Study design and participants
Data were drawn from the Ten to Men Study, the Australian Longitudinal Study on Male Health [
16‐
18]. Ethics approval was obtained from the University of Melbourne Human Research Ethics Committee and the Australian Government Department of Health. The target population was boys and men (aged 10–55 years) who were Australian citizens or permanent residents, living in private households in urban and regional areas across Australia. The study design and sample information are detailed elsewhere [
16,
17]. Briefly, a stratified multi-stage cluster random sample design was used. Separate cluster samples were drawn from geographical strata representing major cities, and inner and outer regional areas, with oversampling of regional areas to increase the representation of males living in these areas. Over 45,510 males living in the sampled households were invited to participate between 2013 and 2014. A total of 15,988 boys and men participated at wave 1 (baseline), and they were followed up at wave 2 (2015–2016). The response rate at wave 2 was 75%. The current study used data from men in the adult cohort who indicated that they had become a father within the 12 months prior to wave 2 data collection (
n = 205).
Ten to Men fieldworkers made three in-person visits to each sampled household to recruit eligible males. Interested males were provided with study information, a consent form, hardcopy questionnaires, and privacy envelopes. Written informed consent was obtained from all participants and paper questionnaires were completed. The questionnaires covered broad topics including socio-demographic characteristics, mental health and wellbeing, physical health, health behaviours, socio-contextual information, and knowledge and use of health services.
Measures
Demographic information at study enrolment included age, Aboriginal and/or Torres Strait Islander origin, country of birth, main language spoken at home, highest level of education, in paid employment, and relationship status (partnered or not). Additionally, the Modified Monash Model geographical classification was used to define whether the area in which a participant lived was city, rural, remote or very remote based on the size of the local town or city [
19]. This was dichotomised into metropolitan areas and rural, remote or very remote areas in analyses. Age, English spoken at home, and education were included in the analyses as indicators of economic status. Metropolitan and rural/remote status was also included given disparities in access to health care in rural areas and some evidence of higher mental health difficulties and suicidality among men living in rural areas [
20].
Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9; [
21] at both waves 1 and 2. Nine items asked about the extent to which participants were bothered by depressive symptoms (e.g., little pleasure in doing things, feeling depressed/hopeless) in the last 2 weeks on a scale ranging from 0 = ‘not at all’ to 3 = ‘nearly every day’. The scores are summed, with scores between 0 and 4 indicating no or minimal depression, 5–9 mild depression, 10–14 moderate depression, 15–19 moderately severe depression, and 20–27 severe depression. The PHQ-9 has well established validity [
22], and good internal consistency in the analytic sample (Cronbach’s α = 0.83 at wave 1 and 0.89 at wave 2).
Preconception estimates of suicidal ideation and behaviours was assessed at wave 1 using two self-report items from the Youth Risk Behaviour Survey [
23]; ‘Have you ever tried killing yourself?’ and ‘Have you ever made a plan about how you would kill yourself?’, and a single item from the Longitudinal Study of Australian Children [
24]; ‘Have you ever seriously thought about killing yourself?’. To obtain postnatal estimates of suicidal ideation and behaviours at wave 2, all three questions were asked again but in regards to the past 12 months. Participants were asked to respond ‘yes’ or ‘no’ to each item.
Stressful life events were assessed at both waves using 24 items selected from either the Australian Longitudinal Study on Women’s Health (ALSWH; [
25] or the Social Readjustment Rating Scale (SRRS; [
26]. Participants were asked to indicate ‘yes’ or ‘no’ to if they had experienced any of the listed stressful life events in the past 12 months. Events included personal injury or illness, divorce or marital separation, conflict between family, death of spouse, moving house, difficulty finding a job, natural disaster or house fire, legal troubles or court case. Items were scored by summing all the ‘yes’ responses, where higher scores indicate the presence of more stressful life events.
Financial difficulties were assessed using items from the Australian Bureau of Statistics [
27]. Participants were asked if any of the following had happened over the past 12 months because they were short of money: (a) could not fill or collect prescription medicine, (b) could not get a medical test, treatment or follow-up that was recommended by a doctor, (c) limited how much fruit and vegetables you ate, (d) could not pay electricity, gas or telephone bills on time, (e) could not pay the mortgage or rent on time, and (f) asked for financial help from friends and family. Respondents were asked to indicate ‘yes’ or ‘no’ to each item, and endorsement of one or more was identified as experiencing financial difficulties (0 = no financial difficulties; 1 = one or more financial difficulties).
Engagement in violence or experience of violence with a past or present partner was assessed using two items from the Comparing Heterosexual and Same Sex Abuse in Relationships survey instrument [
28]. Participants were asked if they had ever (a) made a past or present partner feel frightened or anxious, (b) forced a partner to have sex, and (c) hit, slapped, kicked or otherwise physically hurt a partner when angry. They were asked to indicate ‘yes’ or ‘no’ to each item, and endorsement of one or more was identified as having used partner violence (0 = not used violence; 1 = used violence). Participants were also asked if a past or present partner had ever engaged in these three behaviours toward them. Similarly, they were asked to indicate ‘yes’ or ‘no’ to each item, and endorsement of one or more was identified as having experienced partner violence (0 = not experienced partner violence; 1 = experienced partner violence).
Social support at waves 1 and 2 was assessed using the Emotional/Informational support subscale of the Medical Outcomes Study Social Support Survey (MOS; [
29]. Eight items asked about the extent to which different kinds of support (e.g., “someone you can count on to listen to you when you need to talk”) are available if needed on a scale ranging from 1 = ‘None of the time’ to 5 = ‘All of the time’. The items are summed with higher scores indicating higher availability of social support. The MOS has excellent validity [
29], and excellent internal consistency in the analytic sample (Cronbach’s α = 0.97).
Lifestyle behaviours, physical health conditions and health service use were assessed at waves 1 and 2 using self-reports. Life style behaviours were captured by asking participants if they currently smoke tobacco cigarettes, and have used cannabis or other illicit substances (i.e., ecstasy, cocaine or opiates) for non-medical purposes at least once in the past 12 months (‘yes’ or ‘no’). Harmful alcohol use was assessed using the Alcohol Use Disorders Identification Test (AUDIT; [
30]. Participants were also asked if they had symptoms or been treated for a range of health conditions (e.g., asthma, diabetes, high cholesterol, high blood pressure) in the past 12 months (‘yes’ or ‘no’). Endorsement of one or more was identified as having a physical health condition (0 = no physical health condition; 1 = physical health condition). Health service use was assessed by a single item from the National Health Survey [
31], where men were asked if they had consulted with a family doctor/General Practitioner (GP) for their own health in the past 12 months (‘yes’ or ‘no’).
Data analysis
All analyses were performed using SPSS Version 22. Descriptive statistics for the sample demographics, preconception variables, depressive symptoms and suicidal thoughts, plan and attempts were generated (Aim 1). Bivariate and multivariate standard linear regression analyses were performed to identify preconception risk factors of men’s postnatal depressive symptoms (continuous scores on the PHQ-9; Aim 2). To select variables for inclusion in the multivariate models, we examined each potential predictor variable in a series of bivariate regressions. Predictor variables with significant bivariate associations with postnatal depressive symptoms at p < 0.05 and the demographic characteristics (father age, language spoken, high school education, metropolitan vs. rural or remote) were entered into the multivariate model. We were unable to conduct regression analyses to assess risk factors for suicidal ideation in the postnatal period due to the small number of fathers reporting suicidal ideation (n = 10).
Missing data across all variables were minimal (5.2%) and missing completely at random as evidenced by Little’s MCAR test,
χ2 (16) = 20.18,
p = 0.212. Missing data for all descriptive and regression analyses were handled using multiple imputation. Fifty complete datasets were imputed using a multivariate normal model using all variables used in the analyses. Pooled estimates for all proportions and model parameter estimates were obtained using Rubin’s rule [
32]. Multiple imputation in SPSS provides the following regression model estimates for pooled data: unstandardised B,
t and p-values. For the remaining estimates (R
2, F-statistic), the ranges across the multiply imputed datasets are reported.