Traditionally, the prevalence of health-risk behaviours is high among university students. Although these behaviours are often analysed in isolation, there is likely evidence for the co-occurrence of multiple risk behaviours. In this study, a latent class analysis (LCA) was conducted with cross-sectional data from 840 Portuguese students (55.4% female) to explore patterns of multiple risk behaviours across seven behavioural dimensions (alcohol consumption, smoking, unhealthy eating, sedentary behaviour, risky sexual practices, illicit drug use, and self-medication). Additionally, a latent class regression was performed to explore predictors (perceived well-being and sociodemographic and academic characteristics) for each behavioural pattern. A three-class model emerged with different probabilities of risk: Low-risk behaviours (51.4%), Moderate-risk behaviors (14.9%), and High-risk behaviours (33.7%). Students with better perceptions of well-being and health were likelier to be in the healthier class. Students in the low- and moderate-risk classes were more likely to be in their first year of study, not in a romantic relationship and to be full-time students. Students who had not changed residence at the beginning of their studies and were female were more likely to be in the healthiest class. This study provides essential strategies for health promotion among university students, offering crucial insights for the design of effective health promotion interventions, especially targeted at specific groups of students with similar patterns of multiple risk behaviours.
Downloads
Download data is not yet available.
##plugins.themes.bootstrap3.article.details##
How to Cite
Alves, R. (2024). A latent class analysis of multiple health-risk behaviours among Portuguese college students. Revista de Estudios e Investigación en Psicología y Educación, 11(1). https://doi.org/10.17979/reipe.2024.11.1.10028
Research Centre on Child Studies (CIEC), Institute of Education, University of Minho: https://www.ie.uminho.pt Braga – Portugal
Alves: A latent class analysis of multiple health-risk behaviours among Portuguese college
students
Emerging adulthood is a critical period in the lives of young people when health-risk
behaviours increase significantly. These behaviours correlate positively with the
risk of developing chronic diseases (i.e. cardiovascular disease or cancer) that tend
to last throughout life. Globally, more attention should be paid to university students,
as empirical evidence shows that they are more likely to engage in health-risk behaviours
than non-university young adults (). The most highly correlated health-risk behaviours include an unbalanced diet due
to low consumption of fruits and vegetables (; ), the use of psychoactive substances such as tobacco, alcohol and illicit drugs () and lack of physical activity (). Many factors contribute to health-risk behaviours, especially distance from the
family of origin ; ) and peer influence ().
Empirical studies typically analyse different health-risk behaviours as isolated measures,
although the co-occurrence of multiple risk behaviours is more likely. In this sense,
numerous cross-sectional and cohort studies among university students in several countries
have relied on latent class analysis (LCA) to show the clustering of health-risk behaviours
(; ; ; ; ; ; ; ; ; ), that is, the likelihood that an individual belongs to a particular class, depending
on their response profile and the characteristics that distinguish them from others
(). The first study to use LCA was developed in 2009 by Laska et al. with US college
students to examine lifestyle patterns of health-risk behaviours. In 2014, a study
using NCHA data (n = 39 colleges, n = 30,093 students) was published that identified four classes among the five behaviours
examined (smoking, excessive alcohol consumption, unhealthy diet, sedentary lifestyle,
and overweight or obesity), with clustering of risk behaviours (i.e. the high prevalence
of two or more health-risk behaviours) in three of the classes (). Similarly, conducted an LCA of eight risk behaviours (fruit and vegetable consumption, alcohol
use, drug use, smoking, marijuana use, sexual health, physical activity and sleep)
among Canadian university students. This study found that most students (66%) fell
into the ‘typical’ class (low likelihood of smoking and using illicit drugs and low
likelihood of eating enough fruits and vegetables and being physically active), followed
by the ‘high risk’ class (20%; a low likelihood of eating enough fruits and vegetables,
getting enough sleep and being physically active and an increased likelihood of using
illicit drugs, smoking and binge drinking).
This person-centred approach is more useful than a variable-centred approach as it
allows the identification of multidimensional behavioural patterns () and the provision of personalised interventions after taking into account the characteristics
of subgroups categorised on the basis of behavioural patterns (). Despite the importance of this type of study, there is a lack of research in Portugal
aimed at classifying university students based on patterns of risk behaviours and
analysing how perceptions of well-being may vary based on the grouping of health-risk
behaviours. To advance the development of relevant intervention strategies for Portuguese
university students, the aim of this study is to examine the patterns of multiple
health-risk behaviours (physical activity, eating habits, tobacco use, alcohol use,
illicit drug use, self-prescribed medication use and risky sexual behaviour) in a
sample of Portuguese university students and to determine the predictors associated
with each latent class.
Methods
Participants
This cross-sectional study involved 840 students enrolled at a university in northern
Portugal during the 2018-2019 academic year. The participants’ ages ranged from 18
to 54 years (M = 20.8 ± 4.22), only 3% of students were 30 years or older. The majority of participants
were female (55.4%), not in a romantic relationship (58.3%), had not changed residence
upon entering higher education (64.9%), had a BMI corresponding to normal weight (73.1%)
and were full-time students (88.8%). In addition, approximately one-third of the students
surveyed were enrolled in engineering (36.0%) or social sciences and humanities (32.1%)
programmes, and most of them were in their first year of study (55.2%) (Table 1).
Table 1Sample characteristics (N = 840)
Characteristics
n
%
Year of study
1st year
464
55.2
3rd year
376
44.8
Scientific area
Engineering sciences
302
36.0
Exact and natural sciences
136
16.2
Law and economic sciences
132
15.7
Human and social sciences
270
32.1
Sex
Male
375
44.6
Female
465
55.4
Love relationship
Yes
347
41.7
No
486
58.3
Current residence
Displaced
291
35.1
Not displaced
537
64.9
BMI
Low weight
58
7.1
Normal weight
599
73.1
Overweight
162
19.8
Professional situation
Full time student
739
88.8
Worker/Student
93
11.2
Instruments
The students completed a self-report questionnaire consisting of three sections: sociodemographic
questions, a health-risk behaviour questionnaire and the Well-Being and Health-Perception
Scale (WbHPS).
Sociodemographic and academic questions
While maintaining the anonymity of the research instrument, it was essential to collect
some sociodemographic data about the students, including gender, age, marital status,
current residence, weight and height (to calculate BMI) and professional situation.
In addition, the researcher recorded the year of attendance and the subject area of
the course attended at the time of data collection.
Health-risk behaviour questionnaire
This section was divided into seven categories, each corresponding to the health-risk
behaviours studied. This questionnaire was developed in three stages (scale construction,
content validity, psychometric validity), following the procedures outlined by Bowling
() and the Guide to Conducting Knowledge, Attitude, and Practice (KAP) Surveys (). A detailed description can be found in previous studies ().
Tobacco consumption was assessed with the question, ‘Do you currently smoke?’ For
the purposes of this study, we dichotomised the ‘smoking’ responses into ‘yes’ (2)
and ‘no’ (1).
The alcohol consumption category included the AUDIT-C (frequency of drinking in the
past year, amount of alcohol consumed on a typical day and binge drinking) (). The AUDIT-C has a 5-point scale coded from 0 to 4 (range 0-12). A score above 4
points (≥5) for men or above 3 points (≥4) for women is classified as risk consumption:
‘yes’ (2) and ‘no’ (1). In addition, the following question on drunkenness was included:
‘In the last 12 months, how often did you get drunk?’ The response options were ‘never’,
‘once a month or less’, ‘2-4 times a month’, ‘2-3 times a week’ and ‘4 or more times
a week’. The answer to the question ‘get drunk’ was also dichotomised into ‘yes’ (2)
and ‘no’ (1).
Illicit drug consumption included three questions on the use of cannabis, cocaine
and hallucinogens (‘In the last 12 months, how often have you used...?’). Users of
illicit drugs were categorised as ‘yes’ (2) and ‘no’ (1).
The prevalence of self-medication was analysed using the question, ‘In the past 12
months, how often have you taken antidepressants, sedatives, relaxants or tranquillisers
(without a medical prescription); painkillers or anti-inflammatories (without a medical
prescription) or vitamins or food supplements (without a medical prescription)?’ Consumers
of antidepressants, painkillers and vitamins for self-medication were categorised
as ‘yes’ (2) and ‘no’ (1).
Sexual risk behaviours were divided into four ‘yes’ (2) and ‘no’ (1) response categories:
early age (first sexual intercourse at age 16 or younger); multiple partners (two
or more sexual partners in the past 12 months); inconsistent condom use (did not use
a condom for all sexual intercourse in the past 12 months); and sex, alcohol and drugs
(sexual intercourse after alcohol or drug use in the past 12 months).
Healthy eating habits were assessed by the consumption of fruit, vegetables, foods
with added sugar and fast food in the last seven days (e.g. ‘In the last seven days,
how often did you eat fruit (other than natural fruit juices or drinks)?’ on a scale
from 0 = ‘never’ to 5 = ‘3 or more times a day’). In addition, two questions were
asked about skipping meals (breakfast, lunch and/or dinner) in the last seven days.
For the analyses, the questions were categorised as 1 using a binary method according
to the healthy eating guidelines (), taking into account the following: skipping meals (skipping breakfast and skipping
lunch and/or dinner at least once a week); low fruit consumption (two or fewer times
a day); low vegetable consumption (two or fewer times a day); higher consumption of
sweets (four or more times a week) and higher consumption of fast food (four or more
times a week).
The Godin Leisure-Time Exercise Questionnaire () assessed physical activity, including total time (in minutes) spent walking and
doing moderate or vigorous activity in the previous week. The scores followed the
information provided by the authors of the questionnaire. Subsequently, ‘being sedentary’
was classified into ‘yes’ (2) and ‘no’ (1) using a binary method. The following question
on sedentariness was also included: ‘On a normal day, how many hours do you spend
sitting?’ Again, sitting for more than eight hours per day was classified as high.
Well-being and Health Perception Scale (WbHPS)
Well-being was assessed using the Well-being and Health Perception Scale (WbHPS) (). The WbHPS was answered on a 5-point Likert scale and included the following dimensions:
satisfaction with life, satisfaction with oneself, felt happiness, perceived health
and satisfaction with physical fitness. The scores for each item were summed, with
the highest scores representing a higher level of well-being. The psychometric properties
of this scale have been described in detail in previous studies ().
Procedures
The target population consisted of students from a university in northern Portugal
enrolled in the first and third years of the undergraduate and integrated master's
programmes during the 2018-2019 academic year. The criteria for exclusion were defined
as follows: classes from master's or postgraduate programs, as well as classes held
in post-working hours, were excluded to limit the population to traditional students;
courses without classes in the first and third years were excluded to ensure representation
from students across all years of study within the same courses; and classes related
to health sciences were excluded due to potential biases, especially concerning their
knowledge about health ().
To create a stratified sample based on the year of study and the scientific area,
were considered 5447 university students enrolled in the first- and third-year courses,
distributed across 47 courses and four subject areas (as defined by the Foundation
for Science and Technology): Engineering, Exact and Natural Sciences, Judicial and
Economic Sciences, Social Sciences and Humanities. Considering the number of students
in each subject area, 16 courses were identified by simple random selection, resulting
in 32 first- and third-year classes.
Using G*Power 3.1.9.2 analysis software, the minimum sample size required for this
study was 592 students (margin of error = 5%, confidence level = 99% and distribution
of responses = 50%). A total of 873 questionnaires were distributed, reflecting the
total number of students who were present in the classroom and had agreed to participate
in the study. However, 33 questionnaires were considered invalid due to missing or
incongruent responses that could call into question the validity of the information
collected. The response rate was 96.2% (95% CI: 94.8-97.6).
Data collection was conducted in the classroom in May 2019. Before the paper questionnaires
and the informed consent forms were distributed, the research objectives were presented,
confidentiality and anonymity of responses were ensured, the voluntary nature of participation
was guaranteed, and the right to decline participation or withdraw at any time was
explained. All participants gave their consent before completing the questionnaire,
which took about 15 minutes to complete.
This study was approved by the Ethics Committee for Research in the Social Sciences
and Humanities of the UMinho Ethics Council under protocol CEICSH 009/2019.
Statistical analyses
In the first step, descriptive analysis was conducted for the different variables
by analysing frequencies, means and standard deviations using the Statistical Package
for the Social Sciences (SPSS) version 28.0 for Windows. In the second step, LCA was
conducted with seven categorical observable variables (risky drinking, smoking, unhealthy
diet, sedentarism, risky sexual practices, illicit drugs and self-medication practices)
to identify patterns of health-risk behaviour. A range of LCA models with two to five
classes was explored to determine the number of classes that best represented patterns
of health-risk behaviour (). The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC)
were considered to determine the optimal number of conceptually significant classes
(). In addition, the G2 and log-likelihood (LLIK) values of the estimated model were calculated to select
the best-fitting model. Thus, the lowest G2, AIC and BIC values and the highest LLIK value indicated the best-fitting model ().
Entropy was used as an additional criterion to determine the final model. Entropy
is a quality criterion for classification, with higher values between 0 and 1 indicating
a better fit (). However, as these indices are not uniformly indicative of a single model specification
(and may overestimate or underestimate the number of latent classes), model-specific
estimates for two-, three- and four-class models were examined to select a final model
specification based on the interpretability of the results as well as theory and previous
results in the literature, as suggested by . Based on these criteria, we determined that the three-class model was the best fit
for the data. All LCA analyses were performed using the GLCA R package in Jamovi software,
version 2.3.2. After finalising the model, we performed multinomial logistic regression
estimated by odds ratio and the corresponding 95% confidence intervals to analyse
the analyses of the predictors (well-being and demographic characteristics) of the
classes of health-risk behaviours.
Results
Multiple health-risk behaviours
Table 2 shows the prevalence of the 14 health-risk behaviours in the seven behavioural dimensions
under analysis. The prevalence of these behaviours ranged from 7.4% (high consumption
of fast food) to 90.4% (insufficient consumption of vegetables), indicating that the
prevalence of some risk behaviours (e.g. poor diet and alcohol consumption) was much
higher than others. Statistically significant differences were found in the prevalence
of certain risk behaviours between male and female students. Drunkenness, use of illicit
drugs, early sexual initiation, sexual activity after alcohol and drug use and consumption
of sugar products are more common among male students. Conversely, the self-medication
with analgesics and sedentary behaviour are more common among female students.
Table 2Multiple health-risk behaviours by sex (N = 840)
Total
Male
Female
n
%
n
%
n
%
Smoking
168
20.0
85
22.7
83
17.8
3.01
Risky alcohol consumption
337
40.1
152
40.5
185
39.8
.05
Getting drunk
368
43.8
186
49.6
182
39.1
9.23**
Use of illegal drugs
178
22.2
114
31.7
64
14.5
33.75***
Self-medication with antidepressants
111
13.9
27
7.5
84
19.0
3.01
Self-medication with analgesics
326
40.8
116
32.2
210
47.7
19.72***
Self-medication with vitamins
211
26.4
92
25.6
119
27.0
.21
Early age
213
25.4
110
29.3
103
22.2
5.66*
Multiple partners
103
12.3
52
13.9
51
11.0
1.62
Inconsistent condom use
329
39.2
148
39.5
181
38.9
.03
Sex, alcohol and drugs
177
21.1
94
25.1
83
17.8
6.50*
Skipping meals
428
51.0
188
50.3
240
51.6
.15
Insufficient fruit consumption
753
89.9
344
92.0
409
88.1
3.34
Insufficient vegetable consumption
784
94.0
354
95.7
430
92.7
3.30
Higher consumption of fast food
62
7.4
29
7.7
33
7.1
.11
Higher consumption of sweets
465
55.5
222
59.4
243
52.4
4.09*
Sedentariness
298
35.7
106
28.3
192
41.7
16.11***
High sitting time
436
51.9
204
54.4
232
49.9
1.69
*
p < .05 ; **p < .01 ; ***p < .001
Latent classes of risk behaviours
Based on the 14 health-risk behaviours analysed, a three-latent-class model was selected
after comparing the fit, parsimony and interpretability indices between models with
two to five latent classes, as described in the section on statistical analyses and
presented in Table 3.
Table 3Summary of information for selecting number of latent classes of multiple health-risk
behaviours (N = 840)
Number of Latent Classes
LLIK
AIC
BIC
Entropy
G2
df
G2 p
2
-7583
15239
15451
.77
4683
802
.08
3 SM
-7516
15144
15409
.84
4549
783
.14
4
-7458
15066
15421
.76
4434
764
.10
5
-7416
15020
15464
.79
4349
745
.16
6
-7385
14996
15531
.77
4258
726
.34
Note: SM = Selected model; LLIK = Log-likelihood; AIC = Akaike Information Criterion; BIC
= Bayesian information criterion; G2 = Likelihood ratio test statistic.
The probabilities of a 'yes' response for each of the 14 risk behaviours are shown
in Table 4 for each latent class. The likelihood of a 'no' response can be calculated by subtracting
the item response probabilities from 1. These probabilities indicate the possibility
that a student in a particular class will engage in a specific risk behaviour and
can be used to characterise the three latent classes in the model.
Table 4Item-response probabilities from the three-latent-class model of multiple health-risk
behaviours (N = 840)
Label
1 Low-risk behaviours
2 Moderate-risk behaviours
3 High-risk behaviours
Latent class prevalences
.51
.15
.34
Item-response probabilities Corresponding to a ‘yes’ response
Smoking
.05
.15
.45
Risky alcohol consumption
.09
.62
.77
Getting drunk
.03
1.00
.81
Use of illegal drugs
.04
.18
.51
Self-medication with antidepressants
.12
.15
.16
Self-medication with analgesics
.38
.38
.46
Self-medication with vitamins
.23
.13
.36
Early age
.17
.06
.46
Multiple partners
.29
.00
.32
Inconsistent condom use
.29
.17
.65
Sex, alcohol and drugs
.02
.00
.59
Skipping meals
.44
.43
.66
Insufficient fruit consumption
.89
.96
.89
Insufficient vegetable consumption
.93
.99
.94
Higher consumption of fast food
.05
.48
.13
Higher consumption of sweets
.55
.52
.58
Sedentariness
.38
.42
.29
High sitting time
.58
.48
.44
Note: Major probabilities are in bold to highlight the overall pattern.
Class 1 (labelled ‘low-risk behaviours’) represents the majority of students in the
sample (51.4%) and is considered the healthiest class. These students were more likely
to be physically active (61.7%), although they were more likely to sit for more than
eight hours a day (58.2%) and more likely not to smoke (94.7%), not have risky alcohol
consumption (90.0%) and/or not use illicit drugs (96.1%). However, students in this
class were not likely to consume the recommended amounts of fruit (88.7%) and vegetables
(92.7%).
Class 2 (labelled 'moderate risk behaviours’) represents the students (14.9%) at moderate
risk of engaging in health-risk behaviours. Students in this class were more likely
to have been drunk in the past year (100%), to have consumed too few fruits (96.4%)
and vegetables (98.9%), and to have eaten fast food several times a week (48.3%).
In addition, students with a high probability of alcohol consumption classified as
risky belonged to this class (62.4%). Notwithstanding this, and similar to the ‘low-risk
behaviour’ class, the moderately healthy students were less likely to smoke (14.8%),
use illicit drugs (17.6%) and take non-prescribed medications such as antidepressants
(14.9%), painkillers (37.7%) and vitamins (13.0%). Of the three classes identified,
these students were least likely to engage in the following risky sexual behaviours:
early age (6.3%), multiple partners (0.16%), inconsistent condom use (16.6%), sex,
alcohol and drugs (< .001%).
Class 3 (labelled ‘high-risk behaviours’; 33.7%) exhibited high-risk behaviours in
all dimensions except for sedentariness. For example, students in this class were
more likely to smoke (44.7%), have risky alcohol consumption (77.4%), get drunk (81.2%)
and use illicit drugs (50.8%) but less likely to be physically inactive (29.4%) and
sit too long during the day (44.1%). In addition, this class was the most likely to
take without prescription medications such as antidepressants (16.4%), painkillers
(46.2) and vitamins (36.3%) and engage in the following risky sexual behaviours: early
age (46.2%), multiple partners (31.9%), inconsistent condom use (65.2%) and sex, alcohol
and drugs (59.1%). In addition, the members of this group reported unhealthy eating
habits, such as a high likelihood of skipping meals (65.8%), consuming sugary products
(58.1%), and eating too little fruit (88.7%) and vegetables (93.8%).
Sociodemographic and well-being predictors as a function of latent classes of health-risk
behaviours
Based on the class with the higher probability of risk behaviours, the probability
of belonging to the healthiest class ('low-risk behaviours’) increased by 1.06 times
for each additional point in the WbHPS. Similarly, being female increased the likelihood
of belonging to Class 1 by 24.1% compared to Class 3 and by 10.3% compared to Class
2. Students who had not changed their residence when beginning their university studies
were 1.98 and 1.83 times more likely to be in Class 1 than Class 3 and 2, respectively.
Being in the first year of study, not being in a romantic relationship and being a
full-time student increased the likelihood of being in the healthiest and moderately
healthy classes compared to Class 3 (‘high-risk behaviours’) (Appendix).
Discussion
This study revealed distinct patterns of multiple health-risk behaviours among Portuguese
university students, namely the existence of three recognisable classes. These classes
included students prone to engaging in health-risk behaviours (33.7%) and students
with low (51.4%) and moderate (14.9%) risks of engaging in risky behaviours. These
classes were similar to previous studies with university students regarding class
division (; ; ; ; ; ). However, this study shows a higher proportion of students in the high-risk class,
whereas previous studies showed a higher proportion of students with moderate-risk
behaviour patterns.
Of concern is that the vast majority of students, regardless of class, were very likely
to consume too few fruits and vegetables and too many sugary products. Similarly,
in the study conducted by , nursing students were found to be the most likely to have unhealthy diets across
all latent classes. These results demonstrate that while public health has traditionally
focused on reducing certain risk behaviours, such as tobacco, alcohol and illicit
drugs (; ), other risk behaviours associated with non-communicable diseases, such as dietary
habits, also deserve attention. These behavioural patterns may reflect difficulties
such as the cost and accessibility of fruit and vegetables or the difficulty of preparing
healthy meals.
Sedentary lifestyles and the amount of time students spend sitting during the week
may be related to the number of consecutive lessons and their duration. Therefore,
teachers should encourage students to get up and move around after a prolonged sitting
period (). As this is a new phase of life where students have to balance their personal lives
with academic, family and social commitments, it is not surprising that students may
place less emphasis on physical activity during this stage (; ).
Regarding alcohol consumption, both ‘moderate risk behaviours’ and ‘higher risk behaviours’
were highly likely to involve binge drinking and drunkenness. Other studies have also
found that binge drinking is among the highest health-risk classes (). Thus, alcohol consumption should also be discouraged through university health
promotion programmes.
The multinomial regression analysis indicated that students in the healthiest class
were more likely to be female, in their first year of study, to be full-time students,
not to be in a romantic relationship, and not to have moved when they began their
studies. Similarly, previous research has shown that male and younger students were
likelier to be in unhealthy or moderate lifestyle classes than the healthier ones
(; ). With this in mind, it is critical to intervene with students in their first year
of university study when new experiences significantly influence behavioural choices.
This study also analysed how perceptions of health and well-being were related to
patterns of multiple risk behaviours. Overall, the results showed that students with
a higher level of well-being were less likely to belong to a class with a high number
of risk behaviours. This finding is consistent with other studies showing that higher
risk patterns in multiple behaviours are associated with poorer mental health outcomes,
namely psychological distress and stress (; ; ), with the inverse relationship also demonstrated. In other words, there is a bidirectional
effect between health-risk behaviours and mental health indicators (e.g. happiness,
life satisfaction and depression) (; ; ).
Strengths and limitations of the study
To our knowledge, this is the first study to analyse complex patterns of health-risk
behaviours among Portuguese university students across a wide range of behaviours,
focusing on their relationship with perceived well-being and health. Not only do the
results provide critical information about students' behavioural patterns and help
in the development of health promotion strategies, but this study also demonstrates
the usefulness of sophisticated analytical tools such as LCA to understand complex
behaviours that would not be possible with statistical approaches such as linear regression.
Although a relatively new statistical analysis was used to analyse the clustering
of multiple health-risk behaviours among university students and, because it is a
model-based approach, to estimate the actual probability of an outcome as a function
of latent class membership (), this study has some limitations that should be considered when interpreting the
results. First, it should be noted that this is a cross-sectional study whose data
were only collected at one Portuguese public university. Therefore, it is not possible
to generalise the results to students from other universities or young people in the
same age group or to establish causal relationships. Secondly, a self-report questionnaire
was used, which might have led to a bias in the responses due to social desirability.
However, validated tools were used to minimise this, and confidentiality and anonymity
were guaranteed.
Note that health-risk behaviours were assessed on different time scales (e.g. last
seven days, last 30 days, last 12 months), and seasonal effects were not captured.
Although measures with more extended time scales were intended to capture the less
prevalent health-risk behaviours (e.g. illicit drug use), consistency across all behaviours
would be ideal. Furthermore, even though data collection was carried out during an
academic period that excluded festivities, the described behaviors may not accurately
reflect the pattern of behaviors, especially concerning dietary habits in the past
seven days. Similarly, specific periods of self-medication were not considered, such
as during study and exam preparation. It is noteworthy that the use of electronic
cigarettes or heated tobacco products was not considered, as the use of these tobacco
products was minimal among university students (). Another limitation worth mentioning is the exclusion of students from health science
courses. This decision was made considering the potential bias that could be introduced
(). It is essential to emphasize that this methodological choice by no means implies
a disregard for the importance of investigating the behaviours of students in these
fields in future studies. Additionally, it is crucial to acknowledge that a meta-analysis
conducted by identified a significantly higher prevalence of self-medication practices with antibiotics
among medical students compared to students in other scientific disciplines.
Even though dichotomising risk variables is common in LCA and contributes to the interpretability
and communication of findings, categorising data in this way may lead to some loss
of meaning. Finally, while we included several sociodemographic and academic factors
in the analysis, others may be relevant to the analysis of health-risk behaviours,
such as students' socioeconomic status. Therefore, future studies should include other
variables and conduct longitudinal analyses to assess the same students over time.
Practical implications and future studies
This study highlights the need for a reorientation of health services and higher education
institutions to ensure that efforts are focused on promoting healthy behaviours that
are already commonly targeted, such as smoking and drug use, but also dietary eating
habits. In addition, the Healthy Campus 2020 guidelines () advocate for a greater focus on nutrition and physical activity among young adults.
However, higher education institutions need more initiatives to promote healthy, active
lifestyles to achieve these goals.
Furthermore, there is a need for continued research on this topic to better understand
the patterns of health-risk behaviours and their relationship to well-being among
university students. Future research should therefore include longitudinal studies
to analyse the behavioural changes in students' profiles and their mental health status
throughout their time at university. This further investigation can be used to determine
the timing of interventions to support student health and well-being and to target
interventions at specific classes of students.
Current strategies for health promotion for university students are generally most
effective for the majority of the university population, such as integrating health
promotion activities into curricula. However, the results of this study suggest that
at-risk groups need a tailored intervention. This intervention can be health promotion
materials that target the different health-risk behaviours together rather than addressing
them individually. By simultaneously addressing several target health-risk behaviours,
people can transfer their knowledge and experience from one behaviour to another.
Although it is not clear which approach is most effective (single behaviour or multi-behaviour)
(), it must always be taken into account that, on the one hand, risk behaviours are
interrelated and co-occur, and, on the other hand, separate approaches may be needed
for certain risk behaviours, especially in the case of illicit drug use. Finally,
any strategies developed to promote health and well-being in higher education should
be consistent with the International Charter for Health-Promoting Universities and
Colleges guidelines () and focus on both the academic environment and support for individual students.
Conclusion
This study makes an important contribution to the literature on health-risk behaviours
in higher education by identifying three classes of behaviours in a sample of Portuguese
university students. Common to all latent classes were unhealthy eating habits and
sedentary behaviour. In addition, students who belonged to the classes most likely
to engage in risky behaviours were found to experience lower levels of well-being.
These findings underscore the need for higher education institutions to implement
socio-educational programmes aimed at preventing the adoption of unhealthy habits
upon entering higher education. Such programmes would enable students to make informed
and responsible choices. There is also a call for psychosocial interventions that
facilitate students' reflection on their well-being and happiness, thereby contributing
to their holistic development. Furthermore, these findings highlight the need to support
and target the development of activities for the classes of students most likely to
engage in risky behaviours.
References
1
AFRASHTEH, Sima; GHAEM, Haleh & ABBASI-GHAHRAMANLOO, Abbas (2017). Clustering and
combining pattern of high-risk behaviors among Iranian university students: A latent
class analysis. Journal of Research in Health Sciences, 17(4), 00398. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7189947/
2
ALVES, Regina F. (2022a). Knowledge, attitudes and health-related behaviors among Higher Education students:
assessment and monitoring of a socioeducational intervention program in a non-formal
context [Doctoral thesis, University of Minho]. https://hdl.handle.net/1822/75905
3
ALVES, Regina F. (2022b). The association between subjective well-being and risky
behaviours in university students. Psihologijske Teme, 31(3), 565-583. https://doi.org/10.31820/pt.31.3.5
4
ALVES, Regina F. & PRECIOSO, José (2020). Knowledge, attitude and practice on healthy
diet among university students in Portugal. International Journal of Nutrition Sciences, 5(2), 2-9. https://ijns.sums.ac.ir/article_46568.html
5
ALVES, Regina F. & PRECIOSO, José (2022). A influência dos pares no consumo de Substâncias
Psicoativas entre estudantes universitários/as. Revista de Estudios e Investigación En Psicología y Educación, 9, 5-17. https://doi.org/10.17979/reipe.2022.9.0.8912
6
ALVES, Regina F.; PRECIOSO, José & BECOÑA, Elisardo (2020). Well-being and health
perception of university students in Portugal: the influence of parental support and
love relationship. Health Psychology Report, 8(2), 145-154. https://doi.org/10.5114/hpr.2020.94723
7
ALVES, Regina F.; PRECIOSO, José & BECOÑA, Elisardo (2021). Physical activity among
Portuguese university students and its relation to knowledge and perceived barriers.
Sportis. Scientific Journal of School Sport, Physical Education and Psychomotricity,
7(1), 25-42. https://doi.org/10.17979/sportis.2021.7.1.6924
8
BARRY, Adam E.; CHANEY, Beth H.; STELLEFSON, Michael L. & DODD, Virginia (2015). Evaluating
the psychometric properties of the AUDIT-C among college students. Journal of Substance Use, 20(1), 1-5. https://doi.org/10.3109/14659891.2013.856479
9
BENNASAR-VENY, Miquel; YAÑEZ, Aina M.; PERICAS, Jordi; BALLESTER, Lluis; FERNANDEZ-DOMINGUEZ,
Juan Carlos; TAULER, Pedro & AGUILO, Antoni (2020). Cluster analysis of health-related
lifestyles in university students. International Journal of Environmental Research and Public Health, 17(5). https://doi.org/10.3390/IJERPH17051776
10
BLACK, Tara & STANTON, Alisa (2016, February 4). Final report on the development of
the Okanagan Charter: An international charter for health promoting universities &
colleges. International Conference on Health Promoting Universities and Colleges (7th: 2015). Kelowna, BC, Canada. https://doi.org/10.14288/1.0372504
11
BOWLING, Ann. (1998). Research methods in health: investigating health and health services. Open University Press.
12
CARTER, Ashlee C.; BRANDON, Karen Obremski & GOLDMAN, Mark S. (2010). The college
and noncollege experience: A review of the factors that influence drinking behavior
in young adulthood. Journal of Studies on Alcohol and Drugs, 71(5), 742-750. https://doi.org/10.15288/jsad.2010.71.742
13
CHAN, P., & SHEK, C. (2018). Latent class analysis of student artefacts. Proceedings of the 10th International Conference on Education Technology and Computers, 302-307. https://doi.org/10.1145/3290511.3290521
14
COLLINS, Linda M. & LANZA, Stephanie T. (2010). Latent class and latent transition analysis with applications in the social, behavioral,
and health sciences. Wiley Series in Probability and Statistics. https://doi.org/10.1002/9780470567333
15
DAVOREN, Martin P.; DEMANT, Jakob; SHIELY, Frances & PERRY, Ivan J. (2016). Alcohol
consumption among university students in Ireland and the United Kingdom from 2002
to 2014: a systematic review. BMC Public Health, 16(1), 173. https://doi.org/10.1186/s12889-016-2843-1
16
El ANSARI, Walid & BERG-BECKHOFF, Gabriele (2017). Country and gender-specific achievement
of healthy nutrition and physical activity guidelines: Latent class analysis of 6266
university students in Egypt, Libya, and Palestine. Nutrients, 9(7), 738. https://doi.org/10.3390/nu9070738
17
EL ANSARI, Walid; SSEWANYANA, Derrick & STOCK, Christiane (2018). Behavioral health
risk profiles of undergraduate university students in England, Wales, and Northern
Ireland: A cluster analysis. Frontiers in Public Health, 6. https://doi.org/10.3389/FPUBH.2018.00120
HUTCHESSON, Melinda J.; DUNCAN, Mitch J.; OFTEDAL, Stina; ASHTON, Lee M.; OLDMEADOW,
Christopher, KAY-LAMBKIN, Frances & WHATNALL, Megan C. (2021). Latent class analysis
of multiple health risk behaviors among Australian university students and associations
with psychological distress. Nutrients, 13(2), 425. https://doi.org/10.3390/nu13020425
20
HYSKA, Jolanda; MERSINI, Ehadu; MONE, Iris; BUSHI, Entela; SADIKU, Edite & HOTI, Kliti
(2014). Assessment of knowledge, attitudes and practices about public health nutrition
among students of the University of Medicine in Tirana, Albania. Southeastern European Journal of Public Health, 88, 1-8. https://doi.org/10.12908/SEEJPH-2014-01
21
JAMES, Erica; FREUND, Megan; BOOTH, Angela; DUNCAN, Mitch. J.; JOHNSON, Natalie; SHORT,
Camille E.; WOLFENDEN, Luck; STACEY, Fiona G.; KAY-LAMBKIN, Frances & VANDELANOTTE,
Corneel (2016). Comparative efficacy of simultaneous versus sequential multiple health
behavior change interventions among adults: A systematic review of randomised trials.
Preventive Medicine, 89, 211-223. https://doi.org/10.1016/j.ypmed.2016.06.012
22
JAO, Nancy C.; ROBINSON, Laura D.; KELLY, Peter J.; CIECIERSKI, Christina C. & HITSMAN,
Brian (2019). Unhealthy behavior clustering and mental health status in United States
college students. Journal of American College Health, 67(8), 790-800. https://doi.org/10.1080/07448481.2018.1515744
23
KABIR, Kourosh; BAHARI, Ali; HAJIZADEH, Mohammad; ALLAHVERDIPOUR, Hamid; TARRAHI,
Mohammad J.; FAKHARI, Ali; ANSARI, Hossein & MOHAMMADPOORASL, Asghar (2018). Substance
abuse behaviors among university freshmen in Iran: a latent class analysis. Epidemiology and Health, 40, e2018030. https://doi.org/10.4178/epih.e2018030
24
KANG, Joseph; CIECIERSKI, Christina Czart; MALIN, Emily L.; CARROLL, Allison J.; GIDEA,
Marian; CRAFT, Lynette L.; SPRING, Bonnie & HITSMAN, Brian (2014). A latent class
analysis of cancer risk behaviors among U.S. college students. Preventive Medicine, 64, 121-125. https://doi.org/10.1016/j.ypmed.2014.03.023
25
KWAN, Matthew; ARBOUR-NICITOPOULOS, Kelly P.; DUKU, Eric; & FAULKNER, Guy (2016).
Patterns of multiple health risk-behaviours in university students and their association
with mental health: application of latent class analysis. Policy and Practice, 36(8). https://doi.org/10.24095/hpcdp.36.8.03
26
LANZA, Stephanie T. & RHOADES, Brittany L. (2013). Latent class analysis: An alternative
perspective on subgroup analysis in prevention and treatment. Prevention Science, 14(2), 157-168. https://doi.org/10.1007/s11121-011-0201-1
27
LANZA, Stephanie T.; SAVAGE, Jennifer S. & BIRCH, Leann L. (2010). Identification
and prediction of latent classes of weight-loss strategies among women. Obesity, 18(4), 833-840. https://doi.org/10.1038/oby.2009.275
LASKA, Melissa Nelson; PASCH, Keryn E.; LUST, Katherine; STORY, Mary & EHLINGER, Ed
(2009). Latent class analysis of lifestyle characteristics and health risk behaviors
among college youth. Prevention Science, 10(4), 376-386. https://doi.org/10.1007/s11121-009-0140-2
30
MA, Cecilia M. & LAI, Catie C. (2018). Mental health profile and health-related behavior
among Hong Kong Chinese university students. Health Psychology Open, 5(2), 205510291878686. https://doi.org/10.1177/2055102918786869
31
MACEDO, Tassia T. S.; MUSSI, Fernanda C.; SHEETS, Debra; CAMPOS, Ana C. P.; PATRÃO,
Ana L.; FREITAS, Caren L. M. & PAIM, Melissa A. S. (2020). Lifestyle behaviors among
undergraduate nursing students: A latent class analysis. Research in Nursing & Health, 43(5), 520-528. https://doi.org/10.1002/nur.22064
32
MATHUR, Charu; STIGLER, Melissa; LUST, Katherine & LASKA, Melissa (2014). A latent
class analysis of weight-related health behaviors among 2- and 4-year college students
and associated risk of obesity. Health Education & Behavior, 41(6), 663-672. https://doi.org/10.1177/1090198114537062
33
MOLENDIJK, Marc; MOLERO, Patricio; ORTUÑO SÁNCHEZ-PEDREÑO, Felipe; VAN DER DOES, Willem
& ANGEL MARTÍNEZ-GONZÁLEZ, Miguel (2018). Diet quality and depression risk: A systematic
review and dose-response meta-analysis of prospective studies. Journal of Affective Disorders, 226, 346-354. https://doi.org/10.1016/j.jad.2017.09.022
34
NAZAR, Gabriela; STIEPOVIC, Jasna & BUSTOS, Claudio (2019). Latent class analysis
of lifestyle behavior among Chilean university students. International Journal of Health Promotion and Education, 57(2), 98-111. https://doi.org/10.1080/14635240.2018.1552534
35
NYLUND-GIBSON, Karen & CHOI, Andrew Young (2018). Ten frequently asked questions about
latent class analysis. Translational Issues in Psychological Science, 4(4), 440-461. https://doi.org/10.1037/tps0000176
SANSCARTIER, Matthew D.; EDGERTON, Jason D. & ROBERTS, Lance W. (2018). A latent class
analysis of gambling activity patterns in a Canadian university sample of emerging
adults: Socio-demographic, motivational, and mental health correlates. Journal of Gambling Studies, 34(3), 863-880. https://doi.org/10.1007/s10899-017-9734-y
38
SCHUCH, Felipe B.; VANCAMPFORT, Davy; FIRTH, Joseph; ROSENBAUM, Simon; WARD, Philip
B.; SILVA, Edson S.; HALLGREN, Mats; PONCE DE LEON, Antonio; DUNN, Andrea L.; DESLANDES,
Andrea C.; FLECK, Marcelo P.; CARVALHO, Andre F., & STUBBS, Brendon (2018). Physical
activity and incident depression: A meta-analysis of prospective cohort studies. American Journal of Psychiatry, 175(7), 631-648. https://doi.org/10.1176/appi.ajp.2018.17111194
39
TORQUATO, Sara C. R.; DA SILVA, Emilene F.; RODRIGUES, Aline R.; PIMENTA, Mayara V.
T.; MOURA, Jayne R. A. & DA SILVA, Ana Roberta V. (2016). Sedentary lifestyle and
metabolic disorders among university study / Sedentarismo e alterações metabólicas
entre universitários / Estilo de vida sedentario y trastornos metabólicos entre la
universidad. Revista de Enfermagem Da UFPI, 5(2), 16. https://doi.org/10.26694/reufpi.v5i2.5244
40
WING KWAN, Matthew Y.; BRAY, Steven R. & MARTIN GINIS, Kathleen A. (2009). Predicting
Physical Activity of First-Year University Students: An Application of the Theory
of Planned Behavior. Journal of American College Health, 58(1), 45-55. https://doi.org/10.3200/JACH.58.1.45-55
41
WORLD HEALTH ORGANIZATION. (2008). Advocacy, communication and social mobilization for TB control: a guide to developing
knowledge, attitude and practice surveys. https://apps.who.int/iris/handle/10665/43790
XU, Rixiang; MU, Tingyu; WANG, Guan; SHI, Jing; WANG, Xin & NI, Xiaoli (2019). Self-medication
with antibiotics among university students in LMIC: A systematic review and meta-analysis.
The Journal of Infection in Developing Countries, 13(08), 678-689. https://doi.org/10.3855/jidc.11359
Appendix
Adjusted odds ratios (OR) and 95% confidence intervals (CI) from multinomial logistic
regression model about class on sociodemographic variables and well-being
