Demographic information of the study participants is presented in Table 1. Approximately 94.7% were female, with an average age of 43.12 years (SD = 7.58). Their children with an IDD were on average 11.63 years old (SD = 5.94). Most of the parents had a university degree (61.6%) and were married or in a domestic partnership relationship (76.9%). Nearly half were not employed or unpaid caregivers for their children. They spent on average 114.23 hours per week taking care of their children (SD = 52.39).
A generally high prevalence of barriers was observed (M = 20.93, SD = 9.28); especially on taking caregiving as a priority (i.e., participants perceived this priority affected seeking help for their own mental health challenges) (93.4%), not having enough time (90.4%), and high costs (85.8%). The least experienced barriers were discouragement from people around (18.8%), confidentiality concerns (29.1%), and the fear of losing control/autonomy in a treatment (36.2%); see Table 2 for more information.
Item Reduction and Dimensionality
In the scale development phase of the PHBS, item reduction and extraction of factors were employed. The item reduction technique involves item discrimination tests and item-total correlations. As shown in the Table 3, all items showed good discrimination and moderate to high correlations with the scale total scores. Among all items, item 8 (confidentiality concern), item 5 (no access to healthcare support), and item 15 (parents’ own avoidance) revealed the highest item-scale correlations with the total barriers parents experienced or perceived.
A principal component analysis (PCA) was run on the 15-question PHBS scale on 456 parent participants. The PCA revealed four components that had eigenvalues greater than one and which explained 24.72%, 12.45%, 9.25%, and 7.76% of the total variance, respectively. Visual inspection of the scree plot indicates the four components should be retained. The four components explained 54.17% of the total variance. A direct Oblimin oblique rotation was employed to aid interpretability because correlational relationship between components were observed. The interpretation of the data was consistent with the attributes that the questionnaire was designed to measure, with personal belief items on component 1, support accessibility items on component 2, resource availability on component 3, and emotional readiness on component 4. There were no hyperplane items (items with loadings on no factor). Items 2, 3, 8, and 15 had salient loadings on more than one factor (see Table 4 for details). The component that an item belongs to depends on the magnitude of factor loadings and the concept that the item content conceptually overlaps with. All four components had sufficient items (item n > 3). All communalities were strong (communalities = [.433, .653]). Note that item 8 (“It might not be confidential”) was classified in component 1, although it had higher loading in component 2, for two reasons: (1) the item conceptually overlapped with personal belief more than with support accessibility; and (2) the loading in component 1 was acceptable (component coefficient = 0.436 for component 1 and 0.537 for component 2). Four factor scores were calculated and entered a second PCA to assess whether the components converged into a single factor (i.e., barriers to accessing care). The second PCA confirmed a single attribute (eigenvalue = 1.89); therefore, the use of a single total score to interpret perceived barriers is supported. Component loadings and communalities of the rotated solution are presented in Table 4.
Reliability, Convergent Validity, and Discriminant Validity
The reliability was evaluated by internal consistency. Cronbach’s α was .77 for the PHBS whole scale, .67 for component 1 (i.e., personal belief), .69 for component 2 (support accessibility), .57 for component 3 (resource availability), and .60 for component 4 (emotional readiness). The convergent validity of the PHBS was evaluated by its correlation with physical health, mental health, and social support. There were statistically significant, moderate, positive correlations between barriers and the parental well-being (physical health, r(453) = .276, p < .0005; mental health, r(453) = .325, p < .0005). This means a higher level of barriers in receiving mental healthcare was associated with generally poorer health status. A statistically significant, moderate negative correlation was observed between PHBS and social support scores, r(451) = .273, p < .0005. This means higher barriers in seeking support were associated with lower perceived social support.
Discriminant validity was tested by its correlational relationship with parenting. There was an insignificant and weak correlational relationship between barriers and parenting, r(391) = .092, p = .070, indicating there was no reliable or strong correlation between parents’ experiencing of barriers in seeking support and their parenting styles (see Table 5 for details).
The three subscales of the PHBS (i.e., support accessibility, personal belief, and resource availability) also showed weak to moderate positive correlations with global mental and physical health challenges (r(453) = [.16, .32], p < .05) and moderate negative correlation with social support (r(451) = [-.27, -.15], p < .01). The emotional readiness subscale did not reveal significant correlations with global mental health (r(453) =.05, p = .13), physical health (r(453) = .05, p = .32) or social support (r(451) = -.07, p = .13).
Barriers and Geographic Settings
The total sum score of barriers increased from suburban group (M = 20.56, SD = 9.35), to urban group (M = 20.78, SD = 9.03), to rural and remote group (M = 21.53, SD = 9.49); however, the differences between these geographic groups were not statistically significant, F(2, 451) = .356, p = .701. The three groups did not reveal statistically significant differences in 3 of the 4 subscales either: support accessibility barriers, F(2, 451) = 1.059, p = .348, personal belief barriers, F(2, 451) = 1.665, p = .190, or emotional readiness barriers, F(2, 451) = .367, p = .693. The resource availability barrier was statistically significantly different between different geographic groups F(2, 451) = 3.643, p < .05, η2 = .016. The barriers regarding resource availability increased from the urban group (M = 2.22, SD = 0.87), to suburban group (M = 2.27, SD = 0.93), to rural/remote group (M = 2.51, SD = 0.90). Tukey post hoc analysis revealed that the mean increase from rural/remote to suburban group (0.24, 95% CI [-0.03, 0.51]) was marginally statistically significant, p = .091, and the increase from rural/remote to urban group (0.29, 95% CI [0.03, 0.55]) was statistically significant, p = .023, but not statistically significant between suburban and urban group.