Developmental Trajectories and Predictors of Psychological Well-Being and Distress Across the College Years

ABSTRACT Psychological well-being and distress are critical components of college adjustment that are intricately entwined with student retention and success during and after college. This 5-wave longitudinal study used growth mixture modeling to explore heterogeneous trajectories of psychological well-being (self-esteem) and distress (depression, anxiety, stress) spanning just before college to the end of the fourth year. Students (N = 5,537) most commonly were best characterized by trajectories of stable positive or moderate adjustment, though some were better characterized by trajectories of low or variable adjustment. These latter subgroups may represent the highest-need students, for whom identifying pre-college risk and protective factors is crucial. Some notable differences emerged in trajectories for women versus men. Further, several individual characteristics at the cusp of college predicted these four-year trajectories. The strongest psychological functioning predictors were self-esteem, distress, and stress (less consistently, resilience and self-efficacy). The most predictive cognitive-affective strategy was avoidant emotional coping, followed by cognitive reappraisal and expressive suppression (less consistently, problem-focused and active emotional coping). Social well-being factors that best differentiated adjustment trajectories were general social support, followed by support from family and then from friends. These findings have implications for targeting at-risk students upon university arrival to promote optimal long-term adjustment.

adjustment, so that institutions can focus their programming and resources to maximize student well-being and in turn engagement, retention, and success.
Decades of research from hundreds of institutions of higher education -representing a wide range of institutional types and student characteristics -have documented extensive psychosocial challenges, stressors, and mental health issues in college students (American College Health Association [ACHA], 2021; Center for Collegiate Mental Health [CCMH], 2021; Healthy Minds Study [HMS], 2021). Yet, research that illuminates longitudinal patterns of adjustment is relatively limited. Recent research demonstrates that college students generally experience challenges in domains of psychological functioning (e.g., psychological distress and well-being), cognitive-affective strategies (e.g., emotion regulation and coping styles), and social well-being (e.g., social support) during the first two years of college, followed by some limited improvement in the next two years (Conley et al., 2020). However, those who work with highereducation students know well that not all follow the same path of adjustment throughout their college years: Some may consistently flourish or flounder, whereas others may experience improvements or setbacks over time. Indeed, an emerging body of research using growth mixture modeling demonstrates that students experience heterogeneity in both academic (e.g., grades; Bowers & Sprott, 2012;Hodis et al., 2011) and developmental (e.g., [academic] identity; Luyckx et al., 2013;Robinson et al., 2018) trajectories, and these have been linked to wellbeing indicators such as depression and self-esteem. Given the connections between student success and well-being, further research is needed to explore potentially divergent trajectories of psychological adjustment across the college years, and to understand risk and protective factors for such divergent pathways. Understanding heterogeneity in student adjustment over time, and early predictors of these trajectories, can help students, higher education practitioners, and universities thrive.

Trajectories of psychological adjustment across the college years
Much of our understanding of college student adjustment over time is based on group trends, with cross-sectional data. These studies indicate that college students' rates of psychological distress are climbing to unprecedented levels (e.g., CCMH, 2021;HMS, 2021). Some research has charted change in adjustment over time, tracking students over the first year (Conley et al., 2014) or multiple years of college (Chung et al., 2014;Conley et al., 2020;Edwards et al., 2010;Sher et al., 1996). Further, a few studies have explored important nuances in these adjustment patterns, such as by identifying broad gender differences. For example, women's distress during the first year of college steadily worsened whereas men's distress plateaued (Conley et al., 2014).
Meanwhile women's social support improved more consistently in the latter half of college as compared to men (Conley et al., 2020), and women evidenced more effective coping skills than men across time (Conley et al., 2020;Gall et al., 2000). Statistical models have provided strong evidence that men and women had different patterns of adjustment on average across the college years and that data should not be aggregated based on gender when evaluating change (Conley et al., 2020). However, these studies have only examined mean adjustment across students overall, or by gender. Less is known about potential heterogeneity in change, such as patterns of stable versus variable adjustment, for female and male students.
Longitudinal modeling techniques such as growth mixture modeling (GMM) can be used to assess heterogeneity in change over time, accounting for the fact that individuals differ and that there may be subgroups of individuals characterized by different patterns of change (Jung & Wickrama, 2008;Ram & Grimm, 2009). GMM can elucidate patterns and subgroups that mean trajectories alone would miss, and GMM findings can be important for understanding proportions of individuals best characterized by differing types of change. A small body of research has used longitudinal modeling to explore potential variations in developmental trajectories of psychological distress or broader adjustment in college students. These studies span from less than a year (Hirai et al., 2015;Nightingale et al., 2013) to 4 years of college (Galatzer-Levy & Bonanno, 2013;Salmela-Aro et al., 2008), but only rarely include a pre-college baseline assessment (Duchesne et al., 2007;Larose et al., 2019). Of note, none of these studies both spanned four years of college and included a pre-college assessment.
These studies generally confirm heterogeneity in growth patterns during the college years. In particular, their data tend to represent two or three stable subgroups, or classes, characterized by fairly stable levels of adjustment at different levels (e.g., high, moderate, and/or low stable adjustment), as well as one or two smaller classes characterized by more variable adjustment over time (e.g., moderate decliners, Duchesne et al., 2007;distressed-recovered;Galatzer-Levy & Bonanno, 2013;low-increasing adjustment, Nightingale et al., 2013;high and increasing depression;Salmela-Aro et al., 2008). These latter cases are particularly important to identify, as they represent subgroups of students who do not follow the average patterns of moderate stable adjustment that mean change trajectories would suggest. When analyses only look at average patterns of change across students, such subpopulations are missed and those students who may be especially likely to struggle are not captured. Further, although many of the above studies explored the effect of gender on class membership or distribution, none separately identified trajectories by gender despite gender non-invariance in trajectories of adjustment documented in prior research (Conley et al., 2014(Conley et al., , 2020Gall et al., 2000).

Predictors of college adjustment
The college years witness changes in numerous aspects of psychological, cognitive-affective, and social adjustment (Conley et al., 2020), each of which might play a role in predicting well-being over time. College students report increasing levels of distress and perceived stress (CCMH, 2021;HMS, 2021), both of which can place students at risk for worse functioning during their collegiate years. Positive indicators of psychological functioning, including self-esteem, self-efficacy, and resilience, also are important to consider since they can buffer against negative, and enhance positive, outcomes (Toews & Yazedjian, 2007). For example, students who reported higher levels of resilience subsequently felt more adjusted to college, had higher grade point averages [GPAs], and were more likely to persist in college after academic setbacks (Caporale-Berkowitz et al., 2022;Haktanir et al., 2021;Hartley, 2011). A comprehensive view of adjustment also requires consideration of individual resources, such as cognitive-affective strategies (Billings & Moos, 1982). Research indicates that emotion regulation and coping strategies such as cognitive reappraisal, problem-focused coping, and active emotional coping might support college adjustment (Crockett et al., 2007;Leong et al., 1997), whereas expressive suppression and avoidant emotional coping are linked to worse outcomes (Cousins et al., 2017;Noble et al., 2014). Importantly, coping styles have been linked to college retention for racially and ethnically minoritized students, with active emotional coping associated with academic persistence whereas avoidant emotional coping is associated with drop-out (LeSure-Lester, 2003). Coping strategies also have important implications for academic success, as students who engage in more problem-focused coping strategies had greater levels of academic achievement (e.g., GPA and completed credits; Alarcon & Edwards, 2013;Gustems-Carnicer et al., 2019). Finally, given that college attendance brings considerable shifts in students' social networks, social well-being is an important consideration as well (Lane, 2015). While perceived support from friends is associated with better college adjustment (Brissette et al., 2002;Salmela-Aro et al., 2008), relationships with family remain important as well (Wintre & Yaffe, 2000). Thus, these aspects of psychological functioning, cognitive affective strategies, and social well-being may be important early predictors of student functioning across the college years.
Some researchers have examined select aspects of psychological functioning (e.g., anxiety, Larose et al., 2019), cognitive-affective strategies (e.g., emotion management and emotional self-efficacy, Nightingale et al., 2013), and social well-being (e.g., attachment style, Galatzer-Levy & Bonanno, 2013) as predictors of adjustment trajectories. However, more comprehensive assessments are needed to determine what student characteristics predict different patterns of change over time, which in turn can help identify the most vulnerable students, and the best intervention targets, for improving student well-being.

The current study
Synthesizing findings from prior research, it is apparent that students do not all follow the same path of adjustment in college, and that there are important differences in trajectories by gender that indicate further evaluation of meaningful subgroups of students should be conducted separately. In addition, research indicates that it is possible to predict which students may experience greater difficulty over time. Given that students' well-being is so closely connected to their success (Howard et al., 2010;Leppink et al., 2016;Wilks et al., 2020), more research is needed to evaluate subpopulations based on change in adjustment from pre-college baseline over four years of college, separately by gender, and to examine various aspects of psychological functioning, cognitive-affective strategies, and social well-being as early predictors of subgroup membership.
The current study builds on previous research that evaluated the rate and shape of change in psychological well-being (specifically, self-esteem) and distress (specifically, depression, anxiety, and stress) across four years of college, separately for women and men (Conley et al., 2020). Whereas the prior study investigated average developmental patterns of adjustment by gender, the current study extended such research by (a) exploring the possible heterogeneity in those trajectories of adjustment (specifically, psychological distress and well-being) by gender, and (b) examining baseline predictors of trajectory subgroup membership. Building on other research, this study included a pre-college baseline and additional assessments spanning four years of college, as well as a broader array of baseline predictors.
Based on the above-reviewed literature, we expect that there will be at least two distinct subgroups, or classes, for distress and well-being represented by the data -a generally higher-functioning and a generally lower-functioning class for both college women and men -with greater distress and lower wellbeing for women compared to men across subgroups. We also speculate that in some cases there might also be a moderate-functioning class or a class that exhibits change over time, such as an improving or declining adjustment. Further we expect adjustment trajectory subgroups to be predicted by various aspects of psychological functioning (i.e., resilience, perceived stress, selfefficacy, distress, and self-esteem), cognitive affective strategies (i.e., cognitive reappraisal, expressive suppression, avoidant emotional coping, problemfocused coping, and active emotional coping), and social well-being (i.e., friend support, familial support, general social support, and number of close friends at the university).
For each cohort, all entering first-year students were invited at T0, and those who responded at T0 and were still enrolled as students were eligible and invited to participate in each respective follow-up point (T1-T4). Participation varied across timepoints: N = 5551 of the 8528 (65.1%) invited at T0; N = 2407 of the 5324 (45.2%) invited at T1; N = 1394 of the 4719 (29.5%) invited at T2; N = 1081 of the 4462 (24.2%) invited at T3; N = 1071 of the 3998 (26.8%) invited at T4. Participants who reported their gender and completed at least one timepoint were included in order to capture a more comprehensive and inclusive picture of student functioning. Importantly, neither of the study outcomes (psychological well-being and distress) nor the additional twelve predictor variables (listed below) were related to missing data, ts < 1.86, ps > .05. However, men compared to women had a greater proportion of missing data, χ2(1) = 30.82, p < .001.

Demographic information
Participants reported their gender, sexual orientation, and estimated parental income. With participant's permission, age, race/ethnicity, commuter status, and first-generation college student status (based on parental education) were gathered from university records.

Class membership variables
Psychological well-being. The 10-item Rosenberg Self-Esteem Scale (Rosenberg, 1965; RSE) is a measure of self-esteem that was used to represent psychological well-being. Each item (e.g., "On the whole, I am satisfied with myself;" "I feel that I am a person of worth, at least on an equal plane with others") is answered on a Likert-type scale from 0 (Strongly disagree) to 3 (Strongly agree), yielding total scores between 0 and 30 (α = .89 at T0).

Psychological distress.
The 21-item total scale of the Depression Anxiety Stress Scale-21 (Lovibond & Lovibond, 1995) was used to represent psychological distress (α = .91 at T0). Participants rated items that capture experiences of depressed mood, anxiety, and stress (e.g.,"I couldn't seem to experience any positive feeling at all," "I felt I was close to panic," and "I found it difficult to relax") on a Likert-type scale from 0 (Did not apply to me at all) to 3 (Applied to me very much, or most of the time). Scores ranged from 0 to 63, with scores of 0-15 reflecting a normal range of general psychological distress, 16-18 mild, 19-22 moderate, 23-25 severe, and 26 and above extremely severe distress (Yusoff, 2013).

Predictors of class membership
Variables related to psychological functioning (resilience, perceived stress, self-efficacy, distress, and self-esteem), cognitive affective strategies (cognitive reappraisal, expressive suppression, avoidant emotional coping, problemfocused coping, and active emotional coping), and social well-being (friend support, familial support, general social support, and number of close friends at the university) at baseline (T0) were assessed as predictors of class membership. Information on these measures appears in Table 1.

Trajectories of change in well-being and distress
Building on a prior study that identified non-linear best-fitting latent growth curve models (LGCMs) for the average pattern of change and significant variance in intercept and slope factors (Conley et al., 2020) separately for men and women, the present research uses growth mixture modeling (GMM) in MPlus Version 8.1 (Muthén & Muthén, 1998 to explore heterogeneity within these previously identified models (Psychological well-being [i.e., self-esteem] LGCM: quadratic for both men and women based on fit indices and parsimony; Psychological distress LGCM: quadratic for men, cubic for women based on fit indices and parsimony; see Conley et al., 2020 for more information on the original LGCMs). The prior study assessed gender differences using invariance testing and found that LGCMs were non-invariant across gender. Given the significant gender differences in the original models, the current study maintained this separation for further analysis, evaluating models separately for women and men.

0-30
Psychological Distress a Depression Anxiety Stress Scale (Lovibond & Lovibond, 1995), Cronbach alphas are assessed at T0/baseline when predictors were measured. All scales are scored such that higher scores indicate greater levels of the constructs measured. a Self-esteem, used in this study as proxy of psychological well-being, was only used to predict trajectories of psychological distress; psychological distress was only used to predict trajectories of psychological well-being.
The size and uniqueness of each group also was considered for selecting the best-fitting models. Slope parameters estimates (e.g., linear, quadratic, cubic slopes) were inspected for each class to assess the rate and shape of change in psychological well-being and distress over the course of college.

Predictors of class membership
Once the best-fitting model was identified for psychological well-being and distress in women and men, predictors of latent class membership were examined using the three-step method for auxiliary predictor variables (Asparouhov & Muthén, 2014). To identify predictors, T0 (i.e., at the cusp of college entry) variables related to (a) psychological functioning, (b) cognitive affective strategies, or (c) social well-being were entered simultaneously as covariates into three separate GMM analyses for each domain. Psychological functioning variables included resilience, perceived stress, self-efficacy, distress (included in psychological well-being, but not distress models), and selfesteem (included in psychological distress, but not well-being models).
Cognitive-affective strategy variables included cognitive reappraisal, expressive suppression, avoidant-emotional coping, problem-focused coping, and active-emotional coping. Finally, social well-being variables included friend support, familial support, general social support, and number of close friends at the university. The class representing the trajectory with the most positive adjustment was used as the reference group for evaluating predictors of subgroup membership.

Results
Fit statistics for each model, separated by gender, are presented in Table 2 and parameter estimates for the intercept and slope are presented in Table 3. Latent class trajectories of the best-fitting models are depicted in Figures 1, 2. Predictors of latent class membership are presented in Tables 4, 5.

Trajectories of change
Among women (N = 3,679), the four-class solution provided better fit compared to one-through three-class solutions across indices (see Table 2). The five-class solution did not provide significantly better fit based on BLRT, VLMR, or ALMR. Two of the class sizes for the four-class solution were small (less than 5% of the sample), but practically meaningful based on research finding low self-esteem among college-aged women (Conley et al., 2020;Lawrence et al., 2006), and were therefore retained. As shown in Figure 1 (a), 49.2% of women were best characterized by the largest class, moderate self-esteem, in which moderate levels of self-esteem remain stable over time Covariance between the intercept and linear slope factors was fixed to one for men. c Variance of the intercept factor was fixed to zero for men. d Variance of the quadratic factor was fixed to zero and covariance between the linear and quadratic slope factors was fixed to one for men.
(see Table 3). Meanwhile, 43.4% of women were best characterized by a high self-esteem class, with high levels of self-esteem at each timepoint (see Figure 1(a)) even amid significant curvilinear decline over time (see Table 3). A smaller subgroup (3.8%) of women were best characterized by a low self-esteem class, with the lowest initial and final levels of self-esteem despite small improvements over time. Finally, 3.5% of women were best characterized by a high-declining self-esteem class that started with high initial levels of self-esteem that worsened substantially over time and rebounded slightly by the end of college, with final levels that remained low compared to baseline. Among men (N = 1,848), the two-class solution provided better statistical fit compared to the one-class solution across fit indices. The three-class solution did not fit better than the two-class solution based on BLRT, VLMR, or ALMR (see Table 2). Thus, the two-class solution was chosen as the best-fitting model despite its comparatively lower entropy value ( Table 2). As depicted in Figure 1(b), the majority (62.4%) of men were best characterized by the high self-esteem class, which declined slightly then plateaued over time (see Table 3). The remaining men (37.6%) fit best in a moderate self-esteem class, which had moderate levels of self-esteem overall, even amid slight improvements over time.

Predictors of class membership
As detailed in Table 4, all psychological functioning variables at baseline predicted women's self-esteem trajectory membership over time. Relative to the high self-esteem class, women in the moderate self-esteem and low self- esteem classes had higher levels of distress and perceived stress, as well as lower levels of resilience and self-efficacy, as they were about to start college. None of the predictor variables differentiated between the high self-esteem and high-declining classes. Results indicated that each cognitive-affective variable predicted self-esteem trajectory membership. As compared to the high self-esteem class, women in the moderate self-esteem class reported lower levels of cognitive reappraisal, problem-focused coping, and active emotional coping, in addition to higher levels of expressive suppression and avoidant coping. These same differences were found between women in the low self-esteem class, as compared to the high self-esteem class. None of the predictors differentiated between women in the high self-esteem and high-declining groups. Friend and familial support, as well as general social support, significantly predicted trajectory membership for women; however, number of close friends at the university did not. Women with lower levels of familial and general support were more likely to be in the moderate than the high selfesteem class. Interestingly, in comparisons between the high self-esteem and low self-esteem classes, women with high self-esteem reported more support from people in general, whereas women with low self-esteem class reported more support from friends. None of the social adjustment variables differentially predicted women in the high self-esteem versus the high-declining class. Among men, results indicated that all examined psychological functioning variables predicted self-esteem trajectory membership (see Table 5). As compared to the high self-esteem class, men in the moderate self-esteem class reported lower levels of resilience and self-efficacy, as well as higher levels of perceived stress and distress. Similarly, all examined cognitive-affective strategies significantly predicted self-esteem trajectories. Men who demonstrated lower levels of cognitive reappraisal, problem-focused coping, and active emotional coping, and those who displayed higher levels of expressive suppression and avoidant coping, were more likely to be in the moderate, than the high, self-esteem class. Finally, among social well-being predictors, only familial and general social support differentiated between the two classes of self-esteem for men; friend support and number of close friends at their university did not. Men who reported lower levels of both familial and general social support were more likely to be in the moderate, than the high, self-esteem class.

Trajectories of change
Among women (N = 3,686), the four-class solution provided the best statistical fit across indices ( Table 2). One of the classes in the four-class solution was very small (less than 5% of the sample), but the distinct pattern of this class was deemed to be substantively meaningful and consistent with prior research on the rates of elevated psychological distress in national samples (ACHA, 2021;HMS, 2021). As seen in Figure 2(a), the largest class, low-distress, included 77.5% of the sample and evidenced the consistently lowest levels of psychological distress with a curvilinear worsening in distress that then improved slightly over time. An additional 13.4% of women were represented by the moderate-distress class, which had stable, moderate levels of distress over the four years. Next, the increasing-unstable distress class included 6.5% of women, evidencing an s-shaped curve with distress rapidly worsening into the extremely severe range, recovering slightly into the moderate range, and then worsening again to extremely severe. Finally, 2.6% of women were best characterized by the high-declining distress class, showing a curvilinear improvement in distress (from extremely severe to moderate), followed by slight worsening in distress toward severe. As indicated in Table 2, for men (N = 1,847) the three-class solution provided the best statistical fit and clinical relevance. VLMR and ALMR indicated that adding a fourth class did not improve model fit over the three-class solution. In addition, models with more than three classes had extremely small sample sizes (e.g., <1% when a fourth class was added). Thus, the three-class model was chosen as the best-fitting model for men's psychological distress (see Figure 2(b)). The largest class, low distress, best characterized 77.8% of the sample and showed curvilinear worsening with a plateau in distress over time while still remaining in the normal range (see Table 3). The next largest class, moderate distress, best represented 17.5% of the men who had a stable pattern of moderate levels of distress over time. Finally, the high-declining distress class best characterized 4.7% of the men who had the highest initial levels of distress, which dramatically improved followed by slight worsening in the last year. This pattern is broadly similar to the women's high-declining class, but the women's high-declining class displayed generally higher levels of distress.

Predictors of class membership
All of the psychological functioning variables significantly predicted women's class membership (Table 4). As compared to the low-distress group, women in the high-declining group had lower levels of self-esteem and higher levels of perceived stress. Further, women who had lower levels of self-esteem and selfefficacy, and those with higher levels of perceived stress, were more likely to be in the moderate-distress, than the low-distress, class. Counterintuitively, women in the increasing-unstable distress class, compared to the lowdistress class, had higher levels of resilience at the start of college.
Cognitive-affective strategies of cognitive reappraisal, expressive suppression, and avoidant coping significantly predicted distress trajectory membership, whereas problem-focused and active emotional coping did not. As compared to the low-distress class, women in the high-declining distress class reported lower levels of cognitive reappraisal and higher levels of avoidant coping. Women who exhibited lower levels of cognitive reappraisal and those with higher levels of expressive suppression and avoidant emotional coping were more likely to be classified in the moderate-distress, than the lowdistress, class. Finally, those in the increasing-unstable distress class reported higher levels of avoidance relative to the low-distress class.
For social well-being, familial support, general social support, and number of close friends at university predicted distress trajectory membership, whereas support from friends did not. Women with less social support from people in general were more likely to be in the high-declining than the low-distress class. Meanwhile, women in the moderate-distress class had lower levels of familial and general support compared to women in the low-distress class. Finally, relative to the low-distress class, women in the increasing-unstable class reported having more close friends at their university upon starting college.
For men, findings illustrated that all of the psychological functioning variables significantly predicted class membership (Table 5). Relative to the lowdistress class, men in the high-declining class reported more perceived stress and lower levels of self-esteem and self-efficacy. Further, men in the moderatedistress class experienced higher levels of perceived stress and lower levels of self-esteem as compared to men in the low-distress class.
In terms of cognitive-affective strategies, cognitive reappraisal, expressive suppression, avoidant coping, and problem-focused coping predicted trajectories of men's distress, whereas active emotional coping did not. Men who demonstrated higher levels of expressive suppression and avoidant coping, in addition to lower levels of cognitive reappraisal, were more likely to be in the high-declining than the low-distress class. As compared to the low-distress group, men in the moderate-distress class had higher levels of expressive suppression and avoidant coping, as well as lower levels of problem-focused coping.
Friend, familial, and general social support differentiated between the three trajectories of psychological distress for men, but the number of friends at their university did not. Men in the high-declining distress group, compared to those in the low-distress group, reported higher baseline levels of friend support, but lower levels of familial and general support. Further, lower levels of general social support predicted membership in the moderate-distress, as compared to the low-distress, class.

Differential trajectory classes of well-being and distress
The present findings extend prior research on general trends in college student adjustment to build a deeper, more nuanced picture of ways in which students' psychological well-being change in distinct ways over four years. Whereas students as a whole experience moderate levels of psychological distress and well-being that worsen into the first two years and then improve slightly later in college (Conley et al., 2020), the present findings illuminate multiple varied pathways that different subgroups of college women and men might experience. Further, this study also reveals important pre-college psychosocial predictors of those distinct pathways.
Extending beyond prior research, the current study reveals meaningful differences in types and distributions of trajectories by gender. Women appeared to experience greater heterogeneity in trajectories of well-being and distress compared to men, with some women best characterized by stable low levels of self-esteem, some experiencing declining self-esteem over time, and some best characterized by variability in distress. Although this research is consistent with other research on mean gender differences in college students' psychological adjustment (Eisenberg et al., 2009;Garett et al., 2017;Lawrence et al., 2006), it is novel in revealing that, when modeling trajectories for women and men separately, it appears that they might experience different patterns of heterogeneity in well-being and distress. Specifically, models for women, but not men, displayed some pathways of low or declining well-being, and volatile shifts in distress, over the course of college. These findings have important implications for directing prevention and early intervention efforts at subgroups of college women, in particular.
It is encouraging that the largest classes for well-being and distress generally represent positive to moderate adjustment, and that the proportions of students indicating more distress and less well-being are fairly small (each less than 7% of the relevant sample). These findings are generally consistent with prior research on trajectories of college students' adjustment over time (Duchesne et al., 2007;Galatzer-Levy et al., 2012;Hirai et al., 2015;Larose et al., 2019). Notably, modeling of heterogeneous trajectories by gender also revealed that subsets of female and male students experience difficult pathways of well-being and distress across the college years, and such patterns would not otherwise be captured by mean-level longitudinal analysis.
In terms of psychological well-being (specifically self-esteem), college men are most likely to fall into two basic trajectory classes: approximately two thirds fit best in a high self-esteem group, and a third in a moderate self-esteem group. Although both trajectories included some movement to the middle across the four years, they maintained high and moderate self-esteem, respectively. While it is important to note that a lower entropy value (as was the case with the two-class solution) can suggest poor distinction between classes, meaningful differences in self-esteem levels were illustrated. The picture of psychological well-being appears to be more complicated for college women. In contrast to men, slightly more women were best characterized by moderate self-esteem than the high selfesteem trajectories. Further, estimated means of self-esteem in the moderate class for women were slightly higher than those in the analogous class for men. In addition, there were two small classes, with poor self-esteem overall, that each provided the best fit for less than 4% of women in the current sample. Women's estimated means in the low self-esteem class were initially low with small improvements over time. Women's estimated means in the high-declining selfesteem group began equivalent to those in the high self-esteem group at the cusp of college, but dropped to levels similar to the low self-esteem class in the latter college years. Although these specific trajectories might not reflect all possible pathways that students might follow, they do indicate that students are likely to vary in terms of how their well-being changes across the college years.
For psychological distress (captured by depression, anxiety, and stress), women and men appeared to experience similar types of change, with an additional small class that appeared for women only. Figure 2(a,b) displays a range of distress across trajectory classes and over time, encompassing clinical categories of normal through extremely severe distress (Yusoff, 2013). Reassuringly, over three-fourths of women and men fell into a low distress class, exhibiting low levels of distress over the four years of college. Although these levels increased over the first two years, overall distress in these classes remained low. Approximately 15% of women and men fell into the next largest class, moderate distress, displaying mild (men) to moderate (women) distress that generally remained stable over the four years of college.
Less than 5% of women and men in this study were best characterized by a high-declining distress group, initially displaying extremely severe distress that declined over time. Notably, distress levels in the women's high-declining distress group were generally higher than those in the men's high-declining distress group. Whereas men's distress dropped into the mild range in year three and then increased to moderate, women's distress only decreased to the moderate range and then increased back into the extremely severe range by the fourth year of college. Women and men in their respective high-declining groups likely included those who were prone to distress, and perhaps were particularly impacted by the impending transition to college. Finally, a smaller increasing-unstable distress group of women displayed initially low levels of distress that increased quite severely later in college. This is very striking, and represents a group of students who are crucial to identify as early as possible to effectively target mental health prevention and promotion efforts.
In sum, these analyses reveal that there are small subgroups of female and male students who experience persistent or fluctuating distress over time, even in the context of existing supports such as effective and available college counseling (CCMH, 2021). A critical question for higher education practitioners is how to reach and help these at-risk students, such as through early identification and intervention efforts. The second set of findings, on predictors of trajectory classes of well-being and distress, points toward potential solutions.

Predictors of trajectory classes of well-being and distress
In examining potential predictive (risk and protective) factors of well-being (specifically, self-esteem) over the course of college, some notable patterns emerged. First, all baseline psychological and cognitive-affective predictors differentiated the high from the moderate and low well-being trajectories (the latter, notably, emerged in women only). However, among the social predictors measured at the start of college, what most differentiated those who maintained high versus moderate well-being was social support from family and others in general. In contrast, no psychosocial variables differentiated between women's high well-being versus high-declining well-being trajectories. At the cusp of starting college, there might be distinct subgroups of women who appear to have similar psychosocial resources, but subsequent stressors during college may instigate or exacerbate declines in well-being for a small minority.
Compared to psychological well-being, psychological distress showed more variability in predictors (i.e., risk and protective factors). Across women and men in this sample, avoidant emotional coping, self-esteem, stress, and social support from people in general emerged as the strongest predictors. Cognitive reappraisal, expressive suppression, and familial social support were fairly consistent in differentiating trajectories as well. Other factors, such as selfefficacy, problem-focused coping, resilience, and friendship less consistently predicted distress patterns. Notably, three predictors critically differentiated the low distress from the increasing-unstable distress trajectories in women, which looked similar in terms of baseline distress but diverged over time (see Figure 2(a)). Membership in the latter trajectory, in which initially low levels of distress increased over time, was predicted by greater use of avoidant emotional coping, higher initial resilience, and more close friends at university upon arriving. Prior literature establishes avoidant coping as a risk factor for increased psychological distress (Dyson & Renk, 2006;Noble et al., 2014), but the other two findings were unexpected. Perhaps a subset of women experience an initially over-inflated view of their abilities to "adapt," "cope," and "deal with whatever comes my way" when faced with challenges and difficulties (Campbell-Sills & Stein, 2007), and thus suffer an increase in distress as they encounter the challenges of college. As for the counterintuitive friendship finding, there is some evidence that students' concern for the loss of precollegiate friendships is linked to lower psychosocial adjustment in college (Paul & Brier, 2001). Thus, perhaps women in the increasing-unstable distress class reported numerous friends at the university upon arrival but held back from forming new friendships, contributing to increased distress over time. Alternatively, perhaps students who choose a college where they do not have many friends beforehand have certain personal and social traits that help them thrive emotionally over time.
Additionally, there were two cases in which higher levels of social support from friends, but, importantly, lower levels of social support from family and/ or from others in general, predicted worse adjustment (i.e., trajectories of low self-esteem versus high self-esteem in women, and trajectories of highdeclining distress versus low distress in men). This finding might illustrate that students who rely disproportionately on friends when starting college, with a relative lack of support from family and people in general, may be vulnerable to lower psychological adjustment over time. It also is important to consider that in some cases, separation from family relationships that are negative or non-supportive might be a healthy step for some students. Further, given that women tend to seek support from personal connections (Rickwood et al., 2005;Zimmermann & Iwanski, 2014), women who experience lower well-being may enter college with greater levels of friend support because they are harnessing friend support to help cope with their low psychological health. Of note, each of these psychosocial risk and protective factors at the start of college were investigated as predictors of trajectory subgroup membership. Future research involving higher education practitioners and programs can evaluate prevention programs that target these modifiable factors to potentially change an individual student's adjustment trajectory, though causal links between baseline predictors and adjustment trajectory subgroups are yet to be evaluated.

Contributions and limitations
Although the present study made many significant contributions, there are several limitations that future research should address. First, although the present sample was much larger than prior investigations of this type, the representativeness of this sample is similarly limited by a single university sample, attrition, missing data, and use of only self-report data, all of which are common in this type of longitudinal research. Although using FIML to include the maximum number of students mitigated the limitations of missing data (Curran et al., 2010;Enders, 2010;Little et al., 2014), findings should be interpreted with caution as these limitations still may have affected the number of classes identified, the types of trajectories of change in well-being and distress, and psychosocial adjustment, cognitive-affective strategies, and social support as prospective predictors of trajectory membership. A summary of this paper's analyses performed on a narrower sample of 3,135 students (2,172 women, 963 men) who had data available at least two timepoints, is available as an online supplement. Trajectories for well-being for men and women were consistent with the original results, except for the best-fitting model for women combining high and moderate trajectories. For distress, the bestfitting models showed different patterns than the original findings though fit statistics were not consistent and demonstrated poorer fit to the data in the restricted sample as compared to the original more inclusive sample. Future research should continue to evaluate change over time in distress, and include broader samples across multiple institutions, including historically Black colleges and universities (HBCUs), two-year and four-year institutions, and campuses of various sizes and geographic locations, and consider ways in which institutional and student characteristics (e.g., first-generation college status, work and commuter status, age while in college) might impact findings.
Second, this study utilized a rare longitudinal design that spanned precollege baseline into the end of each of four years, but it is important to consider that additional timepoints within each academic year or semester might reveal a different, more nuanced pattern of adjustment (e.g., Galatzer-Levy et al., 2012;Garett et al., 2017), and that assessment points beyond four years might be warranted for some students who do not graduate in that timespan (Shapiro et al., 2016). Although scholars have cautioned against over-reifying trajectory classes (Sher et al., 2011), their consideration as depicting heterogeneity with potential common trajectories that students might follow, along with predictive models of risk and protective factors, can certainly add benefit to how we understand and serve college students. Although the inclusion of the pre-college baseline assessment is a strength of the current study, the timing of that assessment (i.e., two weeks before classes started) may have affected students' reports of their well-being. The immediate transition to college is a unique time wherein some students may have improvements in their psychological health as they leave home-and schoolbased stressors and experience college life as exciting, whereas other students may experience the transition as stressful, isolating, and anxiety-provoking. Thus, students' reports of their well-being at the cusp of college may look different than even a few weeks earlier or later.
Third, this study is the only of its kind to examine heterogeneous trajectories including separate trajectories by gender. This approach, building on prior research, allowed for identification of additional heterogeneity in patterns of longitudinal adjustment for women in both domains of psychological well-being and distress that were not also reflected in the sample of men. This variability is notable, especially in light of concerns that GMM often yields the same trajectory patterns (Sher et al., 2011); the current study demonstrates that this is not the case. Still, future research would benefit from replicating and extending the present study's findings related to gender differences, as well as considering non-binary gender expressions. Similarly, future research might benefit from exploring differences in trajectories and predictors based on other demographic variables such as socioeconomic status, first generation status, and living arrangements (e.g., on campus versus commuters), as well as behaviors such as attending class and engaging with social media (Bowman et al., 2019).
Fourth, although this research investigated several predictors of the trajectory classes spanning the broad domains of psychological functioning, cognitive-affective strategies, and social well-being, there might be other predictors worthy of future research as well. Finally, data for this study were collected before the COVID-19 pandemic hit the United States in the early 2020s. As early research suggests the pandemic influenced student well-being (Healthy Minds Network and American College Health Association, 2020) and might have a lasting impact on student adjustment, replication of this study postpandemic will be an important area for future study.

Implications and applications
Identifying that there is heterogeneity in trajectories of adjustment in collegeand being able to predict which students are most likely to follow what trajectories based on pre-college factors -can help higher education scholars and practitioners identify the subgroups of college women and men who are at risk for challenges and can inform the development of preventive interventions for at-risk students. Further, the present findings indicate numerous risk and protective factors, which can serve as fruitful targets for such interventions. Hundreds of institutions already participate in annual assessments of student well-being (e.g., ACHA, 2021;HMS, 2021;Stolzenberg et al., 2020), and routine assessments like these -combined with knowledge from more comprehensive longitudinal models like those in the present study -could help higher education practitioners and institutions address specific student and institutional goals.
In terms of predicting and promoting positive well-being, nearly all of the predictors differentiated trajectories of high self-esteem from those of moderate and low self-esteem, for both women and men. This suggests that many types of psychosocial interventions might help foster adaptive psychological well-being. For students who exhibit moderate to high levels of psychological distress at the cusp of college, interventions might best focus on stress-management, emotion regulation (specifically, increasing cognitive reappraisal and decreasing expressive suppression), and adaptive (specifically, reducing avoidant) coping strategies, as well as promoting healthy self-esteem and general social support. For women who display initially low levels of distress, it appears to be particularly critical for prevention efforts to focus on reducing avoidant coping.
It is notable that several cognitive-affective strategies, including coping and emotion-regulation styles, predicted adjustment trajectories. Importantly, many of the student attributes that predicted subsequent well-being trajectories are malleable, teachable skills that are targeted in many evidence-based treatments and can also be incorporated into prevention and promotion programs on campuses. For example, many campuses aim to build students' social support and sense of community at the beginning of college through orientation programs (Larmar & Ingamells, 2010;Wolfe & Kay, 2011) and some also do so through learning communities (Inkelas et al., 2007;Zhao & Kuh, 2004). Going further, universities can employ required or elective seminars that are specifically focused on improving facets of psychosocial wellness (Conley, 2015;Conley et al., 2015). In these seminars, higher education practitioners can help students learn ways of developing active coping skills (e.g., problem-solving), stress-reduction strategies, mindfulness, and social and emotional skills, all of which improve adjustment and well-being (Baghurst & Kelley, 2014;Danitz & Orsillo, 2014;Ramler et al., 2016). Given the present study's findings of heterogeneity in trajectories of female and male student adjustment over four years of college, and identification of particularly at-risk subgroups, targeted interventions are key as well. High school and college practitioners and administrators would do well to identify the most atrisk students based on the predictors identified in the present study, and employ evidence-based psychosocial skill-building programs for these students as they make the important and potentially risky transition to college.