Roles Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing * E-mail: yzhao11@fordham.edu Affiliation Computer and Information Sciences Department, Fordham University, New York, New York, United States of America
Roles Conceptualization, Formal analysis, Investigation, Methodology, Resources, Supervision, Validation, Writing – original draft, Writing – review & editing Affiliation Graduate School of Education, Fordham University, New York, New York, United States of America ⨯
Roles Data curation, Formal analysis, Investigation, Software, Visualization, Writing – original draft Affiliation Computer and Information Sciences Department, Fordham University, New York, New York, United States of America ⨯
Roles Data curation, Formal analysis, Validation, Writing – original draft, Writing – review & editing Affiliation Graduate School of Education, Fordham University, New York, New York, United States of America ⨯
The COVID-19 pandemic has presented unprecedented challenges for university students, creating uncertainties for their academic careers, social lives, and mental health. Our study utilized a machine learning approach to examine the degree to which students’ college adjustment and coping styles impacted their adjustment to COVID-19 disruptions. More specifically, we developed predictive models to distinguish between well-adjusted and not well-adjusted students in each of five psychological domains: academic adjustment, emotionality adjustment, social support adjustment, general COVID-19 regulations response, and discriminatory impact. The predictive features used for these models are students’ individual characteristics in three psychological domains, i.e., Ways of Coping (WAYS), Adaptation to College (SACQ), and Perceived Stress Scale (PSS), assessed using established commercial and open-access questionnaires. We based our study on a proprietary survey dataset collected from 517 U.S. students during the initial peak of the pandemic. Our models achieved an average of 0.91 AUC score over the five domains. Using the SHAP method, we further identified the most relevant risk factors associated with each classification task. The findings reveal the relationship of students’ general adaptation to college and coping in relation to their adjustment during COVID-19. Our results could help universities identify systemic and individualized strategies to support their students in coping with stress and to facilitate students’ college adjustment in this era of challenges and uncertainties.
Citation: Zhao Y, Ding Y, Chekired H, Wu Y (2022) Student adaptation to college and coping in relation to adjustment during COVID-19: A machine learning approach. PLoS ONE 17(12): e0279711. https://doi.org/10.1371/journal.pone.0279711
Editor: Thiago P. Fernandes, Federal University of Paraiba, BRAZIL
Received: July 30, 2022; Accepted: December 12, 2022; Published: December 30, 2022
Copyright: © 2022 Zhao et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Data cannot be shared publicly because of restrictions governed by IRB #1517 from Fordham University. Data are available from Fordham’s Institutional Data Access / Ethics Committee (contact via irb@fordham.edu) with an executed data usage agreement (DUA) for researchers who meet the criteria for access to confidential data.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
The COVID-19 pandemic created numerous unprecedented challenges for university students in particular. For example, thousands of U.S. universities suspended their in-person classes and activities, with some institutions announcing that all classes would be held online for the remainder of the Spring 2020 semester and through the summer session. A large number of university residence halls were closed in response to the suspension of in-person classes, requiring its students to vacate their dorm rooms with only a few days’ notice [1, 2]. Because of the uncertainties imposed by the pandemic [3], students were concerned about the trajectory of their academic careers. Such concerns stemmed from inefficient online learning; suspended fieldwork, internships, and clinical rotations; financial burdens; uncertain living situations; and, for international students, possible changes in visa status [4, 5]. As a result, many students faced heightened levels of psychological distress and changes in behavioral patterns, which called for increased engagement in coping strategies. In terms of coping styles, it is believed that individuals’ coping styles and adjustment processes play important roles in their responses to stress and overall well-being [6], especially during disruptive circumstances such as the COVID-19 pandemic. We were interested in examining how college students’ adjustment processes and coping styles impacted their perceived stress and responses to the COVID-19 pandemic.
The overarching goal of this study is to investigate how participants’ responses to Ways of Coping (WAYS; [6]), the Student Adaptation to College (SACQ; [7]), and the Perceived Stress Scale (PSS; [8]) predict the level of challenges students encountered during COVID-19. In particular, WAYS is an instrument to explore participants’ cognitive and behavioral patterns in managing stressful events. SACQ is an instrument to evaluate students’ adaptation to the university experience. PSS is an instrument to examine participants’ perceived stress levels. For our study, we administered an extensive survey containing questions measuring WAYS, SACQ, PSS, and five COVID adjustment domains (i.e., academic adjustment, emotionality adjustment, social support adjustment, general COVID-19 regulations response, and discriminatory impact) to U.S. students during the initial peak of the pandemic. Many studies focusing on coping and adjustment among college students have used traditional statistical approaches such as correlation and regression analyses. In recent years, social scientists are increasingly interested in utilizing novel machine learning techniques to accurately predict or detect patterns in real-word phenomena [9]. Machine learning offers a wide range of alternative models that might provide substantial improvement in accuracy [10].
In this study, we develop machine learning models to classify well-adjusted and not well-adjusted students in each of the five COVID-19 study domains based on a student’s characteristics in WAYS, SACQ and PSS. We further analyze the most relevant risk factors associated with each classification task. Our findings could help universities to establish systemic and individualized strategies to support their students in coping with stress and adjustments in this era of challenges and uncertainties.
A person’s cognitive and behavioral efforts to manage stressful events can be described as ways of coping [11, 12]. Lazarus and Folkman [12] proposed a transactional model of stress and coping to explain individuals’ use of conceptualizations and behavioral responses to manage perceived stressors. The model includes three major categories of coping strategies that individuals generally engage in, namely emotion-focused coping, problem-focused coping, and avoidance-focused coping [12]. Emotion-focused coping focuses on strategies to control emotional responses whereas problem-focused coping involves active efforts to alter the stressful event. Avoidance-focused coping includes strategies to escape from the situation to avoid the stress [6, 13, 14].
To better explore different coping styles and their psychological aftermath, Folkman and Lazarus [6] developed the Ways of Coping Questionnaire (WAYS), which includes a total of eight coping strategies stemming from the three major categories: confrontive coping, distancing, self-controlling, seeking social support, accepting responsibility, escape-avoidance, planful problem solving, and positive reappraisal. Studies have shown that those who engage in proactive problem-focused coping and positive emotion-focused coping generally perceive less stress during a stressful event whereas those who engage in more reactive emotion-focused coping perceive more stress [15].
It is believed that students generally experience a series of changes and adjustments during their transitions from high school to college [16]. This adjustment process is multifaceted and is closely related to students’ overall university experiences [17]. Areas involved in the adjustment process include emotional well-being, different social expectations, and novel academic requirements [16]. Research has shown that university success is best measured through a combination of students’ cognitive capacity and academic achievement, and those who experience difficulty with university adjustment may choose to drop out of school [16, 18, 19].
To assess how well students were adapting to their novel university experiences, Baker and Siryk [18] developed a quantitative measure, later known as the Student Adaptation to College Questionnaire (SACQ; [7]), and identified four factors closely related to university adjustment: academic adjustment, social adjustment, personal-emotional adjustment, and attachment to the institution. Academic adjustment refers to students’ ability to adhere to the diverse educational demands stemming from university expectations. It is closely related to students’ academic self-efficacy and skills, self-appraisal, motivation to learn, general satisfaction with the academic environment, and educational goals [7, 16]. Social adjustment refers to students’ ability to cope with the interpersonal-societal demands of college life. It is closely related to students’ sense of self-confidence in social situations and ability to cope with stressors, and can be used to predict students’ level of persistence in their university experiences. Personal-emotional adjustment refers to students’ psychological and physical well-being during their university adjustment process and examines students’ coping skills, distress level, and emotional reliance on others. It is related to students’ general school performance, ability to cope with stressors, and overall functioning [7]. Institutional attachment refers to students’ level of commitment to their university and assesses the relationship quality between students and their universities [7]. Studies have shown that students who report higher attachment to and satisfaction with their institutions tend to perceive better social connection, acceptance, and academic competence; possess more coping strategies; and experience fewer negative psychological states [20].
Individuals perceive stress when they interpret the presented situational demands as beyond their own capacity to navigate [12]. The level of stress perceived is determined by the individual’s personal conceptualization about the general stressfulness of their life, their ability and confidence to cope with the stress, and their current functioning in a given period of time [8, 12]. In other words, each individual perceives the same stressor, such as the COVID-19 pandemic, differently based on their personal circumstances and beliefs.
This study utilizes a proprietary survey dataset collected from March to June 2020 during the initial peak of the COVID-19 pandemic. Institutional Review Board (IRB) approval was obtained from Fordham University for data collection and sharing protocols. Eligible participants were at least 18 years of age and enrolled as undergraduate or graduate/professional students at colleges or universities in the United States. All participants completed the SACQ, PSS, WAYS, and the COVID-19 Adjustment questionnaire. The following sections describe the details of these four questionnaires whose data are used in the current study.
In addition, Cronbach’s alpha (α) was used as a reliability measure of internal consistency. It is commonly used to measure how closely related a set of items are as a group in an instrument [21]. Formally, Cronbach’s alpha is defined as: where N is the number of items, is the average inter-item covariance among the items and equals the average variance across all items. In practice, α ≥ .70 is considered acceptable for most social science research instruments.
The Student Adaptation to College Questionnaire (SACQ; (7]) was administered to evaluate the students’ process of adaptation to the university experience during the first wave of COVID-19. The SACQ is a 67-item self-reported questionnaire with a 9-point Likert scale, ranging from doesn’t apply to me at all to applies very closely to me, and consists of four subscales: Academic Adjustment (α = 0.88), Social Adjustment (α = 0.91), Personal-emotional Adjustment (α = 0.87), and Institutional Attachment (α = 0.90). Participants’ responses were summed and then were converted into T-scores, with higher scores indicating higher levels of adjustment [7, 18].
The Perceived Stress Scale (PSS; [22]) was administered to examine participants’ perceived stress levels during the initial peak of the COVID-19 pandemic. The PSS is a 10-item self-report questionnaire with a 5-point Likert-scale ranging from never to very often ([8]; α = 0.87). According to Cohen et al. [8], higher summed scores indicate higher levels of stress and lower scores indicate lower levels of stress. The questions on the PSS are context free (i.e., questions were not worded to fit specific circumstances and they were generic questions), enabling its usage with any subpopulation group. By focusing on the participants’ current thoughts and feelings, the PSS is intended to explore the participants’ perceptions of the degree of unpredictability, uncontrollability, and overwhelmingness of their life experiences during the past 30 days. In this study, two scores were derived upon analysis, with the total score including all 10 items and the short score including four selected items. The total score was summed with appropriate items scored as indicated by Cohen [22]. The short score was derived from the four items that showed the highest correlation with the full-scale items examined by Cohen et al. [8], and it generates general inquiries about the respondents’ experiences of relative current levels of stress.
The Ways of Coping Questionnaire (WAYS; [6]) was used to explore participants’ coping strategies and processes during the initial peak of the COVID-19 pandemic. WAYS is a 66-item self-reported questionnaire with a 4-point Likert scale ranging from does not apply or not used to used a great deal ([23]; α = 0.78). Of note, the WAYS was developed with the specific focus on participants’ actual and/or potential actions in response to a stressful situation rather than their thoughts and feelings about the situation [23–26]. Eight subscales were generated based on factor analyses: Confrontive, Distancing, Seeking Social Support, Accepting Responsibility, Positive Reappraisal, Planful Problem Solving, Escape Avoidance, and Self-Controlling [11]. In the current study, participants’ responses were accumulated to obtain a score for each subscale, with higher scores suggesting their inclination to use the coping behaviors defined by that subscale when encountering COVID-19-related stressors [6].
A self-report questionnaire with a 5-point Likert-scale ranging from strongly disagree to strongly agree was created for the original larger-scale study to measure the effects of COVID-19 on participants and their adjustments [27]. This questionnaire was adapted from an unpublished instrument created to measure university students’ experiences and mental health during the initial COVID-19 outbreak in China [27]. Factor analyses on the original questionnaire yielded five subdomains: Academic Adjustment, Emotionality Adjustment, Social Support, General COVID-19 Regulations Response, and Discriminatory Impact Related to COVID-19. The academic adjustment subscale (7-item, α = .85) measured the degree to which participants’ felt prepared and motivated to complete academic work and ability to adjust to remote education as a result of the COVID-19 pandemic (e.g., “I have a virtual-learning supportive atmosphere at home (e.g., computer, wifi, quiet space)). The emotionality subscale (4-items, α = .71) measured participants’ ability to deal with emotional thoughts and behaviors towards COVID-19 related stimuli and experiences (e.g., “I feel like the Coronavirus is far from me.”). The social support subscale (4-items, α = .69) measured participants’ level of satisfaction with received support during the COVID-19 pandemic (e.g. “I feel supported by my professors and university”). The general regulation reaction subscale (3-items, α = .61) measured participants’ agreement with regulations and restrictions imposed due to the COVID-19 pandemic (e.g., “I feel relieved that schools are closed and classes have moved online.” The discriminatory impact adjustment subscale (3-items, α = .78) measured participants’ acknowledgement and impact of racism as related to COVID-19 (e.g., “I am aware of Asians’ experience with discrimination due to the coronavirus.”) Participants were believed to be adjusting more positively during the COVID-19 pandemic if they reported a high score on these subdomains.
To build our machine learning models, we first labeled each participant as well-adjusted (class 1) or not well-adjusted (class 0) in each of the five COVID-19 adjustment domains. They serve as the output of our predictive models. To accomplish this, we identified the set of questions Q in the survey pertinent to each domain and computed the average score of answers to these questions. A participant was labeled as a class 1 instance if their total score for Q was above the average. Otherwise, the participant was labeled as a class 0 instance. For example, the academic adjustment domain consisted of seven questions, each with a Likert scale from 1 to 5. Thus, the average score for this domain was 7 x 3 = 21, where 7 was the number of questions and 3 was the middle score of each question. Class 1 instances for this domain were those participants whose total score for the seven questions was above 21. Table 1 presents the distribution of participants for each of our classification tasks.