The study was approved by the University of Reading Research Ethics Committee (UREC 15/04), and participants provided informed consent.
The study aimed to recruit a community sample of parents of children with an anxiety disorder with a diverse range of experiences. As there is regional variation in available support for child mental health difficulties, and mental health support provided within schools is also likely to vary, we aimed to include families from a range of geographic locations and schools within England. A two-stage screening process was used. Firstly, during the period March to October 2015, 102 primary/junior schools from different geographic locations in England and with a varied demographic profile were invited to participate. Ten schools were recruited across seven geographic locations (Buckinghamshire, East Sussex, Hampshire, Middlesex, Northumberland, Surrey, Worcestershire) and included nine state schools (2.1–57.5% of children on the roll receiving free school meals) and one fee-paying school.
The screening process within recruited schools is detailed in Fig. 1. Within recruited schools, study information and consent/questionnaires were distributed to all parents with a child in years 3–6 (aged 7–11 years). Parents were asked to complete a questionnaire to assess their child’s anxiety symptoms (SCAS-P), and one or two researchers attended the school to administer the corresponding questionnaires (SCAS-C and SCAS-T) with the children and class teachers of those children whose parent provided consent. In cases where the child scored above the designated cutoffFootnote 1 on either the SCAS-P, the SCAS-C, or the SCAS-T, the family was invited to take part in a follow-up diagnostic assessment (ADIS-IV-C/P). Following the diagnostic assessment, parents were sent a report summarising the assessment findings, which where applicable described the particular difficulties with anxiety that the child was experiencing. Parents of the children who met the DSM-5 criteria for a current anxiety disorder formed a pool of potential participants for the qualitative interviews. Schools and families were reimbursed for giving their time to participate in the study. Schools were given £5 for each set of complete parent/child/teacher questionnaires, and families were given a £20 gift voucher for taking part in the diagnostic assessment, and a further £20 gift voucher for taking part in the qualitative interview.
Spence Children’s Anxiety Scale-Child (SCAS-C/P)
The SCAS-C/P are child and parent report questionnaires and comprise 38 items addressing symptoms of DSM anxiety disorders (and 6 filler items in the child report version). Items are rated on a four-point scale (never = 0; sometimes = 1; often = 2; always = 3), and total scores reflect the sum of responses to the 38 anxiety items. The SCAS-C/P are widely used measures of child anxiety symptoms, with good evidence in support of their reliability and validity [18,19,20,21], and excellent internal consistency in the current sample (SCAS-C α = 0.95; SCAS-P α = 0.91).
Adapted Spence Children’s Anxiety Scale-Teacher (SCAS-T)
The research team developed a teacher-report version of the SCAS-C/P. Eight items that appear on the SCAS-C/P were considered inappropriate for teachers (e.g. ‘I would feel afraid of being at home on my own’) and omitted. The wording of the remaining items was amended to reflect the reporter (e.g. ‘I worry about things’ was changed to ‘Worries about things’). The SCAS-T comprised 30 items, with the same four-point response scale as the SCAS-C/P, and total scores reflect the sum of responses to all items. Internal consistency for the SCAS-T in the current sample was excellent (α = 0.90).
Anxiety Disorders Interview Schedule-IV-Parent and Child Interview (ADIS-C/P)
The ADIS-C/P consists of structured parent and child interview schedules consistent with DSM-IV anxiety diagnoses and common comorbid diagnoses, and has strong psychometric properties . Minor amendments were made to the interview schedules to enable diagnoses consistent with DSM-5 diagnostic criteria. The child and parent interviews were administered independently, with diagnoses and clinical severity ratings (CSRs) assigned independently for each interview. As per the guidelines, a child met diagnostic criteria where the required symptoms were reported and a CSR from 4 to 8 was assigned. In cases where there were discrepancies between the child and parent report, the higher CSR was assigned as the overall CSR. With the exception of two interviews, all assessments were administered by one assessor (TR), and for the first 20 assessments, interviews were discussed with an experienced clinician (CC) and a consensus reached. Assessor reliability was checked at this stage (with a minimum kappa/ICC of 0.85 required), and subsequently one in six interviews were discussed and consensus reached. Overall inter-rater reliability within the assessment team was excellent (child report diagnoses, kappa = 0.95; CSR ICC = 0.97; parent report diagnoses, kappa = 0.98, CSR ICC = 0.96).
We used purposive sampling to ensure that those families invited to be interviewed (from the pool of 32 eligible families) varied in their experiences surrounding seeking help for their child. Invited families varied on the following characteristics, judged by the researchers to be relevant to their help-seeking experiences: (1) child age and gender; (2) type and severity of child anxiety disorders, (3) prior help seeking for the child’s anxiety difficulties, (4) socioeconomic status, and (5) geographic location.
Parents of 20 children were identified. Interviews were conducted with 16 of them; at which point analyses indicated that theoretical saturation  had been reached, as new data ceased contributing to the development and refinement of new codes and concepts. Participant characteristics are provided in Table 1. Interviews were conducted with parents of 11 girls and 5 boys, aged 7–11 years (median age 8.9 years). Interviewees were typically mothers (14 families), but in two cases the father was also interviewed. The sample spanned socioeconomic status; with nine families categorised as higher/professional and three received free school meals for their children. The sample was predominantly white British (n = 13), and three children were from other white backgrounds. Children had between one and three anxiety disorder diagnoses, including separation anxiety disorder (n = 3), social anxiety disorder (n = 5), generalised anxiety disorder (n = 10) and specific phobias (n = 7), with clinical severity ratings (CSRs) for primary disorders ranging from 4 to 6. Children had a non-anxiety comorbid disorder in one-quarter of cases (depression, n = 1; attention deficit hyperactivity disorder, n = 2; oppositional defiant disorder, n = 1). There was variation in prior help seeking reported across cases (sought help/advice from a professional, n = 9; not sought help/advice from a professional, n = 7).
Semi-structured topic-guided interviews were conducted with parents, with 4 interviews conducted face to face and 12 on the telephone. Interviews were conducted by a doctoral researcher (TR) and lasted from 33 to 79 min. Interviews were audio-recorded and explored parents’ views and experiences of recognition and help seeking for their children’s anxiety difficulties. The topic guide explored parents’ (1) knowledge and beliefs surrounding child anxiety; (2) recognition of their child’s anxiety difficulties; (3) knowledge and beliefs surrounding help seeking; (4) beliefs and experiences of services; and (5) suggestions for improvements to the help-seeking process; but it was used flexibly, allowing for variation in the order and wording of questions and ensuring participants had the opportunity to discuss issues that departed from the prepared areas of questioning. Interviews were transcribed verbatim, with all information that could identify participants removed, and participants’ names replaced with pseudonyms.
Coding and analysis
Analysis of the transcribed interviews was guided by the six phases of a thematic analysis described by Braun and Clark . The analysis was inductive, in that codes and themes were data driven, rather than according to a pre-existing theoretical framework and focussed on the specific research questions of the study, rather than attempting to encompass the entire content of the dataset. Codes were generated through an iterative process, whereby as each new transcript was coded, earlier transcripts were revisited and codes were constantly reviewed and refined. Codes were gradually organised into candidate themes and sub-themes, paying particular attention to linkages and distinctions between key ideas and concepts, and commonalities and discrepancies within and across transcripts. The analysis was led by one researcher (TR) who met regularly with other research team members to discuss codes and candidate themes, and alternative interpretations of the data. Following team discussions, candidate themes and sub-themes were reviewed and refined to ensure the thematic structure provided a coherent and credible interpretation of the data that did not only reflect a single researcher’s perspective.