Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 May 1.
Published in final edited form as: Psychol Addict Behav. 2024 Jan 8;38(3):323–333. doi: 10.1037/adb0000979

Associations Between Day-Level Affect Profiles and Same-Day Substance Use Among Young Adults

Natalia Van Doren 1, Bethany C Bray 2, José A Soto 3, Ashley N Linden-Carmichael 4
PMCID: PMC11262887  NIHMSID: NIHMS1943374  PMID: 38190198

Abstract

Objective:

Emotions play a critical role in health risk behaviors, including substance use. However, current research often focuses exclusively on average levels of positive and negative affect, neglecting the complexity of daily emotional patterns. By capturing multiple dimensions of affect, including arousal and discrete states, we can improve our understanding of proximal predictors of substance use. The present study demonstrates the utility of a novel methodological approach for assessing affect patterns in daily life in relation to alcohol and cannabis use.

Methods:

Daily diary data from N=154 young adults who reported recent heavy episodic drinking and simultaneous use of alcohol and cannabis were analyzed using a mixed-indicator latent profile analysis to identify and describe day-level affective patterns and outcomes.

Results:

Results revealed five distinct day-level profiles of affect: undifferentiated negative affect days, undifferentiated positive affect days, high arousal positive affect days, mixed affect days, and low reactivity days. Undifferentiated positive affect days, high arousal positive affect days, and low reactivity days were associated with significantly greater odds of same-day alcohol use compared to days characterized by undifferentiated negative affect (χ2 = 10.55, p =. 032).

Conclusions:

Findings suggest that daily affect patterns differentially impact alcohol use and can inform the development of interventions for problematic substance use. Additionally, the innovative methodological approach employed herein could be applicable for investigating the role of emotion in other health behaviors.

Keywords: emotion, heavy episodic drinking, cannabis use, young adults, latent profile analysis, health behaviors

Emotions represent a key mechanism through which behavior is linked to health and are thus an important target for intervention. However, despite decades of research and theorizing regarding the impact of emotions on health behaviors, findings are often mixed as to whether, when, and what kinds of emotion are linked to health risk or promoting behaviors, challenging intervention scientists to know what aspects of emotion to target and when. To develop more effective interventions, we first need methods that can help elucidate what aspects of emotion are important predictors of daily health risk behaviors, such as substance use.

While many theories posit a role for emotion in motivating substance use, a key health risk behavior—such as negative reinforcement theories including self-medication (Kahntzian, 1997) or stress-response dampening (Levenson et al., 1980)—evidence regarding the role of negative affect (NA) is elusive, particularly for alcohol use. While individuals with higher negative-affect linked traits, such as depression and distress intolerance, tend to be more likely to use substances (e.g., Howell et al., 2010), these relationships often do not bear out in everyday life. For example, intensive longitudinal designs (ILD) to examine daily or momentary associations between NA and alcohol use has found no relationship (Dvorak et al., 2016; Patrick et al., 2016), a negative relationship (Bresin & Fairbairn, 2019; Patrick et al., 2016), and a positive relationship (Mohr et al., 2013; Simons et al., 2005). A recent meta-analysis from 69 intensive longitudinal studies across 12,394 participants and 353,762 days showed no association between NA and use (Dora et al., 2023). Such findings may suggest that the relationship between NA and substance use is more complex, underscoring the need for further examination of the relationship between emotion and substance use using ecologically valid methods.

While positive affect (PA) has been considered with respect to the rewarding properties of substances in the context of reinforcement learning theories, less attention has been paid to the role of PA as an antecedent of substance use. Cooper’s motivational theory of alcohol use is one exception, which maintains that individuals drink not only to cope with negative emotions, but also to enhance positive emotions (Cooper et al., 1995; Cooper et al., 2016). Findings regarding PA and substance use at the trait-level suggest that individual differences in the average level or amount of PA is not associated with use (e.g., Gustafson, 1991; Rankin & Maggs, 2006), while others have found that higher trait PA is linked to greater use (e.g., Dora et al., 2023). In particular, studies using ILD suggest that increases in PA are a proximal predictor of alcohol use (Simons et al., 2010). Similar patterns at the daily level have been observed for cannabis use (Testa et al., 2019), suggesting that PA and NA are both important emotional predictors of substance use.

Most affect work has adopted a valence-based approach by examining whether and how mean levels of affect predict substance use. However, affective science theories, such as the Circumplex Model (Russell, 1980), posits that emotions are more than valence—they vary not only in terms of how positive and negative they are, but also in their relative levels of activation (i.e., arousal; Barrett, 1998). In addition, discrete emotion theories suggest that specific emotions are associated with distinct appraisals and behavioral signatures (Ekman, 1992). Thus, emotions that fall within the same valence category could result in markedly different behavioral outcomes. For example, “excited” and “calm” may both be considered positive emotional states in terms of their valence, yet they differ in terms of their relative levels of arousal. While “excited” is a high arousal positive emotional state, “calm” is a lower arousal positive emotional state. This difference is critical as they are likely to diverge in terms of the behaviors those states may motivate. Consequently, when we aggregate across valence we may miss important information regarding emotions’ links to health behaviors, particularly in daily life contexts.

In addition to valence, the importance of accounting for arousal when examining links between emotion and substance use has also been noted (Goldman, 1994; Rather et al., 1992). For example, expectancy theories of alcohol use propose that assessing both valence and arousal is important (Goldman, 1994). Some empirical work has begun to take this into account, such as Kuntsche & Kuntsche (2017) who validated an alcohol expectancy task using the four quadrants of the affect circumplex and found that expectancies for high arousal positive states were present in those youths who already initiated alcohol use. However, there is a paucity of studies examining links between daily affect and substance use that also include levels of arousal. One notable exception is Jones et al. (2021), which found that high arousal positive states predicted greater odds of alcohol use and heavy episodic drinking (HED). On the other hand, other work (e.g., Peacock et al., 2015) suggests that arousal predicts drinking regardless of valence. One possibility for such mixed findings is that prior work has failed to consider more complex patterns of affect, which could account for such discrepancies.

Methodological and analytical consideration of emotional valence, arousal, and discrete emotional states simultaneously may provide more fine-grained and nuanced information about the role of emotion in predicting substance use in daily life and may enable more accurate predictions of proximal emotional triggers for substance use in daily life, thereby aiding intervention efforts. The Positive and Negative Affect Schedule (PANAS; Watson et al., 1999) is a commonly used assessment tool in studies of emotion and substance use. While this instrument is typically used to capture only valence, it also contains items that differ on arousal. For instance, while “excited” and “calm” are both considered PA, excited denotes a high level of arousal, while calm denotes a low level of arousal. The PANAS also contains discrete emotions, such as guilt and pride, suggesting that this instrument could be used to examine links between such discrete emotional states and substance use (as opposed to just global PA/NA or dimensional valence-arousal). It is thus possible to use the PANAS in a nuanced manner to assess how valence, arousal, and discrete emotions may cluster together in various ways to impact substance use. Analytically, it is also critical to consider the heterogeneity of the affect structure between and within individuals. For example, while the 2-factor structure of the PANAS has been well-replicated at the between-person level, studies examining the within-person structure of affect have shown that this structure does not always generalize to state-level affect. Multi-level factor analysis of affect at the day-level has revealed a four-factor (Dornbach-Bender et al., 2020) or seven-factor structure at the within-person level (Jacobson et al., 2020). The bipolarity assumption (i.e., the idea that the latent correlation between PA and NA is invariant) has also been challenged given that the structure of affect has been shown to be variable across persons and occasions (e.g., Barrett, 1998; Dejonckheere et al., 2021; Kuppens et al., 2013). More generally, the literature on emotion and substance use fails to consider various theoretical models of emotion, such as the circumplex model (Russell, 1980), though several studies within the smoking literature have begun to differentiate between valence and arousal (e.g., Dvorak et al., 2018; Padovano et al., 2020). However, the alcohol and cannabis use literature rarely does so, leaving an important gap.

The predominating approach to the study of the experiential domain of emotion and substance use, to date, has utilized a variable-centered approach (Bergman & Magnusson, 1997). For example, studies using the PANAS have either focused on associations with a single dimension (e.g., the association between NA and substance use) or explored the role of each dimension separately (e.g., effects of each using simple regressions) or simultaneously (e.g., using multiple regressions). A weakness of these approaches stems from the way they utilize the PANAS: they do not account for the way these dimensions are related within individuals or within occasions. Given that research suggests heterogeneity in affect structure, including the relationship between PA and NA within individuals and across occasions (Eisele et al., 2021), assumptions of homogeneity in the effects of PA and NA on substance use may not be tenable. Models that do not adequately account for heterogeneity in affect structure may hinder understanding of the complex relationship between affect and substance use, which is theorized to occur within a short time frame and are thus typically measured using daily or momentary designs (Mohr et al., 2005).

In contrast, person-centered approaches (Bergman & Magnusson, 1997) capture similarities and differences among individuals with respect to how variables (e.g., affect terms) relate to each other (Bauer & Shanahan, 2007). For instance, while a variable-centered analysis could reveal that on average PANAS PA is higher than NA in a given population (e.g., college students), a person-centered approach in this same population may uncover two qualitatively different subgroups: the first characterized by people having high PA and very low NA, and the second characterized by people having high scores on both PA and NA. These two qualitatively different subgroups (or emotion typologies) may be obscured by approaches that use mean scores and that assume homogeneity across individuals.

Latent profile analysis (LPA) can be considered a person-centered approach and provides the opportunity to comprehensively characterize emotional experiences and their associations with substance use in daily life. LPA is a method commonly used to categorize individuals into unobserved classes or profiles based on shared responses to a set of nominal, ordinal, or continuous observed variables (Lanza et al., 2013). While LPA is typically applied to persons to identify groups of individuals that differ on a latent trait, such person-centered methods have recently been applied to identify patterns of behavior in daily life, where the unit of analysis is occasion (e.g., a day, a moment) rather than persons (Linden-Carmichael et al., 2022). Such nuanced applications of LPA to daily diary data could help identify affective patterns that emerge in daily life and to link these patterns to health risk behaviors. Compared to using a PANAS mean or factor score, using LPA applied at the level of occasions can identify how aspects of emotional experience (e.g., valence, arousal, discrete emotions) relate to one another within a given day or moment. In other words, this approach allows examination of the within-occasion relationships among multiple affective indicators, which could provide qualitative information about distinct types of emotion days or emotion moments that are linked with substance use. Thus, this approach could uncover two different types of emotion days: the first characterized by days that involve high PA and low NA, and the second characterized by days with high PA and high NA. These types of days, in turn, may be differentially linked with substance use. In sum, using LPA applied to occasions can uncover how affective indicators cluster together in unique patterns at the day-level to predict health behaviors, such as substance use.

The increased specificity in terms of identifying daily affective patterns and their links to substance use may be particularly advantageous compared to typical approaches (e.g., mean levels of affect) in several ways. First, examining affective patterns enables the simultaneous modeling of multiple indicators of affect within the same model. Second, this approach also allows naturalistic patterns to emerge across items that capture valence, arousal, and discrete emotions, rather than artificially restricting analyses to one or the other index. Together, LPA is an ideal method to advance our understanding of features of affect in daily life and their associations with substance use behavior. The present study leveraged daily diary methodology and a novel application of LPA to (1) identify day-level affective patterns and (2) examine their day-level associations with the most commonly used substances among young adults (alcohol, cannabis; Patrick et al., 2022).

Method

Transparency and Openness

All procedures were approved by the institutional review board of the Pennsylvania State University, protocol number STUDY00009870. Data and code are available via the Open Science Framework (OSF; Van Doren et al., 2023).

Participants

Participants (n=161) were recruited as part of a larger study investigating young adult substance use from the Northeastern region of the U.S. (Linden-Carmichael et al., 2020). Participants were recruited through flyers and the university’s StudyFinder website. To be eligible, participants had to have 1) been between ages 18–25; 2) endorsed simultaneous use of alcohol and cannabis (use so that both effects overlapped) at least once in the past month; and 3) endorsed heavy episodic drinking (4+/5+ drinks for females/males) on 1+ occasions in the past two weeks. M age was 20.3 (SD = 1.5) years. Most participants identified as women (57.2%) and non-Hispanic/Latine (NHL) white (72.7%). Specific demographics are summarized in Table 1.

Table 1.

Demographic and baseline characteristics of participants (n = 154)

Variable
Gender/Sex (% female) 57.24%
20.26 (1.45)
M (SD) Age
Race/Ethnicity
White 72.70%
Asian 11.70%
Black 6.50%
Multiracial 3.30%
Hispanic/Latine White 3.90%
College Status
 Currently Attending 88.30%
 Not Attending/Never Attended 11.70%
Year in College (for current students, n = 136)
 First Year 14.00%
 Second Year 25.00%
 Third Year 27.90%
 Fourth Year or Later 31.60%
Typical Substance Use (reported at baseline)
 DDQ - M (SD) Number of Drinks/Week 17.23 (11.81)
 DDQ - M (SD) Number of Cannabis Joints or Equivalent/Week 6.53 (7.97)
Open in a new tab

Note. If not indicated race/ethnicity group is non-Hispanic/Latine. DDQ = Daily Drinking Questionnaire (Collins, Parks, & Marlatt, 1985). Gender/Sex was self-identified.

Procedure

After providing informed consent, interested participants completed a brief, online screener to determine eligibility. Eligible participants were automatically routed to complete a 15–20-minute online baseline survey. Brief (3–4 min) follow-up assessments for the daily diary portion of the study were also collected online and accessible on any device. For 14 consecutive days participants were sent an e-mail and text message with the survey link at 9am and a reminder at 11:30am. Of surveys completed, roughly 95% were completed between 9am-12pm. The average time of day that surveys were completed was approximately 10:21am. Participants received up to $48: $10 for baseline survey, $2/daily survey, plus a $10 bonus if they completed 12+ surveys. Of eligible participants (n=161), 95.7% (n = 154) completed 1+ daily surveys with an average of 13.1 (SD = 1.95) out of 14 daily surveys completed. Although diary surveys were completed once per day, positive and negative affect were based on current reports, while substance use was based on yesterday’s reports (see daily measures and statistical analyses for details). Our analytic sample resulted in 154 individuals and 1878 person-days.

Daily Measures

To assess affect each day, participants completed the Positive and Negative Affect Schedule—Short Form (PANAS-SF; Watson et al., 1997). Participants were asked to indicate the extent to which they presently felt each of 10 positive (e.g., “excited”, “proud”) and 10 negative (e.g., “scared”, “guilty”) affective states ranging from 1 = Not at all to 5 = Extremely. To assess substance use each day, participants were asked to indicate which substances they used the day prior from a checklist of substances. Participants indicated the number of standard alcoholic drinks and number of hits used on days they used alcohol and cannabis, respectively.

Statistical Analyses

Prior to conducting analyses, substance use variables were lagged to allow for examination of same-day associations between affect and substance use, as affect was based on current (morning) experiences (e.g., “How do you feel right now?”) whereas substance use was based on yesterday’s behaviors (e.g., “Which of the following substances did you use yesterday?”). To ensure the multivariate normality assumption in LPA was met, we examined distributions of indicator variables and transformed those that were not approximately normally distributed (Table S1 and S2). The 10 PA indicators were positively skewed and were normalized via a square-root transformation. The transformed indicators were used for analysis and then back-transformed to facilitate interpretation. However, the 10 NA indicators remained skewed even after a square-root transformation, due to a high number of “not at all” responses. To preserve information and avoid violations of the multivariate normality assumption, we transformed NA indicators into binary variables where 0 indicated 1 = “not at all”, and ratings from 2–5 were all coded as 1, indicating that they experienced any amount or degree of that emotion.

Day-level LPAs were modeled in Mplus version 8.7. All latent profile models were estimated using full information maximum likelihood estimation (Widaman, 2006). The average number of daily surveys per person was 13.13 (SD = 1.95). All daily diary records were complete—there were no instances of missingness (beyond missingness on days when participants did not complete a survey). A mixed indicator LPA was conducted on 10 continuous indicators of PA and 10 binary indicators of NA. Following procedures of identifying day-level latent classes in Linden-Carmichael et al. (2022), to account for the nesting of days within persons, we included a clustering statement to produce robust standard errors and used a pseudo-maximum-likelihood approach to model estimation. Because models were fit to occasion-level nested data, identified latent profiles were representative across all person-days. Model identification was examined by comparing solutions obtained across 100 sets of random starting values. Model selection was guided by the Akaike information criterion (AIC), Bayesian information criterion (BIC), sample-size adjusted BIC (a-BIC), a bootstrapped likelihood ratio test (BLRT), the Vuong–Lo–Mendell–Rubin likelihood ratio test (VLMRT), and entropy, as well as model stability, interpretability, and parsimony. Lower values for the AIC, BIC, and a-BIC indicated better model fit, and higher entropy values indicate higher classification utility. Significance testing in the BLRT and VLMRT were used to compare models with k classes relative to a model with k + 1 classes.

To examine links between affect profiles and daily substance use, we fit a series of LPAs with a distal outcome using the “BCH approach” (Bakk & Vermunt, 2016). The following models were conducted separately: 1) binary outcome indicating any alcohol use; 2) binary outcome model indicating any cannabis use; 3) continuous outcome indicating number of alcoholic drinks consumed on days with any alcohol use; 4) continuous outcome indicating number of hits on days with any cannabis use. The latter two analyses were limited to days with any use to avoid zero-inflation.

Results

Identifying Day-Level Profiles of Affect

Model fit information and model selection criteria for models with 1–8 profiles are shown in Table 2. The BLRT and VLMRT suggested only the 2-profile model provided significantly better fit compared to the 1-profile model. In contrast, the AIC, BIC, and a-BIC were not minimized and continued to decrease across models, but relative reductions decreased across models with 4 to 6 profiles. Thus, we carefully considered models with 4 to 6 profiles based on theoretical interpretation and parsimony. Specifically, we found that when moving from 4 to 5 profiles, the new profile that emerged was distinct, interpretable, and continued to appear in models with additional profiles. However, the new profile that emerged in the 6-profile solution was redundant with another profile, suggesting the 6-profile solution was over-extracted. Thus, the 5-profile model was selected as optimal for interpretation and further analysis.

Table 2.

Model fit information and selection criteria for models with 1–8 profiles

No. of Profiles df AIC BIC a-BIC Entropy BLRT VLMRT
1 30 39422.49 39588.63 39493.32 -- -- --
2 51 31214.57 31497.01 31334.98 0.92 0.04 0.04
3 72 27998.62 28397.35 28168.61 0.91 0.12 0.12
4 93 26519.68 27034.71 26739.25 0.90 0.25 0.25
5 114 25125.04 25756.37 25394.20 0.90 0.14 0.14
6 135 24674.99 25422.61 24993.72 0.88 0.32 0.32
7 156 24233.95 25097.88 24602.27 0.87 0.51 0.51
8 177 23860.14 24840.36 24278.04 0.87 0.22 0.22
Open in a new tab

Note. df = degrees of freedom; AIC = Akaike information criterion; BIC = Bayesian information criterion; a-BIC = sample size adjusted BIC; BLRT = bootstrap likelihood ratio test; VLMRT = Vuong–Lo–Mendell–Rubin likelihood ratio test. Dashes indicate criterion was not applicable; bold indicates selected model.

Parameter estimates for the day-level, 5-profile model are shown in Table 3. This table displays the item means (PA indicators) and item-response probabilities (NA indicators), conditional on profile membership. Profile 1 (14% of days across all participants) was characterized by low levels of PA and high probabilities of experiencing NA; this profile was labeled “undifferentiated negative affect days.” We use the term “undifferentiated” to denote the fact that all NA indicators clustered together within a day. Profile 2 (20%) was characterized by high levels of PA and low probabilities of experiencing NA; this profile was labeled “undifferentiated positive affect days.” Profile 3 (25%) was characterized by high levels of high-arousal positive affect (“excited”, “enthusiastic”, “strong”) along with low probabilities of NA; this profile was labeled “high arousal positive affect days.” Profile 4 (20%) was characterized by high levels of both PA and NA; this profile was labeled “mixed affect days.” Profile 5 (22%) was characterized by low levels of affect across the board; this profile was labeled “low reactivity days.” 1

Table 3.

Parameter estimates for the selected five-profile model

Undifferentiated negative affect days Undifferentiated positive affect days High arousal positive affect days Mixed affect days Low reactivity days
Profile Prevalences 0.14 0.20 0.25 0.20 0.22
(ndays = 265) (ndays = 373) (ndays = 461) (ndays = 374) (ndays = 405)
Affect Indicators Sample mean / probability
Within-Profile Item Means / Probabilities of ‘Yes’ Response
Interested 2.46c 1.96 3.51 2.53 3.01 1.50
Excited 2.38c 1.55 3.56 2.52 2.86 1.51
Strong 2.16c 1.53 3.38 2.21 2.77 1.18
Enthusiastic 2.33c 1.44 3.61 2.52 2.97 1.29
Proud 2.03c 1.32 3.22 2.06 2.61 1.16
Alert 2.20c 1.89 3.10 2.17 2.81 1.28
Inspired 2.02c 1.36 3.28 1.92 2.68 1.14
Determined 2.35c 1.81 3.64 2.32 3.05 1.28
Attentive 2.26c 1.78 3.34 2.28 2.86 1.28
Active 2.10c 1.50 3.33 1.99 2.80 1.19
Distressed 0.58b 0.93 0.43 0.42 0.94 0.34
Upset 0.39b 0.76 0.20 0.17 0.83 0.18
Guilty 0.25b 0.50 0.08 0.09 0.63 0.10
Scared 0.31b 0.70 0.13 0.07 0.74 0.08
Hostile 0.22b 0.46 0.09 0.04 0.59 0.06
Irritable 0.48b 0.80 0.26 0.27 0.87 0.36
Ashamed 0.23b 0.50 0.05 0.05 0.64 0.06
Nervous 0.51b 0.89 0.36 0.30 0.96 0.21
Jittery 0.39b 0.55 0.32 0.24 0.82 0.11
Afraid 0.29b 0.70 0.09 0.04 0.77 0.04
Open in a new tab

Note.

c

= continuous indicator

b

= binary indicator. Inline graphic Dark shade = .5 or more standard deviations higher than the sample mean / probability; Inline graphic Light shade = .5 or more standard deviations lower than the sample mean / probability; Inline graphic No shade = no difference from sample mean / probability. The top half of the table contains the positive affect indicators; the bottom half contains the negative affect indicators. Total number of days in the analytic sample was 1878.

Associations Between Day-Level Affect Profiles and Substance Use

Overall, 605 (58%) days involved any alcohol use and 584 (56%) days involved any cannabis use. Descriptive statistics for each profile’s probability of use and level of use are shown in Figure 1 (Panels A and B, respectively).

Figure 1.

Figure 1.

Open in a new tab

Graphical representation of descriptive probabilities (Panel A) and level (Panel B) of alcohol and cannabis use observed in each profile.

Probabilities of Alcohol and Cannabis Use

Conditional probabilities and means based on profile membership and pairwise differences across profiles are shown in Table 4. Probability of drinking alcohol differed significantly across day-level profiles (χ2 = 10.55, p = .032). Examination of pairwise comparisons showed that compared to undifferentiated negative affect days, days characterized by undifferentiated positive affect (χ2 = 6.91, p = .009), high arousal positive affect (χ2 = 7.24, p =. 007), and low reactivity (χ2 = 7.67, p = .006) were more likely to be drinking days. Descriptively, low reactivity days and high arousal positive affect days had the highest probabilities of cannabis use (.39 and .34, respectively; Figure 1, Panel A), while mixed affect days had the lowest probability of cannabis use (.27). However, these differences were not statistically significant (χ2 = 1.44, p = .84).

Table 4.

Effects of profile membership on outcomes--mean (SE) / probability (SE)

Profile 1: Undifferentiated negative affect days Profile 2: Undifferentiated positive affect days Profile 3: High arousal positive affect days Profile 4: Mixed affect days Profile 5: Low reactivity days Significant pairwise comparisons
Substance Variable
Alcohol
Any Use 0.22 (0.03) 0.34 (0.03) 0.34 (0.03) 0.31 (0.03) 0.36 (0.04) 1 < 2, 3, 5
Level of Use 5.17 (0.50) 6.73 (0.42) 6.48 (0.69) 6.17 (0.49) 6.32 (0.52) --
Cannabis
Any Use 0.29 (0.05) 0.31 (0.06) 0.34 (0.05) 0.27 (0.05) 0.39 (0.05) --
Level of Use 7.92 (1.37) 7.48 (1.12) 9.74 (1.77) 6.74 (1.43) 9.95 (2.60) --
Open in a new tab

Note. Binary variables were coded as 0 = no use, 1 = any use. For binary variables values indicate probability of ‘any use’ for each profile and for continuous variables values indicate mean for each profile.

Levels of Alcohol and Cannabis Use

Regarding level of alcohol use on drinking days, the overall test of differences across profiles was non-significant (χ2 = 5.87, p = .21), suggesting that the number of alcoholic drinks consumed was similar across profiles. Profiles did not significantly differ in terms of level of cannabis use on cannabis use-days (χ2 = 2.16, p = .71). Examination of descriptive patterns (Figure 1, Panel B) shows that on days with any alcohol use, number of drinks was highest on undifferentiated positive affect days and high arousal positive affect days and lowest on undifferentiated negative affect days. For cannabis use, the number of hits used was highest on low reactivity and high arousal positive affect days and lowest on mixed affect days.2

Discussion

Emotion plays an important role in substance use in daily life, yet few studies have examined multiple components of affect in predicting health behaviors. The present study used a nuanced, person-centered approach to assess daily patterns of affect and examined patterns as risk factors for alcohol and cannabis use among young adults. Results revealed five distinct typologies of days characterized by different patterns of PA and NA, defined by valence, arousal, and discrete emotional states. Profiles were differentially linked with day-level alcohol use but did not differ for day-level cannabis use.

The findings regarding the pattern of day-level profiles of affect have several important implications. First, using an occasion-based approach allowed for the identification of heterogeneity in the relations between PA and NA at the day level. Specifically, while some subtypes of days were characterized by negative correlations between PA and NA (i.e., undifferentiated positive affect days, undifferentiated negative affect days), other types of days displayed positive associations between PA and NA (i.e., mixed affect days, low reactivity days), suggesting that PA and NA at the day level represent independent factors (Watson & Clark, 1997), rather than a bipolar structure as proposed by Barrett & Russell (1999). In addition, the emergence of a mixed affect days profile that was evident in around 20% of days in our sample adds to the literature on mixed emotional states, suggesting that they are not a rare occurrence, but instead are relatively common (Larsen, et al., 2017). Our results align with prior findings that individuals experience mixed (i.e., both positive and negative) emotional states in everyday life on about 33% of days (Trampe et al., 2015), and suggest that LPA may be a helpful approach to modeling these day-level relationships.

Regarding links between affect profiles and daily substance use, our findings revealed that while day-level profiles were differentially linked to probability of alcohol use, daily patterns of affect were unassociated with cannabis use. Specifically, undifferentiated negative affect days were linked to lower odds of using alcohol compared to undifferentiated positive affect days, low reactivity days, and high arousal positive affect days. Importantly, the finding that day-level profiles with higher levels of NA were linked to lower odds of alcohol use is contrary to several influential theories of alcohol use positing that a primary reason for alcohol use is to alleviate or avoid negative emotion (e.g., tension reduction models, self-medication). Findings add to a growing body of work that suggest limited support for the link between daily NA and alcohol use (e.g., Dora et al., 2023; Dvorak et al., 2016; Patrick et al., 2016). Accordingly, greater attention to boundary conditions by specifying when (i.e., situations) and for whom (i.e., person characteristics) NA is expected to be related to alcohol use will be critical for theory advancement (Busse et al., 2017). While some theories point to person-level characteristics, such as drinking motives (e.g., drinking to cope), it is notable that negative affect does not reliably predict drinking even in those who endorse greater coping motives (e.g., Dora et al., 2023). Thus, it seems important that theory address not only person-level moderating variables but add further specificity in terms of timing (e.g., proximity to use), situational variables (e.g., social setting, stressful events), cognitive processes (e.g., decision making), and other contextual factors (e.g., availability, peers). Greater integration across theories that address each of these factors separately may be a fruitful avenue for better situating the role of NA in substance use in future work.

Undifferentiated PA days and high arousal PA days were more likely to be drinking days compared to undifferentiated NA days. These findings build on prior literature suggesting an important role for PA in predicting alcohol use in young adults (e.g., Jones et al., 2021) and add nuance by demonstrating the importance of considering relative levels of NA. Specifically, even though mixed affect days had higher than average levels of PA, this profile was not linked to greater odds of alcohol use. Instead, only day-level profiles that were characterized by higher levels of PA and lower than average NA were linked to greater odds of alcohol use, suggesting that it is the unique combination of greater PA and lack of NA that is linked to greater risk for alcohol use in daily life.

Findings are consistent with theories suggesting that PA plays an important role in alcohol use, particularly among individuals who drink socially (e.g., via enhancement motives; Cooper et al., 2016), and may suggest that personalized, real-time interventions targeting positive emotional processes should also take into account relative levels of daily NA when determining daily risk for alcohol use. Future studies can build upon the present work by examining mechanisms that may explain these associations, such as emotion-relevant impulsivity (e.g., positive urgency; Cyders & Smith, 2008). These findings can also be considered within the context of literature on PA and health outcomes more broadly (Pressman & Cohen, 2005), and calls for further research on potential negative impacts of positive emotions on health and further theoretical explication of these processes in daily life.

Intervention Implications

By enabling a more nuanced understanding of the nature of the relationship between affect and substance use in daily life, the present study advances prior research by identifying not only who is at greatest risk for alcohol use, but what affective states may put one at heightened risk, which may aid in future theory building and intervention efforts. Given that profiles characterized by high PA and lower NA were more likely to be linked to daily alcohol use compared to days high in just PA, this may suggest that future efforts to enhance prediction of substance use in daily life (e.g., to develop Just-In-Time adaptive interventions [JITAIs]) should account for relative levels of both PA and NA to most accurately capture moments of greatest risk. For example, at present, some JITAIs just use stress without assessing other negative states or any positive emotional states (e.g., Battalio et al., 2021). In addition, some JITAIs have used a bipolar scale to assess daily affect (e.g., one scale that ranges from positive to negative; Coughlin et al., 2021), whereas the present study findings suggest that better measurement would include at least two separate items—one for PA, one for NA—given the daily structure of affect identified. Testing the relative predictive utility of each measurement approach would be a fruitful next step. Furthermore, future app-based digital health interventions targeting daily alcohol use could use morning affect profiles to provide risk “feedback” to the user and provide alternative activities to enhance positive emotion via a list of substance-free reinforcing social activities and/or to send reminders to students regarding safe drinking practices (e.g., spacing, hydration). In addition, emotion regulation interventions that target impulsivity when experiencing positive states may be warranted.

Finally, because higher NA was linked to lower substance use this may suggest that when targeting substance use specifically in young adults, targeting negative emotional states may be less optimal. However, given the limits of generalizability of our sample (e.g., predominantly White young adults who engaged in binge drinking and co-use of alcohol and cannabis), findings may not extend to other samples or populations. For example, there is some evidence that young adults with attention deficit hyperactivity disorder (ADHD) showed significant associations between NA and alcohol use, whereas these associations were not present in those without ADHD (Kennedy et al., 2022). Moreover, recent work suggests that NA and alcohol craving were more strongly associated in Black individuals facing high discrimination stress (Pederson et al., 2022). Thus, future work that accounts for person-level variables (e.g., clinical diagnoses, demographics) and environmental factors (e.g., racial discrimination) is important to continue adding to the knowledge base as to whether and for whom targeting NA to reduce alcohol use is appropriate. The findings of the present study (in combination with prior literature ; e.g., Dora et al., 2023), suggest that for the majority of young adult drinkers, NA may not be the best place to start.

Methodological Applications of Latent Profile Models in Daily Life

The present study advances statistical methods relevant to the study of affective science and health psychology through the application of a novel methodological approach. Specifically, the latent profile application used herein that can account for both day-level and person-level factors can address similar questions about other critical day-level patterns of proximal risk-factors for substance use, thereby aiding in our understanding of how to best intervene in moments of greatest risk. For example, adding additional indicators of affect, such as physiological arousal assessed via wearable sensors, could be pursued in future studies. In addition, the present modeling approach could be applied to study the relationships between emotion and other types of health behaviors in daily life shown to be impacted by emotion (e.g., diet, exercise) and could enhance our understanding of how complex patterns of affect impact health more broadly across different kinds of health promoting and health risk behaviors. Another potential application lies in understanding daily patterns of health behaviors themselves to better understand risk. The health psychology literature often does not account for multiple risk behaviors (e.g., sleep, exercise, diet, smoking) in the same study, which are often examined in isolation. The present methodology could be applied to identify what patterns of daily health behaviors are linked to greatest risk, as well as how these may vary across demographic characteristics. Such an approach could inform future intervention efforts by elucidating what combinations of daily behaviors/habits matter most and for whom.

Limitations and Future Directions

There are several important limitations of the present work that warrant mention. First, the results of the present study are tempered by limits on generalizability with respect to sample characteristics. Our sample consisted of young adults (primarily college students) who reported recent simultaneous alcohol and cannabis use and recent binge drinking. Thus, findings concerning the patterns of affect may not necessarily be typical of all young adults and should be replicated in additional samples, including a more representative sample of non-college-attending young adults and treatment-seeking clinical samples. In addition, due to the racial homogeneity in a predominately white sample, results may not generalize to other racial groups. This is particularly important considering that Hispanic/Latine individuals report higher levels of alcohol use problems (Vaeth et al., 2017), so future research that identifies patterns of affect and links to daily use will be important to identify any potentially unique patterns of emotional risk factors in more vulnerable populations.

Second, while the use of a daily diary design provides some insight into daily processes, future studies using ecological momentary assessments would allow for greater granularity of affect measurement, as well as greater proximity of emotional experiences to substance use behavior. Participants responded to daily diary surveys in the morning about current affect and yesterday’s substance use. Therefore, we cannot rule out the possibility substance use may have preceded the morning affect measurement. Given that that average survey completion time was 10:21am and the relative infrequency of day-drinking and daytime cannabis use amongst young adults (estimated at 9% of drinking days and 11% of cannabis days; Calhoun et al., 2023; Calhoun & Maggs, 2021), it is likely that affect reports preceded substance use. However, fine-grained ecological momentary assessment studies that more precisely assess the timing of substance use could help to tease apart these effects. In addition, as some research suggests that morning affect tends to be overall more positive than evening affect (English & Carstensen, 2014), future studies should assess evening affect in addition to morning affect to understand how time of day may impact emotional patterns and links to use.

Third, the present study only assessed affect as a predictor of daily substance use, which does not provide insights into other factors known to impact young adult alcohol and cannabis use, such as motivational factors, context, social norms, and beliefs. There may be important interactions between these trait-level variables that may moderate the associations between affect and substance use (e.g., coping motives; Skrzynski & Creswell, 2020). Nevertheless, the present methodology offers a way to incorporate such factors in future studies. For example, instead of just using emotional indicators in the model, one could use motives, perceived norms, and affect scores as indicators to build a more comprehensive understanding of these variables in relation to both affect and substance use in terms of their daily patterns and associations. Moreover, the present study cannot speak to consequences of use. Given that prior work suggests that NA-linked drinking is associated with more negative consequences in college students (Prince et al., 2018), examining consequences is an important future direction.

Finally, while the PANAS can provide some insights into valence-arousal relationships, future studies that employ a more expansive set of emotion terms (e.g., Harmon-Jones et al., 2016; Yik et al., 2011) could help to further elucidate nuanced relationships between specific emotional states and substance use.

Conclusions

The current study extended prior work on the role of emotion in substance use by identifying daily profiles of affect and examining their relationships with daily alcohol and cannabis use. Together, findings suggest that daily emotional patterns may meaningfully predict daily alcohol use and level of use, such that experiencing PA in the absence of NA may be particularly risky for alcohol use, while high arousal positive states may have stronger associations with greater levels of alcohol use. By considering emotion’s role using an occasion-level analysis and in a multidimensional, holistic way, the findings from this study reveal important and nuanced differences among college students’ daily emotional risk factors for substance use, highlighting that conceptualizations of emotion in relation to substance use may be aided by classifications that go beyond mean PA and NA. In addition, the methodological approach used herein may provide a useful framework for conceptualizing daily affect in future studies linking affect to health behaviors in daily life. Our hope is that this work will catalyze ongoing investigations into building daily process models relevant to a variety of distinct health behaviors and thereby contribute to building meaningful theories that can aid in advancing translational work on interventions targeting emotional processes relevant to health outcomes.

Supplementary Material

Supplemental tables

Public Health Significance Statement:

This study introduces a novel application of latent profile analysis for nested data and demonstrates its utility in capturing affect patterns in daily life in relation to alcohol and cannabis use. Findings suggest that daily affect patterns differentially impact alcohol use such that positive affect in the absence of negative affect may be particularly risky for use. By enabling a more nuanced understanding of the nature of the relationship between affect and substance use in daily life, the present study advances prior research by identifying not only who is at greatest risk for alcohol use, but what affective states may put one at heightened risk, which may aid in future theory building and intervention efforts.

Funding:

The research was supported by a Doctoral Student Small Grant Program award from the Research Society on Alcoholism, by award P50 DA039838, R01DA057588, and T32DA007250 from the National Institute on Drug Abuse (NIDA), and by award K01 AA026854 by the National Institute on Alcohol Abuse and Alcoholism (NIAAA). NIDA and NIAAA did not have any role in study design; collection, analysis, and interpretation of the data; writing the report; or the decision to submit the report for publication.

Footnotes

The authors have no conflicts of interest to report.

CRediT Taxonomy: NVD: conceptualized the project; analyzed the data; wrote the paper. BCB: developed the methodology; oversaw the data analyses; collaborated on writing the paper, particularly the sections pertaining to methods and data analysis. JAS: collaborated on writing the paper. ALC: designed the study; collected the data; oversaw the analyses; collaborated on writing the paper.

Disclosures and Acknowledgements: Data and code are available at https://osf.io/tjg92/.

1

Odds of profile membership did not differ by gender (see Table S3).

2

In response to a reviewer comment, we ran exploratory analyses to examine whether day-level affect profiles differed in terms of probability of simultaneous alcohol and marijuana use. Probability of simultaneous alcohol and marijuana use did not significantly differ by profiles (χ2 = 0.78, p = .94). Results are shown in Table S4 of the supplementary materials.

References

  1. Bakk Z, & Vermunt JK (2016). Robustness of stepwise latent class modeling with continuous distal outcomes. Structural Equation Modeling: A Multidisciplinary Journal, 23, 20–31. [Google Scholar]
  2. Barrett LF (1998). Discrete emotions or dimensions? The role of valence focus and arousal focus. Cognition & Emotion, 12, 579–599. [Google Scholar]
  3. Barrett LF, & Russell JA (1999). The structure of current affect: Controversies and emerging consensus. Current Directions in Psychological Science, 8, 10–14. [Google Scholar]
  4. Battalio SL, Conroy DE, Dempsey W, Liao P, Menictas M, Murphy S, Nahum-Shani I, Qian T, Kumar S, & Spring B (2021). Sense2Stop: a micro-randomized trial using wearable sensors to optimize a just-in-time-adaptive stress management intervention for smoking relapse prevention. Contemporary Clinical Trials, 109, 106534. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bresin K, & Fairbairn CE (2019). The association between negative and positive affect and alcohol use: an ambulatory study. Journal of Studies on Alcohol and Drugs, 80, 614–622. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Busse C, Kach AP, & Wagner SM (2017). Boundary conditions: What they are, how to explore them, why we need them, and when to consider them. Organizational Research Methods, 20, 574–609. [Google Scholar]
  7. Calhoun BH, Graupensperger S, Fairlie AM, Walukevich-Dienst K, Patrick ME, & Lee CM (2023). “Wake-and-bake” cannabis use: Predictors and cannabis-related outcomes of use shortly after waking. Drug and alcohol dependence, 248, 109–937. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Calhoun BH, & Maggs JL (2021). Day drinking among college students and its association with risky substance use behaviors. Alcoholism: Clinical and Experimental Research, 45(12), 2546–2559. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cooper ML, Frone MR, Russell M, & Mudar P (1995). Drinking to regulate positive and negative emotions: a motivational model of alcohol use. Journal of personality and social psychology, 69, 990–210. [DOI] [PubMed] [Google Scholar]
  10. Cooper ML, Kuntsche E, Levitt A, Barber LL, & Wolf S (2016). Motivational Models of Substance Use: A Review of Theory and Research on Motives for Using Alcohol, Marijuana, and Tobacco. The Oxford Handbook of Substance Use and Substance Use Disorders, 1, 375–385. [Google Scholar]
  11. Coughlin LN, Nahum-Shani I, Bonar EE, Philyaw-Kotov ML, Rabbi M, Klasnja P, & Walton MA (2021). Toward a Just-in-Time Adaptive Intervention to Reduce Emerging Adult Alcohol Use: Testing Approaches for Identifying When to Intervene. Substance use & misuse, 56, 2115–2125. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cyders MA, & Smith GT (2008). Emotion-based dispositions to rash action: positive and negative urgency. Psychological Bulletin, 134, 807. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. DeCoster J (2001). Transforming and restructuring data. University of Alabama, Department of Psychology, 7, 8–15. [Google Scholar]
  14. Dejonckheere E, Mestdagh M, Verdonck S, Lafit G, Ceulemans E, Bastian B, & Kalokerinos EK (2021). The relation between positive and negative affect becomes more negative in response to personally relevant events. Emotion, 21, 326–330. [DOI] [PubMed] [Google Scholar]
  15. Dora J, Piccirillo M, Foster KT, Arbeau K, Armeli S, Auriacombe M, Bartholow BD, Beltz A, Blumenstock S, & Bold K (2023). The daily association between affect and alcohol use: A meta-analysis of individual participant data. Psychological Bulletin, advanced online publication. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Dornbach-Bender A, Ruggero CJ, Smith P, Schuler K, Bennett CB, Neumann CS, & Callahan JL (2020). Association of behavioral activation system sensitivity to lower level facets of positive affect in daily life. Personality and Individual Differences, 152, 109570. [Google Scholar]
  17. Dvorak RD, Waters AJ, MacIntyre JM, & Gwaltney CJ (2018). Affect, craving, and cognition: An EMA study of ad libitum adolescent smoking. Psychology of Addictive Behaviors, 32(6), 583. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dvorak RD, Pearson MR, Sargent EM, Stevenson BL, & Mfon AM (2016). Daily associations between emotional functioning and alcohol involvement: Moderating effects of response inhibition and gender. Drug and alcohol dependence, 163, S46–S53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ekman P (1992). An argument for basic emotions. Cognition & emotion, 6(3–4), 169–200. [Google Scholar]
  20. English T, & Carstensen L (2014). Emotional experience in the mornings and the evenings: consideration of age differences in specific emotions by time of day. Frontiers in Psychology, 5, 185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Eisele G, Lafit G, Vachon H, Kuppens P, Houben M, Myin-Germeys I, & Viechtbauer W (2021). Affective structure, measurement invariance, and reliability across different experience sampling protocols. Journal of Research in Personality, 92, 104094. [Google Scholar]
  22. Gustafson R (1991). Does a moderate dose of alcohol reinforce feelings of pleasure, well-being, happiness and joy? A brief communication. Psychological reports, 69, 220–222. [DOI] [PubMed] [Google Scholar]
  23. Harmon-Jones C, Bastian B, & Harmon-Jones E (2016). The discrete emotions questionnaire: A new tool for measuring state self-reported emotions. PloS one, 11(8), e0159915. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Howell AN, Leyro TM, Hogan J, Buckner JD, & Zvolensky MJ (2010). Anxiety sensitivity, distress tolerance, and discomfort intolerance in relation to coping and conformity motives for alcohol use and alcohol use problems among young adult drinkers. Addictive behaviors, 35, 1144–1147. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Jacobson NC, Evey KJ, Wright AG, & Newman MG Integration of discrete and global structures of affect across three large samples: Specific emotions within-persons and global affect between-persons. Emotion, 101–107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Jones DR, Allen HK, Lanza ST, & Graham-Engeland JE (2021). Daily associations between affect and alcohol use among adults: The importance of affective arousal. Addictive behaviors, 112, 106623. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kennedy TM, Molina BS, & Pedersen SL (2022). Change in Adolescents’ Perceived ADHD Symptoms Across 17 Days of Ecological Momentary Assessment. Journal of Clinical Child & Adolescent Psychology, 1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Khantzian EJ (1997). The self-medication hypothesis of substance use disorders: A reconsideration and recent applications. Harvard review of psychiatry, 4(5), 231–244. [DOI] [PubMed] [Google Scholar]
  29. Kuntsche E, & Kuntsche S (2017). Development and initial validation of the alcohol expectancy task. Alcoholism: Clinical and Experimental Research, 41(8), 1461–1470. [DOI] [PubMed] [Google Scholar]
  30. Kuppens P, Tuerlinckx F, Russell JA, & Barrett LF (2013). The relation between valence and arousal in subjective experience. Psychol Bull, 139(4), 917–940. [DOI] [PubMed] [Google Scholar]
  31. Larsen JT, Hershfield H, Stastny BJ, & Hester N (2017). On the relationship between positive and negative affect: Their correlation and their co-occurrence. Emotion, 17, 323. [DOI] [PubMed] [Google Scholar]
  32. Linden-Carmichael AN, Van Doren N, Masters LD, & Lanza ST (2020). Simultaneous alcohol and marijuana use in daily life: Implications for level of use, subjective intoxication, and positive and negative consequences. Psychology of addictive behaviors, 34(3), 447. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Linden-Carmichael AN, Van Doren N, Bray BC, Jackson KM, & Lanza ST (2022). Stress and affect as daily risk factors for substance use patterns: An application of latent class analysis for daily diary data. Prevention science, 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Levenson RW, Sher KJ, Grossman LM, Newman J, & Newlin DB (1980). Alcohol and stress response dampening: pharmacological effects, expectancy, and tension reduction. Journal of Abnormal Psychology, 89, 528. [DOI] [PubMed] [Google Scholar]
  35. Mohr CD, Armeli S, Tennen H, Temple M, Todd M, Clark J, & Carney MA (2005). Moving beyond the keg party: a daily process study of college student drinking motivations. Psychology of addictive behaviors, 19(4), 392–403. [DOI] [PubMed] [Google Scholar]
  36. Mohr CD, Brannan D, Wendt S, Jacobs L, Wright R, & Wang M (2013). Daily mood–drinking slopes as predictors: A new take on drinking motives and related outcomes. Psychology of Addictive Behaviors, 27(4), 944. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Patrick ME, Yeomans-Maldonado G, & Griffin J (2016). Daily reports of positive and negative affect and alcohol and marijuana use among college student and nonstudent young adults. Substance use & misuse, 51(1), 54–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Patrick ME, Schulenberg JE, Miech RA, Johnston LD, O’Malley PM, & Bachman JG (2022). Monitoring the Future Panel Study Annual Report: National Data on Substance Use among Adults Ages 19 to 60, 1976–2021. Institute for Social Research. [Google Scholar]
  39. Pedersen SL, Kennedy TM, Holmes J, & Molina BS (2022). Momentary associations between stress and alcohol craving in the naturalistic environment: differential associations for Black and White young adults. Addiction, 117(5), 1284–1294. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Pressman SD, & Cohen S (2005). Does positive affect influence health? Psychological Bulletin, 131(6), 925. [DOI] [PubMed] [Google Scholar]
  41. Prince MA, Pearson MR, Bravo AJ, & Montes KS (2018). A quantification of the alcohol use‐consequences association in college student and clinical populations: A large, multi‐sample study. The American Journal on Addictions, 27(2), 116–123. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Rankin LA, & Maggs JL (2006). First-Year College Student Affect and Alcohol Use: Paradoxical Within- and Between-Person Associations. Journal of Youth and Adolescence, 35(6), 925–937. [Google Scholar]
  43. Russell JA (1980). A circumplex model of affect. Journal of personality and social psychology, 39(6), 1161. [DOI] [PubMed] [Google Scholar]
  44. Simons JS, Dvorak RD, Batien BD, & Wray TB (2010). Event-level associations between affect, alcohol intoxication, and acute dependence symptoms: Effects of urgency, self-control, and drinking experience. Addictive behaviors, 35(12), 1045–1053. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Simons JS, Gaher RM, Oliver MN, Bush JA, & Palmer MA (2005). An experience sampling study of associations between affect and alcohol use and problems among college students. Journal of studies on alcohol, 66(4), 459–469. [DOI] [PubMed] [Google Scholar]
  46. Skrzynski CJ, & Creswell KG (2020). Associations between solitary drinking and increased alcohol consumption, alcohol problems, and drinking to cope motives in adolescents and young adults: a systematic review and meta‐analysis. Addiction, 115(11), 1989–2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Testa M, Wang W, Derrick JL, Brown WC, & Collins RL (2019). Does morning affect contribute to daily Cannabis use? Addict Behav, 95, 64–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Trampe D, Quoidbach J, & Taquet M (2015). Emotions in Everyday Life. PLoS One, 10(12). [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Treloar Padovano H, Merrill JE, Colby SM, Kahler CW, & Gwaltney CJ (2020). Affective and situational precipitants of smoking lapses among adolescents. Nicotine and Tobacco Research, 22(4), 492–497. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Vaeth PA, Wang‐Schweig M, & Caetano R (2017). Drinking, alcohol use disorder, and treatment access and utilization among US racial/ethnic groups. Alcoholism: Clinical and Experimental Research, 41(1), 6–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Watson D, Wiese D, Vaidya J, & Tellegen A (1999). The two general activation systems of affect: Structural findings, evolutionary considerations, and psychobiological evidence. Journal of personality and social psychology, 76(5), 820. [Google Scholar]
  52. Van Doren N, Linden-Carmichael A, Bray BC, & Soto JA (2023, November 3). Characterizing the role of affect in substance use in daily life: A novel application of latent profile analysis for daily diary data. Retrieved from osf.io/tjg92.
  53. Widaman KF (2006). Best practices in quantitative methods for developmentalists: III. Missing data: What to do with or without them. Monographs of the Society for Research in Child Development, 71(3), 42–64. [DOI] [PubMed] [Google Scholar]
  54. Yik M, Russell JA, & Steiger JH (2011). A 12-point circumplex structure of core affect. Emotion, 11(4), 705. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental tables
NIHMS1943374-supplement-Supplemental_tables.docx (28.8KB, docx)

RESOURCES