 0.30, 95% CI 0.09-0.97), while older participants reported higher intensity of cigarette smoking (AOR 1.52, 95% CI 1.22-1.89). Hispanic participants (vs non-Hispanic White peers) had higher odds of reporting higher intensity of cigarillo smoking (AOR 11.46, 95% CI 1.82-72.36); however, this estimate was very wide due to a small number of cigarillo-use reports.

This study is one of very few examining young adult co-use of tobacco and cannabis within shorter timeframes (ie, a day) than the typical survey measure of past-30-day use, and it is among the first to examine same-day co-use of tobacco and cannabis products that are not smoked (ie, e-cigarettes and vaporized cannabis), including day-level intensity of co-use. The main findings were as hypothesized and showed that the more cannabis participants reported using on a given day, the greater the intensity of tobacco product use (cigarettes, cigarillos, and e-cigarettes). Notably, participants reported smoking cigarettes or cigarillos on the days they smoked cannabis, and vaping e-cigarettes on the days they vaped cannabis, indicating the same routes of administration may play a role in facilitating same-day co-use.

Since traditional measures are insufficient to fully capture and monitor co-use of tobacco and cannabis, recent research called for more accurate measures of this behavior [6] and highlighted the potential of digital health applications for collecting fine-grained data and specifying co-use patterns [7]. As a methodological example, our study used a daily-diary design and smartphone-based data collection to generate intensive longitudinal data on co-use patterns on a daily basis over 30 consecutive days, providing a nuanced understanding of the extent of co-use within a day. Another strength of this study was an examination of use of a variety of tobacco and cannabis products, including co-use of vaporized products (ie, e-cigarettes and vaporized cannabis), for which more evidence is needed. In addition to our smartphone-based daily assessment method, future research should also consider using other mobile-data collection methods (eg, ecological momentary assessments and mobile sensors) that may more comprehensively assess co-use of tobacco and cannabis [6,7]. Furthermore, while there are only a handful of studies, including this study, that have directly examined co-use as the focal outcome, many prior studies indirectly addressed co-use by adjusting for use of both tobacco and cannabis in the same analytic models. Systematic review or meta-analysis of both direct and indirect evidence may be warranted to provide comprehensive insights on co-use and its effects.

The positive associations between use of the same types of tobacco and cannabis products on the same day indicate that there may be behavioral cues from shared routes of administration that may facilitate co-use of tobacco and cannabis (eg, smoking or vaping one substance triggers smoking or vaping the other) [22]. Indeed, a combination of cigarettes and combustible cannabis was the most common same-day co-use pattern in our sample of young adult cigarette smokers. It was also the most common pattern of past-30-day and past-year co-use found in other samples of young adults [8,11]. In addition to well-documented co-use patterns via smoking (eg, cigarettes/cigarillos and combustible cannabis), we also found that participants reported vaping e-cigarettes more on the days when they vaped cannabis. This finding, coupled with the high prevalence of vaping among young populations, indicates that more attention to emerging co-use patterns via vaping is needed [12,23]. Previous research has reported vaping-related harms among covapers, such as lung impairments [24] and increased odds of having COVID-19 symptoms and diagnoses [25]. Further investigation of tobacco and cannabis covaping among young adults and its health consequences is warranted. Moreover, our participants also reported using other product combinations across the spectrum of tobacco and cannabis products, underlining the heterogeneity of co-use patterns. Further exploration of unique reasons and contexts for different patterns of co-use would help to identify targets for tailored prevention and treatment strategies.

While one might expect a potential drug substitution effect, in which people use cannabis as a substitute for tobacco [26], our finding of positive associations between intensity of tobacco and cannabis use on the same day suggests the substitution effect did not occur in our sample of young adult smokers. Instead, as explained by the theory of synergistic effects, individuals may use 2 substances at the same time or use 1 substance under the effect of the other to amplify positive effects or counteract negative effects between nicotinic and endocannabinoid systems [22]. Relatedly, shared contexts (eg, being with friends and socializing) may also facilitate intensity of same-day co-use of tobacco and cannabis [27,28]. In addition, our participants with higher average intensity of cannabis use also reported higher average intensity of cigarillo use. This finding could be due to our participants using cigarillos for blunt smoking. Although we did not directly ask about blunt use in the daily assessments, previous studies indicated that young adults perceived cigarillos were frequently used for blunts [29,30]. We also found that participants smoked more cigarettes on the days when they drank alcohol. This finding could be explained by well-known rewarding effects when cigarettes and alcohol are used together [13,31-33].

In addition, several subgroups in our sample demonstrated greater average intensity of tobacco use. Participants who were older and had less than a college education reported greater intensity of cigarette smoking, whereas Hispanic participants reported a greater intensity of cigarillo use. These findings are consistent with previous research documenting high prevalence of tobacco use in these subgroups [17,34,35]. Interestingly, participants’ use of cigarettes and cigarillos decreased over the study period. This may be due to a Hawthorne effect or other impacts of research participation, as the process of reporting on their own behaviors may induce reflection and influence participants’ behaviors [36]. To our knowledge, this reactivity effect was rarely observed in previous research. Future studies using experience-sampling methods (eg, ecological momentary assessments or daily diaries) should explore reactivity effects and potential impacts on behavioral outcomes.

Collectively, our study has implications for efforts to support smoking cessation among young adults. As co-use of tobacco and cannabis was common, and this may increase harm and addiction, smoking cessation programs may need to address co-use of multiple tobacco products or tobacco and cannabis to improve efficacy with this age group. Most available interventions to reduce tobacco use in young people may not address engaging in co-use [7,37]. A recent study found that when young people reduced their tobacco use, their cannabis use also decreased, suggesting the potential benefits of dual cessation treatment for co-users [38]. In addition, treatment strategies should be expanded to include co-use of nonsmoking products to meet cessation needs of covapers [7,39]. As such, tailored interventions that adapt supports to individuals’ co-use patterns may be more effective for reducing the use of both substances. Moreover, tailored interventions may be needed to reach those with high rates of co-use, such as those without college education or those who identify as Hispanic.

Several limitations ought to be considered. The data were collected during 2016 and 2017. Since then, there have been rapid changes in public policy related to both tobacco and cannabis, in patterns of use (eg, increasing use of vaporization devices), in cannabis legalization, and in product availability in the marketplace. As such, more recent data are needed to replicate our findings. The convenience sampling procedure via online recruitment in California and the oversampling of young sexual minority adults limit our study’s generalizability to other young adult samples or geographic regions. While co-use of tobacco and cannabis is common among smokers, our sample included a minority of young adults who did not report past-30-day cannabis use at baseline; further research should examine co-use of tobacco and cannabis among young adults who report recent use of both substances. Although using categories for intensity of use of tobacco products provided a general measure of increasing intensity, this may result in limitations to interpretation of actual effects for each product, given that the increase of using e-cigarettes from, for example, one time per day to 2 to 5 times per day may be different from the increase of smoking from one to 2 to 5 cigarillos per day. Data on cigar use were not collected, and simultaneous use of tobacco and cannabis and their overlapping effects were not directly assessed in our study. Moreover, data on timing or ordering in use of tobacco and cannabis were not collected; thus, we could not identify temporal relationships in use of these substances. Likewise, we did not collect data on cannabis concentrations and specific intensity of use by type of cannabis. The use of concentrated cannabis could impact the same-day co-use of tobacco and cannabis in meaningful ways depending on complementing versus supplementing behaviors. In addition, our participants were not trained in defining intensity of cannabis use and the meaning of “times of cannabis use” may vary depending on personal definitions of use sessions and types of cannabis use. Future research should consider collecting these data and developing more accurate measures of daily use of tobacco and cannabis in order to provide a better understanding of co-use. Missing data due to participants’ compliance with daily assessments may impact the study’s internal validity; however, our compliance rate is within the range of previous studies using the same data collection methods [10,13-15] and the models in our analysis are generally robust to missing data under the missing-at-random assumption [40,41].

By using smartphone-based daily assessments, this study identified a substantial correlation of product types and intensities of tobacco and cannabis co-use at the day level, with young adults reporting more tobacco use on days when they used more cannabis, including same-day co-use of e-cigarettes and vaporized cannabis. Future research and interventions should address co-use in all forms, especially co-use via new products and in short timeframes, to better prevent and reduce use of both tobacco and cannabis and related health impacts among young people.