The separate and joint effects of recent interpersonal abuse and cannabis use on psychotic experiences: findings from students in higher education in the United States

We analyzed data from the 2020–2021 Healthy Minds Study (HMS), a non-probability web-based survey examining health and wellness among students enrolled in higher education in the United States [37]. The HMS survey is administered annually as a repeated cross-section of schools, with a different set of schools every year, including community colleges, four-year colleges, and professional schools. The HMS survey uses several validated measures to provide information about the prevalence of mental health outcomes, knowledge and attitudes about mental health, and service utilization. The survey was administered at 37 institutions of higher learning (N = 34,168) between September through December of 2020, and then administered again at 103 institutions (N = 103,748) between January through June 2021. The response rate was 14%, which is comparable to other response rates from online surveys using convenience samples and panels. We restricted the sample by age (18–29) to isolate young adults and further excluded individuals who were missing data on any of the variables of interest; we used complete-case analysis, resulting in a final analytic sample of 94,722. We used sample probability weights to adjust for non-response using administrative available data on full student populations at each institution, consistent with prior studies [37]. Using multivariable logistic regression, response propensity was estimated based on gender identity, race/ethnicity, academic level, and grade point average. We then assigned response propensity weights to each student who completed the survey. Students who were less likely to have completed the survey were assigned a larger weight in the analysis. Sample weights gave equal aggregate weight to each school in the national estimates rather than assigning weights in proportion to school size, so that overall national estimates were not dominated by schools in our sample with large enrollment. The HMS was approved by the Institutional Review Board Advarra, and the Institutional Review Boards at all participating campuses (IRB number: Pro00028565). Further, the secondary analysis presented in this study was deemed exempt under the approval of USC (UP-22–00068).

Psychotic experiences. Psychotic experiences were measured using an abbreviated version of the World Health Organization Composite International Diagnostic Interview Psychosis Screen, which has been used in large global epidemiology studies [38]. Respondents were asked four questions about the following experiences: (1) A feeling something strange and unexplainable was going on that other people would find hard to believe; (2) A feeling that people were too interested in you or that there was a plot to harm you?; (3) A feeling that your thoughts were being directly interfered or controlled by another person, or your mind was being taken over by strange forces?; and (4) An experience of seeing visions or hearing voices that others could not see or hear when you were not half asleep, dreaming, or under the influence of alcohol or drugs? Respondents were then asked a single item (yes/no) about whether any of these four experiences occurred over the past 12 months. This variable was treated dichotomously in accordance with prior studies to signify the presence of psychotic experiences (i.e., hallucinatory experiences and/or delusional ideations) [38]. We focused on 12-month psychotic experiences to minimize recall bias.

Sociodemographic characteristics and mental health (covariates). We restricted the sample to focus on young adults further controlled for age as a continuous variable. We also adjusted for gender (cis-gender man, cis-gender woman, transgender/nonbinary/other), and race/ethnicity (White, Black, Latinx/Hispanic, Asian American/Pacific Islander, multiracial, and other). Education is a common proxy for socioeconomic status, and looking within a single socioeconomic stratum (i.e., students in higher education) can help isolate the effects of interpersonal abuse and cannabis use on psychotic experiences. While students in higher education represent a relatively high socioeconomic stratum, there is still a socioeconomic gradient within the stratum. Other common proxies (such as income and employment) may be inadequate measures of the gradient given that many students receive financial support from parents and do not have to work, and therefore do not earn any income. Food insecurity, however, has proven to be informative given that it is prevalent and a useful proxy for economic/financial instability [39], especially among young adults and college populations [40], and given its associations with mental health [41‐43]. Food insecurity was assessed using two items, which asked: (1) Within the past 12 months I was worried whether our food would run out before we got money to buy more; (2) Within the past 12 months the food I bought just didn’t last and I didn’t have money to get more. Respondents could answer: never true, sometimes true, often true. Individuals were identified as food insecure with an affirmative answer (sometimes true or often true) to either question, following the two-item screen for families at risk of food insecurity [44]. We also adjusted for mental health using measures of depression and anxiety. Depression was measured using the Patient Health Questionnaire-9 (PHQ-9; Kroenke & Spitzer, 2002). The scale ranged from 0 to 27, which was dichotomized (scores 15 and higher) to reflect moderately severe to severe depression. Anxiety was measured using the General Anxiety Disorder – 7 (GAD-7; Spitzer et al. [46]). The scale ranged from 0 to 21 and was dichotomized (scores 11 and higher) to reflect moderately severe to severe anxiety [45, 46].

We calculated the prevalence of abuse, cannabis use, and all covariates (total and stratified by psychotic experiences). We opted to model predictions on an additive scale because multiplicative models assume that risks multiply in their effects, while additive models assume that psychosis can develop from either of the risk factors acting alone and synergistically (i.e., positive deviations from additivity). Additive models are particularly meaningful in psychiatric epidemiology because it is believed that psychiatric conditions generally have complex multifactorial etiologies. Following prior studies on the topic [35, 36], we tested for additive interaction [47], and depict the synergy between abuse and cannabis use by creating the following categorical variable: (1) no interpersonal abuse or cannabis use; (2) interpersonal abuse only; (3) cannabis use only; and (4) both interpersonal abuse and cannabis use. We adjusted for age, gender, race/ethnicity, food insecurity, depression, and anxiety. We calculated the interaction contrast ratio (ICR), which allows use of odds ratios derived from logistic models to estimate the relative excess risk resulting from the synergy of combined exposures (i.e., ICR = OR_{interpersonal abuse & cannabis use} — OR_{interpersonal abuse only} — OR_{cannabis use only} + 1). An ICR greater than zero signifies a positive deviation from additivity. We used nlcom command in Stata SE 15 to calculate confidence intervals and p-values for the ICR. To assess the potential influence of multicollinearity, we calculated the variance inflation factor (VIF) value for each independent variable. The highest VIF was 1.49, indicating that multicollinearity was unlikely to be a problem in our analyses [35].

In this study, we sought to extend previous research by examining the separate and joint effects of interpersonal abuse (emotional, physical, sexual, over the past year) and cannabis use (past 30 days) on odds of psychotic experiences (past year), adjusting for age, gender, race/ethnicity, food insecurity, anxiety, and depression. In terms of separate effects, we found evidence that interpersonal abuse and cannabis use were each associated with significantly greater odds of psychotic experiences. These findings comport with meta-analyses and systematic reviews showing the association between interpersonal abuse, cannabis use, and psychosis [14, 21, 22]. In terms of joint effects, we found evidence that exposure to interpersonal abuse and using cannabis together increased odds of psychotic experiences beyond either exposure individually. This aligns with a prior study among a general population sample of adults in the United Kingdom. Morgan and colleagues found that interpersonal abuse (OR: 2.04) and cannabis use (OR: 2.11) in the past year combined synergistically to increase odds of psychotic experiences (OR: 5.54), with an ICR of 2.40 (95% CI  – 0.17 to 4.97), though this was only marginally significant [36]. Our study found a smaller but statistically significant synergistic effect (p = 0.015). To our knowledge, our study was the first to test these main and synergistic effects in a large sample of students in higher education in the United States.

Our findings support and contribute to extant literature on psychotic experiences by examining the impact of environmental exposures and their synergies. Prior studies have focused on childhood adversities, since exposures that occur early in the developmental life course tend to shape later health outcomes [31, 32], and could set the trajectory for future exposures to risk factors, including revictimization [48] and cannabis use [49]. Our study examined recent exposures to interpersonal abuse, as studies have also noted recent interpersonal abuse can be particularly impactful on psychotic experiences [36]. The pathways by which interpersonal abuse and cannabis use interact to synergistically increase odds of psychotic experiences are not well understood. The separate and combined effects of interpersonal abuse and cannabis use align with social defeat [30] and other socio-developmental models [50]. It is possible that exposure to interpersonal abuse may lead to stress sensitization and cognitive biases, and in turn produce psychotic experiences [51]. Moreover, interpersonal abuse can be traumatic and lead to the formation of negative schemas that underlie hypervigilance and suspiciousness [52, 53], which can be further exacerbated by cannabis use [54].

Findings should be interpreted considering several limitations. First, in terms of design, the data were cross-sectional and did not allow us to establish the temporal order of events to make causal inferences. The relationships among interpersonal abuse, cannabis use, and psychotic experiences are difficult to disentangle; it is possible that interpersonal abuse can lead to cannabis use (as a form self-medication), in which case, cannabis could serve as a partial mediator, as indicated in prior studies [55, 56]. However, cannabis use may also occur before exposure to interpersonal abuse. And while socio-environmental exposures can precede psychotic experiences, studies also show that psychotic experiences can also occur throughout childhood before interpersonal abuse or cannabis use. Second, in terms of the sample, the study only examined students in higher education in the United States, and findings cannot be generalized beyond this population. The HMS employed a convenience sampling strategy that yielded a large sample but with a relatively low response rate (14%). The response rate is to be expected for online convenience samples [57, 58], and we attempted to account for non-response using survey weights, as done in prior studies using the dataset [37]. However, sampling bias remains a major concern. In terms of measurement, all measures used in the study failed to elicit adequate information about severity, frequency, and context of experiences. Notably the measures cover a relatively short timeframe. Further, the HMS asked whether individuals had used ‘marijuana’, and this item may not have captured all forms of cannabis use. Additionally, there may have been some social desirability bias in that students may have been reluctant to disclose interpersonal abuse, cannabis use, or psychotic experiences, given that the survey was administered through the institutions of higher learning in which the students were enrolled.

Our findings show that the effects of interpersonal abuse and cannabis use are synergistic rather than overlapping, which may suggest the possibility that interpersonal abuse and cannabis use are linked to psychosis via different mechanisms. Moreover, socio-environmental risk factors for psychosis can combine synergistically to shape the expression and persistence of psychosis, potentially shaping one’s health trajectory toward need for care [59]. Since interpersonal abuse, cannabis use, and psychotic experiences may co-occur and coalesce over time, our study highlights the importance of examining the impact and interaction of multiple exposures and their underlying (and potentially distinct) mechanisms. While more research is needed, the co-occurrence of interpersonal abuse and cannabis use may nonetheless signal a higher clinical risk profile; prevention efforts may identify high-risk individuals and groups based on joint exposures and explore the utility of cannabis cessation support for people exposed to interpersonal abuse for psychosis prevention.