Loneliness in the college student population: prevalence and associations with substance use outcomes

Abstract

Objectives:

We assessed the prevalence of loneliness by gender and gender-specific associations between loneliness and substance use in university students.

Participants:

Participants were students enrolled in colleges and universities throughout the United States (n = 84,481).

Methods:

Gender was self-reported and categorized as male, female, and transgender/nonbinary. Descriptive analyses were conducted, and logistic regression was used to measure associations between loneliness and substance use by gender and substance.

Results:

Loneliness prevalence was 58% in the sample, and loneliness was most prevalent among transgender/nonbinary students (79%). Loneliness was associated with reduced odds of alcohol use and increased odds of benzodiazepine use in all genders, and increased odds of marijuana, stimulant, and opioid misuse in females and males.

Conclusions:

Loneliness is prevalent in college students, especially in transgender/nonbinary students, and it is associated with increased odds of substance use. College administrators should consider implementing interventions to reduce loneliness in this population.

Introduction

Loneliness is an emerging public health concern and the subject of significant recent discussion. In 2023, United States Surgeon General issued an extensive report urging action on loneliness, the subjective feeling that one lacks sufficient social connections.¹ Notably, loneliness has been linked to premature mortality,² and mortality due to cardiovascular disease and stroke,³ hypertension,⁴ and type 2 diabetes.⁵ Furthermore, research has shown that loneliness is associated with depression and anxiety,^6,7 suicidal ideation in adolescents,⁸ and suicide mortality in men.⁹

For numerous reasons, the negative health consequences of loneliness are particularly critical to recognize among young adults, including that young adults are the age group most likely to report experiencing loneliness.^10,11 Given that roughly 15 million young adults are enrolled in U.S. higher education,¹² colleges and universities represent an important setting to understand and address loneliness. The American College Health Association National College Health Assessment (ACHA-NCHA) has reported the prevalence of loneliness to be approximately 50% in the full college population in the last two years and generally higher in transgender/nonbinary students (66–67%) than females (50–52%) and males (49–50%).^13–16 In contrast, a recent publication of Canadian students found that females had a slightly lower risk of loneliness than males, who had similar outcomes to nonbinary respondents.¹⁷

While associations between loneliness and various negative mental health outcomes, including depression,^18,19 anxiety,^18,19 stress,¹⁹ overall mental distress,¹⁷ suicidal ideation,²⁰ and eating disorder risk¹⁹ have been found among college students, the relationship between loneliness and substance use has been understudied in this population. Existing evidence suggests that associations between loneliness and substance use in this population vary according to type and measure of substance use, prompting need for national assessment of several substances in a comparable manner.

Substance use is common in college students,²¹ and students who report substance use are also more likely than students without substance use to screen positively for depression and anxiety and to report suicidal ideation.²² Furthermore, while there is evidence that the association between loneliness and substance use may differ by gender,^23–25 studies of college students have not explicitly examined gender differences in these associations while including transgender/nonbinary students.

Understanding how the relationship between loneliness and substance use differs by gender in this population is essential, because loneliness is a potentially mutable risk factor and one addressable through public health interventions in university settings.²⁶ Targeting these interventions towards populations most at risk of the negative outcomes of loneliness can help direct limited resources most appropriately. Given the limitations of previous research, studies using validated measures of loneliness and examining gender-specific associations can make an important contribution to knowledge in this area.

To fill these gaps in the literature, we first examined the proportion of students reporting loneliness by gender using a valid²⁷ and reliable measure in a large sample of students enrolled in colleges and universities across the United States. Second, we estimated the gender-specific associations between loneliness and substance use in these students. Given that no previous studies have examined the relationship between loneliness and several different types of individual substances by gender specifically in college students using a validated and reliable measure of loneliness, this is a crucial first step in understanding how the high prevalence of loneliness may affect the college student population.

Materials and methods

Data source

This paper used data from the 2021–2022 Healthy Minds Study (HMS),²⁸ a web-based, cross-sectional survey. HMS examines mental health, substance use, and health service utilization among students enrolled in postsecondary institutions throughout the United States. All participants provided informed consent, and this study was approved by the Institutional Review Board of all relevant campuses.

Details regarding administration and recruitment of the study have been described elsewhere.^28,29 In brief, at universities with more than 4,000 students, a random sample of 4,000 students was invited to participate in the survey. At institutions with fewer than 4,000 students, all students were invited to participate. Students must have been 18 years or older to participate in the survey. Invitations were sent out via email, and all invited students were eligible to receive one of several prizes. The total prize value was $2,000 per annual wave.

Participants

The complete data set from HMS 2021–2022 includes 95,860 students at 133 colleges and universities. After removing individuals with missing data on one or more of the variables of interest (gender, loneliness, age, financial stress, and substance use measures), the result for all analyses was a final analytic sample of 83,471 students.

Measures

Age in years was measured as a continuous variable. Race was measured by having participants “select all that apply” from the following categories: “African American/Black”, “American Indian or Alaskan Native”, “Asian American/Asian”, “Hispanic/Latin(x)”, “Native Hawaiian or Pacific Islander”, “Middle Eastern, Arab, or Arab American”, “White”, and “Self-identify”. Gender was measured by having participants “select all that apply” from the following categories: “Male”, “Female”, “Trans male/Trans man”, “Trans female/Trans woman”, “Genderqueer/Gender non-conforming”, “Self-identify”, and “Gender non-binary”. Participants were considered female if they selected “Female”, male if they selected “Male”, and transgender/nonbinary if they selected any other gender. Participants who selected both a transgender/nonbinary category and a cisgender category were classified as transgender/nonbinary. Those who selected both male and female, but not a transgender/nonbinary gender, were removed from the primary analysis (n = 14).

Loneliness was measured with the UCLA 3-item loneliness scale, a widely used,³⁰ reliable, and valid²⁷ measure of loneliness. There are three questions in the scale: “How often do you feel that you lack companionship?”; “How often do you feel left out?”; and “How often do you feel isolated from others?”.²⁷ A score of 1 was assigned to each question if the participant responded, “Hardly ever”, 2 if they responded, “Some of the time”, and 3 if they responded, “Often”. A total score was calculated by summing the scores for the three items. For descriptive and regression analyses, we used a score of greater than or equal to 6 to dichotomize high and low levels of loneliness³¹ based on the existing literature in this population.^13–16 We conducted additional descriptive analyses examining differences in the raw, continuous score (ranging from 3–9) by gender identity.

Alcohol use in the past two weeks (any/none) was determined by asking, “Over the past 2 weeks, did you drink any alcohol?”. All other substances were asked as a “select all that apply” term in response to: “Over the past 30 days, have you used any of the following drugs?”. Marijuana use was considered if a participant selected yes to “Marijuana”. Prescription stimulant misuse was defined as selecting yes to “Other stimulants (such as Ritalin, Adderall) without a prescription or more than prescribed”. Benzodiazepine use was defined as selecting yes to “Benzodiazepines (such as Valium, Ativan, Klonopin, Xanax, Rohypnal (Roofies)”. Opioid misuse was considered if a participant selected yes to “Opioid pain relievers (such as Vicodin, OxyContin, Percocet, Demerol, Dilaudid, codeine, hydrocodone, methadone, morphine) without a prescription or more than prescribed” and/or yes to “Heroin”.

Financial stress was included in descriptive analyses and regression models due to potential to cause bias due to confounding. The financial stress variable included in this analysis was self-reported by participants as the response to: “How would you describe your financial situation right now?” Potential answers included always, often, sometimes, rarely, and never stressful.

Data analysis

First, we conducted descriptive analyses by loneliness status. To examine differences in loneliness between gender groups, we stratified the data into those identifying as female, those identifying as male, and those identifying as transgender/nonbinary. We calculated the number of individuals in each category (lonely or not lonely, by gender) and respective weighted percentages of the total group sample size for categorical variables, or the mean and standard deviations for continuous variables. We additionally examined differences in both the prevalence of loneliness and the continuous loneliness scores with chi-squared test with Rao & Scott’s second-order correction and Wilcoxon rank-sum test respectively. Survey weights, based on estimated nonresponse using administrative data on sex, were applied using the “survey” package and its svydesign() function in R.³² The response rate for the 2021–2022 survey was 12%.³³

Next, we conducted separate multiple logistic regression analyses to examine the association between loneliness and substance use for each substance within gender strata, adjusting for age, financial stress and, in all analyses except that for marijuana use, marijuana use in the last 30 days. Use of non-marijuana substances was not included as confounders in regression models due to low prevalence in the sample (stimulants, benzodiazepines, and opioids) or due to lack of a bivariate association with loneliness (alcohol), which eliminated the possibility of confounding. We checked for multicollinearity between predictors using adjusted generalized standard error inflation factor (aGSIF) and ensured that values in all models were below 2.2 (square root of 5; see Fox & Monette, 1992 and Nahhas, 2024 for details^34,35). Sensitivity analyses were conducted in which individuals missing substance use and financial stress data were included in loneliness prevalence calculations (n = 85,159) to ensure that substance use-related nonresponse did not influence these results. Data were analyzed in R version 4.3.0.³⁶

Results

Descriptive analysis

The descriptive characteristics of the sample are presented in Table 1. There were more female (52%) than male (41%) and transgender/nonbinary (7%) participants in the study. Loneliness was endorsed by the majority of participants in the sample (58%). Loneliness was more common in transgender/nonbinary students (79%) than females (59%) and males (54%). Average loneliness score was also higher in transgender/nonbinary participants (m = 6.9, SD = 1.8) than in females (m = 5.9, SD = 2.0) and males (m = 5.7, SD = 2.0). Tests examining both binary and continuous loneliness measures noted differences by gender. In sensitivity analyses including all participants with complete loneliness and gender data, but not necessarily complete substance use data, proportions of loneliness in the full population and by gender were generally equivalent to those reported above. With the exception of alcohol use, loneliness was also associated with greater proportions of substance use, including marijuana use, benzodiazepine use, opioid (including heroin) misuse, and prescription stimulant misuse (Table 1).

Table 1.

Characteristics of the analytic sample by loneliness status and gender

	Female (weighted n = 59,459 unweighted n = 56,232 weighted % = 52%)		Male (weighted n = 46,730 unweighted n = 22,016 weighted % = 41%)		Transgender/Nonbinary (weighted n = 8,098 unweighted n = 5,223 weighted % = 7%)
	Lonely (59%)	Not Lonely (41%)	Lonely (54%)	Not Lonely (46%)	Lonely (79%)	Not Lonely (21%)
Age	23.5 (7.4)	25.5 (9.4)	23.4 (7.5)	25.1 (9.5)	21.5 (4.5)	22.4 (7.3)
African American/Black	9.9%	9.0%	7.5%	8.1%	8.9%	5.9%
American Indian or Alaskan Native	1.9%	1.3%	1.9%	1.5%	2.2%	2.5%
Asian American/Asian	11.5%	11.1%	12.5%	11.3%	10.4%	11.0%
Hispanic/Latin(x)	18.6%	19.4%	15.8%	17.4%	14.6%	16.8%
Native Hawaiian or Pacific Islander	0.7%	0.5%	0.6%	0.6%	0.7%	0.6%
Middle Eastern, Arab, or Arab American	1.6%	1.3%	2.0%	1.9%	2.1%	1.3%
White	66.8%	66.6%	68.4%	67.5%	77.4%	73.3%
Self-identified race	1.4%	1.1%	2.2%	2.3%	3.7%	6.6%
Alcohol Use Last 2 Weeks	49.3%	48.7%	47.1%	46.2%	43.0%	47.9%
Marijuana Use Last 30 Days	22.6%	15.4%	23.7%	16.4%	35.1%	33.1%
Prescription Stimulant Misuse Last 30 Days	1.2%	0.7%	1.7%	0.9%	2.4%	2.3%
Benzodiazepine Use Last 30 Days	0.7%	0.4%	0.6%	0.4%	1.5%	0.3%
Opioid (including heroin) Misuse Last 30 Days	0.5%	0.2%	0.6%	0.3%	0.9%	0.8%

Data presented as mean (SD) or column percent, applying survey weights.

Regression analyses

In adjusted models, association between loneliness and substance use varied by type of substance and gender. Loneliness was associated with a slightly reduced likelihood of alcohol use in females (OR: 0.94, [0.88, 0.99]), males (OR: 0.95, [0.87, 1.05]), and transgender/nonbinary students (OR: 0.78, [0.63, 0.97]) (Table 2).

Table 2.

Logistic regression examining loneliness (UCLA-3 item loneliness ≥ 6) and substance use by gender and substance in a large sample of college students throughout the U.S. from the 2021–2022 Healthy Minds Study

		Odds Ratio	95% CI
Alcohol	Female	0.94	0.88–0.99
	Male	0.95	0.87–1.05
	Transgender/Nonbinary	0.78	0.63–0.97
Marijuana	Female	1.44	1.35–1.55
	Male	1.45	1.30–1.61
	Transgender/Nonbinary	1.05	0.85–1.30
Prescription Stimulant	Female	1.27	0.95–1.69
	Male	1.56	1.11–2.18
	Transgender/Nonbinary	0.82	0.41–1.65
Benzodiazepine	Female	1.64	1.00–2.70
	Male	1.37	0.75–2.51
	Transgender/Nonbinary	6.69	0.72–62.54
Opioid	Female	2.15	1.05–4.42
	Male	1.67	0.88–3.18
	Transgender/Nonbinary	1.44	0.31–6.78

Analyses controlled for use of marijuana in the last 30 days (except for marijuana analysis), self-reported financial stress, and age

Loneliness was associated with increased odds of marijuana use in females (OR: 1.44, [1.35, 1.55]) and males (OR: 1.45, [1.30, 1.61]), but not in transgender/nonbinary students (OR: 1.05, [0.85, 1.30]). Loneliness was also associated with increased odds of prescription stimulant misuse in females (OR: 1.27, [0.95, 1.69]) and males (OR: 1.56, [1.11, 2.18]), but not in transgender/nonbinary students (OR: 0.82, [0.41, 1.65]). Additionally, strong associations were observed for loneliness and benzodiazepine use in females (OR: 1.64, [1.00, 2.70]) and transgender/nonbinary students (OR: 6.69, [0.72, 62.54]), and, while the association was not statistically significant, an increase was also observed in males (OR: 1.37, [0.75, 2.51]). Finally, loneliness was associated with increased likelihood of opioid misuse in females (OR: 2.15, [1.05, 4.42]) and, although not statistically significant, increased odds was observed among males (OR: 1.67, [0.88, 3.18]) and transgender/nonbinary students (OR: 1.44, [0.31, 6.78]).

Discussion

In this large cross-sectional survey of college students, the prevalence of loneliness was high, especially among transgender/nonbinary participants. Loneliness was associated with a range of substance use outcomes. Whereas loneliness was generally associated with a lower likelihood of alcohol use across gender identities, loneliness was associated with a higher likelihood of benzodiazepine use among all gender identities. For marijuana, opioid, and prescription stimulant misuse, loneliness was associated with greater use in males and females but not in transgender/nonbinary students. Importantly, though, the sample size of transgender/nonbinary participants was small, which may affect statistical significance among this group, particularly for the association between loneliness and opioid use, where the effect size was relatively high. Overall, these findings fill a current gap in the literature by expanding upon evidence supporting mental health concerns in college students to include characterizing differential associations between substance use and loneliness in this population. Our results are important, as loneliness has increased in prevalence since the start of the COVID-19 pandemic.³⁷ This is highly relevant to the current study given the timing of data collection (2021–2022).

Of particular importance is the high proportion of transgender/nonbinary students reporting loneliness in this sample. A few other studies have documented relatively greater loneliness among transgender/nonbinary individuals compared to cisgender peers,^38,39 including in Norwegian university students,⁴⁰ while others have found a similar prevalence of loneliness.¹⁷ These estimates of the prevalence of loneliness provide an important marker and a call to action for colleges and universities to address a mutable risk factor for mental health and substance use, especially for a growing population with unique vulnerabilities due in large part to structural discrimination. There has never been a more important time to prioritize inclusive campus environments that protect the safety and wellbeing of transgender/nonbinary students, including gender-inclusive housing and restrooms, accessible name change policies, student health insurance coverage for transgender healthcare, and others outlined in the Consortium of Higher Education LGBT Resource Professionals’ Promising Policies and Practices for Supporting Trans and Nonbinary People in Postsecondary Education, released in June 2024.⁴¹ While exploring the exact mechanisms of the greater loneliness among transgender/nonbinary students is outside the scope of this paper, it is established that transgender/nonbinary individuals often experience stigma and discrimination across several domains,⁴² which is associated with worsened mental health (e.g., depression and anxiety) in this population.⁴³

Transgender/nonbinary students in the U.S. face health inequalities across several areas of health, including greater levels of suicidality, experiences of violence, and mental health diagnostic history than their cisgender peers.⁴⁴ These statistics warrant public health action to improve transgender/nonbinary students’ mental health, including through the implementation of protective institutional policies. More research is necessary to establish and assess specific interventions, including at the system and community levels, for groups at high risk of loneliness, including transgender/nonbinary students.

In our regression analyses, we found associations between loneliness and substance use vary by gender and type of substance. These results add to the existing literature on substance use and loneliness in college students, which has varying findings depending on the exact definition of the substance use outcome, study population, and study design. First, we found that loneliness was associated with reduced odds of alcohol use in the past two weeks regardless of gender. Past studies on the relationship between loneliness and alcohol use or excessive drinking in the college student population have not established a direct association in the absence of a mediator and/or a moderator^45–47 or have found a protective association.¹⁷ In the current study, the lower odds of alcohol use associated with loneliness may simply reflect the social nature of drinking on university campuses,⁴⁸ rather than a reduced use of alcohol to cope with loneliness or its associated mental health problems. Therefore, given the cross-sectional nature of this study, the reduced likelihood of alcohol use associated with loneliness should not be interpreted as causal.

The strongest associations found in this study were between loneliness and benzodiazepine use, especially in transgender/nonbinary students. However, it is important to note that, while the wording for questions on prescription stimulants and opioids specified whether use was “without a prescription or more than prescribed,” thus indicating misuse, the survey item on benzodiazepines did not make this distinction. Therefore, this substance use outcome has been classified as benzodiazepine use throughout this manuscript, in contrast with opioid and prescription stimulant misuse. However, benzodiazepine use was measured directly after a survey item on opioid misuse. As a result, there may be some misclassification in which benzodiazepine use, as prescribed, and benzodiazepine misuse, without a prescription or not as prescribed, were measured together. Thus, while the associations between loneliness and benzodiazepine use are the strongest calculated effect sizes, they are the most likely to be explained by uncontrolled confounding due to preexisting anxiety disorders. This confounding may explain why our result differs past findings on a null relationship between loneliness and benzodiazepine misuse²⁵ yet concurs with other results on a cross-sectional (but not prospective) association between loneliness and general prescription medication use.⁴⁷ Our results should be interpreted with caution given these important limitations.

Loneliness was associated with increased odds of using many other substances, with some variability across gender and type of substance. To date, previous research on the associations between loneliness and substances other than alcohol has been mixed. Analyses in a general student population have found no association between loneliness and prescription medication misuse regardless of gender.²⁵ In contrast, analyses of American Indian, Alaska Native, and Native Hawaiian college students⁴⁹ and of exclusively students who use drugs⁵⁰ have observed significant relationships between loneliness and substance use, which were more consistent with the associations we observed for many substance- and gender-combinations in this study. Reasons for discrepancies between studies, particularly between studies in general U.S. student populations like the current study and Zullig and Divin,²⁵ may be attributed to using different measure of loneliness. In Zullig and Divin,²⁵ researchers exclusively asked whether individuals had ever “Felt very lonely” in their lifetimes: loneliness was considered positive even if they had felt loneliness at some point, but “not in the past 12-months.” In the current study, however, we used the UCLA 3-item loneliness scale, which does not specify a time frame when measuring loneliness. Instead, the UCLA 3-item loneliness scale assesses loneliness in the present tense as well as the current frequency of loneliness-related occurrences. This difference in the assessment timeframe between loneliness measurements is important because the associated difference in results may indicate that current, but not necessarily lifetime, loneliness is related to several substance use outcomes.

The associations between loneliness and substance use in this analysis add to the body of literature on negative mental and physical correlates of loneliness,^6–9 within a particularly lonely college student population. These results indicate that, in addition to the concerning prevalence of loneliness among students enrolled in colleges and universities, loneliness is also related a greater likelihood of several substance use outcomes. While mechanisms underlying this relationship require more exploration, emerging evidence indicates that stress and anxiety may serve as mediators on the pathway between loneliness and substance use.^51,52

Our findings indicate that college and university administration should intervene to reduce loneliness on college campuses. Several recent reviews have identified effective interventions for loneliness,^26,53,54 including among college students specifically.²⁶ These programs include support groups with student peers to discuss common concerns and opportunities for structured social interaction such as through sports or creative activities.²⁶ Such interventions may help reduce the prevalence of loneliness and its impact on substance use outcomes in this population. Additionally, we recommend that colleges and universities measure loneliness as part of the standard assessment process in campus mental health clinics given the high prevalence in students and the negative health-related consequences of loneliness. Systematic assessment can provide opportunities to connect students with social supports and reduce negative implications of loneliness.

Strengths and limitations

This study has several limitations. First, it is important to consider that these data are self-reported, and thus may be affected by social desirability biases; participants may underreport substance use due to its associated stigma. Additionally, because of limited sample size among transgender/nonbinary individuals, the many diverse genders underlying this group had to be combined into a single transgender/nonbinary category for the purpose of analysis. More research is needed on those who identify as trans male, trans female, genderqueer, nonbinary, and self-identified separately to examine the prevalence of loneliness within these gender categories more specifically. Furthermore, a complete case analysis was conducted in this study, which may induce biases in the regression analyses due to differences in non-responders compared to responders.³³ However, we are confident that calculations of the prevalence of loneliness are robust to substance use-related nonresponse, as complete case and sensitivity analyses produced the same proportions. Survey nonresponse was also accounted for using weighting methods, partially addressing concerns over response rates. Additionally, as noted above, potential misclassification bias may impact the results on benzodiazepine use in this study. Further, while analyses of students on college campuses can be strengthened from considering of institutional clustering through statistical methods such as hierarchical modeling, our analyses were not sufficiently powered to allow these complex models to converge. Finally, as it pertains to the secondary regression analyses, the data are cross-sectional and the association between loneliness and substance use may be reciprocal, so future longitudinal analysis is needed.

In addition to these limitations, this study also has several strengths. The HMS cohort is a large, national sample of college students in the United States, and it includes data from both four-year universities and community colleges. Additionally, this study includes a diverse sample of students, allowing us to examine transgender/nonbinary students, who are an understudied population, particularly as it pertains to loneliness. These results can therefore provide new and important data on this population to motivate future public health interventions.

Conclusion

Loneliness is prevalent and should be considered a significant public health concern in the college student population. Substance use and loneliness have complex relationships depending on the gender and substance studied. Colleges and universities should consider implementing evidence-based interventions to reduce loneliness in college students, including support groups and opportunities for structured social interaction.

Data Availability Statement:

The data that support the findings of this study are available upon request from the Healthy Minds Study: https://healthymindsnetwork.org/research/data-for-researchers/