An 8-Week Peer Health Coaching Intervention among College Students: A Pilot Randomized Study

This study explored the effects of an 8-week peer coaching program on physical activity (PA), diet, sleep, social isolation, and mental health among college students in the United States. A total of 52 college students were recruited and randomized to the coaching (n = 28) or the control group (n = 24). The coaching group met with a trained peer health coach once a week for 8 weeks focusing on self-selected wellness domains. Coaching techniques included reflective listening, motivational interviews, and goal setting. The control group received a wellness handbook. PA, self-efficacy for eating healthy foods, quality of sleep, social isolation, positive affect and well-being, anxiety, and cognitive function were measured. No interaction effects between time and group were significant for the overall intervention group (all p > 0.05), while the main effects of group difference on moderate PA and total PA were significant (p < 0.05). Goal-specific analysis showed that, compared to the control group, those who had a PA goal significantly increased vigorous PA Metabolic Equivalent of Task (METs) (p < 0.05). The vigorous METs for the PA goal group increased from 1013.33 (SD = 1055.12) to 1578.67 (SD = 1354.09); the control group decreased from 1012.94 (SD = 1322.943) to 682.11 (SD = 754.89); having a stress goal significantly predicted a higher post-coaching positive affect and well-being, controlling the pre-score and other demographic factors: B = 0.37 and p < 0.05. Peer coaching showed a promising effect on improving PA and positive affect and well-being among college students.


Introduction
According to the Centers for Disease Control and Prevention [1], approximately one in two people in the United States have at least one chronic health condition, such as heart disease, cancer, hypertension, diabetes, or obesity. One in four adults have two or more chronic health conditions. It has been estimated that a major portion of chronic conditions could be prevented by behavior-related lifestyle interventions [1].
College is a critical lifetime transition period for young adults to develop and independently practice healthy behaviors [2]. Poor health behavior practices and associated mental health challenges have emerged as critical risks among undergraduate students in the United States [3]. College students are at risk of gaining weight due to a lack of review study, An and colleagues noted that very few RCTs have evaluated the effects of health coaching and called for more RCTs of the health coaching studies [22].
In response to the nationwide call for more community-and evidence-based programs to improve population health, the research team designed this pilot program to gather scientific evidence to evaluate the effectiveness of the peer health coaching program in the college campus setting. Therefore, the current study aimed to assess the effectiveness of a randomized, 8-week peer health coaching program on PA, nutrition, sleep, social isolation, and mental health, among college students.

Study Design
The study used a randomized, 8-week interventional design to evaluate the efficacy of a peer health coaching intervention delivered in a college setting. The study was conducted at a midsize private college (i.e., student population~4000 to 5000 students) located in New England, USA. The baseline and post-intervention assessment were conducted in January 2022 and May 2022, respectively. The Institutional Review Board (IRB) at the college approved all study procedures.

Participants
Student participants. Freshmen and sophomore students 18 years or older were eligible to participate. Participants were recruited through campus flyers, recruitment tables in front of the student center building, and classroom visits to a course that all freshmen are required to take. We also purposely recruited from programs serving historically underrepresented populations, including first-generation students and students from minoritized races/ethnicities, since such students were disproportionately affected by the pandemic and generally experienced greater barriers to participation in health programming. A total of 52 participants were recruited. After completing the pre-assessment, they were randomly assigned into the coaching group (n = 28) or the control group (n = 24).

Health Coaches
Student coaches were health science major undergraduate students who were enrolled in a series of two health coaching courses that prepared them for basic coaching skills and improving skills and self-efficacy through practice. The courses were taught by a group of faculty. Two of those faculty have received WellCoaches ® health coaching certification; other faculty had related expertise (e.g., cultural studies, counseling, mental health). In the first coaching course, students had received the basic training on coaching theories and techniques. They all passed a mock health interview exam by the end of the first coaching course and before they started the second coaching courses, in which they needed to complete 50 coaching sessions. This study was conducted as part of their 50-session training. During this study and throughout the course, student coaches met with the course instructor and other coaches weekly to discuss their coaching progress and challenges. In addition, each coach met with the course instructor individually every other week to receive additional feedback and support.

Procedures
After completing the baseline assessment, participants were randomized to either the intervention (i.e., coaching group) or the control group. Covariate adaptive randomization was adopted to balance the participants between the intervention and control groups on gender, first-year students, and students of color. Participants in the coaching group met with their assigned 1:1 peer health coach once a week for 8 weeks. Coaching meetings were scheduled for 30-40 min. Coaching meetings were in-person but zoom coaching meetings were allowed if students were sick or had safety concerns related to COVID-19. See Appendix A Figure A1 for the flow chart of the intervention.

Intervention
The health coaching program was designed to facilitate and emphasize several aspects. The first coaching session focused on self-goal identification. At the first coaching session, the health coaches assisted the students to identify 2-3 areas within the topics of PA, nutrition, sleep, and social support that they would like to improve on. During each coaching meeting, student coaches evaluated their previous weekly goals and discussed their gains and challenges and areas that they would like to work on in the coming weeks. Second was "Peer support". In each session, coaches facilitated the discussion using coaching techniques, including reflective listening, affirmation, motivational interviewing, etc. Lastly, the health coaching program facilitated and emphasized "Goal setting." By the end of each session, coaches assisted student participants to come up with two SMART goals (i.e., specific, measurable, action-based, realistic, time-limited) that they felt ready to work on. Notably, participants were encouraged to set behavioral goals that they would like to work on. That said, no specific behavior goal was pre-set for participants. For instance, one participant may set a weekly physical activity goal of walking more, whereas another participant may work on doing more moderate-vigorous exercise. In the following week, students then conducted a self-evaluation on the completion rate from 0 percent to 100 percent on each of the wellness goals.
Students in the control group received a wellness handbook that was created by the research team. This handbook provided information related to how to improve physical activity, nutrition, sleep, stress, and social support in the college setting. Students in the control group were asked to use this handbook as an information source if they would like to improve their health behavior practices.

Demographics
Gender, age, race, and whether they were first-generation students or student athletes were collected. In addition, socioeconomic status was measured by the MacArthur Scale of Subjective Social Status [35]. Students indicated their perception of their social status by rating a 10-point Likert-type scale from 1 (lowest standing in the community) to 10 (highest standing in their community). Previous studies indicated good evidence of reliability and validity on scores of the measure [36,37].

Physical Activity (PA)
The International Physical Activity Questionnaire (IPAQ)-short form was used as the subjective measure of physical activity. A total of 7 items were structured to provide separate METs (min/week) on walking; moderate-intensity activity; vigorous-intensity activity, total physical activity, and time spent on sitting (walking MET-minutes/week = 3.3 × walking minutes × walking days; moderate MET-minutes/week = 4.0 × moderateintensity activity minutes × moderate days; vigorous MET-minutes/week = 8.0 × vigorousintensity activity minutes × vigorous-intensity days). This questionnaire has demonstrated acceptable reliability and validity on the total score in previous studies [38]. Exercise selfefficacy was also assessed using a five-item measure to assess the confidence to perform physical activity in five different situations (i.e., vacation, feeling tired, bad mood, not having enough time, and bad weather [39]. This measure has been shown to predict physical activity in previous studies [40].

Diet
Healthy eating self-efficacy was assessed as a summary of nine five-point Likert-scaled questions about self-confidence for eating healthy foods at the mall, after school, with friends, under stress, feeling down, bored, at a fast-food restaurant, alone, and at family dinner (not confident at all = 1 to very confident = 6; (Cronbach a = 0.83)) [41].

Sleep
The Pittsburgh Sleep Quality Index (PSQI) [42] was used as a subjective measure of sleep quality. It differentiates "poor" from "good" sleep quality by measuring seven areas (components): subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medications, and daytime dysfunction over the last month. A Global PSQI score was calculated by adding scores from the 7 areas together. A total score of 5 or higher indicates a poor sleep quality.

Social Isolation
NIH Item bank V2.0 Social Isolation (Short Form) from PROMIS was used to measure social isolation. Participants responded to a four-item scale. An example question was: "In the last 7 days, how often do you feel (e.g., I felt left out)." Responses ranged from Never (1) to Always (5). The sum score ranged from 4 to 20, with a higher score indicating higher social isolation.

Mental Health
For the purposes of this study, we utilized validated measures developed as part of the NIH-funded Patient-Reported Outcomes Measurement Information System (PROMIS). Positive affect and well-being were measured using the Neuro-QOL Item Bank v1.0-Positive Affect and Well-Being (Short Form). An example question was "Lately, my life was satisfying . . . " Responses range from Never (1) to Always (5). The instrument is 9 items with a score range of 9 to 45, where higher scores indicated a higher overall positive affect and well-being. Anxiety was measured by the Neuro-QOL Item Bank v1.0-Anxiety (Short Form) from the PROMIS. An example question was "In the past 7 days, I felt nervous." Responses ranged from Never (1) to Always (5). The instrument features 8 items with a score range of 8 to 40. A higher score indicates a higher anxiety level. Cognitive function was measured by NIH Neuro-QOL Item Bank v2.0-Cognition Function (Short Form) from the PROMIS. Four questions started with "In the past 7 days . . . (e.g., I had to read something several times to understand it)." Answering options ranged from "Never (5)" to "Very often (several times a day) (1)." Another four questions started with "How much DIFFICULTY do you currently have . . . (e.g., learning new tasks or instructions?)", and answering options range from "None (5)" to "Cannot do (1)." The total score ranged from 8 to 40, with higher scores indicating better cognitive function.

Data Analyses
Intention-to-treat analysis was performed. In addition to descriptive statistics, repeated MANOVA was used to explore the overall impact of the coaching intervention on physical activity, sleep, nutrition, mental health, and social isolation between participants in the coaching group and the control group. In addition, participants in the coaching group were further divided into subgroups based on the wellness goals they identified. Repeated ANOVA and MANOVA analyses were used to explore the coaching effect on the goal-specific domain. For instance, participants who identified physical activity as one of their coaching goals were compared to the control group on the physical activity variables. Regression analysis was performed to examine the effects of the demographic factors on the post-assessment scores, controlling the pre-assessment scores. All data were analyzed using SPSS 21 (IBM, Armonk, New York, NY, USA).

Results
Demographics: Two participants (7%) in the coaching group (n = 28) dropped the study during the intervention. Among the rest of the 26 students who completed the 8-week intervention, 23 (82%) completed the post-assessment. For the remaining 24 participants in the control group, 20 (83%) of them completed the post-assessment. Table 1 also showed that there was no significant difference in the demographic characteristics between the coaching and control groups. Intervention Effect Analysis. Means and standard deviations for all variables are shown in Table 2

Goal-Specific Analysis
Physical Activity. A total of 15 participants in the coaching group (65%) identified PA as their coaching goals and discussed PA-related goals for at least one coaching meeting. The average goal meeting rate was 73%.
Repeated MANOVA results showed that the interaction effects between time and group on Vigorous MET was significant F(1,32) = 6.42, p = 0.017. For those who identified PA as one of their coaching goals, vigorous MET increased from 1013. 33  We also explored whether having PA as a coaching goal, as well as the demographic factors, would predict more PA changes. The regression analyses showed that, controlling for pre-PA level and the demographic factors, coaches who had PA as a coaching goal had a marginal significant higher post-total PA MET (B = 0.35, p = 0.053) and a significant higher post-vigorous PA MET (B = 0.44, p < 0.01) than those who did not have PA as a coaching goal and those who were in the control group. See Tables 3 and 4.   Having diet as a coaching goal, as well as other demographic factors, did not predict the healthy eating efficacy score. See Table 5 for details.
Sleep. A total of 13 participants in the coaching group (57%) identified sleep as one of their coaching goals, with the goal completion rate of 79.96%. The interaction between group and intervention was not significant: F(1,28) = 3.60 and p > 0.05. The group differences were significant: F(1,28) = 5.18 and p < 0.05. Among the 13 participants who identified sleep as one of their coaching goals, the average Global PSQI score decreased from 6.18 (2.35) in the pre-test to 4.36 (2.20) in the post-test; as for the control group, the score changed from 7.74 (2.76) to 7.26 (3.58). Regression analysis showed no significant intervention effect on predicting post-PSQI scores. See Table 5 for the details.
Social Isolation. Seven students (30%) identified improving social-related support as one of their coaching goals, with the self-evaluated goal completion rate of 75%. There were no significant intervention or group effects, all p values > 0.05. The social isolation score of the intervention group changed from 9.33 (5.84) to 10.33 (6.12), while the control group changed from 9.74 (3.79) to 9.15 (3.67) from the pre-and post-assessment. Compared to those in the control group, having the social goal as a coaching goal did not predict the post-social isolation score; however, females had significantly less social isolation than males, controlling all other variables: B = −0.42 and p < 0.01.
Mental Health. Although no student directly identified "mental health" as one of the coaching goals, seven participants in the coaching group identified stress management as one of their coaching goals (30%) and discussed stress-related goals during at least one coaching session. The overall self-evaluated goal completion rate was 83.09%. The interactions between group and intervention on anxiety, positive affect and well-being, and cognitive function were not significant, all p values > 0.05. The main effect of the group was significant on anxiety (F(1,24) = 5.18, p < 0.05). The anxiety score for the goal-specific group changed from 14.14 (6.04) to 14.71 (7.18), while for the control group it changed from 22.05 (7.82) to 20.52 (7.29). All other main effects were not statistically significant. Compared to those who were in the control group, those who had a stress goal had a significantly higher post-positive affect and well-being score, controlling pre-positive affect score and other demographic factors, B = 0.37, p < 0.05. No relationships were identified between a stress goal and post-anxiety and cognitive function.

Discussion
Although the concept of a health coach and peer education model has been well developed, there has been very little empirical research on the effectiveness of peer health coaching programs. Even less is known regarding the effectiveness of health coaching among non-clinical young adult populations in the higher education setting. To the authors' knowledge, this is the first published study in which multiple health behavior outcomes were evaluated via a health coaching RCT among college students.
The most important findings were that the students in the coaching group who worked on PA showed significant intervention effects on vigorous PA and marginally significant improvement on total PA over the 8-week period. This is consistent from previous studies. For instance, a meta-analysis with 27 randomized trials shows a small, significant effect size (SMD = 0.27) in PA improvement achieved by health coaching among people aged 60 years or older [43]. In another meta-analysis study [22] that examined health coaching intervention among adults with cardiovascular disease, the effect size of health coaching intervention is small (effect size < 0.20) for physical activity and diet. The current study provided preliminary evidence of the effectiveness of health coaching on physical activity among healthy young adults.
Although there was no significant intervention effect on walking and moderate PA, the improvement of vigorous PA and marginally significant improvement on total PA among the participants indicated that they engaged in more planned exercise behaviors after the intervention. Interestingly, while vigorous and total PA increased among the participants, there was a decrease in vigorous and total PA among the control group. This may be due to the fact that post-assessment was one week before the final examinations, and the increased stress and study time may have made students in the control group less likely to engage in PA [44]. This also demonstrates that peer health coaching not only buffered the negative impact of increased stress and lack of time, but also provided additional support for students to engage in PA during a challenging time.
We also found that individuals who worked on stress-management-related goals had an improved positive affect and well-being, compared to those in the control group. This is supported by the previous research indicating that stress evokes a negative affect [45] and successful stress management could improve an individual's positive affect [46]. In addition to discussing stress-management-related issues, previous research also suggested that the peer support provided by the health coaches may play a role in improving participants' positive affects [47]. Future path analysis and qualitative studies may provide additional evidence for this relationship.
Another strength of the study was that it provided the participants with the antonym to self-select the wellness topics and behavioral goals, and the health coaches were in the supporting role in the coaching relationship. This is different from other health coaching intervention studies that involved adults with health conditions (e.g., diabetes, cardiovascular diseases) in which the topics and behavioral goals were often pre-determined, and health outcomes (vs. behavioral outcomes) were often assessed [11]. Considering that our participants were healthy young adults, they usually did not have health conditions that required them to achieve specific behavioral outcomes. In this case, providing autonomy and peer support would invoke their internal motivation [14,48]. This approach may make the assessment of implementation fidelity more difficult, but it has gained empirical value by supporting participants with on-going, real-world problems. For example, while final exams were approaching towards the end of the study, more student participants chose to work on stress-management-related topics with their coaches. Although there is no direct evidence from the current study, it is possible that the improved stress management skills may benefit other health behavioral practices for the participants. In addition, more flexibility in coaching topics provides practical value for colleges and universities who may consider adopting peer coaching programs in future.
Although not statistically significant, there was a trend towards the improvement of sleep quality among participants who worked on sleep-related goals, compared to the control group. Type II error due to the small sample size may play a role here. No intervention effect was detected among participants who worked on diet-related goals. This may be because, compared to other behavioral goals, diet-related goals were very diverse, from eating breakfast to drinking more water. This makes the intervention effect very challenging to measure without significant noise. We suggest future studies consider personalized diet measures based on the specific diet goals.
Given that first-year students, students of color, students from low-income families, and first-generation students are often disproportionately affected by the pandemic and have a more difficult time transitioning to the academic and social demands of college [49], we purposely recruited students from those populations with success; over 23% of the students were first-generation and 37% were non-white, as compared to 20% and 16% in the college student body as a whole. There were no significant differences in intervention uptake or outcomes by demographic group; however, additional adaptations could be made to the health coaching approach to improve engagement among specific groups and potentially decrease health disparities disproportionately experienced by historically marginalized and under-represented student groups. We suggest future studies continue to explore the potential moderating or mediating effect of demographic factors on health coaching engagement and effectiveness.
There were several limitations of this study. First, the study began in January and ended in May 2022, during which seasonality became a challenge for the participants. That is, the end-of-semester stress may cause college students to have increased stress, worse nutrition, less PA, and poor sleep quality [44,49]. Second, although having student participants choose the health topics that they were interested in provided them with autonomy, working on several topics may have lowered the intervention dosage they received for each topic. For example, participants who chose to work on PA, sleep, and stress had less exposure to each of those three topics than those who only worked on PA or sleep throughout the intervention period. This type of study design also imposes challenges on statistical analyses as it is impossible to know the exact topics that participants would work on. Furthermore, when clients chose the coaching topic of their interest, it may have posed selection bias as they were more motivated to work on this topic than participants in the control groups. Future studies may consider screening participants' well-being interest and then place them into different coaching groups. In addition, we did not perform a priori power analysis due to limited research on this topic and the pragmatic constraint of the limited trained coaches we had with whom to conduct the study. However, our power, calculated based on the number of participants enrolled, was 50% to detect a 0.5 effect size. Although this is much lower than the ideal power of 80%, we still generated some significant results, indicating a trend towards effectiveness that could be fully ascertained with a full-effectiveness trial. Finally, the current study did not follow up on the participants' post-intervention. The fidelity of health coaching intervention should be further explored in future studies.
The current study was conducted in a college campus in which intervention was delivered by students majoring in health science who went through training via two health coaching courses. Although they had limited coaching experiences compared to the professional health coaches, they provided unique peer support that other professional coaches would not be able to provide. This study also set an example of an innovative health promotion program in the higher education setting. Additionally, from the education experiential learning perspective, the peer health coaching programs also served as an education and practice opportunity for students with related majors. The experience and skills they developed through this intervention will further prepare them for their future career, especially those who will work in client-facing settings.

Conclusions
The post-COVID-19 era poses new challenges for colleges to think creatively to assist young adults with their overall well-being [50,51]. The current pilot study was set in a real-world setting where college students chose their health coaching topics and received support from their peers. It provided important empirical evidence on the effectiveness of peer health coaching in the college setting. That is, an 8-week peer health coaching intervention in a college setting showed preliminary evidence of improving PA and positive affect and well-being for participants who had worked on those topics with their peer health coaches. With more evidence from future studies that are conducted with larger sample sizes and more rigorous methods, the peer health coaching approach has the promise to be scalable and feasible to promote health and well-being among students at institutions of higher education. In addition, colleges and universities may consider adopting the peer health coaching model as an education and training opportunity for health sciences students or those in related majors, such as human development and psychology.