Measures
Receipt of and perceived need for help. The survey instrument included the Perceived Needs for Care Questionnaire [22] which asked about need for different types of help for mental health problems over the past year, including PIs (i.e., CBT, psychotherapy, and counselling) and ‘other’ interventions (i.e., information, medications, and assistance with life skills with regards to money, work, looking after yourself, and meeting other people). A mutually exclusive, hierarchical ‘type of intervention’ classification was derived: PIs only; PIs plus other interventions; and other interventions only. Reported use of each type of help was followed by questions on whether enough help was received. Categories of perceived need were ‘met’ or ‘partially met;’ those whose needs were not met indicated barriers to receiving help that were grouped as ‘structural’ (social, environmental or economic systems) or ‘attitudinal/knowledge’ (thoughts, feelings or ideas) [23]. Respondents with a mental disorder diagnosis who did not receive help indicated whether they needed each type of help and barriers to receiving help.
Respondents indicated if they had consulted any of eight professionals for their health. These were recoded into four binary (yes/no) ‘type of professional’ variables: GP; psychologist; other mental health professional (i.e., psychiatrists, mental health nurses, and other professionals providing specialist mental health services); and other health professional (i.e., specialist doctors or surgeons, complementary/alternative therapists, and other professionals providing general services). If ‘yes,’ respondents indicated how many consultations with each professional were for mental health. Responses were collapsed into four groups (i.e., one, two to four, five to nine, and 10 or more). Respondents were also asked if they incurred out-of-pocket costs for each type of professional seen; a binary (yes/no) variable was derived.
Clinical characteristics. The World Mental Health Composite International Diagnostic Interview Third Edition (WMH-CIDI-3.0), was used to assess the presence of 12-month mental disorders according to the International Classification of Diseases, Tenth Revision (ICD-10) [24]: anxiety disorders (agoraphobia, social phobia, panic disorder, generalized anxiety disorder, obsessive compulsive disorder, and post-traumatic stress disorder); affective disorders (depression, dysthymia, and bipolar disorder), and; substance use disorders (harmful use and dependence derived separately for alcohol, cannabis, sedatives, stimulants and opioids).
Mental disorder severity (none, mild, moderate, and severe) was determined using an algorithm that calculated the impact of the disorder on functioning (accounting for comorbidity) [21]. The presence of any chronic physical condition in the past 12 months was categorised as a binary (yes/no) variable.
Disability. The World Health Organisation Disability Assessment Schedule (WHODAS) [25] assessed difficulties with performing tasks due to poor health over 30 days prior to the interview (0 = no disability to 100 = full disability). Respondents also indicated how many of the past 30 days they were unable to work or complete usual activities due to poor health.
Socio-demographic characteristics. Respondents’ age, sex, highest level of education, labour force status, marital status, geographical location (i.e., urbanicity), and Index of Relative Socioeconomic Disadvantage (IRSD) was recorded. Household financial problems in the past 12 months (i.e., could not pay electricity, gas or telephone bills on time; could not pay car registration or insurance on time; pawned or sold something; went without meals; unable to heat home; sought assistance from welfare/community organisations; or sought financial help from friends or family) was also recorded; a binary (yes/no) measure was derived.
Data analysis
Data from the 2007 NSMHWB Basic Confidentialised Unit Record File (2009) (cat. no. 4326.0.30.002) were analysed using Stata MP version 13, accounting for the complex survey design and weighting procedures [26]. The jackknife method was employed to compute standard errors. Estimates with a relative standard error (RSE) of 0.25–0.50 were to be interpreted with caution; estimates with an RSE of > 0.50 were not reported [27]. A p-value < .05, and non-overlapping 95% confidence intervals, indicated statistical significance [27].
Weighted percentages and confidence intervals described help received and perceived need for help. Chi-square analyses examined the association between type of intervention, sociodemographic and clinical factors, and type of professional consulted. Multinomial logistic regression models examined predictors of type of intervention(s) received. The dependent variable was the hierarchical ‘type of intervention;’ the independent variables were sociodemographic and clinical and treatment factors. Given the relationship between provider and intervention received [28], two models were run with and without the type of professional consulted variable (Model 1 and Model 2). Independent variables were from previous studies [11, 29–32] and were considered if associated with the dependent variable (Wald p < .200) in univariate analyses [33]. Where correlations between independent variables were ≥ 0.40, the variable with better model fit was retained.