Clustering of socio-demographic factors and their association with the mental health of Australian children and adolescents: A latent class analysis

doi:10.21203/rs.3.rs-1821998/v1

Download PDF

Research Article

Clustering of socio-demographic factors and their association with the mental health of Australian children and adolescents: A latent class analysis

https://doi.org/10.21203/rs.3.rs-1821998/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: Previous studies have shown a relationship between socioeconomic and demographic variables with the mental health of children and adolescents. However, while utilizing a nationally representative sample, no study has yet been conducted on a cluster-based connection or model-based cluster analysis of socio-economic and demographic characteristics with mental health. This study aims to identify the cluster of the items representing the socio-demographic characteristics of Australian children and adolescents aged 11-17 years by using latent class analysis and examining the associations with their mental health.

Methods: This study involved children and adolescents aged 11–17 (n = 3152) from the Young Minds Matter (YMM): The Second Australian Child and Adolescent Survey of Mental Health and Wellbeing. The YMM is a nationwide cross-sectional study in which data was obtained from children and adolescents, as well as from their parents/carers. Latent class analysis was performed based on nine socio-demographic factors related to Individual level (age, sex), Household level (regions, family blending, and parents living status in the household), and Socio-economic level (primary carer education, income, occupation, Index of relative socio-economic advantage and disadvantage [IRSAD]) variables. Bivariate logistic regressions were used to examine associations between identified classes and mental disorders.

Results: Based on various model selection criteria, this study identified five classes. Class 1 (working/poor class), 2 (lower middle class), 3 (mobile middle class), 4 (growing wealthy class) and 5 (wealthy class) represents 18.7%, 16.7%, 27.7%, 13.5% and 23.4% of the total sample respectively. Most of the households of class 1 and class 5 were from the underprivileged and most privileged groups respectively. Results of binary logistic regression model showed that, in unadjusted latent classes, children and adolescents from the wealthy class (OR: 0.2938, 95% CI: 0.2187–0.3947) were 71 percent less likely to suffer from mental illness than their working-class counterparts. Lower middle class (OR: 0.4455, 95% CI: 0.3170–0.6260), mobile middle class (OR: 0.3227, 95% CI: 0.2405–0.4329) and growing wealthy class (OR: 0.5079, 95% CI: 0.3520–0.7329) also had lower odds of mental disorder than working class. The majority of the members of the ‘growing wealthy class’ were from the socio-economically advantaged group than the lower and mobile middle class, but their family functioning (interactions and relationships within the family) was very poor. As a result, mental disorders of children and adolescents of that growing wealthy class were respectively 1.14 and 1.57 times higher compared to that of lower and mobile middle class.

Conclusions: Among five latent classes, adolescents with lower socio-economic status and poor family functioning were exposed to a higher risk of developing mental health problems. The findings suggest that the reduction of socioeconomic inequalities and improving family interactions might help to reduce the mental health problems of children and adolescents’.

Australia

Children and adolescents

Family functioning

Latent class analysis

Mental health

Socio-demographic factors

Mental disorders among adolescents have been identified as one of the major public health concerns worldwide [1] and these are among the significant predictors of suicide attempts. According to the World Health Organization (2013), 20 percent of adolescents globally may develop mental illness in any one year. For people, their families, and society as a whole, mental illness has a wide range of short and long-term negative consequences [2]. These problems are also viewed as a source of concern for the quality of life of adolescents. Children and adolescents are susceptible to mental illness and many mental disorders have direct and indirect consequences on children's health and academic performance [3, 4]. Adolescents’ numerous physical, psychological, and behavioral changes can lead to psychosocial and mental health issues, which can have major effects on their growth, productivity, and quality of life [5]. Furthermore, mental illnesses that arise throughout childhood and adolescence are well-known risk factors for mental disorders later in life [6–8].

Several studies have the way that, socio-demographic characteristics such as age, regional status, gender, parental education, parental employment status, and household income can predict mental health and aggressive behaviors in children and adolescents[9–11]. Heshmat et al. (2016) found that psychiatric issues and violent behavior were more common in children and adolescents with poor socioeconomic status [12]. A cross-sectional study in Indonesia found that adolescent mental health problems were associated with their educational level and area of residence [13]. Another study revealed that children and adolescents from low-income families were two to three times more likely to suffer mental health issues and this low socioeconomic status is substantially related to greater rates of mental health issues over time [14].

To determine the prevalence of mental disorders, Australian epidemiological research considered labor force status, industry sector, socio-economic index for areas (SEIFA), and income as socio-economic factors, and age and gender as demographic factors [15]. The research showed the relationship of individual socio-economic and demographic variables with mental health [15].

However, this study considered those factors independently of one another, meaning the influence of a second variable on the first one is ignored in the investigation. On the other hand, Park and Lee (2020) in their studies showed that family structure has a significant impact on adolescent health behavior, mental health, and perceived academic achievement [16]. Therefore, instead of analyzing each socio-demographic factor separately, in our study, we considered latent class analysis (LCA) which allows for a person-centered approach to understand how a wide range of socio-demographic factors may cluster together to reflect distinct social classes within a large and heterogeneous population.

The literature describes diverse classifications of the cluster analysis of influencing factors of health. Kino et al. (2017) identified different clusters of health-related behaviors and examined their relationships with the different indicators of socioeconomic variables[17]. A US study identified classes according to varying combinations of health lifestyles using latent class analysis and examined the associations between these classes and suicidal behaviors among US adolescents[18]. Similarly, Hoogstoel et al. used LCA to identify different profiles from a list of risky behaviors among adolescents in Mauritius and then determined the associations of suicidality among adolescents with these profiles [19]. Conry et al. (2011) examined how health behaviors clustered in a nationally representative sample of Irish adults and discovered the links between these clusters with mental health, self-rated health, and quality of life [20]. An Australian National University report revealed a greater stratification in Australian society. Using LCA the report showed that Australian society is best explained by five classes and these classes are based on respondents’ economic, social and cultural capital [21].

Research on the application of model-based cluster analysis on socio-demographic variables is very rare in Australia. According to our knowledge, only one social survey research document has used this clustering approach to determine the different social classes in Australia [21]. The variables selected covered three different capitals, economic capital (measured by respondents’ household income and the estimated value of the property they own), social capital (measured by the total number of occupations and the mean occupation status score of respondents’ social contacts) and cultural capital (measured as the number of ‘highbrow’ and ‘emerging’ cultural activities that respondents report having undertaken in the past year). However, some studies based on YMM data have documented the link between individual socio-demographic characteristics with the mental health of adolescents [22, 23]. But no study showed the cluster-based association or model-based cluster analysis of socio-demographic factors with mental health by using a nationally representative sample. To present a more comprehensive picture of the mental health of Australian adolescents, the current study identified the different classes by considering various socio-demographic variables. These variables were taken based on some literature which shows that these variables have a significant impact on the mental health of adolescents [12, 13, 15].

Therefore, the primary objective of this study is to identify the cluster of the items representing socio-demographic characteristics of Australian adolescents aged 11–17 years by using latent class analysis. The additional objective is to define each cluster’s characteristics and examine their differences to identify the most privileged and underprivileged clusters and then investigate the potential associations between identified clusters with adolescent mental disorders. Lastly, this study explores the link between adolescent mental health with class membership and adjusted controlling factors.

Data source and study design

Young Minds Matter (YMM): The Second Australian Child and Adolescent Survey of Mental Health and Wellbeing, conducted by the University of Western Australia (UWA) through the Telethon Kids Institute in collaboration with Roy Morgan Research and the Australian Government Department of Health during 2013–2014, was used to analyze cross-sectional data [24, 25]. Ethical approval has been obtained from the Human Research Ethics Committee of the University of Southern Queensland to conduct this research.

YMM used a multi-stage, area-based random sample technique that was representative of Australian households with children and adolescents aged 4–17 years. If the household had more than one qualified child, the sample included a single child at random. In total, 6310 parents of children and adolescents aged 4–17 years (55 percent of eligible households) voluntarily participated in the survey through a face-to-face interview using a structured questionnaire, and 2967 adolescents aged 11–17 years (89 percent of eligible youth) privately completed computer-based self-reported questionnaires to provide information on risk behaviors (e.g., suicidality, self-harm, substance use, bullying) and service use. The survey does not include the most rural places, homeless teenagers, adolescents in any type of institutional care, or homes where interviews could not be performed in English. The survey details and methodology used in the survey will be found in Hafekost et al. (2016) [25].

Measurements

Mental disorders

To maintain consistency with the first child and adolescent national survey and ensure maximum comparability among prevalence estimates, the DISC-IV scale was used [26]. DISC-IV diagnostic criteria are referred to in the Diagnostic and Statistical Manual of Mental Disorder (DSM-IV) formulated by the American-psychiatric association [26]. Worldwide DISC-IV remained the best tool for measuring the 12-month incidence of mental disorders in 2013. Primary carers completed the DISK-IV modules. Mental disorders included major depressive disorder, attention deficit hyperactivity disorder (ADHD), conduct disorder, and four types of anxiety disorders—social phobia, separation anxiety disorder, generalized anxiety disorder, and obsessive-compulsive disorder [27].

In this study, only parental data were used to create a binary variable to detect the presence of any mental disorder, as it only provided diagnostic information about each type of mental disorder in children and adolescents. Here we consider a variable ‘mental disorder’ in our analysis: whether the adolescent has had any of the following types of mental disorder- major depressive disorder, ADHD, conduct disorder, or anxiety disorders which includes social phobia, separation anxiety disorder, generalized anxiety disorder, and obsessive-compulsive disorder in the past 12 months. Responses included ‘Yes’ (if a child has at least one of these issues, the code is 1.) or ‘No’ (coded as 0 if otherwise).

Socio-demographic factors

In this study to identify different classes, our main variables of interest were socio-demographic factors of the adolescents and their corresponding households. Depending on income, regional status, employment, family history, and other variables, every country, and its people are categorized into numerous social classes or levels. While class distinctions are not completely evident or defined by any formal organization, they do exist in terms of how people live, conduct themselves, and spend money [21]. There are five main social classes in Australia (Figure 1) [21]. The following Figure 1 represents the five social classes of Australia with the mean characteristics of the different classes.

Each class is defined by its members' possession of economic, social, and cultural capital. There are also significant differences in the average ages, education, and occupational status of the members in each class. There is evidence of intergenerational mobility in occupational prestige across all groups, but particularly in the working and middle classes. This is most obvious in the two middle classes (mobile middle class and established middle class), where members' prestige scores are more than their parents' [21].

For this study, socio-demographic variables were represented by age (11-14 years vs. 15-17 years), Sex (male vs. female), income (less than $52000 as low, $52000–$129999 as medium, and more than $130000 as high), regions (metropolitan vs. non-metropolitan), primary carer education (bachelor, diploma, and year-12/below), primary carer occupation (employed vs unemployed), index of relative socio-economic advantage and disadvantage (IRSAD) quintile [lowest (most disadvantaged), second, third, fourth and highest (most advantaged)] of the parents, family blending (intact family vs other families) and both parents living status in the household, that is whether both parents live in the household or not (yes vs no). The last two variables indicate the family functioning of a household. In 2015, a study found that adolescents from non-intact households had a lower perception of family functioning, father’s and mother’s behavioral control, paternal psychological control, and parent-child relational traits than adolescents from intact households [28]. All of these socio-demographic variables were taken from three different levels. Figure 2 represents the framework for selecting these variables to determine the significant number of classes by using latent class analysis.

Study participants and data analysis

To determine the appropriate number of classes and to examine their probable associations with adolescent mental disorders data were taken from parent data on the DISC-IV. Age is an important factor in the mental health of children and adolescents, and children and adolescents aged more than 11 have a higher risk of mental health disorders [22]. A study by the Telethon Kids Institute found that children and adolescents belonging to the age group 12-17 years are almost three times more likely than the age group 4-11 years to experience a severe mental disorder [29]. Taking these into account, in our study, the sample is restricted to children and adolescents aged 11–17 years. In parent data, there are 6310 parents/carers involved with children aged 4–17 years. Since our study sample was restricted to adolescents 11–17 years and all the ‘don’t know’ responses were omitted therefore study participants were 3152 parents/carers.

Descriptive analysis

Descriptive analyses were carried out taking into account the variables that were used to find the latent classes. All variables used in this analysis are categorical. Proportions for each of the categorical variables are calculated using R. Bivariate analyses are conducted to examine the distribution and association of the socio-demographic characteristics with the mental health of the adolescents. Chi-square tests of significance are used to describe and compare the sample characteristics of adolescents with mental disorders.

Latent Class Analysis (LCA)

The LCA is a cluster analysis method that uses a probability model [30–32]. LCA, like all cluster analysis methods, finds homogeneous clusters of data from varied demographic and socioeconomic data by maximizing similarity within the cluster while minimizing similarity across cluster elements. LCA assumes the data is from a mixed model with several probability distributions[33]. It presumes that the data is divided into mutually exclusive homogeneous subgroups by a latent variable. LCA is said to have numerous distinct advantages over traditional cluster analysis approaches [31, 34]. Different statistical criteria are available in the LCA output, which can be used to identify the most appropriate number of clusters; and different types of variables (for instance, counts, continuous, categorical, nominal) can be used in LCA directly without any further standardization process [33].

In this study, using R (version 4.1.3) a latent class analysis was performed to identify and describe the classes associated with adolescents’ mental health. To determine the exact number of latent classes, an empirical approach was used [18]. Starting from a two-class model, the analysis was carried out several times consecutively, increasing the number of subgroups by turn and replicating each of the models 5 to 10 times for greater precision. Then, statistical model adjustment indices, such as the Bayesian information criterion (BIC), the Akaike information criterion (AIC) [35], the Likelihood ratio/deviance statistic [G^2], and Chi-square goodness of fit were used to identify the final number of classes. Indeed, the selected classes had to have enough observations to provide a representative class of a population [22]. In practice, subgroups with a size of less than 5% were not retained, as in similar studies [18]. The clusters were named in a way that best represented the most notable discoveries in the data. While naming the clusters made it easier to communicate them to the audiences [36], it is difficult to capture the level of difference across clusters with labels, according to the argument. The best potential name for each cluster's identifying traits was determined. The clusters were not meant to be displayed in a linear fashion.

Bivariate analyses and binary logistic regression

After determining the necessary number of latent classes, a series of cross-tabulations and bivariate analyses (using Chi-square tests) were used to investigate the distribution of and relationships between adolescents’ mental disorders. Two binary logistic regression models were fitted to observe the association between class membership and the mental health status of children and adolescents. The first model is considered without controlling any covariate and thus gives an unadjusted effect and the second model is based on considering the controlling variables; ‘household size’, ‘household type’, ‘inability to pay gas, electricity, or telephone bills on time’, ‘could not pay the mortgage or rent on time’, ‘adults/children have gone without meals’, and ‘households receive a carer benefit or pension in relation to child’ and this gives adjusted effects.

Socio-demographic characteristics of adolescents

The Socio-demographic characteristics of the study participants are presented in Table 1. According to the age group, 55.8% of the total samples were between ≤15 to 17 years old and the mean age was 14.6. The male/female distribution was equitable, with 51.9% males and 48.1% females. Regional status shows that 62.7% of adolescents live in the metropolitan area and 63.3% comes from intact families. Moreover, the study results showed that 68.1% of parents of adolescents live in households. In socio-economic level variables, around fifty percent (46.3%) of household belongs to the medium income group ($52000–$129999), 29.6% fall in the high-income group, and 24.1% in the low-income group. In parental education from the socio-economic level, 31% of parents have the highest level of education (bachelor), around 37% have a diploma and 32.1% have the lowest level of education which is year 12/below. Most of the parents of the adolescents are employed (75.2%) and 24.8% are unemployed. Sample data also showed that 23.6% of household belongs to the highest level in the IRSAD quintile and 15.6% is in the lowest level. 63.5% of adolescents in this study population have mental disorders.

Table 1 Socio-demographic characteristics of the study participants

Characteristics			Frequency	%
Individual-level	Age (mean ± SD:14.6 ± 2.04)	11 to <15 years	1393	44.2
	Age (mean ± SD:14.6 ± 2.04)	≤15 to17 years	1459	55.8
	Sex	Male	1636	51.9
	Sex	Female	1516	48.1
Household-level	Regions ^a	Metropolitan	1977	62.7
	Regions ^a	Non-metropolitan	1175	37.3
	Family blending ^b	Intact family	1996	63.3
	Family blending ^b	Other families	1156	36.7
	Both parents live in the household	Yes	2147	68.1
	Both parents live in the household	No	1005	31.9
Socio-economic level	Household income/year	Low	760	24.1
		Middle	1460	46.3
		High	932	29.6
	Parental education	Bachelor	976	31.0
		Diploma	1165	36.9
		Year 12/below	1011	32.1
	Parental employment	Employed	2370	75.2
	Parental employment	Unemployed	782	24.8
	IRSAD quintile ^c	Lowest	493	15.6
		Second	557	17.7
		Third	666	21.1
		Fourth	691	21.9
		Highest	745	23.6

^aRegions: Regional status indicates whether adolescents come from the metropolitan area or non-metropolitan area.

^bFamily blending: Intact families refer to those adolescents who come from families in which their parents are married and live together. Other families include step, blended, and single-parent families.

^cIRSAD: “The Index of Relative Socio-economic Advantage and Disadvantage (IRSAD): “It presents statistics on people's and households' economic and social circumstances in a given area, including relative advantage and disadvantage indicators. A low score suggests a relative lack of advantage, whereas a high score shows a relative lack of disadvantage and greater advantage.

Distribution of socio-demographic characteristics with adolescents’ mental health (measured by mental disorders)

Accompanying Table 2 represents the association of socio-demographic characteristics of the study participants with the adolescents’ mental health. The study result reveals that there is a significant association between the sex of the study child with mental disorders. All the socio-economic level variables considered in this study are significantly associated with adolescents’ mental disorders. In household-level variables, family blending and both parents’ living in the household are significantly associated with adolescents’ mental disorders. The regional status and mental disorders are statistically associated with a 10% level of significance.

Table 2 Association of socio-demographic characteristics with mental disorders of adolescents.

Socio-demographic characteristics Mental disorder ^a			Total	p-value
	yes n (%)	no n (%)	n (%)
Individual-level
Age
11 to <15 years	883(63.39)	510(36.61)	1393(44.19)	0.882
≤15 to17 years	1120(63.67)	639(36.33)	1759(55.81)	0.882
Sex
Male	1005(61.73)	631(38.57)	1636(51.90)	0.011^*
Female	998(65.83)	518(34.17)	1516(48.10)	0.011^*
Household-level
Regions
Metropolitan	1232(62.32)	745(37.68)	1977(62.72)	0.056
Non-metropolitan	771(65.62)	404(34.38)	1175(37.28)	0.056
Family blending
Intact family	1164(58.32)	832(41.68)	1996(63.32)	0.000^***
Other families	839(72.58)	317(27.42)	1156(36.68)	0.000^***
Both parents live in the household
Yes	1251(58.27)	896(41.73)	2147(68.12)	0.000^***
No	752(74.83)	253(25.17)	1005(31.88)	0.000^***
Socio-economic level
Household income
Low	549(72.24)	211(27.76)	760(24.11)	0.000^***
Medium	922(63.15)	538(36.85)	1460(46.32)
High	532(57.08)	400(42.92)	932(29.57)
Parent’s educational status
Bachelor	578(59.22))	398(40.78)	976(30.96)	0.000^***
Diploma	735(63.09)	430(36.91)	1165(36.96)
year 12/below	690(68.25)	321(31.75)	1011(32.07)
Parent’s occupational status
Employed	1446(61.01)	924(38.99)	2370(75.19)	0.000^***
Unemployed	557(71.23)	225(28.77)	782(24.81)
IRSAD
Lowest (most disadvantages)	366(74.24)	127(25.76)	493(15.64)	0.000^***
second	378(67.86)	179(32.13	557(17.67)
third	399(59.91)	267(40.09)	666(21.13)
fourth	427(61.79)	264(38.21)	691(21.92)
highest (most advantages)	433(58.12)	312(41.90)	745(23.64)

^a The number of children with any of the following mental disorders: attention deficit hyperactivity disorder (ADHD), major depressive disorder (MDD), anxiety disorder (general anxiety/separation anxiety/OCD/Social phobia), or conduct disorder.

Identification of latent classes and associated characteristics

Model selection by latent class analysis

The findings of the latent class analysis are presented in Table3. The BIC and G^2 values suggest a minimum for a five-class distribution. Thus, the 5-class model was selected as the best fitting model based on the entire set of data, even though the AIC and Pearson values decreased as the number of classes increased.

Table 3 Information for determining the latent class model and the fit statistic.

No. of Classes	AIC	BIC	G^2		Estimated class population shares
No. of Classes	AIC	BIC	G^2		class 1	class 2	class 3	class 4	class 5	class 6
1	43801.8	43880.6	6153.7	15355.5	100%
2	40328.2	40491.7	2652.0	3752.2	33.8%	66.2%
3	39722.3	39970.6	2018.1	2655.1	27.7%	39.3%	33.0%
4	39248.9	39582.0	1516.7	1730.9	13.4%	27.0%	19.6%	40.0%
5	39073.7	39491.6	1313.5	1506.0	18.7%	16.7%	27.7%	13.5%	23.4%
6	39007.0	39509.7	1318.8	1430.2	16.8%	10.8%	27.6%	8.9%	23.3%	12.6%

Note: AIC-Akaike Information Criterion; BIC-Bayesian Information Criterion; G^2 [Likelihood ratio/deviance statistic]; [ Chi-square goodness of fit]. The identified model represents by a bolded row.

Characteristics of the identified laten classes

Associated Table 4 represents the characteristics of each class. Among the 5 classes, adolescents who belong to the most privileged group are classified as class 5 and named as “wealthy class”, demonstrating 23.4% which is 738 of the total sample. In this wealthy class, more than 70% of household belongs to the high-income group (against 3.6% in class1, 6.0% in class 2, 28.7% in class 3, and 22.9% in class 4) and around 87% of them live in the metropolitan area. More than 59% of the parents have the highest level of education and about 85% of the parents of this class are employed. Most of the parents of the adolescents (around 98%) in this cluster live in households and about 95% of the adolescents come from intact families. Another important characteristic of this class is that about 64.4% of households belong to the most advantaged level of the variable ‘IRSAD’ quintile and there is no one in this group belonging to the most disadvantaged level.

On the other hand, adolescents who belong to the underprivileged group are classified as class 1 and named as ‘poor/working class’, representing 18.7% (n=589) of the total sample. In this class, about 55% of adolescents live in the non-metropolitan area and around 77% of the household belongs to the low-income group. More than 50% (53.3%) of parents of this class are unemployed and the same percentage (50%) of parental educational status belongs to the lowest level which is year12/below. The most important characteristic of this cluster is that only 4% of the household belongs to the most advantaged group and around 40% of the household belongs to the most disadvantaged group in the IRSAD quintile. In this class, only 6% of cases, both parents of the adolescents live in the household and fewer than 2% of adolescents come from intact families.

Class 2 is defined by those adolescents who belong to a group, which is better than the underprivileged group and named as ‘lower middle class’. This class represents 16.7% (n=526) of the total sample, where only 6% of household belongs to the high-income group and about 24% of household belongs to the most disadvantaged group in the IRSAD quintile. The parental education and employment status of this class contains the worst level compared to the other four classes.

Table 4 Characteristics of class obtained in adolescents aged 11-17 in Australia.

		Class 1 (%)	Class 2 (%)	Class 3 (%)	Class 4 (%)	Class 5 (%)
	Total proportion	18.7	16.7	27.7	13.5	23.4
	n (total=3152)	589	526	873	426	738
Socio-demographic characteristics
Individual-level	Age
	11 to <15 years	46.76	46.12	44.96	37.52	43.72
	≤15 to17 years	53.24	53.88	55.04	62.48	56.28
	Sex
	Male	53.10	50.60	52.67	49.13	52.55
	Female	46.90	49.40	47.33	50.87	47.45
Household-level	Remoteness
	Metropolitan	44.43	63.85	49.83	71.03	86.99
	Non-metropolitan	55.57	36.15	50.17	28.97	13.01
	Family blending
	Intact family	1.72	93.39	91.05	000	94.92
	Other families	98.28	6.61	8.95	100	5.08
	Both parents living in the household
	Yes	5.77	98.91	99.5	000	98.24
	No	94.23	1.09	0.5	100	1.76
Socio-economic level	Household income/year
	Low	77.33	34.78	5.97	13.71	1.47
	Medium	19.04	59.17	65.35	63.31	26.61
	High	3.63	6.06	28.68	22.98	71.92
	Parental education
	Bachelor	5.35	21.00	24.86	42.41	59.14
	Diploma	44.37	23.08	48.92	41.42	24.20
	Year 12/below	50.28	55.92	26.22	16.17	16.65
	Parental employment
	Employed	46.49	37.49	100	93.63	84.96
	Unemployed	53.51	62.51	000	6.37	15.04
	IRSAD
	Lowest(most disadvantages)	39.56	23.90	12.09	6.71	000
	Second	25.46	23.77	25.41	14.13	000
	Third	20.51	27.87	24.43	22.85	11.92
	Fourth	9.97	17.28	29.55	25.48	23.69
	Highest (most advantages)	4.50	7.17	8.52	30.84	64.39

Class 3 and Class 4 were characterized by those individuals who belong to moderate privileged and privileged groups and named as ‘mobile middle class’ and ‘growing wealthy class’ respectively. Mobile middle class [class 3] and growing wealthy class [class 4] consists of 27.7% (n=873) and 13.5% (426) of the total sample correspondingly. More than 70% of the household of the growing wealthy class lives in the metropolitan area and this proportion was reduced to 50% in the case of the mobile middle class. Although the percentage of high and middle-income levels is higher for the mobile middle class compared to the growing wealthy class, but in the IRSAD quintile about 31% of the household of the growing wealthy class belongs to the most advantaged level, while in the mobile middle class, only 8% household belong to this most advantages level. More than 42% of the parents of the growing wealthy class have the highest level of education and about 94% of the parents of this class are employed. In the mobile middle class, around 50% of parents have a medium level of education (diploma) and 100% of them are employed. One of the important demerits of the growing wealthy class is that the family functioning of this class is worst. Study results show that no adolescents in this class come from intact families and at least one or no parents of the adolescents’ lives in the household.

Regression modeling

In this study the binary logistic regression model was used which was conducted in two steps. In the first step, we explored the association of unadjusted class membership with mental disorders of adolescents. In the second step, we have determined the association of adjusted class membership with mental disorders of adolescents where six socio-demographic covariates were used as control variables. Table 5 presents the results of binary logistic regressions examining associations between identified classes and mental disorders, adjusting for covariates. Children and adolescents belonging to wealthy class are about 71% less likely to be suffering from mental illness compared to their working-class counterparts (OR: 0.2938, 95% CI: 0.2187–0.3947 for wealthy class). Children and adolescents in lower middle class and mobile middle class, also showed lower odds of mental disorders compared to respondents belonging to working class (OR: 0.4455, 95% CI: 0.3170–0.6260 for lower middle class, OR: 0.3227, 95% CI: 0.2405–0.4329 for mobile middle class). Although the growing wealthy class reported a lower chance of occurring mental disorders (OR: 0.5079, 95 percent CI: 0.3520–0.7329) than in the poor/working class, mental disorders of children and adolescents belonging to the growing wealthy class are respectively as 1.14 and 1.57 times higher compared to their counterparts belonging to lower middle class and mobile middle class.

Table 5 Binary logistic regression of mental disorder on class membership (unadjusted and adjusted) and control variables.

Laten classes (ref. poor/working class	Unadjusted		Adjusted^b
Laten classes (ref. poor/working class	Coefficient	OR (95% CI of OR)	Coefficient	OR (95% CI of OR)
Lower middle class	-0.8087^***	0.4455 (0.3170-0.6260)	-0.6608^***	0.5164 (0.3835-0.6954)
Mobile Middle class	-1.1311^***	0.3227 (0.2405-0.4329)	-0.8722^***	0.4180 (0.3241-0.5391)
Growing wealthy class	-0.6774^***	0.5079 (0.3520-0.7329)	-0.5624^***	0.5698 (0.4234-0.7668
Wealthy class	-1.2249^***	0.2938 (0.2187-0.3947)	-1.0272^***	0.3580 (0.2738-0.4680)

Note: OR-Odds ratio; CI-Confidence interval; *: p<0.05; **: p<0.01; ***: p<0.001

^bControl variables: Household size”, “Household type”, “Have not been able to pay gas, electricity, or telephone bills on time”, “Could not pay the mortgage or rent on time”, “Adults/child have gone without meals”, and “Household receive a carer benefit or pension in relation to [child]”

In addition, a study relating to controlling six variables revealed that children and adolescents in lower middle class (OR: 0.5164, 95% CI: 0.3835–0.6954), mobile middle class (OR: 0.418, 95% CI: 0.3241–0.5391), growing wealthy class (OR: 0.5698, 95% CI: 0.4234–0.7668) and wealthy class (OR: 0.358, 95% CI: 0.2738–0.4680) have lower odds of mental disorder compared to the cohort belonging to poor/working class. Comparing both unadjusted and adjusted effects, the binary logistic regression model showed almost similar results which means children and adolescents belonging to the working class are the worst cohort in respect of mental illness.

The major goal of this study is to use latent class analysis to find the classes of items representing socio-demographic characteristics of Australian children and adolescents. First, this study considered adolescents aged from 11 to 17 years where more than fifty percent of adolescents’ are aged between ≤ 15 to17 years. As mentioned in a previous study [36], our results show a fair distribution of males and females, and this is significantly associated with adolescents’ mental disorders. Family blending and parental living status in the household were significantly associated (p-value < 0.001) with the mental health of adolescents. The current study discovered that adolescents who belong to a family with good family functioning suffer less from mental disorders. A study conducted in Hong Kong discovered that adolescents from non-intact households had lower perceptions of family functioning, paternal and maternal behavioral control, paternal psychological control, and parent-child relational traits than adolescents from intact households [28], and these figures support our study results. Parental education is significantly associated with the mental disorders of adolescents. These figures show that if parental education belongs to the highest level (bachelor’s degree: 30.96%), then mental disorders in adolescents of those fathers are less frequent which is consistent with multiple earlier previous findings [13, 23]. A German study by Franziska Reiss et.al (2019) also found that, in a challenging life situation, children and adolescents with more educated parents had fewer mental health issues [37]. Current research revealed that the adolescents whose parents are employed are suffer less from mental health issues than their counterparts whose parents are unemployed. Household income and the IRSAD quintile are significantly associated with adolescents’ mental health. Adolescents who belong to high-income and the highly advantaged group are suffer less from mental health. This result is also consistent with preceding studies [13, 15].

Second, this study identified five distinct social classes using LCA based on nine socio-demographic factors of Australian adolescents aged 11–17 years. The Poor/working class represents the characteristics, of the lowest level of income, education, and occupation. Most of the household of this class belongs to the most disadvantaged level in the IRSAD quintile. An Australian study invented five social classes using LCA based on some socio-demographic characteristics [21]. Of these 5 classes, the lowest class is the established working class bearing the main characteristics lowest incomes, lowest occupational prestige, and lowest educational attainment of the five classes. The characteristics of this identified class were consistent with our study findings. In the lower middle class, the second-highest percentage of adolescents belong to the most disadvantaged level in the IRSAD quintile and more than fifty percent of parents in this class have the lowest level of education. Out of five classes, the mobile middle class is the largest class which contains the maximum number of samples. Australians in the mobile middle class, consisting of twenty-five percent of the population, have higher levels of educational qualifications and consequently have higher household income [21]. But in our study, the educational attainments of the parents are not higher in this class, as only 24% of the parents have the highest level of education. In this class, most of the yearly income of the households belongs to the medium level of the income category. The growing wealthy class shares many of the characteristics of the mobile middle class, with three key differences: they are- family blending, both parents living in household status, and the IRSAD quintile. The main disadvantage of this class is that all the children and adolescents came from non-intact families, which indicates relatively poor family functioning. Class 5 is the most privileged class represented by the wealthy class. This class possesses the highest level of income, education, and good family functioning. Most of the households of this class belong to the most advantaged level in the ‘IRSAD-quintile’. In our study, we found that the working/poor class and growing wealthy class have very poor family functioning, and the other three classes possess good family functioning which has a positive impact on the mental health of children and adolescents [28, 37]. These findings indicate that for the better mental health of children and adolescents were needed to improve not only the socioeconomic status of the parents but also their family structure.

Third, the binary logistic regression model is used to determine the association of latent classes (unadjusted and adjusted) with the mental health of children and adolescents. In unadjusted latent classes, children and adolescents from the wealthy class were 71 percent less likely to suffer from mental illness than their working-class counterparts. The lower middle class, mobile middle class, and growing wealthy class also showed lower odds of mental disorders than the poor/working class. Adolescents belonging to these four classes (classes 2, 3, 4 and 5) were from socio-economically advantaged groups compared to the adolescents belonging to the poor/working class. These findings are consistent with previous research which revealed that adolescents from the socio-economically disadvantaged group were two to three times more likely to suffer mental health issues than their counterparts who were from the socio-economically advantaged groups [14]. In their studies, Fryers et.al (2003) found that the link between socioeconomic discrepancy and common mental health issues can be seen not only in adults but also in children and adolescence [38].

In our study, we found that more than 90% of children and adolescents of lower middle class, mobile middle class and wealthy class were from intact families, and in all of these three classes most of the situations, parents of the adolescents live in the household, which indicates good family functioning. Daniel et al. (2015) revealed that adolescents in intact families evaluated better family functioning and parental–child relationship qualities than adolescents in non-intact families [28]. In the present study, most of the children and adolescents of poor/working class and growing wealthy class were from non-intact families, and their parents did not live in the household, which indicates poor family functioning. As a result, even though members of the growing wealthy class are from the socioeconomically advantaged group, mental disorders among children and adolescents in this class are respectively 1.14 and 1.57 times higher than those in the lower middle class and mobile middle class. This result indicates that, in addition to better socio-economic status, good family functioning has a positive impact on children’s and adolescents’ mental health. Moreover, controlling some socio-demographic variables, study results showed similar results that children and adolescents from the working class are the most vulnerable.

Strengths and limitations

As per our knowledge, this study is the first to explore the clustering of socio-demographic factors which have a significant impact on the mental health of children and adolescents in Australia. It used latent class analysis on YMM data, a robust approach to identify adolescent classes over a wide range of samples. This study has a few limitations that need to be mentioned. First, this study explored the association of adolescents’ mental health (indicated by mental disorder) with latent classes, but we did not show the relationship of latent classes with different kinds of mental disorders in adolescents. ‘Don’t know responses’ were omitted in our study, which may have a significant impact on the outcome of the study. Lastly, the analysis was based on a cross-sectional study and self-reported responses, where causal relationships cannot be established through a cross-sectional study, and self-reported responses compromise with the reliability and validity of measurement.

The present study identified five classes based on socio-demographic characteristics and highlighted the association of mental health with these clusters of Australian children and adolescents. The poor/working class presented the worst class carrying characteristics of the lowest income, education, occupation, and poor family functioning. Among the five clusters, the poor/working class showed higher odds of mental disorder in children and adolescents than the other four clusters (lower middle class, mobile middle class, growing wealthy class, and wealthy class) for both unadjusted and adjusted (adjusting some of the covariates) models. On the other hand, due to the lack of good family functioning, the growing wealthy class has higher odds of mental disorders in children and adolescents than the lower and mobile middle class, despite the members of that clusters are from socioeconomically advantaged groups. The associations between these clusters and mental health highlight the importance of screening the patterns of socio-demographic factors as a promising step to detect and prevent mental disorders among at-risk children and adolescents more effectively.

Acknowledgments

The authors would like to thank the Telethon Kids Institute, The University of Western Australia, Roy Morgan Research, the Australian Government Department of Health for conducting the survey, and the Australian Data Archive for providing the YMM data set. The authors also would like to like to express their gratitude to Dr Barbara Harmes for proofreading the manuscript before submission.

Authors’ contributions

NA: Conception, Design of the work, Data acquisition, curation and analysis, Writing- Original draft, Writing- Reviewing and Editing.

EK: Supervision, Writing- Reviewing and Editing.

KA: Supervision, Writing- Reviewing and Editing.

The author(s) read and approved the final manuscript.

Funding

No specific grant was given to this research from any funding organizations in the public, private, or nonprofit sectors.

Availability of data and materials:

There are few restrictions on the use of this data, and the approval of the data application is depending upon a signed confidentiality agreement. Those interested in accessing this data should contact the Young Minds Matter: The second Australian Child and Adolescent Survey of Mental Health and Wellbeing Dataverse through email or lodge an online application from the following web link: http://www.youngmindsmatter.org.au/information/for-researchers/

Ethics approval and consent to participate

This study is approved by Human Research Ethics of University of Southern Queensland and The Australian Data Archive. All methods were carried out in accordance with relevant guidelines and regulations. As the nature of the study is retrospective Human Research Ethics of University of Southern Queensland and The Australian Data Archive have waived informed consent statement for the study.

Consent for publication

Not applicable

Competing interests

The authors state that they have no conflicts of interests

WHO - The Global Burden of Disease. The global burden of disease 2004. Updat World Heal Organ. 2004;:146.
Davies SC. Annual Report of the Chief Medical Officer. Public Ment Heal priorities Invest Evid. 2013;:259.
Busch SH, Barry CL. Marketwatch - Mental health disorders in childhood: Assessing the burden on families. Health Aff. 2007;26:1088–95.
DeSocio J, Hootman J. Children’s mental health and school success. The Journal of school nursing : the official publication of the National Association of School Nurses. 2004;20:189–96.
Sánchez-García M de los Á, Lucas-Molina B, Fonseca-Pedrero E, Pérez-Albéniz A, Paino M. Emotional and behavioral difficulties in adolescence: Relationship with emotional well-being, affect, and academic performance. An Psicol. 2018;34:482–9.
Fryers T, Brugha T. Childhood Determinants of Adult Psychiatric Disorder. Clin Pract Epidemiol Ment Heal. 2013;9:1–50.
Kessler RC, Davis CG, Kendler KS. Childhood adversity and adult psychiatric disorder in the US National Comorbidity Survey. Psychol Med. 1997;27:1101–19.
Macmillan HL, Fleming JE, Streiner DL, Lin E, Boyle MH, Jamieson E, et al. Childhood abuse and lifetime psychopathology in a community sample. Am J Psychiatry. 2001;158:1878–83.
Magklara K, Bellos S, Niakas D, Stylianidis S, Kolaitis G, Mavreas V, et al. Depression in late adolescence: A cross-sectional study in senior high schools in Greece. BMC Psychiatry. 2015;15:1–11.
Robert E, Catherine R, Richard Y. Ethnocultural differences in prevalence of adolescent depression Reproduced with permission of the copyright owner . Further reproduction prohibited without permission . 1997.
Basch CE. Aggression and Violence and the Achievement Gap Among Urban Minority Youth. J Sch Health. 2011;81:619–25.
Heshmat R, Qorbani M, Ghoreshi B, Djalalinia S, Tabatabaie OR, Safiri S, et al. Association of socioeconomic status with psychiatric problems and violent behaviours in a nationally representative sample of Iranian children and adolescents: The CASPIAN-IV study. BMJ Open. 2016;6:1–8.
Pandia V, Noviandhari A, Amelia I, Hidayat GH, Fadlyana E, Dhamayanti M. Association of Mental Health Problems and Socio-Demographic Factors Among Adolescents in Indonesia. Glob Pediatr Heal. 2021;8.
Reiss F. Socioeconomic inequalities and mental health problems in children and adolescents: A systematic review. Soc Sci Med. 2013;90:24–31.
Hashmi R, Alam K, Gow J, March S. Prevalence of Mental Disorders by Socioeconomic Status in Australia: A Cross-Sectional Epidemiological Study. Am J Heal Promot. 2021;35:533–42.
Park H, Lee KS. The association of family structure with health behavior, mental health, and perceived academic achievement among adolescents: A 2018 Korean nationally representative survey. BMC Public Health. 2020;20:1–10.
Kino S, Bernabé E, Sabbah W. Socioeconomic inequality in clusters of health-related behaviours in Europe: latent class analysis of a cross-sectional European survey. BMC Public Health. 2017;17:1–8.
Xiao Y, Romanelli M, Lindsey MA. A latent class analysis of health lifestyles and suicidal behaviors among US adolescents. J Affect Disord. 2019;255 May 2019:116–26.
Hoogstoel F, Samadoulougou S, Lorant V, Kirakoya‐samadoulougou F. A latent class analysis of health lifestyles in relation to suicidality among adolescents in Mauritius. Int J Environ Res Public Health. 2021;18.
Conry MC, Morgan K, Curry P, McGee H, Harrington J, Ward M, et al. The clustering of health behaviours in Ireland and their relationship with mental health, self-rated health and quality of life. BMC Public Health. 2011;11.
SOCIAL CLASS IN AUSTRALIA BEYOND THE ‘ WORKING ’ AND ‘ MIDDLE ’ CLASSES ANUPOLL SEPTEMBER 2015 ANU College of Arts & Social. 2015; September.
Vu XB (Benjamin), Biswas RK, Khanam R, Rahman M. Mental health service use in Australia: The role of family structure and socio-economic status. Child Youth Serv Rev. 2018;93 August:378–89.
Irteja Islam M, Khanam R, Kabir E. The use of mental health services by Australian adolescents with mental disorders and suicidality: Findings from a nationwide cross-sectional survey. PLoS One. 2020;15:1–17.
Lawrence D, Hafekost J, Johnson SE, Saw S, Buckingham WJ, Sawyer MG, et al. Key findings from the second Australian Child and Adolescent Survey of Mental Health and Wellbeing. Aust N Z J Psychiatry. 2016;50:876–86.
Hafekost J, Lawrence D, Boterhoven De Haan K, Johnson SE, Saw S, Buckingham WJ, et al. Methodology of Young Minds Matter: The second Australian Child and Adolescent Survey of Mental Health and Wellbeing. Aust N Z J Psychiatry. 2016;50:866–75.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th edition. Washington, DC; 2000.
Lawrence D, Johnson S, Boterhoven De Haan K, Ainley J. The Mental Health of Children and Adolescents: Report on the second Australian Child and Adolescent Survey of Mental Health and Wellbeing The physical health of people with mental illness View project Child Health CheckPoint View project. 2015.
Shek DTL, Xie Q, Lin L. The impact of family intactness on family functioning, parental control, and parent-child relational qualities in a Chinese context. Front Pediatr. 2015;2 JAN:1–7.
TELETHON KIDS INSTITUTE. Anxiety and depression.
Depaire B, Wets G, Vanhoof K. Traffic accident segmentation by means of latent class clustering. Accid Anal Prev. 2008;40:1257–66.
Vermunt, J.K., Magidson J. Applied Latent Class Analysis. Cambridge University Press; 2002.
Collins, L.M., Lanza ST. Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral, and Health Sciences. Newyork: Wiley; 2010.
Mohamed MG, Saunier N, Miranda-Moreno LF, Ukkusuri S V. A clustering regression approach: A comprehensive injury severity analysis of pedestrian-vehicle crashes in New York, US and Montreal, Canada. Saf Sci. 2013;54:27–37.
Hair Jr., J.F., Anderson, R.E., Tatham, R.L., Black WC. Multivariate Data Analysis. 5th edition. New york: Macmillan Publishing Company; 1998.
Nicola Reavley, George C. Patton, Susan M. Sawyer, Elissa Kennedy PA. Health and Disease in Adolescence. 3rd edition. 2017.
Reedy J, Wirfält E, Flood A, Mitrou PN, Krebs-Smith SM, Kipnis V, et al. Comparing 3 dietary pattern methods-cluster analysis, factor analysis, and index analysis-with colorectal cancer risk. Am J Epidemiol. 2010;171:479–87.
Reiss F, Meyrose AK, Otto C, Lampert T, Klasen F, Ravens-Sieberer U. Socioeconomic status, stressful life situations and mental health problems in children and adolescents: Results of the German BELLA cohort-study. PLoS One. 2019;14:1–16.
Fryers T, Melzer D, Jenkins R. Social inequalities and the common mental disorders - A systematic review of the evidence. Soc Psychiatry Psychiatr Epidemiol. 2003;38:229–37.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Clustering of socio-demographic factors and their association with the mental health of Australian children and adolescents: A latent class analysis

Status:

Version 1

Abstract

Figures

Background

Methods

Results

Discussion

Strengths and limitations

Conclusions

Declarations

References

Additional Declarations

Status:

Version 1