|Year : 2022 | Volume
| Issue : 2 | Page : 41-45
Predictive value of veterinary student application data for assessing adjustment to year 1 of veterinary school
Samuel Karpen1, Robert M Gogal2, Steven D Holladay2
1 Department of Faculty Experience--Data Services, Western Governors University, Salt Lake City, UT, USA
2 Department of Biomedical Sciences, College of Veterinary Medicine, University of Georgia, Athens, GA, USA
|Date of Submission||02-Mar-2022|
|Date of Acceptance||16-Jun-2022|
|Date of Web Publication||09-Sep-2022|
Dr. Samuel Karpen
Department of Faculty Experience–Data Services, Western Governors University, Salt Lake City, UT
Source of Support: None, Conflict of Interest: None
Background: The veterinary education literature warns of psychological distress among doctor of veterinary medicine (DVM) students. Despite its impact on veterinary education, there is little research on predictors of DVM student wellbeing, as most researchers have focused on predictors of academic performance. Methods: We used best subset regression to examine the relationship between application variables and student’s self-reported first year well-being. Objective: To determine whether variables available in students’ applications to veterinary school predicted self-reported well-being in their first year. Results: Age, graduate record exam (GRE) score, pre-DVM grade point average (GPA), hometown population, and paid employment experience were all significant predictors of various facets for self-reported first year well-being and involvement among DVM students. Conclusions: The predictive value of these variables, while significant, was quite low; consequently, veterinary educators should not base policy decisions on this information alone. We hope that our model serves as a useful tool to predict which applicants may need additional wellness resources during their first year.
Keywords: Applicant file data, depression, regression, stress, student admissions, veterinary college, veterinary education, well-being, wellness
|How to cite this article:|
Karpen S, Gogal RM, Holladay SD. Predictive value of veterinary student application data for assessing adjustment to year 1 of veterinary school. Educ Health Prof 2022;5:41-5
|How to cite this URL:|
Karpen S, Gogal RM, Holladay SD. Predictive value of veterinary student application data for assessing adjustment to year 1 of veterinary school. Educ Health Prof [serial online] 2022 [cited 2023 Jan 28];5:41-5. Available from: https://www.ehpjournal.com/text.asp?2022/5/2/41/355839
| Introduction|| |
The veterinary education literature warns of psychological distress among doctor of veterinary medicine (DVM) students., Indeed, the largest survey, to date, of veterinary student mental health (n = 1245) found a correlation between student stress levels and depression, with 66.4% of respondents reporting mild-to-moderate depression. More disconcertingly, recent research suggests that DVM students’ well-being does not differ markedly from the general population before they enter veterinary school. Not only is student well-being an important concern in its own right, it also influences academic performance and attrition., Despite its impact on veterinary education, there is a little research on predictors of DVM student well-being, as most researchers have focused on predictors of academic performance.,, It would be useful for student affairs personnel and admissions committees to have tools to proactively identify students who may struggle socially and emotionally; hence, we attempt to identify application variables that predict DVM student well-being. We are not suggesting that students who may struggle with well-being should be barred from admission. Rather, we are examining whether there is a way to identify students who may benefit from support, such as counselors or peer groups, if they are admitted. Furthermore, patterns in student well-being may point to deficits in the DVM program, not deficits in the students.
The Veterinary Medical College Application Service (VMCAS) is a centralized service sponsored by the American Association of Veterinary Medical Colleges that provides considerable data to DVM applicants. This report evaluates whether VMCAS data predict scores on a09modified version of the college adjustment test (CAT)—a Likert scale-based measure of veterinary student well-being in their first year. Because the CAT was designed for entering college freshmen, some of the phrasing was minimally changed to make it more relevant to DVM students (VCAT). The literature from the outside of veterinary medicine suggests several admissions variables that may predict well-being as measured by VCAT scores.
Previous studies indicate that the most common predictors of undergraduate student distress during the first year of college include the level of social involvement,, distance from home, and parental education. Distance from home and parental education are both captured in the VMCAS application, and social involvement can be inferred from the involvement in preadmissions extracurricular activities. Consequently, we expected these variables to have the highest likelihood of significantly predicting VCAT. Additionally, we expect that age may play a more limited role in well-being. Some studies have indicated that older students experience a lower sense of belonging than younger students, but others have indicated older students dedicate more time to study and are more satisfied with their overall educational experience than younger students. Thus, we suspect that age will play a role in adjustment, but we are unsure whether it will be positive, negative, or mixed. This study endeavors to build on these previous reports by creating a model that can predict first-year veterinary student well-being based on information contained in the student’s application materials.
| Materials and Methods|| |
Selection of the VMCAS predictor variables
Predictor variables were either downloaded from VMCAS or calculated from VMCAS variables [Table 1]. Sixteen predictor variables were included in this study. Thirteen variables were downloaded from VMCAS, and some were split when calculating predictor variables, whereas others were combined, resulting in 16 predictor variables for further analysis. The downloaded information included all applicants who matriculated into the classes of 2023 or 2024 at one southeastern college of veterinary medicine (n = 233).
Administration of the VCAT
In the late April to early May of their VM1 year, the classes of 2023 and 2024 completed the VCAT in Qualtrics. Because the VCAT was designed for entering college freshmen, the researchers altered some of the phrasing to make it veterinary school-relevant CAT (VCAT). After the initial e-mail invite, nonresponders received one reminder per week for 3 weeks. The VCAT contained 14 five-point Likert items (never to almost always) related to social adjustment, academic adjustment, and emotional adjustment. Based on VM1 students who completed the VCAT during this 4-week fielding period, the reliability coefficients were: α = 0.67 for social adjustment, α = 0.78 for academic adjustment, and α = 0.77 for emotional adjustment [Table 2]. The researchers also asked respondents to report the number of college of veterinary medicine (CVM) clubs and organizations to which they belonged and the number of leadership positions that they held in those clubs. These two items were included at the end of the VCAT to provide an additional measure of student adjustment to veterinary school. One hundred and forty of the eligible 233 students completed the VCAT (response rate = 60.1%).
Best subset regression analysis of VMCAS predictors versus VCAT subscales
The best subsets regression in R was used to determine which combination of independent variables best predicted respondents’ scores on each of the three VCAT subscales and the two involvement items. Thus, the researchers ran five separate best subsets regression algorithms—one for each outcome variable. Best subsets regression evaluates all possible combinations of predictor variables for each outcome variable and identifies the best models of varying sizes (i.e., the best model with one predictor; the best model with two predictors; the best model with three predictors, etc). The researcher further evaluates the models of varying sizes to determine the one best model. The “best” model is determined by the researchers’ priorities. For example, the researchers prioritize predictive power over parsimony, and then adjusted r2 should matter more than Bayesian Information Criterion (BIC).
As there were 16 predictors in our dataset [see [Table 1]], the best subset algorithm generated 16 models for each of the five outcome variables (three VCAT subscales and two involvement items, see [Table 2]). We only considered models with five or fewer predictors, however, given the relatively small sample size (n = 140). Our samples’ demographics were broadly representative of our CVM. The mean age at application was 22.4 (22.3 for the entire CVM), 18.0% were male (18.7% for the entire CVM) and 86.3% were white (82.1% for the entire CVM). To choose the one best model from the five candidate models for each outcome variable, BIC, CP, and adjusted r2 were evaluated. If one model outperformed all other models on two or more of the three evaluation criteria, it was chosen as the best model. If none of the five candidate models outperformed any other model on at least two criteria, the model with the highest adjusted r2 was chosen. Because the main purpose of the analyses was to determine which models best predict student adjustment, the index of predictive power (adjusted r2 ) was made as the deciding criteria when there was no clear best model. Best subsets tests were also used for the number of clubs/organizations and number of leadership positions, though the underlying models were different than those used to predict VCAT scores. Because the VCAT subscales are close enough to continuous data, ordinary least squares regression models were used for the three VCAT analyses. Because the number of clubs/organizations is a count variable, negative binomial models were used for clubs/organizations. The number of leadership positions was transformed into a binary variable (leader versus not) and analyzed using logistic regression. The leadership variable was transformed from count to binomial because (A) we were more interested in whether or not a student attained a leadership position than how many leadership positions they attained, and (B) most students (90%) had either 0 or 1 leadership position.
| Results|| |
The mean VCAT scores for each subscale and overall were academic adjustment (M = 3.22, standard deviation [SD] = 0.72), social adjustment (M = 2.84, SD = 0.62), emotional adjustment (M = 3.14 SD = 0.63), and overall adjustment (M = 3.17, SD = 0.55). Additionally, participants were involved in an average of 3.68 clubs (SD = 2.19) and held an average of 1.31 leadership positions (SD = 1.21).
To evaluate the impact of admissions variables on adjustment, the researchers determined the best model for each of the five outcome variables [see [Table 3]]. In the case of student academic adjustment, it was a three-variable model (employment hours, age, and graduate record exam [GRE] score). With student social adjustment, it was also a three-variable model (GRE score, hometown population, and pre-DVM grade point average [GPA]). The best model for student emotional adjustment contained two variables (employment hours and age). Club membership and number of leadership positions were also best predicted by two variable models (GRE score and pre-DVM GPA).
|Table 3: Each cell contains the beta for the corresponding model. The rightmost column contains the model-adjusted r2|
Click here to view
The best model in terms of our selection criteria contained employment hours, age, and GRE score. The direction of the coefficients indicates that the more students work for pay before coming to veterinary school, the better academic adjustment they report. Similarly, there was a positive relationship between academic adjustment and highest GRE score. In contrast, there was a negative correlation between age and academic adjustment, indicating that older students reported lower levels of academic adjustment than did their younger classmates.
Both hometown population and total pre-DVM GPA negatively predicted social adjustment. That is, respondents with lower pre-DVM GPAs reported higher social adjustment than respondents with higher pre-DVM GPAs, whereas respondents from small hometowns reported higher social adjustment than respondents from larger hometowns. Conversely, GRE positively correlated with the social adjustment such that respondents with higher GRE scores reported higher levels of social adjustment than respondents with lower GRE scores. GRE’s relationship with social adjustment, however, was not significant (P = 0.16).
Employment hours was positively related to emotional adjustment such that respondents who worked more paid hours before veterinary school reported higher levels of emotional adjustment than those who worked fewer hours. Conversely, age was negatively correlated with emotional adjustment such that older students reported lower levels of emotional adjustment than younger students.
Higher GRE scores positively predicted club/organization membership such that respondents with higher GRE scores were involved in more veterinary clubs and organizations than respondents with lower GRE scores. Total pre-DVM GPA correlated negatively with club and organization involvement, but was not significant (P = 0.184).
Highest GRE score positively predicted holding a leadership position such that respondents with higher GRE scores had a higher likelihood of obtaining a leadership position than respondents with lower GRE scores. Total pre-DVM correlated negatively with leadership attainment, but was not significant (P = 0.077).
| Discussion|| |
Predicting well-being has received much less attention in the veterinary education literature than predicting academic performance. Given the prevalence of mental health issues among DVM students, we have attempted to address this oversight. Accordingly, we identified several application variables that were consistent (albeit weak) predictors of the first-year well-being: GRE score, age, and previous employment experience, and to a lesser extent, GPA and hometown population. Younger students with higher GRE scores and substantial work experience tend to fare best. Growing up in a small town is weakly related to higher levels of social adjustment, but further research is needed to determine whether the effect is reliable and to explain the reason for the relationship. Finally, pre-DVM GPA shows a negative relationship with well-being in models where it is included—though the relationship is weak or nonsignificant. We do not have the data necessary to explain GPA’s negative relationship, nor would such an endeavor be withing the scope of this article, but we suspect that students who dedicate more time to study and coursework may have less time to dedicate to socialization and organizational involvement.
Even if our models significantly predicted student well-being, all of the relationships were quite weak. No model in the article accounted for more than 5% of the variance in the first-year well-being. r2 values are somewhat better when predicting academic performance, but it is still rare for a model built on admissions variables to account for more than 30% of the variance in DVM academic performance. Perhaps this suggests that the DVM application process needs to be revised. We believe, however, that admissions data and the humans that produce it are complex. Not all GPAs are equal; not all biology degrees are equal; not all anatomy classes are equal; not all veterinary technician positions are equal. Furthermore, not all DVM students’ first-year experiences are equal. All else being equal, different students will make different friends, form relationships with different professors, and experience different life events. Given the complexity of both university education and humans, it is unlikely the educational researchers will ever see models with predictive power that approach those in the basic sciences. Although additional data would decrease our model’s P values, it would have a little effect on the r2 value. The best way to create a stronger model is to collect data in VMCAS that is predictive of well-being in the VM1 year.
| Conclusion|| |
Despite their relatively weak predictive power, statistical models can be useful tools for admissions committees and student affairs professionals. Incomplete information is better than no information, and linear models often outperform expert judges in terms of predictive power. We hope that our model serves as a useful tool to predict which applicants may need additional wellness resources during their first year. In the meantime, we encourage other CVMs to examine the relationship between their admissions variables and student well-being. Perhaps they will find predictors that perform better than the ones that we currently collect. Additionally, this study was predictive rather than explanatory. That is, we know which admissions variables predict the first-year well-being, but we cannot explain why. It may be useful for future researchers to explore why older students and students from urban areas tend to experience lower levels of well-being than their younger rural counterparts. It is also unclear why GPA and GRE often had discrepant effects in our model. Although the correlation between these two variables was low in our dataset (r = 0.12), they are both considered measures of potential academic performance. The most important avenue for future research, however, is to identify or create admissions variables that reliably predict student well-being. Admissions committees, however, should avoid overinterpreting any predictors of student well-being. Even the best predictors—if they are someday uncovered—will be weak. As such, any predictors of well-being should be treated as relatively minor aspects of the overall student evaluation. Admissions committees should also consider whether variables associated with student well-being point to a problem with a student or a problem with their DVM program. If students from urban areas, international students, or students over 25 report relatively low well-being, admissions committees should ask what their institution can do differently rather than avoid applicants in these categories. Given the recent emphasis on student well-being, we believe that it is useful to highlight early indicators of student struggle; however, our results should be considered in light of their predictive power.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Siqueira Drake AA, Hafen M Jr, Rush BR, Reisbig AM Predictors of anxiety and depression in veterinary medicine students: A four-year cohort examination. J Vet Med Educ 2012;39:322-30.
Reisbig AM, Danielson JA, Wu TF, Hafen M Jr, Krienert A, Girard D, et al
. A study of depression and anxiety, general health, and academic performance in three cohorts of veterinary medical students across the first three semesters of veterinary school. J Vet Med Educ 2012;39:341-58.
Killinger SL, Flanagan S, Castine E, Howard KA Stress and depression among veterinary medical students. J Vet Med Educ 2017;44:3-8.
Waters, A Edinburgh University under pressure to admit unhealthy culture in vet school. Vet Rec 2020;187:334-5.
Drake AA, Hafen M Jr, Rush BR A decade of counseling services in one college of veterinary medicine: Veterinary medical students’ psychological distress and help-seeking trends. J Vet Med Educ 2017;44:157-65.
Karpen SC, Brown SA Building and validating a predictive model for DVM academic performance. Educ Health Prof 2019;2:55-8.
Molgaard LK, Rendahl A, Root Kustritz MV Closing the loop: Using evidence to inform refinements to an admissions process. J Vet Med Educ 2015;42:297-304.
Holladay SD, Gogal RM, Moore PC, Tuckfield RC, Burgess BA, Brown SA Predictive value of veterinary student application data for class rank at end of year 1. Vet Sci 2020;7:120.
Pennebaker JW, Colder M, Sharp LK Accelerating the coping process. J Pers Social Psychol 1990;58:528-37.
Brooks, JH, DuBois DL Individual and environmental predictors of adjustment during the first year of college. J Coll Stud Dev 1995;36:347-60.
Friedlander LJ, Reid GJ, Shupak N, Cribbie R Social support, self-esteem, and stress as predictors of adjustment to university among first-year undergraduates. J Coll Stud Dev 2007;48:259-74.
Mooney SP, Sherman MF, LoPresto CT Academic locus of control, self-esteem, and perceived distance from home as predictors of college adjustment. J Counsel Dev 1991;69:445-8.
Toews ML, Yazedijan A College adjustment among freshmen: Predictors for White and Hispanic males and females. Coll Stud J 2007;41:891-900.
Gonclaves SA, Trunk D Obstacles to success for the nontraditional student in higher education. Psi Chi J Psych Res 2014;19:164-72.
Laanan FS Does age matter? A study of transfer students’ college experience and adjustment process. Paper presented at the Annual Forum of the Association for Institutional Research, Seattle, WA, May 30–June 3, 1999. p. 35.
Dawes RM, Faust D, Meehl PE Clinical versus actuarial judgment. Science 1989;243:1668-74.
[Table 1], [Table 2], [Table 3]