'Student mental health – sociodemographic disparities'
(AsPredicted #102,929)
Download PDF

---

---

1) Have any data been collected for this study already?
It's complicated. We have already collected some data but explain in Question 8 why readers may consider this a valid pre-registration nevertheless.

2) What's the main question being asked or hypothesis being tested in this study?
* Worse internalising and overall mental health will be observed in 1) female compared to male students; in 2) students who do not identity as heterosexual compared to heterosexual students; in 3) Dutch students with a migration background compared to Dutch students with no migration background; in 4) international students compared to domestic students; in 5) students with compared to students without disabilities; and in 6) first-generation students compared to continuing-generation students.
* Worse externalising mental health will be observed in 1) male students compared to female students; and in 2) students without disabilities compared to students with disabilities.

3) Describe the key dependent variable(s) specifying how they will be measured.
The study includes four dependent variables, which are all measured with the use of the Brief Problem Monitor for ages 18-59 (BPM/18-59):
* Internalising mental health problems – the sum score of the 6 internalising items of the BPM
* Externalising mental health problems – the sum score of the 6 externalising items of the BPM
* Attentional mental health problems – the sum score of the 6 attentional items of the BPM
* Overall mental health problems – the sum score of all 18 items of the BPM
The BPM will be scored as proposed by its developers, Achenbach and Ivanova (2018): not true = 0, somewhat true = 1, and very true = 2.

Note that in Question 2 no hypotheses are specified with regard to attentional mental health problems, while it is included as a dependent variable. The reason for this is that the previous literature on sociodemographic disparities in attentional mental health is too limited to derive a hypothesis from.

4) How many and which conditions will participants be assigned to?
The study is observational, so participants will not be assigned to conditions. Mental health (the dependent variables) will be compared between sociodemographic groups (the independent variables), which we will define as follows:

* Gender – In the dataset we use, participants are asked to select their gender, choosing from "Male", "Female", "Other", and "I would rather not say". We will compare the mental health of students who select "Male" to that of those who select "Female".

* Sexual orientation – Participants are asked to indicate their sexual orientation, choosing from "Heterosexual", "Homosexual", "Bisexual", "Asexual", 'Other", and "I would rather not say". We will group students who identify as homosexual, bisexual, asexual, or other together (labelling them as "Non-heterosexual") and compare their mental health to that of students who select "Heterosexual".

* Background – Participants are asked to select their background, choosing from "Dutch background", "Dutch with a migration background", "International student from an African country", "… an Asian country", "… a European country", "… a North-American country", "… a South-American country", "… an Oceanian country", "Other", and "I would rather not say". We will compare the mental health of students who select "Dutch background" to that of those who select "Dutch with a migration background".

* Internationality – Using the answer options described for background, we will group students who select "Dutch background" and "Dutch with a migration background" together (labelling them as "domestic") and group students who select that they are an international student (from any region) together, and compare the mental health of domestic to that of international students.

* Disability – Participants are asked to select disabilities that apply to them (regardless of diagnosis) from a list of 53 disabilities, and may enter up to three more disabilities that are not on the list. We will compare the mental health of students who select/report one or more disabilities to that of students who do not select/report a disability.

* Parental education – Participants are asked to report the highest level of education attended or completed by their parent(s)/caregivers(s), choosing from "No formal education", "Primary school", "Secondary school", "Vocational education", "University of applied sciences / polytechnic", "Research university", and "Other / I don't know". We will compare the mental health of students who report at least one parent/caregiver that attended or completed a bachelor's programme or higher (i.e., a university of applied sciences or a research university programme) to that of students who report parental education levels lower than this (labelling the former "continuing-generation" and the latter "first-generation" students).

5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
We will conduct two sets of linear (non-hierarchical) multiple regression analyses:
* One regression predicting the BPM total score from the six independent variables (gender, sexual orientation, migration background, internationality, disability, and parental education). As migration background and internationality are measured within the same question, these will be included via two contrast variables: one comparing the mental health of Dutch students without a migration background to that of Dutch students with a migration background, and another comparing the mental health of all Dutch students to that of international students (a reverse Helmert contrast).
* Three regressions predicting respectively the internalising BPM score, the externalising BPM score, and the attentional BPM score, again from the six independent variables. For these regressions the critical p value will be set at 0.017 (0.05/3).

6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
Participants will be excluded when showing a careless (i.e., insufficient effort) answering pattern on both the BPM and the Autism Spectrum Quotient (AQ; Baron-Cohen et al., 2001), of which the latter is outside the scope of this pre-registered study. The AQ and BPM were presented consecutively in the survey from which the data are extracted. Separate from the currently pre-registered study, the responses of all survey participants to the AQ and BPM were explored to define a set of response patterns that reflect carelessness. Two pattern types were defined: high-frequency responses, i.e., high use of one particular answer option; and patterned responses, i.e., answer series that form a particular pattern.
* For the AQ, a high-frequency response was defined as answering 35 items or more (out of 50 items in total) the same. Patterned responses were defined as 12341234, 12343212, 12121212, 2323232323, 34343434, 13131313, 24242424, 11223344, and 44332211.
* For the BPM, a high-frequency response was defined as answering all 18 items the same. Patterned responses were defined as 012012, 012101, 01010101, 12121212, 02020202, 001122, and 221100.
These definitions will be used as exclusion criteria for this pre-registered study: participants will be excluded if they show a high-frequency response or patterned response on both the AQ and BPM.

In case of duplicate student IDs, entries made after the first entry will be excluded.

7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined.
The data we will use have already been collected. They come from the entry survey of a longitudinal project that was fielded at one Dutch university in the fall of 2018, and for which students were invited if they were 16 years or older; had just started their degree programme that year; and had not been registered as a student at this university before. For the currently pre-registered study, we will select all participants from this dataset who are younger than 30 years of age; completed all 18 items of the Brief Problem Monitor (BPM); and have data on at least one demographic from the hypotheses (gender, sexual orientation, migration background and internationality, disability, or parental education). Therefore, the sample size will be determined by the number of participants from the entry survey who meet these criteria, minus potential duplicate entries, and minus the participants who will be excluded based on the criteria described in Question 6.

8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)
In reference to Question 1, an explanation of why this is a valid pre-registration even though the data have already been collected:
As explained in Question 7, the data that we will use for the currently pre-registered study come from the entry survey of a longitudinal project for which the data collection plan was not pre-registered. We now pre-register the data analysis plan for a first paper using this existing dataset.

Secondary analyses: in addition to the analyses examining the main hypothesis (see Question 5), we will
* Examine the psychometric properties of the BPM using Principal Axis Factoring (PAF), extracting three fixed factors with oblique rotation. We will do this for the full sample and for the subsample who filled out the (bespoke) Dutch translation of the BPM.
* Benchmark the BPM total score and subscale scores from the sample against age-, gender-, and society-specific normative data from Achenbach and Ivanova (2018).
* Evaluate the data for nonresponse bias by comparing the BPM total score and subscale scores of 1) participants who responded before the reminder emails, 2) those who responded after the first but before the second reminder email, and 3) those who responded after the second email. This will be done with one-way ANOVAs and Games-Howell post-hoc tests. We will also examine the associations between respondent group (i.e., when participants responded) and the independent variables for those independent variables that are not distributed in line with what we would expect based on previous research or population data (for example, 50-50 for gender). This will be done using chi-square tests.