Participants
Ethical approval for the study was granted by the University of Adelaide Human Research Ethics Committee on the 5th of June, 2019 (approval code HREC-2018-273). Participants were Australian men who had experienced the loss of a baby at any stage of gestation to miscarriage, ectopic pregnancy, medical termination of pregnancy for nonviable foetal anomaly (TOPFA), stillbirth or neonatal death. Inclusion criteria were that participants were aged 18 years of age or older and had experienced pregnancy loss or neonatal death in Australia within the last 20 years. Although potentially open to recall bias, this timeframe was selected to maximise the potential pool of eligible respondents. Of 277 participants who commenced the survey, 228 completed all items and were included in the final sample reported here (completion rate = 82%). There were no apparent differences between completers and non-completers on demographic characteristics. At the time of survey completion, participants were aged between 19 and 60 years (M = 36, SD = 7.4). At the time of loss, they were aged between 18 and 58 years (M = 32, SD = 5.5). See Table 1 for a summary of participant characteristics at the time of survey completion.
Table 1. Participant characteristics
|
Category
|
N (%)
|
Ethnicity
|
Australian
|
194 (85%)
|
Other^
|
34 (15%)
|
Sexual orientation
|
Heterosexual
|
224 (98%)
|
|
Bisexual
|
3 (1.5%)
|
|
Homosexual
|
0 (0%)
|
|
Transgender
|
0 (0%)
|
|
Rather not answer
|
1 (0.5%)
|
Highest level of education
|
High School
|
54 (24%)
|
Technical and Further Education (TAFE)/Trade
|
83 (36%)
|
Undergraduate Degree
|
58 (25%)
|
Postgraduate Degree
|
33 (15%)
|
Marital status
|
Married
|
186 (82%)
|
In a relationship
|
35 (15%)
|
Divorced
|
1 (0.5%)
|
Separated
|
4 (1%)
|
Never married/single
|
2 (2%)
|
Area of residence*
|
Major city
|
131 (58%)
|
Inner regional
|
64 (28%)
|
Outer regional
|
28 (12%)
|
Remote/very remote
|
4 (2%)
|
Number of losses
|
One
|
138 (61%)
|
Two–three
|
15 (7%)
|
Four–five
|
47 (21%)
|
Six or more
|
28 (12%)
|
Loss type reflected on for the survey
|
Ectopic pregnancy
|
5 (2%)
|
Termination of pregnancy for foetal anomaly (TOPFA)
|
30 (13%)
|
Miscarriage
|
69 (30%)
|
Stillbirth
|
77 (34%)
|
Neonatal death
|
47 (21%)
|
Time since loss
|
Less than one year
|
65 (28%)
|
|
1-2 years
|
40 (18%)
|
|
3-5 years
|
59 (26%)
|
|
6-10 years
|
43 (19%)
|
|
11-15 years
|
10 (4%)
|
|
16-20 years
|
11 (5%)
|
^Other ethnicities reported by participants include: European (8%), Asian (2%) and New Zealander (2%).
*Based on Australian Bureau of Statistics classification of remoteness.
Procedure
A web-based survey was developed by the authors (see Appendix 1), hosted by the online platform SurveyMonkey. This survey was developed for the purposes of the current study, and has not been published elsewhere. Extensive piloting was undertaken with members of a reference group (including Australian fathers and mothers who had experienced pregnancy loss/neonatal death, grief counsellors and pregnancy loss/neonatal death support workers and researchers) as part of the broader program of research to form the final survey. The survey was developed with extensive consultation and piloting. Initially, preliminary discussions were held with individual members of the reference group concerning the types of measures used and questions to be asked in line with the socioecological model of men’s grief. With this feedback, the first author drafted a full survey. In the two successive rounds of piloting, members of the reference group reviewed updated drafts of the survey in full and were invited to provide suggestions for revision. Although major concepts remained the same, the ordering, inclusion and wording of questions and final measures selected, were edited and refined according to feedback to ensure both sensitivity and ease of understanding.
Potential participants were invited to take part in the survey via advertisements through Australian pregnancy loss and neonatal death support and advocacy organisations. These included Pillars of Strength, Bears of Hope, Sands Australia, Still Aware, Miracle Babies Foundation, SIDS and Kids SA, and the Australian Perinatal Loss Centre. Following ethics approval, these organisations were contacted by the first author via email or telephone to discuss the study. All organisations agreed to share a study flyer and information through either social media platforms (primarily Facebook), newsletters, and/or organisation websites.
The study flyer contained brief information about the survey and the online survey link, which opened to a covering page with a preamble providing potential participants with detailed information about the study. After reading the study preamble, participants provided passive consent, a method of consent approved by the University of Adelaide Human Research Ethics Committee, by choosing to commence the survey and submit their responses. In recognition of the sensitivity of the topic and potential for participants to experience emotional distress in reflecting on their experience of loss, a comprehensive distress protocol was developed and articulated to participants. This included providing contact details for national pregnancy loss telephone support lines at the beginning and end of the survey. No concerns regarding participant distress were raised during the research.
The survey took approximately 30 minutes to complete. Depending on participant responses, skip logic was incorporated to hide questions which were irrelevant to individual experiences, often resulting in a shorter completion time (M=22 mins). The number of items/questions presented to participants who completed they survey therefore ranged between 110 and 130. Participation in the survey was voluntary and anonymous. Data collection occurred between June and August 2019. Data were exported from the online SurveyMonkey platform and stored on a secure university-approved network at the University of Adelaide.
Measures
Participants completed questions relating to demographic characteristics (age, ethnicity, education, occupation, sexual orientation, marital status, religion, and postcode), along with questions about their pregnancy and loss history. Definitions for the death of a baby during pregnancy or shortly following birth vary, with gestational cut-offs for classification differing between countries. In Australia, a miscarriage is defined as the death of a baby in-utero before 20 weeks’ gestation and occurs for approximately 20% of pregnancies [54]. In 1-2% of pregnancies, an ectopic pregnancy occurs when the fertilised ovum implants outside of the uterus, most commonly in the fallopian tube [54-56]. A stillbirth is defined as the loss of a baby from at least 20 weeks’ gestation or over 400g in weight, occurring for 7.1 per 1,000 births [57]. Neonatal death refers to the death of a newborn infant within the first 28 days of life and occurs for 2.5 per 1,000 live births [57]. A congenital anomaly is diagnosed in approximately one in 22 pregnancies in Australia [58]. National data on TOPFA is not collected in Australia; however, it has been estimated that approximately 10-13% of parents elect to medically terminate a pregnancy diagnosed with foetal anomaly, particularly when the diagnosis is considered life-limiting or fatal [59]. Where more than one type of pregnancy loss or neonatal death had been experienced, participants were asked to reflect on only one type of their choice for the remainder of the survey. An option to comment on other losses was provided at the end of the survey. Participants then completed a mix of questions developed by the authors as well as standardised measures.
In line with the literature on the ‘supporter role’ relating to men’s grief [10, 12, 13, 16, 36, 40, 60], two author-developed measures were included to determine the extent to which men perceived this to be their role, and whether they felt it interfered with their grief. Participants responded to these questions on a five-point Likert scale from 1 = strongly disagree to 5 = strongly agree. Higher scores indicated that men perceived their role to be a supporter to their female partner and family after the loss and that their supporter role had a larger impact on their ability to grieve. Scales were also developed to determine the extent to which participants felt their grief was recognised by others, namely: their partner, family, friends, health professionals, and the wider community. For these, participants responded on a five-point Likert scale from 1= not at all to 5 = extremely. Higher scores were indicative of higher levels of recognition for their grief. Participants were also asked a series of questions about their experiences of returning to work, including whether they were offered leave, and what type of leave they were offered (detailed results under review for publication elsewhere). If they had contact with a hospital as part of their loss experience, they were also asked about the extent to which they felt included (from 1 = not at all to 5 = extremely; where higher scores indicated a greater sense of inclusion), and whether they were offered information on grief for fathers (yes/no). The six included standardised measures are outlined below.
Paternal Antenatal Attachment Scale (PAAS): A modified version of the Maternal Antenatal Attachment Scale (MAAS), the PAAS assesses both the quality and strength of the subjective experience of the father’s attachment to the developing baby [61, 62]. Comprising 16 items forming two subscales (Quality of Attachment and Time in Attachment), the PAAS is answered using five-point Likert scales, where higher scores indicate stronger attachment to the baby. Although only a small number of papers have used the PAAS, relationships have been found with related measures including relationship quality, mental health, increasing gestational age, and father identity [63, 64]. Previous research also supports the reliability and validity of the PAAS, with reports of high internal consistency (Cronbach’s alpha = 0.83) [62]. For this study, the final question of the scale “If the pregnancy was lost at this time (due to miscarriage or other accidental event) I expect I would feel…” was omitted, given that participants had experienced a pregnancy loss or neonatal death. Internal consistency of this 15-item version in this study was also high (Cronbach’s alpha = 0.83).
Perinatal Grief Scale-33 (PGS-33): Designed to quantify bereaved parents’ grief based on emotional responses, the PGS-33 assesses thoughts and feelings associated with perinatal loss [65]. The overall scale comprises three subscales: Active Grief (outward expressions of grief including crying, sadness and missing the baby), Difficulty Coping (difficulties with daily activities and relating to others) and Despair (feelings of hopelessness and worthlessness). Participants rate each item on a five-point Likert scale ranging from 1 = strongly agree to 5 = strongly disagree, with higher scores indicating more intense grief. The PGS-33 is the most common grief scale used among the perinatal loss literature and has been extensively evaluated, with psychometrically sound properties reported (including Cronbach’s alphas between 0.92 and 0.96) [66, 67]. Internal consistency for the full measure was also high in this study (Cronbach’s alpha = 0.94). Although questions remain surrounding the accuracy of using the PGS among men, as it may not be sensitive to instrumental grieving styles [36], given a current lack of alternative grief measures specifically for men following pregnancy loss/neonatal death, we decided that in conjunction with the Grief Patterns Inventory (described below), this was the best available measure to adopt.
Grief Patterns Inventory-Revised (GPI-10): A measure developed to assess an individuals’ general grieving pattern, the GPI indicates a tendency toward either an instrumental or intuitive grieving style. The original measure comprised 24 items containing true-false responses; however, a revised version containing ten items (five items each for the instrumental and intuitive styles) was used in the current study to reduce respondent burden [68, 69]. A pilot study of the 10-item version reported moderate inter-correlations between subscale items, along with a significant negative correlation between the intuitive and instrumental subscales (r = -.525) [68]. Although alpha coefficients were not reported for the 10-item version, research demonstrates acceptable internal consistency for the original version (Cronbach’s alphas ranging between 0.71 and 0.76) [70]. In this study, a similar level of internal consistency was found (Cronbach’s alpha = 0.71). Items are rated on a five-point Likert scale from 1 = Strongly Disagree to 5 = Strongly Agree, with instrumental items reverse-scored. As such, potential total scores ranged from 10 to 50, with lower scores indicating a more instrumental style, and higher scores indicating a more intuitive style. As applied previously [69], categorisation of grief styles was made as follows: 10-23 = instrumental; 24-36 = blended; 37-50 = intuitive.
Crisis Support Scale (CSS): The CSS is a measure of social support received from family and friends following a traumatic event (in this case, pregnancy loss/neonatal death). Comprising seven items relating to the availability of others, emotional support, and practical help, respondents rate their agreement to the items on a seven-point Likert scale, ranging from 1 = never to 7 = always. In the original scale, participants responded to two time points: just following the event (T1) and the present time (T2). However, for this study, participants were only asked to provide responses for the support that was available to them most of the time following their loss. Higher scores indicate higher levels of social support. Validation studies indicate robust psychometric properties for the scale across a range of trauma populations, including bereaved parents of infants (Cronbach’s alphas ranging between 0.67 and 0.82; in this study, Cronbach’s alpha was 0.69) [71].
Conformity to Masculine Norms Inventory (CMNI): Developed based on Mahalik’s model of gender role conformity, the CMNI assesses the extent to which an individual male does or does not conform to the actions, thoughts, and feelings reflected by broad masculinity norms [72, 73]. The original scale consists of 144 items forming 11 distinct factors. However, to reduce participant response burden, only one subscale comprising five items from the overall measure was included for this study, to determine respondents’ tendencies toward Self-Reliance. This subscale was chosen in line with previous literature which suggests men often feel the need to hide their grief from others, preferring to cope in isolation [10, 12, 13, 16, 29, 33]. The questions included: “I never ask for help”, and “It bothers me when I have to ask for help”. Respondents rated the degree to which they agreed with these statements on a four-point Likert scale from 1 = strongly disagree to 4 strongly agree, with higher scores indicating a stronger tendency toward being self-reliant. Widely used in the literature, many studies have reported construct validity for the CMNI, along with discriminant validity between its subscales and high internal consistencies (Cronbach’s alpha of 0.85 for the Self-Reliance subscale; in this study, Cronbach’s alpha was 0.86) [73].
Male Role Norms Inventory Short Form (MRNI-SF): A measure of masculinity ideology developed by Levant et al. [74], the original MRNI comprised 57 items with seven subscales. In 2011, a 39-item revised form was proposed, followed by a 21-item short-form in 2013. For this study, the Toughness subscale from the MRNI-SF was used, as items closely aligned with the recurrent theme of needing to be ‘strong’ or ‘tough’ reported by men following pregnancy loss in previous literature [10, 13, 16, 32, 33]. The subscale comprises three items, including: “When the going gets tough, men should get tough”. Responses are given on a seven-point Likert scale from 1 = strongly disagree to 7 = strongly agree. Higher scores indicate higher levels of endorsement of traditional masculine ideology [75]. Research has demonstrated sound psychometric properties for the MRNI-SF, including subscale alphas ranging from 0.79 to 0.90 [75]. In this study, Cronbach’s alpha for the 3-item Toughness subscale was 0.61.
Data analysis
Analyses were performed using SPSS Statistics (V.25). Data were summarised using descriptive statistics and relationships between the variables were assessed using generalized linear modelling with a multiple stepwise approach, including a backward elimination method outlined by Sainani [76]. The generalised linear model is a flexible form of usual linear regression used to compare the effect of several variables which may have error distributions other than a normal distribution on a continuous outcome variable. Using a link function to relate the response variable to the linear model, it provides a maximum likelihood estimation of the model parameters rather than assuming a linear-response model [78]. As recommended for multivariable modelling [77], a priori selection of variables for this study was guided by the socioecological model of men’s grief identified in our previous systematic review [36]. Given the nested form of the socioecological model, variables were entered into the regression models in four (stepwise) stages. Assumptions required for generalised linear modelling were assessed prior to analysis; all necessary assumptions were satisfied. Individual-level variables were entered first, and a backward elimination process was carried out until all variables were statistically significant at the 0.5 level as recommended by Harrell [77]. This process was repeated with each of the interpersonal, community and policy/system-level variables until all had been entered into the model (see Table 2 for the variables entered at each level).
While we acknowledge the debates surrounding the use of p-values in making decisions regarding variable selection [79,80], the cut-off for inclusion of 0.5 (rather than the traditional choice of a = 0.05) used for variable selection in this study is considered to be a reasonable and conservative estimate for a multivariable model [77]. Also, to reduce the risk of bias from sparse data, backward elimination is recommended to achieve a suitable number of degrees of freedom for the model given the number of observations in the study; a general rule is that degrees of freedom should be no more than the number of observations divided by ten to reduce the risk of bias [77]. Without the use of backward elimination, our models would have violated this rule. Ultimately, our approach resulted in a suitable number of degrees of freedom for each model in this study.
Table 2. Variables entered into the multiple linear regression analyses
Stage entered into the model
|
Variables
|
Stage 1 (individual variables)
|
Loss type; Grief stylea; PAAS Time in Attachment; PASS Quality of Attachment; Whether men attended obstetric appointments; Whether men viewed an ultrasound image of their baby; Age at time of loss; Ethnicity; Importance of religion; Number of previous losses; Number of surviving children at time of loss
|
Stage 2 (interpersonal variables)
|
Marital satisfaction; Extent of agreement to the statement: “My role following the loss was to support my partner and family”; Extent of agreement to the statement: “I was unable to grieve, because I was too busy supporting everyone else”; Total CSS score; Extent of acknowledgement from partner; Extent of acknowledgement from family; Extent of acknowledgement from friends
|
Stage 3 (community variables)
|
Extent of acknowledgement from community; CMNI Self-Reliance subscale total score; MRNI Toughness subscale total score
|
Stage 4 (policy/system variables)
|
Extent of acknowledgement from healthcare professionals; Degree to which participants felt included in the hospitalb; Whether employment leave was offered to men; Whether other psychosocial supports were offered to men
|
Note: aEntered only into model 1 (dependent variable = PGS total score); bClassified into a high/low level of inclusion based on original Likert scale responses (scores 1-3 = low level of inclusion and 4-5 = high level of inclusion)
Statistical power
There are no consistent rules for sample size requirements in linear regression [81]. However, various general recommendations have been made about minimum sample size, or sample size depending on the number of independent variables included in the model. While one general rule recommends a minimum of 100 participants regardless of the number of independent variables [82], others suggest 50 plus the number of independent variables [83], or at least 100 for less than three independent variables or 300-400 for nine or 10 independent variables [84]. Tabachnick and Fidell also suggested a sample size of 50 + 8k, where k is equal to the number of independent variables [85]. Employing recommendations to consider a minimum sample size of 50 plus the number of independent variables [83, 85], with a sample size of 228 and the number of independent variables included in the models at any one stage not exceeding 16, the current study had sufficient statistical power.