Participants
The Australian Rural Mental Health Study (ARMHS) is an Australian longitudinal population study examining the determinants of mental wellness in rural and remote communities. Data was collected in 2006-2009 from adults living in New South Wales (NSW) excluding major metropolitan zones (e.g. Sydney and Newcastle) (Kelly et al., 2010). All invited participants were over 18 years old and randomly selected from the Australian Electoral Roll (a list of all adult Australian residents who are registered to vote). A random number generator was used to identify potential participants. Those who were selected by this random number generator received a phone call to inform them about the study, and ask their interest in participating. A household sampling frame was used, whereby people who were selected for participation were also mailed a survey for each other adult residing in their house; they were also mailed a survey for each child in the house, to be completed by the parent or guardian. People in special dwellings such as hospitals and prisons, those without an identifiable telephone number, non-English speaking members of a household, and those with hearing impairments that made obtaining phone consent difficult were excluded. Participants over 65 years were briefly screened for cognitive status using the Telephone Interview for Cognitive Status (TICS-M) and those with a score of <17 were excluded.
Information about the study and surveys were mailed, with up to five follow up telephone contacts. The data was collected at baseline with follow up at 3 years for children and the parent who completed the child measures Participant numbers in the current analysis are provided in Figure 1. The project was approved by the Human Research Ethics Committees of participating institutions.
Instruments
Current Child Psychological Wellness
The Strengths and Difficulties Questionnaire -Parent Report (SDQ, 39) is a 25-item questionnaire for parents or carers to report on how they perceive their child’s level of functioning as a mechanism to assess psychological wellness and distress in children. There are versions for children aged 4-10 and 11-17 years old. The SDQ has five subscales: emotional symptoms, conduct problems, hyperactivity-inattention, peer problems and prosocial behaviour. Each item is scored 0, 1 or 2, with somewhat true always scoring 1 and not true and certainly true scoring 0 or 2 depending if the item is a strength or difficulty. The absence of prosocial behaviour is indicated by lower scores whereas higher scores on the other domains indicate increased problems. The SDQ also calculates a Total Problem score which includes four of the domains above, excluding prosocial behaviour. The maximum Total Problem score is 40, with a higher score indicating increased problems. The SDQ is psychometrically sound with good internal consistency (Cronbach α: 0.73) and re-test reliability (0.62) (40) and is widely used as a screening tool for psychological wellness in children (41). Further, SDQ scores in the 90th percentile have predictive validity for independent psychiatric disorders for the SDQ parent, teacher and youth scales (40).
Family SES
Family SES was measured by two independent characteristics: parental employment status, and household income. Parental employment status was self-reported by the parent responding to the survey. Household income was also self-reported by the responding parent, as the combined income of all members of the household, from all sources (including benefits, pensions, and superannuation) (42).
Parental Sense of Community
The Sense of Community Index, a 12-item true or false self-report measure that depicts a person’s sense of connection to a place or community (43) was administered to the responding parent. This is a psychometrically adequate tool with higher score indicating stronger attachment and connection to the community (44).
Rurality Factors
Community SES
The Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD, 45) was used to represent the level of advantage and disadvantage for an area through the families’ reported postcode. The IRSAD encapsulates variables such as family income, mortgage levels, education levels, household overcrowding and vehicle access, and then assigns a decile based on these factors. A higher IRSAD score indicates relatively less disadvantage and more general advantage. A lower score indicates relatively more disadvantage and less general advantage. The IRSAD formula used for this study was based on Australian Bureau of Statistics 2006 Census of Population and Housing data to match when the participant data was collected.
Level of Rurality
The Australian Standard Geographic Classification (ASGC) categorises the level of remoteness based on the Accessibility/Remoteness Index of Australia Plus (ARIA+), a measure of the level of remoteness in road distance from required services within Australia (46). The ASGC groups the ARIA+ into five categories: major cities, inner regional, outer regional, remote, and very remote. This study uses the last four categories as major cities were not included in the sample.
Data Analysis
Data entry, cleaning and analysis was performed using Statistical Package for Social Sciences version 25 statistical software (47).
Analysis primarily reported SDQ as a continuous variable as this gave clearer indications of changes. When SDQ scores are grouped it was into clinical significance bands of normal, borderline and abnormal, based on the Australian norms by Mellor (48).
Due to non-normality of the data, the SDQ subscale and total scores, and parental Sense of Community scores were transformed to standardised scores for further analysis.
Household income was grouped into three categories based on exploratory analysis which grouped the participants into equal income thirds and based on appropriate statistical comparison of income groups on SDQ means. This matched the Australian Taxation Office personal income tax brackets for 2010-2011 being split into thirds. Low income includes nil income to $37,000, medium income $37,001 to $80,000 and high income above $80,000.
Employment status of the parent completing the child measures was spilt into three groups based on appropriate statistical comparison of SDQ means across the original five employment categories. There were no significant differences in SDQ scores across those who were not working (unemployed (n=19), not working due to illness/disability (n=13) and retired (n=1); this was therefore collapsed into one group. The remaining groups were employed and unemployed.
Demographic differences were examined using ASGC Chi square, Fisher’s Exact test, independent sample t-tests, ANOVAs and Kruskall-Wallis statistics as appropriate.
Comparisons were conducted of the current rural SDQ sample with Australian norms (48) using independent sample t-tests.
The relationship between child psychological wellness (using the SDQ subscales and total score) and personal factors (sex and age), family factors (employment status and sense of community of responding parent and household income), community SES (IRSAD) and rurality (ASCG) were examined by conducting Pearson’s correlations, t-tests, and Chi square analyses as appropriate. For employment status where no correlation could be conducted, an ANOVA was used.
The amount of influence that each level of factors impacted upon SDQ scores Multiple Linear Regressions were conducted using hierarchical analysis with variables entered in four steps: step 1, personal factors (child age, child gender); step 2, Parental/Family factors (parental employment status, household income, and parental sense of Community); step 3, community SES (IRSAD); and step 4, level of rurality (ASGC). This model was used for each SDQ subscale and total score. The order of factors entered was based on postulated order of influence from the relevant literature on child development, following the method of Kelly, Lewin (29), although separating community SES and level of rurality into separate steps based on the existing literature and hypotheses.
A moderator analysis was conducted to examine whether ASGC interacted with age or parental sense of community in influencing children’s psychological wellness measured through SDQ subscales and total scores.
For all analyses, statistical significance was set at p ≤ 0.05.