Sensory integration therapy for children with autism and sensory processing difficulties: the SenITA RCT

Recruitment and sampling

We were interested in the views of carers and the approaches used by occupational therapists. We chose a mixed-methods approach. Both carers and occupational therapists completed online questionnaires and, to build a more detailed picture of OT support provided, we also carried out qualitative focus groups and subsequent telephone interviews with occupational therapists.

Carer participants

An online questionnaire was distributed to carers of children with autism and SPDs aged 4–11 years and in mainstream education. The questionnaire was made available via the National Autistic Society (NAS) website, local NAS groups, social media [Facebook (Meta Platforms, Inc., Menlo Park, CA, USA; www.facebook.com) and Twitter (Twitter, Inc., San Francisco, CA, USA; www.twitter.com)], flyers at public events and through the Royal College of Occupational Therapists specialist section on children, young people and their families.

Occupational therapist participants

Data collection

Ethics approval and data management

All activities adhered to the Research Governance Framework for Health and Social Care. Ethics approval was granted by a Health Research Authority Research Ethics Committee. All respondents consented to participate in questionnaires, interviews or focus groups.

Online questionnaires

Descriptive analyses were used in the form of frequency tables for the quantitative data, and included summaries of the geographical area of respondents, nature of reported difficulties and contact with services, and information on OT assessment and treatment received. Free-text comments within questionnaires provided context and additional explanation for some responses. Examples have been included to give meaning and carers’ lived experience of services received. Free-text comments within questionnaires completed by occupational therapists were incorporated into the thematic analysis of focus group and interview data.

Focus groups and interviews

Transcripts were analysed using thematic analysis. This involved data familiarisation, generating initial themes, and searching, reviewing and defining themes. 43 The first two steps were carried out as a team by two experienced qualitative researchers and two members of the research team who were experienced occupational therapists. The final analyses were undertaken by the lead qualitative researcher.

We carried out a thematic analysis using deductive methods to identify themes reflecting reality as reported across the whole data set. We initially coded data within six overarching themes (i.e. referral, initial assessment, initial management, further assessment, treatment plan/intervention, feedback and future plans) relating to the different phases of OT support for children with autism and SPDs. In addition, we used an inductive thematic approach to identify occupational therapists’ views, experiences and challenges relating to these processes, moving towards a more contextualist approach and ensuring that subthemes, which had not been pre-empted and initiated from occupational therapists themselves, were identified. These subthemes were factors influencing decisions around which assessment occupational therapists used (at child, OT and system level), factors influencing the location of assessment, the multifaceted nature of interventions, empowering others, and making interventions achievable for parents and schools. This analysis was then integrated with the results of the questionnaires. While interpreting the qualitative data we produced a visual map of the different phases of management of children with autism and SPDs (Figure 1). We developed this map this in a phased way, starting with one account in an interview and building on this with each new transcript, revising and incorporating as we progressed. In this way, data were constantly compared and contrasted. Developments in the analytic process were recorded through researcher memos and version control of the visual maps. The first transcript was discussed in detail and themes developed by a group of qualitative researchers and occupational therapists. Themes were then applied to the next transcript and discussed again. The research team were encouraged to reflect on their own professional role in asking and interpreting interview questions.

FIGURE 1.

Routes of support described by occupational therapists. ITP, individualised therapy programme/treatment plan; SP, sensory processing; SPM, Sensory Processing Measure.

Carers’ views of usual occupational therapy for children with autism and sensory processing difficulties

A total of 159 carers responded to the questionnaire. Most (n = 149, 93.7%) carers were mothers. We received six (3.8%) responses from fathers, two from grandmothers (1.3%), one from a kinship carer (0.6%) and one from an educational visitor (0.6%). The spread of respondents across geographical regions is shown in Table 1. Thirty-two (20.1%) respondents were from Wales, 115 (72.3%) were from England, 11 (6.9%) were from Scotland and one (0.6%) was listed as ‘other’ (location not specified).

Nature of reported difficulties

The most frequently reported difficulties were difficulties with relationships and communication (84.3%) and difficulty making friends/joining in/playing with others (83.6%), followed by refusal to carry out daily activities (73.6%) and outbursts without any obvious reason (73.6%). A total of 77.3% carers reported at least moderate difficulty with sensory hyper-reactivity and 49.6% of participants reported at least moderate hyporeactivity. Being overactive and having repetitive behaviour that interrupts activity were reported by 64.8% of carers. Other difficulties encountered included sleeping difficulties (59%), poor co-ordination (56%), poor performance in school (54.1%) and feeding difficulties (39.6%). Less common difficulties included anxiety (n = 8, 5%), hyperacusis (n = 2, 1%), hypermobility (n = 2, 1%), poor concentration (n = 2, 1) and violent behaviour (n = 2, 1%).

Carers’ views on contact with services

A total of 119 (74.8%) carers reported some contact with services [e.g. children’s centres (health service) and local authorities (social care services)] in relation to difficulties in the last 6 months. Professionals seen within these organisations included occupational therapists (NHS and private), speech and language therapists, paediatricians, psychologists (clinical and educational), dieticians and specialist teachers. However, 40 (25.2%) carers reported that they had received no therapy/support in relation to their child’s difficulties in the same period. Further detail regarding contact with services is given in Table 2.

Ninety (56.5%) carers received information and advice on how to manage their child’s difficulties, such as anxiety, behavioural and sensory difficulties. Although 68 (42.8%) carers received specific advice on sensory information from services, a further 46 (28.9%) carers would have liked more information on these aspects of their child’s difficulties. Where advice and information on sensory issues had been given, this was most commonly by OT services. However, sometimes other NHS and third-sector organisations provided similar information. Information was typically in leaflet form and included information on sensory processing, exercises/programmes to do at home, social stories and contact information/websites for parent courses.

Carers’ views on occupational therapy assessments and treatment

Seventy (44.0%) carers had not received any input from OT services. Those who had received input (n = 89, 56.0%) generally received input in the form of assessment and observation plus advice. Assessment was most frequently through a carer questionnaire and observation of play. A few carers reported more formal testing of co-ordination skills. If direct intervention was offered, then this was typically for a limited block of therapy (e.g. four to eight sessions). Of the carers (89/159) who had received input from OT services, 64 (40.3%) were not currently in contact with an occupational therapist. Of the 44 carers who had seen an occupational therapist in the last 6 months, 10 had received more than five sessions in that period. In terms of length of overall contact with OT services, the majority of carers were discharged within a 6-month period (n = 76, 85%).

Of those carers who had ever received OT input, 15 (16.8%) reported that they had received weekly input and most visits were clinic based (n = 57, 35.8%), with intervention varying from sensory based (not SIT), parent-directed programmes, functional skills training, sensory diets and behavioural programmes. Most carers (n = 58, 36.5%) stated that the intervention that they received focused on sensory modulation. Fewer carers (n = 22, 13.8%) received intervention for both sensory modulation and dyspraxia, and an even smaller number of carers (n = 9, 5.7%) received intervention for dyspraxia alone. Forty-three (27.0%) carers reported that the intervention focused on other aspects, such as anger management, behavioural difficulties, speech and sleep. Carers were asked further details about specific interventions that their child received in terms of whether or not they felt that the therapy helped their child, how satisfied they were with the intervention received and whether or not they would recommend any changes to intervention delivery.

Questionnaire responses by occupational therapists

The questionnaire was completed by 79 occupational therapists. In phase 1, 38 questionnaires were completed by occupational therapists working in trial recruitment areas (i.e. Wales and Cornwall). In phase 2, 41 questionnaires were completed, including by occupational therapists in England and Scotland. Occupational therapists were generally very experienced: 53 (67.1%) occupational therapists reported having at least 10 years’ post-qualification experience and only 10 (12.7%) reported having < 5 years’ post-qualification experience. Eighteen (22.8%) occupational therapists reported postgraduate qualifications, with 11 (13.9%) specifically in sensory integration. Almost all (n = 77, 97.5%) occupational therapists reported seeing one or more children with autism and SPDs in the last 6 months, with 15 occupational therapists (15.2%) seeing more than 10 children each. However, only 11 (13.9%) occupational therapists were based within a specialist autism service. The settings in which children were seen included children’s centres, special and mainstream schools, local authority settings, Child and Adolescent Mental Health Services (CAMHS) and the child’s home. Most occupational therapists saw children referred for one to five contacts/sessions (n = 39, 49.4%), with only three (3.8%) occupational therapists seeing children for between 10 and 20 contacts/sessions. The frequency of contact was very varied, with the most common frequencies being weekly (27.8%), monthly (19%) or twice per month (17.7%). Most occupational therapists maintained contact with families for 3 months (n = 26, 32.9%) or 6 months (n = 20, 25.3%) prior to discharge.

Focus groups and interviews with occupational therapists

Two focus groups and six semistructured telephone interviews were conducted following the online survey, representing the views of 11 occupational therapists working in South Wales, Mid-Wales and Cornwall. Respondents described themselves as occupational therapists, paediatric occupational therapists and occupational therapist senior practitioners, and had experience of working in autism assessment units, core services, in the community, within the NHS, in social services and in the local authority. This section integrates occupational therapists’ views on usual care gathered from questionnaires, focus groups and interviews. Illustrative quotes are provided.

Representativeness of views

The views reflected within our data were collated from survey responses from occupational therapists and carers in Wales, England and Scotland, and from interviews/focus groups conducted with occupational therapists in South Wales and Cornwall. The purpose was to determine how carers and children with autism and SPDs access OT services, the assessment process and what interventions and/or support are commonly provided. Determining the nature of usual care within trial areas and the wider UK context would ensure that our comparator was sufficiently different from SIT delivered in the intervention arm of the trial.

Theoretical approaches to intervention

Survey responses indicated that occupational therapists utilised OT theory to guide assessment and intervention. This core philosophy values human participation in meaningful occupations59 and this is reflected in the criteria that occupational therapists use to choose whether or not to accept a referral for assessment. Within the broader scope of theoretical approaches, sensory integration theory was reported as the most common approach to help to understand and manage SPDs. This is consistent with findings from Kadar et al. ,60 who found that 72.7% of occupational therapists used a sensory integration frame of reference. Likewise, Brown et al. 61 reported that sensory integration and client-centred practice models were most commonly used in a survey of Canadian paediatric occupational therapists.

Accessing services and the referral process

Although services differ, there appeared to be a common trend across the regions surveyed. From both quantitative and qualitative data, it was clear that most children were seen for initial assessment only if they met specific referral criteria, particularly with regard to SPDs affecting daily functional activities. Many services would not accept a referral unless there were clear occupational performance difficulties in self-care, play or school-related activities. It is likely that this more focused approach has been driven by increasing waiting lists for access to OT,62 ensuring that only the most appropriate children received therapy. There have also been government drivers to ensure that children are seen within 14 weeks of referral (e.g. the NHS Wales Delivery Framework and Reporting Guidance 2019–202063). From the carers’ perspective, understanding of what OT services could offer their child was less clear cut. Carers wanted support for their child’s social interaction, help with dealing with aggressive outbursts/frustration, help with understanding autism and help with siblings. Fewer carers listed functional skills, such as toileting, feeding, dressing and sleeping, as behavioural issues seemed to have a greater impact on quality of life. Respondents were, therefore, frustrated because they felt that they could not always access the service that they perceived would help them and their child, and expressed disappointment that services were not always offered by the NHS.

Assessment process

Both occupational therapists and carers reported that initial assessment included observations, interviewing the caregiver, questionnaires and, sometimes, more formal assessments. Observations of the child’s behaviour and performance were typically within the clinic setting and, to a lesser extent, within the school or home environment. Standardised assessments were not used routinely. Although most therapists reported using a sensory integrative approach, only very few reported using the SIPT to formally identify the nature of the sensory integration difficulty.

Some settings used structured goal-setting with carers to establish the child’s needs. There has been a consistent drive in recent years to develop family-centred practice, with services working in partnership to address concerns and goals. 64 From this survey, it appears that some carer concerns were addressed and that most were satisfied with the advice given, but many would have liked more input. After initial assessment, most services generated a report with advice on strategies for managing SPDs in everyday life. For some parents, contact with services ended at this point or they were referred to a workshop/training group. For others, additional OT support was received with further discussion and advice, the offer of loan of suitable equipment to support their child’s needs or some sessions of one-to-one intervention.

Intervention

Therapists appeared to have a two-pronged approach: (1) giving strategies and advice to support carers to manage their child’s SPDs to improve self-regulation, and (2) attempting to address SPDs directly with the aim of remediation of the dysfunction and underlying nervous system. 65

Primary objective

The primary objective was to determine the impact of SIT on irritability and agitation, as measured by the corresponding subscale of the Aberrant Behavior Checklist (ABC). 67

Secondary objectives

Secondary objectives were included to examine:

• the effectiveness of SIT for additional behavioural difficulties (e.g. hyperactivity/non-compliance, lethargy/social withdrawal, stereotypic behaviour and inappropriate speech)

• the impact of SIT on adaptive skills, functioning and socialisation

• sensory processing scores post intervention (i.e. at 6 months) as a potential mediator of any association observed between SIT and the primary outcome at 12 months

• age, severity of SPDs, adaptive behaviour, socialisation and comorbid conditions as potential moderators of any association between SIT and irritability/agitation, adaptive functioning (child) and carer stress

• the impact of the intervention on carer stress and quality of life

• cost-effectiveness (including direct intervention costs, costs of health, social care and education services, carer expenses and lost productivity costs)

• fidelity, recruitment, acceptability, adherence, adverse effects and contamination in a process evaluation conducted alongside the main trial.

Inclusion criteria

Participants were eligible to take part if they:

• had a diagnosis of autism (as documented on medical and/or educational records) or had probable or likely autism (defined as undergoing assessment within the local autism pathway)

• were aged 4–11 years at the start of the trial

• planned to remain in mainstream primary education until the primary outcome time point (i.e. 6 months post randomisation)

• had definite or probable SPDs defined as (1) definite dysfunction on at least one sensory dimension (defined as all domains except social participation) and the total score on the SPM44 or (2) at least a probable dysfunction on two or more sensory dimensions and the total score

• provided carer consent/child assent.

Exclusion criteria

Participants were excluded if they were:

• currently undergoing or had previously undergone SIT

• currently undergoing applied behaviour analysis therapy.

Informed consent

It was likely that potentially eligible children could have a range of impairments, including a degree of intellectual disability (ID). Provided that all inclusion criteria were met and exclusion criteria were not met, no child was excluded for this reason or for other comorbid conditions. In accordance with good clinical practice (GCP), written informed consent was taken from each child’s carer (i.e. their parent or legal guardian) and assent from children before any trial-related activities were undertaken. In signing the consent form, carers also consented to their participation in the trial (including completion of some outcome measures) and for the study team to contact the child’s school to ask for feedback on the child’s behaviour if necessary. Schools were also asked to complete the Aberrant Behavior Checklist – irritability (ABC-I) at the 6-month time point. Consent could be withdrawn by carers at any time. To complement the carers information sheet, age-appropriate information was presented to the child where appropriate and the views of children capable of expressing an opinion considered. Children deemed to have capacity and who were able to write were asked to sign an age-appropriate assent form.

Consent was sought from carers to video-record intervention sessions to assess fidelity of intervention delivery and for supervision feedback. Carers could also consent to video-recordings being used in future research or for training opportunities. In addition to intervention sessions, ADOS53 assessments were video-recorded to evaluate consistency among ADOS assessors. Refusal of consent to any video-recording did not affect the participants’ eligibility to otherwise take part in the trial. Carers received a copy of any consent form that they signed, with the original filed in the site file and a further copy held with participants’ clinical notes.

Once recruited to the trial, treating therapists remained free to give participants alternative treatment to that specified in the protocol if it was in the child’s best interest; however, the child would remain a participant in the trial for the purpose of follow-up and data analysis according to original treatment allocation. Further informed consent was taken from a small number of carers for participation in qualitative interviews. In compliance with Welsh-language requirements, the PIS, consent form and any other required participant documentation were available in Welsh. However, all outcome measures were available in English only, as none was validated in Welsh.

Risk assessment

A trial risk assessment was completed to assess:

• known and potential risks and benefits to participants

• the level of intervention risk compared with standard practice

• how risks were to be minimised/managed.

The SenITA trial was categorised as low risk (i.e. no higher risk than standard medical care). This categorisation was used to inform the level and focus of monitoring activity undertaken during the trial.

Sensory integration therapy

The intervention was delivered in regional clinics by OTs trained in SIT (typically NHS band 7) (see Chapter 5 for a description of the intervention). None of the therapists delivering the intervention delivered therapy to participants in the control arm. Fidelity of intervention delivery was assessed using the ASI intervention fidelity measure (see Chapter 8). 23

Comparator

To define usual care, scoping work in the form of a short survey, interviews and focus groups with therapists and carers was completed before the trial opened to recruitment (see Chapter 2). Carers also recorded usual care using a diary that was specifically designed for the trial. This could be completed either on paper or electronically. Usual care for autism was also recorded more generally, including any contact with NHS services (e.g. SLT, paediatrics and CAMHS). Therapists recorded usual care in accordance with local policy.

Screening and baseline measures

Primary outcome measure

The primary outcome was irritability/agitation at 6 months post randomisation, as measured by the corresponding ABC subscale (i.e. the community version ABC-I with 15 items67,68). The ABC-I was also measured at baseline and at 12 months post randomisation. The primary outcome comparison was based on carer ratings of ABC-I. However, teacher/teaching assistant ratings of ABC-I (assessed at 6 months post randomisation only for intervention and control arms) were also explored.

Secondary outcome measures

Trial oversight and patient and public involvement

The trial was conducted in accordance with Centre for Trials Research standard operating procedures. Oversight was provided by the Trial Management Group (including the investigators and directly employed trial staff), the Trial Steering Committee, the Data Monitoring and Ethics Committee and an advisory group of family carers. The advisory group, including the patient and public involvement co-investigator, provided feedback on study materials, recruitment and retention plans, and continue to advise on dissemination of results to stakeholders.

Staff training

Good clinical practice training was provided to staff carrying out trial-specific procedures (e.g. recruitment/consent, data collection and intervention delivery). Tailored training materials were created for all aspects of intervention delivery and data collection (see Intervention data collection). Principal investigators (PIs) and intervention therapists received an individualised session on delivering the intervention in accordance with the study protocol, following the detailed description in the intervention handbook (see Chapter 5). Therapists clearly marked participant consent on their clinical notes.

Sensory integration therapy sessions

Safeguarding procedures were in place for reporting or raising any concerns that arose during therapy sessions, and these procedures were detailed in the intervention handbook. It was the responsibility of each therapist to make participant’s session appointments. Future appointments were agreed at the child’s first assessment visit and recorded on an appointment card for carers to keep. The card held contact details for the therapist, including who was the first point of contact for carers and an emergency contact number. It was also the therapist’s responsibility to issue reminders of upcoming appointments and to reschedule appointments as necessary. If a participant could not attend a session as intended, up to two attempts could be made to reschedule within ± 3 days of the original appointment. If a session could not be rearranged to fit in that time frame, then it would be foregone.

For each session, therapists documented and video-recorded (where consent was given) the intervention delivered. Video-recordings were uploaded to password-protected secure OneDrive (Microsoft Corporation, Redmond, WA, USA) folders for fidelity analysis. Once uploaded, video-recordings were deleted from the recording device.

Assessment time points

Outcomes and follow-up time points (Table 6) were the same for both the intervention and the control group and were performed as close as possible to the required time point.

Data collection

A manual (Table 7) was created for research assistants carrying out any data collection. The document was created to ensure consistency of data collection from different research assistants and research network staff. Data were entered directly into the trial SQL (Structured Query Language) database via an online link, with the exception of the ADOS and VABS II, which were scored on paper first and then relevant total scores entered online.

The study team contacted carers to rearrange any missed follow-up data collection appointments as soon as possible. For those carers who were unable to attend appointments in person (or stay for the duration), there was the option to complete any remaining measures at home using a booklet that was to be returned to the study team. Those carers who were unable or unwilling to take the booklet home were given the option of answering key questions over the telephone. All individuals carrying out data collection (i.e. research assistants and clinical study officers from research networks) were appropriately qualified and completed relevant training. Clarification of any clinical questions relating to SIT that arose during appointments was deferred to the co-chief investigator.

Screening and baseline visit

There were two screening/baseline visits, which were split to reduce any burden on the child taking part. At the first meeting, the study team met with only the carer to discuss the study in detail and to take consent. Following consent, the carer completed the screening measure (i.e. the SPM) to confirm eligibility. If eligible, the carer was immediately asked to complete the baseline outcome measures (see Table 6). If not eligible, the carer was thanked for their interest and no further trial involvement took place. If the carer could not complete the baseline assessment immediately, then it was arranged for a convenient time.

The second meeting was held during the following week to complete an ADOS assessment (video-recorded where consent was given) with the child and to carry out randomisation. Carers in both arms were given access to a diary at this point to report health professional contacts and any home programmes or self-initiated activities for their child (Figure 3).

FIGURE 3.

Flow diagram of data collection procedures.

Follow-up data collection

Follow-up assessments for all participants were conducted 6 and 12 months after randomisation with a ± 4-week window. Reminders were sent out to participants/carers to complete the diaries using the text message service Esendex (Commify UK Ltd, Nottingham, UK). The text message system was also used to supplement data collection where necessary.

Intervention data collection

Therapists delivering the intervention were provided with written guidance in the form of a manual on how to collect and record all aspects of trial data (Table 8). The manual was very detailed and gave step-by-step instructions on using trial equipment, including screen shots and pictures to aid explanation. Each intervention therapist was given access to an iPad, which had been preloaded with links to the purpose-built trial SQL database and individual OneDrive accounts. Intervention therapists were also given access to a digital camera and tripod to record the intervention sessions.

Once a participant had been randomised to receive the intervention, the next available therapist in their area was sent an e-mail to notify the therapist that a child had been allocated to them. The therapist then logged into their personal OneDrive space to pick up contact details for the participant and to arrange their first appointment. During the initial assessment, the therapist used their personal log-in details to access the secure trial database. The therapist completed online forms for the COPM48 and clinical observations. Therapists also completed the SIPT54 with each child. The SIPT was not available electronically and, therefore, therapists filled out a paper booklet with the child. To score the test, the booklet was sent to the trial team, who held the scoring software. To ensure that data were sent securely, therapists uploaded scans of each page of the booklet to their OneDrive account for the trial team to score. The SIPT report was then sent back to the therapist via OneDrive.

At the start of each intervention session, the therapist set up the camera and tripod to record the session. All recordings were transferred from the camera to the iPad and were uploaded to OneDrive. Recorded sessions could then be shared via OneDrive with relevant mentors and fidelity assessors, who were also given university-allocated single-use OneDrive log-in details. The manual included a flow diagram that each therapist could refer to (Figure 4).

FIGURE 4.

Flow diagram for therapist data collection. CTR, Centre for Trials Research.

There were occasions when technology failed (e.g. when camera batteries ran out and when the camera failed to record), which led to some missing recordings. Intervention therapists received detailed training alongside the manual; however, there were several occasions where the trial team had to go out to site to help with technical issues. All data transfer and storage strictly adhered to GCP data handling and security specifications.

Qualitative data collection

Analysis population

Participants were analysed based on the arm to which they were randomised. Main analyses were based on a modified intention-to-treat population, with those providing outcome data included in the analysis. A full intention-to-treat analysis set comprised all participants in the group to which they were randomised, with missing outcome data imputed using multiple imputation. This analysis set served as a sensitivity analysis to the primary outcome. Finally, a complier-average causal effect population comprised participants with outcome data in the group to which they were randomised, accounting for those who received the intervention as intended. This analysis set also served as a sensitivity analysis to the primary outcome.

Main analysis

The primary outcome analysis compared trial arms by fitting a linear regression to ABC-I scores at 6 months post randomisation, including a trial arm indicator (i.e. usual care/intervention), adjusting for baseline ABC-I, recruitment region (i.e. South Wales/South England region), severity of SPD (i.e. some problems/definite dysfunction) and sex of the child (i.e. male/female). We found no evidence of therapist clustering and, therefore, fitted single-level models only. Findings are reported as adjusted mean differences with associated 95% CIs and p-values.

Most secondary outcomes were analysed similarly. Teacher-assessed ABC-I was collected at 6 months only and, therefore, no baseline ABC-I adjustment was made for this analysis.

Subgroup analysis

We explored differential intervention effects on the ABC-I at the 6-month post-randomisation time point by age, severity of SPDs and comorbid conditions [i.e. attention deficit hyperactivity disorder (ADHD), ID or other neurodevelopmental/genetic conditions], site (n = 7), region (n = 2) and sex of the child. Our primary model was extended by including subgroup × trial arm interaction terms. These subgroup effects were also explored for APSI scores at 6 months post randomisation.

Sensitivity and exploratory analysis

Health economic analysis

The health economic analyses were carried out on an intention-to-treat basis. The main analyses compared cost and cost-effectiveness of SIT with usual care at the 6-month follow-up, and were conducted from an NHS and Personal Social Services (PSS) perspective. Secondary analyses adopted a societal perspective, adding education services, carer out-of-pocket expenses and lost productivity to NHS and PSS. The ABC-I at 6 months was used, in turn, as a measure of effectiveness in a series of cost-effectiveness analyses. The main cost-effectiveness measure was incremental cost per point improvement on the ABC-I (at 6 months).

Qualitative analysis

Qualitative data were analysed using thematic analysis43 to search across the data set to find repeated patterns of meaning and to identify key themes and subthemes. Double coding was carried out until consensus was reached. The qualitative component was designed using the principles of the Critical Appraisal Skills Programme qualitative checklist to ensure the quality of qualitative research. 79 Additional detail regarding qualitative methods is provided in Chapters 8 and 9.

Process evaluation

A process evaluation (see Chapters 8 and 9) was conducted to examine (1) recruitment, (2) retention, (3) fidelity, (4) acceptability, (5) adherence, (6) adverse effects and (7) contamination of intervention delivery within the trial. Recruitment rates/patterns, attendance and intervention fidelity were assessed quantitatively, whereas the acceptability of study processes and measures, intervention implementation and acceptability, and the role of contextual factors were examined through the qualitative interviews with therapists and carers.

Usual care

Usual care was expected to comprise use of sensory-based strategies not meeting fidelity criteria for SIT and no treatment at all in a significant proportion of cases (established via methods described in Chapter 2 and from carer diaries).

Recruitment of participants

Recruitment feasibility criteria would be met if at least 70% of those approached were eligible and at least 50% of those eligible were willing to be randomised.

Retention of participants at 6 months post randomisation

If dropout at the 6-month follow-up time point exceeded 20%, then we proposed to reassess the sample size calculation and associated implications for feasibility of recruitment.

Contamination (i.e. the extent to which participants allocated to the control arm received treatment/therapy similar to that in the intervention arm)

Carer-completed diaries were qualitatively assessed to determine whether or not usual care was sufficiently different from the intervention. Broadly defined, this criterion would be considered met if carers in the control arm did not receive any intervention meeting criteria for full SIT.

Adherence (i.e. receipt of intervention by participants allocated to the intervention arm)

An ‘effective’ dose for SIT has not yet been established. However, attending 13 of a possible 20 sessions delivered during the intensive intervention phase (two-thirds) was felt likely to indicate sufficient exposure, based on clinical experience and previous trials (see Chapter 5). 13,33,34,80

Fidelity of intervention delivery

A random sample of SIT sessions (lasting at least 15–20 minutes) delivered by individual therapists were evaluated using the fidelity measure. 23 To demonstrate adequate fidelity, therapists were expected to score least 80 out of 100 on the process fidelity measure for at least 80% of sessions rated. Each clinic was also assessed for suitability (of environment and equipment) prior to participant recruitment. 23

Sample size assumptions

To confirm the accuracy of the sample size calculation and other features of our proposed design, we obtained an estimate of the following: the proportion of participants providing primary outcome data, the SD of the ABC-I67 at the primary outcome time point (i.e. 6 months) in both trial arms, and the intracluster correlation coefficient (ICC) of therapists within participants for the ABC-I at the primary outcome time point (intervention arm only).

Adverse effects

We studied any adverse effects of study participation.

Primary outcome

The primary outcome for the internal pilot comprised those with prespecified progression criteria, as set out below.

Secondary outcomes

Secondary outcomes comprised outcomes without formal progression criteria (see Results: secondary outcomes).

Usual care

A specific requirement of the commissioning brief for this trial was to demonstrate that the intervention was sufficiently different from usual care. At the point at which we reported on the pilot evaluation, that is after the first 40 participants were recruited, survey data had been received from 39 therapists and 131 carers with regard to their experiences of delivery and receipt of usual care. At this stage of the trial, we had also undertaken two face-to-face focus groups (group 1 had two participants and group 2 had three participants) and six (five individual and one shared) telephone interviews with therapists (n = 7) in trial regions only. The final survey results and a detailed qualitative exploration of usual care are provided in Chapter 2; however, characteristics of those who responded during the pilot evaluation period are given in Results: secondary outcomes, as is a brief summary of findings.

Recruitment

We planned to recruit participants from a wide variety of sources, namely via secondary care services (e.g. OT, paediatrics, CAMHS and specialist autism/neurodevelopmental services), schools, third-sector organisations and self-referral. We anticipated that most referrals would come from OT and/or specialist neurodevelopmental services. However, most expressions of interest came from third-sector organisations (e.g. local NAS support groups) and via self-referral. Reasons for this are likely to be complex, but might include, for example, the length of time from discharge from OT or specialist autism/neurodevelopmental services. Some specialist teams also expressed concerns about recruiting participants who were yet to receive a diagnosis (i.e. children were deemed eligible once they were on the local neurodevelopmental pathway provided that all other eligibility criteria were met; see Chapter 3 for inclusion criteria). In England, in particular, some OT services expressed concern that referring patients to the trial could have raised expectations of families in the area with regard to availability of interventional services, which they were not currently commissioned to deliver.

Retention

Contamination

We received 58 diaries and 107 SMS text messages (i.e. via the text service used to collect diary information; see Chapter 3 for a full description) from participants regarding their use of other services following randomisation. In the control arm, no participant indicated that they had received any intervention meeting criteria for SIT (i.e. according to fidelity criteria23 and/or face-to-face therapy more than once per week).

Adherence/acceptability

Of the 20 participants randomised to receive the intervention, eight participants were expected to have completed the intervention by the analysis time point, and we received attendance data for seven of these participants. All seven participants attended at least 13 of the 20 intensive sessions.

Fidelity

All trial clinics were assessed as having met structural fidelity criteria23 prior to opening for recruitment (see Chapter 8). Our outcome of interest was the number of therapists scoring at least 80 out of 100 on the process fidelity measure for at least 80% of sessions rated (expressed as a proportion of therapists who provide at least one video session). Fidelity scores were available for a total of 24 session recordings across six therapists. The mean fidelity score was 81.8 (SD 12.84). In total, 15 out of 24 (62.5%) sessions were scored at least 80 out of 100. The general impression score (measured via a single question that indicates whether or not the therapist intentionally applied ASI intervention theory and methods) was scored ‘no’ for one session only. The corresponding fidelity score for this session was 45 out of 100. All other sessions were scored ‘yes’. Complete fidelity ratings across all therapists are presented in Chapter 8.

Other sample size assumptions

Adverse effects

No adverse effects were reported.

Sample size implications

It would have been premature to use the SD and correlation estimates gathered at this time point to make alterations to the target sample size (as prespecified and described in Aims and objectives). Using a deflation factor of 1 – ρ2,76, Table 14 illustrates how the sample size could be reduced by accounting for this correlation (fixing all other assumptions).

At the request of the Health Technology Assessment programme and as advised by the Independent Data Monitoring and Ethics Committee, this correlation was revisited at two further time points: (1) when primary outcome time point data were available for 38 participants (correlation between baseline and 6-month ABC-I scores 0.806, 95% CI 0.654 to 0.895) and (2) when outcome data were available for 53 participants (correlation between baseline and 6-month ABC-I scores 0.69, 95% CI 0.52 to 0.80). This high level of correlation supported a reduction in the target sample size from 216 to 138 (a correlation of ≥ 0.6 indicates a sample size of ≤ 138 with 90% power).

Initial greeting and scene-setting

We began with an initial greeting, which involved scene-setting, a summary of the purpose of the study and the intervention, and checking consent for recording of sessions.

Clinical observations

Unstructured observations, including descriptive assessment of level of arousal, postural control, organisation of behaviour and motor planning ability, were carried out. In addition, a set of structured observations based on sensory integration principles, covering an assessment of performance/difficulties in proprioception, vestibular processing, motor planning and postural and ocular control, were carried out. Specific guidelines on administration, interpretation and scoring were written in the intervention handbook to ensure consistency between therapists.

The Sensory Integration and Praxis Test

The SIPT54 provides detailed information on the sensory processing and practice (i.e. motor planning) abilities that underlie learning and behaviour, and facilitates identification of specific areas of difficulty. There are 17 subtests that fall into four overlapping areas. The tests that are most closely associated with each of the areas are as follows:

• motor-free visual perception (i.e. the ability to visually perceive and discriminate form and space without involving motor co-ordination), including –

  • space visualisation

  • figure-ground perception.

• somatosensory perception (i.e. tactile, muscle and joint perception, encouraging the child to ‘feel’ rather than to ‘see’), including –

  • manual form perception

  • kinaesthesia

  • finger identification

  • graphesthesia

  • localisation of tactile stimuli.

• praxis (i.e. the ability to motor plan in a range of ways), including –

  • design copying

  • postural praxis

  • bilateral motor control

  • praxis on verbal command

  • constructional praxis

  • motor accuracy

  • sequencing praxis

  • oral praxis.

• sensorimotor (i.e. sensory integration at multiple levels), including –

  • standing/walking balance

  • postrotary nystagmus.

Occupational profile

The occupational profile questions were given to the carer to complete between assessment sessions. The occupational profile questions also included information about the child’s play preferences and strengths, their school, their medical and developmental history, and infant and current behavioural patterns. The therapist encouraged the carer to think of functional issues that were affecting their child’s performance for discussion at the next session.

Sensory Processing Measure

The SPM44 was used as a screening measure to ensure a minimum level of SPD as an inclusion criterion. The SPM provides eight standard scores for the following dimensions: (1) social participation, (2) vision, (3) hearing, (4) touch, (5) body awareness (proprioception), (6) balance and motion (vestibular function), (7) planning and ideas (praxis) and (8) a total sensory symptoms score. Scores on each of these dimensions were classified as typical, some problems or definite dysfunction. For the purposes of the SenITA trial, SPD was defined as either (1) a definite dysfunction on at least one sensory dimension (all domains except social participation) and the total score or (2) at least a probable dysfunction on two or more sensory dimensions and the total score. Treating therapists were able to access these scores to aid hypothesis development and delivery of the intervention.

Canadian Occupational Performance Measure

The COPM48 is an individualised measure designed to detect change in a child’s (and carer’s) self-perception of occupational performance over time. The COPM comprises the following four key steps.

Hypothesis development

Using the assessment data as indicated on the interpretation tool, therapists developed a hypothesis about which areas of dysfunction were particularly influencing the child’s performance and participation.

Goal development

Therapists were encouraged to use the FEAST (Functional, Expectation, Action, Specific condition, Timeline) model to assist in articulating the specific goals identified by the child/carer in the COPM. The functional element asks ‘why’ a goal is required (i.e. the area of occupation to be addressed, such as to dress himself/herself). The expectation element specifies ‘who’ (i.e. ‘X will . . . ’). The action element looks at ‘what’ is to be undertaken (e.g. ‘X will put his/her shirt and shorts on’). The specific condition element considers ‘how’ this will be carried out (e.g. with three verbal prompts once the clothes have been laid out). Finally, the timeline element considers ‘when’ this will be carried out (e.g. by the end of session 24).

Summary of strengths, participation challenges and goals

The child’s strengths were summarised from the occupational profile and discussion with the carer. Participation challenges were summarised from the occupational profile and carer discussion. Therapists then produced a summary of their interpretation of the assessment results. From the assessment results and hypothesis of underlying difficulties, the therapist selected specific SIPT items and clinical observations to reassess at the end of therapy (proximal outcomes), in addition to the child or carer re-rating their performance and satisfaction with their goals on the COPM (distal outcomes).

Session 1

Key steps and activities were as follows:

• Goals and hypotheses for underlying sensory–motor difficulties were checked prior to the session.

• The room was set up with provisional equipment and toys that were in keeping with hypothesis from assessment data, considering strengths, challenges and needs.

• The video camera was checked and switched on.

• Carer and participant were greeted. The child was made comfortable with a simple explanation of the activities that they would be doing and how these games would help them.

• The therapist used their professional and clinical reasoning to apply each of the ASI fidelity principles to ensure that the challenges were appropriate for the child’s age, ability levels and needs.

• The carer was encouraged to participate in the session or to watch as they felt comfortable. The level of carer engagement was negotiated with the carer during the telephone call (see Fidelity assessment tool).

• There was no set sequence, manner or format for incorporating all of the key principles and implementing tailored activities. However, the flow of the session could be divided into three phases of beginning, middle and end. Sessions lasted about 50–55 minutes, including 5–10 minutes to have a short discussion with the carer about activities or strategies to try at home. Advice sheets may have been given at this point, if appropriate.

Intervention sessions 2–20

Intervention sessions 2–20 continued as with session 1 above and included a review of any activities carried out since the last session with the carer.

Intervention sessions 21–23

After the twice-weekly intense phase of therapy, instead of sessions ceasing altogether, a tailoring phase of therapy sessions once per fortnight was delivered and continued with planning as previously detailed in Intervention session content, Session 1, Beginning of the session. This was designed to aid transition from therapist-directed intervention to carer-supported management of the child, and provided an opportunity to discuss any issues that arose to jointly think of solutions.

Intervention session 24

For session 24, the SIPT or clinical observation items identified as proximal outcome measures were reassessed. The COPM was re-administered to determine how the carer (and the child if they had capacity to do this) initially rated the five goals for performance and satisfaction with that performance. The therapist then explained to the participant that there would be two follow-up telephone calls to monitor progress and to problem-solve any issues that might arise. Strategies and advice for continuing the progress made during the face-to-face sessions were discussed.

Intervention session 25

Session 25 comprised a telephone call, which was carried out 1 month following the last intervention session. The purpose of the telephone call was to:

• follow up on the child’s performance and behaviour in the past month

• determine whether or not any advice and strategies previously given were successful

• offer support and advice for any ongoing issues.

Possible prompt questions were provided to therapists to discuss with participants.

Intervention session 26

Session 26 was the same as session 25, but with the additional purpose of closure, as this was the final contact with the carer. The purpose of this telephone call was twofold. First, to follow up (as in the previous telephone call) and discuss any changes to behaviour and to determine whether or not any advice and strategies were carried out. Second, to consider the goals chosen at the beginning of therapy and informally discuss the progress of the child, and, finally, to thank the carer for their engagement in the process (prompt questions were provided to be used as needed).

A brief written summary of the intervention was provided to carers/schools if requested. The summary did not include the outcome of intervention, as this was not appropriate within the research trial; however, it did include the goals that were set prior to intervention, including the specific target areas that were concentrated on.

Information regarding fidelity of intervention delivery, participant adherence and therapist experience of therapy and of trial participation is provided in Chapter 8.

Initiation of sensory integration therapy

Of the 69 participants randomised to the intervention arm, 61 (88.4%) initiated SIT and 49 out of 57 (86.0%) initiated prior to their primary outcome follow-up. In those who initiated SIT, the median time from randomisation to initiation was 48 (IQR 32–85) days (Figure 6).

FIGURE 6.

Time from randomisation to initiation of SIT in days (for the 61 participants allocated to and initiating SIT). a, Figure includes 61 out of 69 participants who initiated SIT.

Implementation of sensory integration therapy

Overall, 54 (78.3%) participants allocated to the intervention arm received at least 13 SIT sessions during the intensive stage, with 38 out of 57 (66.7%) participants with primary outcome follow-up data receiving at least 13 SIT sessions during the intensive stage. The median number of SIT sessions received was 20 (IQR 16–21) and the median number of SIT sessions received prior to primary outcome follow-up (in those with primary outcome data) was 18 (IQR 10–21).

Receipt of SIT sessions varied considerably between participants. Figure 7 highlights different patterns of SIT session receipt for six participants, with Table 18 providing summary statistics for each participant. Participants A, B and C display typical patterns of participants who initiated SIT early (i.e. between 20 and 48 days post randomisation) and attended most of their sessions. Participant A received two SIT sessions after their primary outcome data were collected. Participant D initiated SIT late (i.e. at 119 days post randomisation) and shows a typical pattern of someone who attended most of their sessions, but received a substantial number of them after their primary outcome data were collected. Participant E displays a pattern of someone who started SIT but who stopped early on. Participant F is a participant who started SIT 348 days post randomisation and, therefore, all sessions received were after collection of their primary outcome data.

FIGURE 7.

Panel plot of session attendance for six different participants. (a) Participant A; (b) participant B; (c) participant C; (d) participant D; (e) participant E; and (f) participant F.

Outcome variables are summarised in Tables 19 and 20 at baseline and at 6 and 12 months post randomisation. For the primary outcome (i.e. ABC-I), Figure 8 illustrates baseline and 6-month scores by study site.

FIGURE 8.

ABC-I scores at baseline and 6 months, by site. (a) Site 1, Aneurin Bevan; (b) site 2, Cardiff and Vale; (c) site 3, Cornwall; (d) site 4, Cwm Taf; (e) site 5, Hywel Dda; (f) site 6, Pace Centre; and (g) site 7, Swansea Bay.

Our primary analysis included 106 participants. At 6 months post randomisation, the mean ABC-I score was 18.8 (SD 10.48) in the control arm and 18.5 (SD 9.33) in the intervention arm. The correlation between baseline and 6-month ABC-I score was 0.698 (95% CI 0.585 to 0.784). After adjusting for baseline ABC-I, sex of child, SPM category and region, the between-arm difference was 0.40 points higher in the intervention arm than in the control arm (95% CI –2.33 to 3.14; p = 0.77) (Table 21).

Behavioural measures

At 6 months post randomisation, secondary outcomes relating to behaviour were available for 85 participants. There was no evidence of any differences between arms across the lethargy/social withdrawal, stereotypic behaviour, hyperactivity or inappropriate speech domains of the ABC (see Table 21).

At 12 months post randomisation, secondary outcomes relating to behaviour were available for 60 participants. Similarly, we found no evidence of any differences between arms across these measures (Table 22).

Adaptive behaviour measures

The analysis of adaptive behaviour measures at 6 months included between 63 and 72 participants. For the composite measure and communication, socialisation and daily living skills domains, there was no evidence of any differences between arms (see Table 20). At 12 months, the communication domain score was 5.69 points lower in the intervention arm than in the control arm (95% CI –9.87 to –1.51; p = 0.009). There was no evidence of any differences between arms for the composite measure or other domains (see Table 20).

However, an error in administration of this measure for the majority of participants meant that we had introduced an artificial ceiling effect and scores were potentially less sensitive to change. Start points of each subdomain depend on the child’s chronological age, and age ranges are given as indicators of where performance might be expected to lie. Indicative upper limits are not intended to be used as cut-off points during administration, but were, for most participants, potentially limiting the magnitude of change that could be observed. It is uncertain whether or not this error masked any differences between arms with regard to adaptive behaviour domains. However, given the absence of any marginal effects (excepting a negative effect on communication at 12 months, which may also be reflective of multiple testing) and an absence of effect across other outcomes, this is perhaps unlikely.

Parenting stress

At 6 months, APSI scores were available for 83 participants. We found no evidence of between-arm differences for this measure. Similarly, we found no evidence of between-arm differences at 12 months.

Health utility and quality of life

At 6 months, EQ-5D and CarerQoL scores were available for 83 participants. Similar to other outcomes, we found no evidence of any differences between arms with regard to health utility, generic health-related quality of life or carer quality of life (see Table 20). Similarly, we found no evidence of between-arm differences at 12 months (see Table 20).

Teacher-assessed Aberrant Behavior Checklist – irritability

The mean of the difference between carer- and teacher-assessed ABC-I at 6 months was 10.28 (SD 14.48), with a 95% limit of agreement of –18.12 to 38.68. There was strong evidence of difference between the two measures (p < 0.001), indicating that, on average, carer-assessed ABC-I scores were higher (Figure 9).

FIGURE 9.

Carer- and teacher-assessed ABC-I. Mean difference of 10.28; 95% limits of agreement of –18.12 to 38.68.

Canadian Occupational Performance Measure

Carer-rated goal performance and satisfaction were available for 43 and 42 participants, respectively. There was strong evidence of an increase in score between session 1 and session 24 for both measures (p < 0.001 for both). The mean change in score was 2.75 (95% CI 2.14 to 3.37) for carer-rated performance and the mean change in score was 3.34 (95% CI 2.63 to 4.40) for carer-rated satisfaction. Child-rated outcomes were available for between 10 and 12 participants. There was evidence of an increase in score for both performance and satisfaction (p = 0.01 for both) (Table 23).

Subgroup effects for the Aberrant Behavior Checklist – irritability

We found evidence of a differential effect of SIT on the ABC-I at 6 months by region (intervention arm × South England region interaction = 9.77, 95% CI 4.04 to 15.49; p = 0.001), indicating higher levels of irritability in the South England intervention arm at 6 months.

There was some evidence suggesting that there was also a differential effect of SIT on the ABC-I at 6 months by sex of child (intervention arm × female = 6.42, 95% CI 0.00 to 12.85; p = 0.050), with the intervention being less effective in girls than in boys.

There was also some evidence to suggest that the intervention was more effective in those with ADHD (intervention arm × ADHD present = –6.77, 95% CI –13.55 to –0.01; p = 0.050). We found no evidence of a differential effect of SIT on the ABC-I at 6 months by SPM range or general comorbidity, or by ID and other neurodevelopmental or genetic condition specifically (Table 24).

Subgroup effects for the Autism Parenting Stress Index

Similarly, we found some evidence of a differential effect of SIT on carer stress (using the APSI) at 6 months by region (intervention arm × South England region interaction = 7.01, 95% CI 0.45 to 13.56; p = 0.04) (Table 25). There was also some evidence to suggest that there was a differential effect of SIT on APSI at 6 months by the presence of other neurodevelopmental or genetic conditions (intervention arm × neurodevelopmental/genetic condition present = –9.53, 95% CI –18.08 to –0.98; p = 0.030).

We found no evidence of a differential effect of SIT on the ABC-I at 6 months by sex of child, SPM range or general comorbidity, or by ID and ADHD specifically (see Table 24).

Accounting for intervention receipt

Figures 10–12 highlight that scores on the ABC-I at 6 months were generally lower (i.e. better) in those who initiated and received SIT than in those who received usual care.

FIGURE 10.

Distribution of ABC-I scores at 6 months in those who did and did not initiate SIT.

FIGURE 11.

Distribution of ABC-I scores at 6 months in those who did and did not receive at least 13 SIT sessions during the intensive phase.

FIGURE 12.

Box plot of ABC-I scores at 6 months for those who received differing levels of SIT.

Table 26 provides results from the sensitivity analyses accounting for receipt of SIT. Although the point estimates change slightly, the conclusions drawn from the primary analysis remain unaltered.

Impact of the COVID-19 pandemic

Changes of underlying event rates and/or treatment effects may change

Primary outcome data were collected for 11 participants during the intrapandemic period.

Although exclusion of those participants with primary outcome data available during the intrapandemic period led to an increase in our point estimate (i.e. between-arm adjusted mean difference of the ABC-I at 6 months), the CI remained wide and the conclusions drawn on our primary analysis remain unaltered (Table 27). Baseline to 6-month follow-up ABC-I correlation was slightly higher following exclusion of participants who provided 6-month data during the intrapandemic period. Figure 13 illustrates differences in ABC-I scores at 6 months by arm and by time period.

TABLE 27 - Comparison of primary analysis and primary analysis in those with primary outcome data available prior to the COVID-19 pandemic (pre 23 March 2020)

Analysis	n	Estimate	95% CI	p-value	ABC-I correlation between baseline and 6 months (95% CI)
Original primary analysis	106	0.40	–2.33 to 3.14	0.770	0.698 (0.585 to 0.784)
Primary analysis on participants followed up pre pandemic	95	0.50	–2.40 to 3.39	0.733	0.712 (0.596 to 0.799)

FIGURE 13.

Box plots of ABC-I scores at 6 months by arm and by time period.

Intervention delivery

The intervention arm received 24 face-to-face and two telephone sessions, which were used as the basis for calculation of intervention costs.

Usual care

Data on usual-care services and support were recorded using the CSRI,75 and were adapted to reflect the type of data collected and to record service use that would facilitate cost estimation. Usual care involved contact with OT, autism-specific services and any other NHS services.

NHS/Personal Social Services

Information about the use of services by participants and carers was collected from carers at baseline and at 6- and 12-month follow-up, covering a retrospective 6-month period. Carers were asked to record contact with various services and professionals (for themselves and their child). At each time point, service use since the previous assessment was recorded.

Productivity losses

Data regarding the number of half-days taken off work by the main carer to provide support were collected (last 6 months).

Valuation of resources

All resources were costed by applying a unit cost (GBP) for the financial year 2020–21. Total resource use was calculated for each participant.

Sensory integration therapy sessions

The unit cost of an occupational therapist was derived using established unit costing principles. 83 Therapists were NHS band 7 and national insurance, superannuation contributions and overheads were added to the average salary. We used protocol-driven information on intervention sessions. A total of 24 sessions were delivered face to face and two were conducted over the telephone. All sessions lasted approximately 1 hour. To derive the cost of the intervention, we multiplied occupational therapist time spent delivering the intervention by the unit cost of an occupational therapist. The total cost of intervention delivery over 6 months was £1538.16 (Table 29), which is equivalent to a monthly cost of £256.36 per person.

NHS/Personal Social Services

For all health-care and social care service use data, including voluntary sector, nationally applicable unit costs were attached at 2020–21 prices, or at the most recent price level on which costs were available, to reflect long-run marginal opportunity costs from available public sources. Costs older than the chosen financial year were inflated using the NHS Cost Inflation Index. 82

Costs per unit of measurement for each NHS/PSS service type were taken from the Personal Social Services Research Unit Costs of Health and Social Care compendium. 82 NHS reference costs84 were used for inpatient and outpatient attendances and community-based services. Unit costs not obtained from these sources were estimated directly based on salary of the professional delivering a service, plus employer on-costs, overheads, costs of supervision and equipment/consumables.

Productivity costs

Productivity losses were calculated by multiplying time taken off work by carer’s wage rate. We calculated the hourly rate for each carer by dividing their gross annual income by the number of weeks per year and weekly working hours reported. We then multiplied rate per hour by number of hours in the half-days off work reported at follow-up assessments (Table 30).

Service and support use

Differences in the number of participants using each type of service and the mean number of contacts were compared between arms at baseline, 6 months and 12 months. Resource use patterns are presented descriptively, given that the economic evaluation was focused on NHS/PSS and societal costs and cost-effectiveness and to avoid multiple testing.

Analysis of costs

The health economic analysis was an available-case analysis. Costs were calculated using data on the type, number and length of contacts. Mean costs were analysed using regression analysis and bootstrapping. We analysed differences between arms in health-care (i.e. NHS/PSS) or societal costs in secondary analyses at 6 months by regressing the total cost on treatment allocation, baseline ABC-I, recruitment region, severity of SPDs, sex of child and baseline costs. To mitigate effects of data skewness, non-parametric bootstrapping methods were used to estimate 95% CIs for mean costs with 10,000 resamples. No discounting was needed, as we assessed cost over periods of < 1 year.

Cost-effectiveness analysis

Cost-effectiveness analyses were conducted from an NHS/PSS perspective and a wider societal perspective. The primary economic analysis focused on cost-effectiveness of SIT compared with usual care, assessed in terms of cost per point improvement in carer-reported ABC-I. The incremental cost-effectiveness ratio (ICER) was calculated as the difference in mean costs divided by difference in mean ABC-I. 108

Cost-effectiveness acceptability curves (CEACs) were plotted to locate findings in their wider decision-making context. Non-parametric bootstrapping from cost and effectiveness data generated a joint distribution of incremental mean costs and effects for comparators to explore the probability that each is the optimal choice, subject to a range of maximum values (i.e. a ceiling ratio) that a decision-maker might be willing to pay for an additional point improvement in ABC-I. Each CEAC was derived using a net benefit approach. Monetary values of incremental effects and incremental costs for each case were combined, and net monetary benefit was derived as:

where λ is willingness to pay for a 1-point difference in ABC-I, and a and b denote SIT and usual care, respectively. We explored a range of willingness-to-pay values for each outcome.

Sensitivity analyses

An intention-to-treat analysis was carried out (with missing data imputed using multiple imputation) to explore the sensitivity of results using a broader societal perspective, including NHS/PSS costs, education service costs, carer expenses and lost productivity.

Data completeness

At baseline, service use data were available for 138 participants (intervention, n = 69; usual care, n = 69). At 6-month follow-up, data were available for 60% of participants (intervention, n = 48; usual care, n = 35). At 12-month follow-up, 23 and 36 participants provided data in the control arm and intervention arm, respectively.

Resource use

The difference in service use between arms at baseline was broadly similar (Table 31). As shown in Table 32, at 6-month-follow-up, usual-care participants reported considerably higher contact with education services [mean 425.97 (SD 388.37) vs. mean 333.46 (SD 354.10)], slightly greater use of community-based health and social care services and more accident and emergency unit attendances (14.29% vs. 6.25%) than participants in the intervention arm. However, carers in the intervention arm spent, on average, more half-days off work [mean 18.19 (SD 37.22) days vs. mean 4.14 (SD 9.02) days]. At 12-month follow-up, modest but higher utilisation of health-care and social care resources was reported by families in the intervention arm across all categories (Table 33).

Caregiver-reported gross income and the mean annual income was £17,537.31 (SD $15,909.86) for SIT and £18,405.8 (SD £14,956.87) for usual care.

Total costs

The total costs for the child, sibling and carers by sector and perspective were reported as adjusted mean differences with bias-corrected 95% CIs.

NHS/Personal Social Services perspective

At baseline, the main contributor to the total NHS/PSS cost was community-based care. Education-based services constituted the major proportion of societal cost (Table 34). From an NHS/PSS perspective, the mean total health-care and social care cost at 6 months (Table 35) was £1915.2 in the intervention arm and £554.42 in the control arm. The adjusted mean difference was statistically significant (£1325.1, 95% CI £1057.6 to £1592.6). When SIT costs were not included in the total NHS/PSS costs, NHS/PSS costs were significantly higher for usual care than for SIT at 6 months (mean –£177.38, 95% CI –£431.36 to £76.59).

Societal perspective

From a societal perspective, the mean total costs at baseline were £6290.30 and £4862.90 for SIT and usual care, respectively. The adjusted mean difference in costs of £1463.70 was not statistically significant (95% CI –£488.33 to £3415.70) (see Table 34). The mean total societal costs, including intervention at 6 months, were £6784.2 and £6418.9 for the intervention arm and control arm, respectively. The adjusted mean difference was not statistically significant (mean –£353.57, 95% CI –£3518.9 to £2811.7). When intervention costs were excluded, the mean total societal cost at 6 months was statistically higher for usual care than for SIT (mean –£1891.70, 95% CI –£5022.10 to £1238.70). Costs measured at 12 months were higher overall in the intervention arm, with an adjusted difference in total societal care cost of £1141.50 (95% CI £1333.4 to £3616.4); however, the adjusted mean difference was not statistically significant (Table 36).

Cost-effectiveness: 6-month follow-up

In terms of challenging behaviour, NHS/PSS costs were higher in the intervention arm than in the control arm and outcome showed no statistically significant difference, generating an ICER (i.e. the additional cost divided by the additional effects) of £3312.75 per 1-point improvement in the ABC-I score from the NHS/PSS perspective (Table 37). For the secondary analysis, we explored cost-effectiveness from a societal perspective and evaluated incremental cost per point improvement in ABC-I over the same period. It was not necessary to calculate ICERs, as there were no significant differences in costs and effects in the primary outcome. However, there is uncertainty around these results. A Bayesian decision-making approach was adopted, and available data on costs and effects were used to plot a CEAC. Non-parametric bootstrapping generated a joint distribution of incremental mean costs and effects for SIT compared with usual care to explore the probability of SIT being cost-effective subject to a range of values (λ) that a decision-maker might be willing to pay for an improvement in outcomes.

Figures 14 and 15 present scatterplots of the bootstrapped cost and effectiveness of SIT compared with usual care, from NHS/PSS and societal perspectives. Points in the scatterplot for the NHS/PSS perspective fall in the north-east and north-west quadrants of the cost-effectiveness plane, suggesting that SIT is more costly than usual care, with unclear conclusions regarding effectiveness. Points in the scatterplot for the societal perspective fall in all four quadrants of the cost-effectiveness plane.

FIGURE 14.

Bootstrapped mean differences in health and social care costs and effects (6-month ABC-I total score): cost-effectiveness plane – SIT vs. usual care (NHS/PSS perspective).

FIGURE 15.

Bootstrapped mean differences in societal costs and effects (6-month ABC-I total score): cost-effectiveness plane – SIT vs. usual care (societal perspective).

Statistical uncertainty around the ICER was explored using CEACs (Figures 16 and 17). Given the lack of suggested monetary thresholds for assessing cost-effectiveness using ABC-I, we assumed a willingness-to-pay threshold of £500. The probability that SIT is cost-effective from the NHS/PSS perspective was < 10% (see Figure 16). From the wider societal perspective (including education services, carers’ health-care use, out-of-pocket expenses and lost productivity, in addition to health and social care costs), the probability that SIT is cost-effective increased to just under 60% for a willingness-to-pay threshold of £500 (see Figure 17). At 6 months, the probability that SIT is cost-effectiveness rose from 10% under an NHS/PSS perspective to 60% under a societal perspective because of the higher cost of education in the control arm. Higher education costs offset costs in NHS/PSS and in other sectors, closing the gap in societal care costs at 6 months between arms (from £1360.80 under an NHS/PSS perspective to £356.32 under a societal perspective; see Table 35).

FIGURE 16.

Health-care and social care costs and effects (6-month ABC-I scores): CEAC.

FIGURE 17.

Societal costs and effects (6-month ABC-I scores): CEAC.

Cost-effectiveness: 12-month follow-up

At 12 months, it was not necessary to calculate the ICER, as there were no significant differences in costs and ABC-I scores under both the NHS/PSS and the societal perspectives (see Table 37). Figures 18 and 19 present scatterplots of the bootstrapped cost and effectiveness for SIT compared with usual care, from the NHS/PSS and societal perspectives at 12 months. Points in the scatterplot for the NHS/PSS perspective fall in the north-east and north-west quadrants of the cost-effectiveness plane, suggesting that SIT is more costly than usual care, with unclear conclusions regarding effectiveness. Points in the scatterplot for the societal perspective fall in all four quadrants of the cost-effectiveness plane, but are more concentrated in the north-east and north-west quadrants, suggesting that SIT is more costly than usual care, with unclear effectiveness.

FIGURE 18.

Bootstrapped mean differences in health-care and social care costs and effects (12-month ABC-I total score): cost-effectiveness plane – SIT vs. usual care (NHS/PSS perspective).

FIGURE 19.

Bootstrapped mean differences in societal costs and effects (ABC-I total score: 12-month complete-case analysis): cost-effectiveness plane – SIT vs. usual care (NHS/PSS perspective).

Cost-effectiveness acceptability curves for the 12-month follow-up showed similar results to the 6-month follow-up (Figures 20 and 21). The probability that SIT is cost-effective from the NHS/PSS perspective was < 10% (see Figure 20). From the wider societal perspective, the probability that SIT is cost-effective was just over 30% for a willingness-to-pay threshold of £500 (see Figure 21).

FIGURE 20.

NHS/PSS perspective (12-month ABC-I scores): CEAC.

FIGURE 21.

Societal perspective (12-month ABC-I scores): CEAC.

Research design

The process evaluation employed a mixed-methods approach. Quantitative methods were used to assess recruitment rates/patterns, attendance and intervention fidelity. Qualitative interviews with therapists involved in intervention delivery and carers of children enrolled in the trial examined experiences of the trial recruitment process, the acceptability of trial processes and measures, intervention implementation and acceptability, and the role of contextual factors. Methods for collecting data as part of the process evaluation are described in the following sections. Table 40 summarises each set of data collected, the method of data collection and the objectives addressed.

Recruitment of participants and therapists

Baseline data were used to describe the characteristics of the recruited sample (e.g. age, degree of SPD and autism symptoms). The baseline data provide an indication of the trial ‘reach’, that is whether or not the intervention reaches the intended population. This is described here as the proportion of those eligible who were then recruited (see Chapter 10 for a discussion of sample representativeness).

The number of people eligible (expressed as the proportion of those approached) and randomised (expressed as the proportion of those eligible) were utilised within the process evaluation.

Sixteen therapists were recruited to deliver the intervention across eight sites/clinics.

Retention

The clinical database was used to capture retention data, that is the number of people with missing outcome data at 6 and 12 months post randomisation (as a proportion of those who had been randomised and reached the time point).

Fidelity of intervention delivery

Fidelity was assessed using the ASI intervention fidelity measure. 23 This measure includes assessment of both structural and process fidelity. The scale demonstrates high content validity according to expert ratings and high reliability for process elements (total score ICC 0.99; Crohnbach’s α = 0.99). 23 Independent raters were provided with a manual and training on how to score the sessions that they reviewed.

Structural fidelity is defined by:

• the level of therapist training/qualifications, that is the occupational therapist providing the intervention is certified in sensory integration/SIPT (with a minimum of 50 education hours in sensory integration theory and practice)

• supervision, that is the therapist has a history of mentorship, with an equivalent of 1 hour per month from an advanced therapist with 5 years’ experience of providing OT using ASI intervention for a period of 1 year

• a score of 85 out of 110 across the following four areas – (1) safety of the environment; (2) detail and content of therapist-held records, including therapist–carer collaboration in relation to goals set during therapy; (3) physical space and equipment; and (4) communication with carers.

Process fidelity determines whether or not the therapist ensures physical safety; provides sensory opportunities; helps the child to maintain an appropriate level of alertness; challenges postural, ocular, oral or bilateral motor control; challenges praxis and behavioural organisation; collaborated in activity choice; tailors activity to provide appropriate challenge; ensures that activities were successful; supports intrinsic motivation to play; and establishes therapeutic alliance. This produces a process fidelity total score out of 100. Each rating reflects the rater’s global impression of whether or not a particular therapeutic strategy is an intentional part of the therapy being delivered within the observed sample of SIT. Therapist mentors considered initial assessment, goal-setting and delivery of the intervention for each therapist by reviewing and providing feedback on the first two video-recorded face-to-face sessions delivered to any participant.

Where consent was provided, face-to-face sessions were video-recorded. A sample of recorded sessions in the intensive phase were rated for process fidelity by at least one independent SIT-trained therapist (based on a randomly selected sample of the full session, of a minimum of 15–20 minutes in duration). To minimise bias through selecting ‘good’ sessions, the choice of which sessions to rate was made at random. To ensure that participants had both earlier and later sessions rated, the first session to be rated was randomly selected between session 6 and 12, and the second session was randomly selected between session 13 and 20. In the event that a session allocated for fidelity rating was not available (e.g. owing to the participant not attending), the next available session was rated. To ensure consistency of ratings, a selection of sessions were rated by multiple independent SIT-trained therapists. Demonstration of adequate fidelity of intervention delivery was defined as scoring ≥ 80 on the fidelity measure across at least 80% of sessions sampled.

Adherence

This was the level of intervention received by participants allocated to the intervention arm. An ‘effective’ dose for SIT has yet to be definitively established. However, attending 13 of a possible 20 sessions (i.e. two-thirds) delivered during the intensive intervention phase was felt likely to indicate sufficient exposure (based on clinical experience and available literature).

Adverse effects

Serious adverse events were reported from the time of signature of informed consent, throughout the treatment period up to, and including, 1 month after the participant received the intervention. A SAE is defined as any untoward medical occurrence in a participant who receives an intervention that was not necessarily caused by, or related to, that intervention (i.e. an AE) and that:

• results in death

• is life-threatening (as defined by GCP)

• requires hospitalisation or prolongation of existing hospitalisation (as defined by GCP)

• results in persistent or significant disability or incapacity

• consists of a congenital anomaly or birth defect

• is a medically important condition (as defined by GCP).

Adverse events that did not meet the criteria of serious were not systematically recorded and none was reported within the qualitative interviews.

Contamination

The potential for contamination, that is if participants recruited to the control arm received or sought therapy consistent with SIT or received enhanced/additional support from clinicians aware of their participation in the trial, is acknowledged. Therefore, there was an examination as to whether or not the usual care received differed in any way from the expected provision mapped out as a result of the scoping focus groups (see Chapter 2).

Quantitative analysis

The approach taken to analyse quantitative data was described in a trial statistical analysis plan. Recruitment, retention, adherence and safety data are descriptive, and categorical data are presented using counts and percentages. Process fidelity ratings were completed in accordance with the categories described in the fidelity tool. 23 After viewing a short section of video from a session, raters gave a general impression of the intervention delivered by responding with ‘yes’ or ‘no’ to the statement ‘This intervention session is provided by a qualified therapist intentionally applying ASI intervention theory and methods’. Raters then scored each of the 10 process fidelity items according to whether or not the therapist demonstrated each item as a key therapeutic strategy in the session. Scores were from 1 to 4, representing:

1. No, I do not think the therapist intentionally uses this strategy.

2. Doubtful, I do not think the therapist intentionally uses this strategy.

3. Probably, I think the therapist intentionally uses this strategy.

4. Certainly, I think the therapist intentionally uses this strategy.

Scores were then calculated to give a total score (Table 41). If the general impression was that the session was delivered in accordance with ASI, then the expected total score would be ≥ 80%.

Qualitative analysis

Interviews were audio-recorded, transcribed verbatim and anonymised. Interview transcripts were analysed thematically using a framework approach. 66 Two thematic frameworks were created (one for carer interviews and one for therapist interviews), drawing from the aims of the process evaluation and a review of transcripts. Data from carer and therapist interviews were coded separately in NVivo version 12 (QSR International, Warrington, UK). Thematic frameworks were discussed reflectively by the trial qualitative researchers, and a sample of interviews double coded. For the process evaluation, the viewpoints of carers and therapists have been collated and presented thematically (see Chapter 9 for full details of qualitative data analysis procedures).

Recruitment of participants and therapists

The CONSORT (Consolidated Standards of Reporting Trials) flow diagram (see Figure 5) gives a detailed breakdown of recruitment. A total of 258 carers contacted the study team to express an interest in taking part. Of those 258 carers, 24 (9.3%) were not eligible at referral. A further 84 carers did not go any further for the following reasons:

• no further response (n = 61)

• not interested in participating (n = 23).

Screening appointments were held with 150 carers and children. Three children (2%) were screened as ineligible after not meeting inclusion criteria (other than SPD). Nine children did not progress for the following reasons:

• no further response (n = 4)

• not interested in participating (n = 1)

• eligible and consented but then withdrew immediately (n = 4).

A total of 138 children were randomised into the trial (i.e. 92% of those screened and 53.5% of those who expressed an interest in taking part).

Participants were probably representative of the wider population of children in primary education with a diagnosis of autism presenting to services. The majority of participants were male (79%) and white British (88%) (see Chapter 6 for a description of baseline characteristics).

Recruitment of therapists to deliver the intervention occurred in stages as new health boards and clinic sites were opened. From August 2017, just one therapist was delivering the intervention to participants in South Wales (at a clinic in Cardiff University). By November 2017, a further five therapists across three South Wales university health boards (i.e. Hywel Dda University Health Board, Aneurin Bevan University Health Board and Cardiff and Vale University Health Board) started delivering the intervention. One occupational therapist was available to start in Cwm Taf University Health Board in February 2018, with another occupational therapist starting in May that year. A therapist at a private clinic in Cornwall started delivering the intervention in May 2018, with another therapist starting at the same clinic in July 2018. Given that each therapist could deliver the intervention to only one or two children at a time (given their NHS obligations), extra sites and therapists were recruited. In May 2019, a further therapist was recruited in Swansea Bay University Health Board, another therapist in Aneurin Bevan University Health Board and four were recruited at the Pace School in Buckinghamshire. Therapists’ capacity to see multiple children simultaneously depended on their working hours, leave and other commitments.

Therapists were not directly involved in trial recruitment but did refer potential participants to the trial team. Some therapists reported that they had difficulty identifying potentially eligible and willing children to take part, despite some being involved in publicising the trial through schools or presentations to professionals. One therapist thought that the trial criteria were too strict, particularly relating to age, as children either side of the age range may have benefited from the intervention.

Carers were asked during interviews where they had heard about the trial. A variety of sources was reported (Figure 26), most commonly via professional recommendation (note that one carer did not provide this information).

FIGURE 26.

How parents heard about the trial.

The main reasons that carers gave for taking part in the trial were that they hoped that their child would receive the intervention and that they wanted to improve support for children with autism and SPDs in the future. Some parents specifically wanted to pursue SIT for their child (having heard about it on courses, through their own research or via recommendations from friends), as they thought that it would be beneficial. It was often mentioned that carers and their child had received little or no support and, therefore, carers were happy to try anything that might help (carer interviews):

When we went in to [the trial it was] sort of like, not . . . a last resort, but like oh my God, something has got to help . . . I just think it’s really sad that there’s just nothing out there . . . and . . . maybe, maybe you know, in years to come, maybe this might be something that benefits everybody. So although it’s benefitting you, you’re also helping to benefit others.

P219, intervention arm

Most carers said that they had no concerns about their child taking part in the trial, although several worried that their child might be allocated to the control arm, meaning that they would not get the help that they needed. One carer was concerned that the intervention may not benefit their child, that is they would have ‘put him through’ appointments for nothing. Another carer was worried in case they were not allowed to stay with their child during appointments. One carer was concerned about anonymity owing to their profession, but was reassured when they were told that data would be labelled with participant numbers rather than names. Therapists did not report any concerns being expressed directly to them by carers.

Retention

Fidelity of the intervention

Acceptability

Adherence

Overall, 54 (78.3%) participants allocated to the intervention arm received at least 13 SIT sessions during the intensive stage. The median number of SIT sessions received was 20 (IQR 16–21). A full breakdown can be found in Chapter 6.

Adverse events

There were no SAEs during the trial. One AE was reported to the trial team by a therapist as a SAE; however, it was clear this did not meet the criteria of ‘serious’ as laid out in the reporting guidance.

Contamination

Contextual factors

Overview

Semistructured interviews were carried out with a sample of carers and therapists who delivered the SIT intervention. Before taking part in an interview, carer participants were asked to complete a timeline of the support that they and their child had received, which was used to facilitate discussion of their experiences. The aims of the qualitative study were to explore the (1) support experiences of families (outside the SenITA trial, i.e. usual care) and (2) perceived impact and clinical effectiveness of the SIT intervention for children with autism and SPDs.

Participants

Interviews were conducted with 13 occupational therapists who delivered the SIT intervention and carers of 30 children enrolled in the SenITA trial (Table 45).

Sampling and recruitment

For the carer interviews, participants were purposively sampled from the SenITA trial database, aiming for variation in terms of sex of child and age. Two-thirds of children in the interview study were recruited from the intervention arm to enable a more in-depth exploration of the perceived impact of the SIT intervention.

Carers were invited to take part in an interview once their child had reached the 6-month follow-up point of the SenITA trial. A qualitative researcher contacted carers by telephone or e-mail (in accordance with their stated contact preferences) to ask if they would consider taking part in an interview. Those carers who agreed were sent an information pack, which included a PIS, a template timeline with completion instructions and several coloured pens. Potential participants were re-contacted after approximately 1 week to check whether or not they would be willing to participate. Fifty carers were purposively sampled from the trial database and 30 agreed to participate, giving a response rate of 60%. Of the 20 carers who were sampled but did not participate, 15 could not be reached to discuss participation, three declined to take part as they had insufficient time and two said that they did not wish to take part in an interview, but did not give a reason.

All therapists (n = 16) who delivered the SIT intervention were e-mailed by the qualitative researcher to ask if they would be willing to participate in an interview once they had delivered all intervention sessions for at least one child as part of the trial. The e-mail included a participation information sheet with details of the interview study. Thirteen therapists agreed to take part in an interview and three could not be reached to discuss participation, giving a response rate of 81%.

Study materials

Separate sets of interview questions for carers and therapists were developed by qualitative researchers, with input from the SenITA Trial Management Group and parent/carer advisory group. Carer interview questions focused on experiences of the SIT intervention (for participants in the intervention arm only), if and how the intervention had affected their child’s behaviour and the implementation of activities or strategies suggested by the occupational therapist (for carers of children in the intervention group) and experiences of support for their child’s autism outside the SenITA trial (for all carers). Therapist interview questions focused on experiences of delivering the SIT intervention, perceptions of the impact of the intervention and the receptiveness of carers to suggestions of activities or strategies. Both sets of interview questions also covered carers’ and therapists’ experiences of trial processes (see Chapter 8 for related findings).

Before taking part in an interview, carers were asked to complete a template timeline of the support that they and their child had received. Carers were provided with timeline completion instructions that requested them to include any key events in their child’s life and the different forms of support experienced for their child.

Data collection

Interviews were conducted at a mutually convenient time and in accordance with participant preferences, that is either face to face at the participant’s home/workplace or over the telephone. Most therapist interviews took place over the telephone (n = 10), whereas most carer interviews were face to face (n = 19). Therapist interviews lasted between 27 and 58 minutes (mean 40.2 minutes, SD 10.6 minutes). Parent/carer interviews lasted between 12 and 86 minutes (mean 38.0 minutes, SD 17.6 minutes). With the consent of participants, all interviews were audio-recorded. Around half (n = 13) of carers completed the timeline template prior to the interview, whereas others said that they had insufficient time or had not remembered to do so. Following the interview, all carer participants were given a £10 shopping voucher to thank them for taking part.

Analysis

All interviews were transcribed verbatim and analysed using a framework approach66 by two qualitative researchers. First, transcripts were read in full and key themes were identified in accordance with the aims of the qualitative study. These key themes addressed both aspects of the study, that is the support experiences of families (carer interviews only) and the perceived effectiveness of SIT (interviews with therapists and carers of children in the intervention arm). Transcripts were then re-examined and emergent subthemes were identified within each main theme by the first qualitative researcher. After discussion with the second qualitative researcher, a thematic framework was developed and entered into NVivo. All interviews were then coded in NVivo by the first qualitative researcher, with 10% of transcripts coded by the second qualitative researcher to enhance reliability. Tables were created to summarise the experiences of each interview participant, corresponding to each theme and subtheme. The inclusion of illustrative quotes enabled the validity of findings to be reviewed. Both qualitative researchers met to discuss the identified themes and discussion continued until consensus was reached.

Support provision

Usefulness of support

Unmet support needs

Effect on the child

Effect on family and school life

Limitations

Recruitment and retention

The overall recruitment target (revised as per planned sensitivity analysis of sample size assumptions following the internal pilot) was met and 77.5% of randomised participants provided outcome data at the primary outcome time point. Loss to follow-up was slightly higher than the estimated 20%. However, given that all other assumptions relating to the sample size calculation were met, power to detect the target standardised effect size of 0.5 for the primary analysis remained at 90%. Longer-term (12-month) retention was lower at just under 70% (69.6%), although no between-arm difference was evident and this may reflect the complex life circumstances of the population (i.e. carers of young children with neurodisability who may also be working and caring for other children/family members).

Baseline characteristics and reach

Participants were well balanced across arms in terms of baseline characteristics. Most (79%) participants were male and white British (88%), reflecting those likely to have received an autism diagnosis. Autism is underdiagnosed in girls111 and ethnic minority boys112 for a variety of complex reasons, including stereotypical beliefs about presenting symptoms and perceived stigma. Most (71.7%) participants were recruited from South Wales and had a mean age of approximately 7 years and 10 months. Recruited participants showed significant sensory dysfunction. Eligibility criteria required at least probable dysfunction on two or more sensory dimensions of the SPM and the total score, or definite dysfunction on one dimension and total score. All potential participants screened met criteria, in line with available SPD prevalence estimates in autism,9–11 and 77.5% of participants had definite sensory dysfunction. A significant proportion of participants (25%) also had a diagnosis of ADHD.

Usual care and contamination

As detailed in Chapter 2, and confirmed during the internal pilot phase (see Chapter 4) and qualitative interviews with carers (see Chapter 9), usual care for young children with autism and SPDs is highly variable and, in many cases, comprises no formal OT intervention. Sensory-based difficulties are commonly reported and there is a clearly expressed demand from carers for services that specifically address sensory difficulties. Current NHS provision is generally focused on delivery of sensory strategies and advice (e.g. generic written information, bespoke leaflet, parent/carer groups and one-to-one consultations), although evidence to support the use of sensory strategies is limited (see Chapter 1). Some direct therapy may be offered using a sensory integration approach, but this does not generally meet fidelity criteria for SIT, in terms of both content and dose/frequency of contact. Where carers do access SIT, this is generally from private practitioners and, therefore, is inaccessible to most families. In terms of contamination, one trial participant allocated to the control arm received SIT, but sensitivity analysis suggests that this is unlikely to have influenced interpretation of results and no other evidence of sustained contact with OT services was reported. We found no evidence to suggest that families allocated to SIT were in receipt of significant OT intervention.

Fidelity and intervention receipt

Overall, SIT was delivered with good fidelity. All clinics were rated as achieving structural fidelity, meaning that the level of therapist training, documentation and the physical environment (including equipment) met the required standard for SIT. Ten out of 12 therapists met criteria for process fidelity consistently across sessions (i.e. ensured safety and established therapeutic alliance, and provided the appropriate level of challenge and sensory opportunity). It was not possible, however, to rate fidelity for all therapists (as no useable recordings were available for 4 out of 16 therapists). However, therapists did receive mentoring throughout intervention delivery, with a focus on aspects central to fidelity. Rater agreement was generally good, although lower average scores for two therapists may reflect poor inter-rater agreement, particularly where more than two ratings per session were available.

The pattern of intervention receipt was variable, although 78.3% of participants allocated to the intervention arm had received 13 out of 20 SIT sessions during the intensive delivery phase (defined as the target dose, of a possible 26 sessions overall). However, many participants received more than 13 sessions in total, with a median of 20 sessions received.

Acceptability

In terms of acceptability of trial procedures, most carers were very positive about their experience of taking part in the trial. A significant proportion of carers allocated to receive the intervention, however, reported some degree of difficulty or inconvenience in travelling to appointments, particularly given the frequency of contact. Outcome measures were also felt to be acceptable. Diary completion (detailing contact with other services) was poor, although the introduction of an automated text reminder service did increase response rates. Therapists reported a degree of difficulty with some trial processes and technology and were generally unfamiliar with the level of documentation required in clinical trials. In addition, most therapists reported difficulty in completing assessments and paperwork within the time estimated as per excess treatment costs.

Carers of children randomised to receive SIT reported high levels of satisfaction with the intervention, and many carers felt that it had been of significant benefit to their child and often to the wider family, in terms of facilitating activities it had not previously been possible to engage in. Carers and therapists tended to report that children were calmer, more independent and better able to tolerate everyday tasks following receipt of SIT. Carers also reported increased understanding of their child and felt able to implement strategies suggested by the therapist. However, perceived positive changes in behaviour were not necessarily sustained in all cases once therapy was complete.

Therapists’ experience of delivering SIT was also generally very positive. Many NHS therapists reported being unused to offering treatment so intensively and, although some therapists had found the experience challenging, felt that it had been professionally rewarding and was an approach that could offer benefit to some children. However, most therapists viewed SIT as part of a range of treatment options, the appropriate choice of which would depend on the child and their needs.

Behavioural outcomes

There was no evidence of a statistically significant difference between trial arms for the primary outcome, that is irritability/agitation, which is a measure of challenging behaviour (as assessed by the ABC-I67 at either 6 or 12 months). Parent-assessed ABC-I ratings at 6 months were, on average, higher than teacher ratings (where available). It is likely, however, that this indicates a true difference in behaviour and functioning in different settings, rather than a reporting bias on the part of either type of participant. Exploratory subgroup analyses suggest a differential effect of SIT on ABC-I at 6 months by sex of child, indicating that SIT was more effective for male participants. Subgroup analyses also indicate that the intervention was more effective in participants with a comorbid ADHD diagnosis. It is important to note, however, that even where statistically significant subgroup effects suggest clinically relevant change, that these analyses are hypothesis-generating only and may reflect multiple testing. We found no evidence of a differential effect of SIT on 6-month ABC-I scores by level of SPD, general comorbidity, ID or other neurodevelopmental or genetic condition. Likewise, no effect of SIT was observed for other ABC subscales (i.e. lethargy/social withdrawal, stereotypic behaviour, hyperactivity or inappropriate speech) at each time point.

Additional sensitivity analyses suggest that scores on the ABC-I at 6 months were generally better in those who initiated and received SIT. However, adjusting for the initiation of SIT did not alter the conclusions drawn based on the primary analysis (i.e. there was no evidence of a difference between the intervention arm and the control arm). Sensitivity analyses did not indicate any pandemic-related effects on outcome. No between-arm differences were observed at 6 or 12 months for adaptive behaviour, socialisation or daily living skills (using VABS domains). As detailed in Chapter 6, an error in administration of the VABS measure resulted in the introduction of an artificial ‘ceiling effect’ for these subdomains, meaning that it was less sensitive to change. A statistically significant between-arm difference was observed for the communication domain (indicating a negative intervention effect), although this is felt likely to reflect a combination of multiple testing, ceiling effects and an outlier in the control arm, rather than a clinically meaningful difference. Further sensitivity analyses would be unlikely to add to the interpretation of results, given that no other marginal effects were observed and the pattern of results was similar to that observed for all other outcome measures.

Parenting stress, quality of life and health utility

No between-arm differences were evident for parenting stress, quality of life or health utility. However, subgroup analyses suggest a differential effect of SIT on carer stress (using the APSI) at 6 months by region. A similar effect was observed for the presence of other neurodevelopmental or genetic conditions on carer stress, although this may indicate type I error.

Functional outcomes: occupational performance

There is strong evidence to suggest that carer-rated goal performance and satisfaction increased significantly with SIT exposure. This finding is in line with two previous trials of SIT,33,34 which reported significant effects for GAS. However, as detailed in Chapter 1, both trials were underpowered and did not assess maintenance of skills (i.e. did not include longer-term follow-up), and descriptions of comparators/usual care were limited. 33,34 It is possible that, given the individualised nature of SIT as an intervention, individualised goal-based outcome measures would be more sensitive to change. However, it is uncertain whether or not improvements in performance can be maintained in the longer term. Furthermore, data were available (COPM) for a subset of intervention participants only. It was not possible to collect these data from participants in the control arm, as participants in the control arm were not routinely seen by occupational therapists and not all intervention therapists recorded goals consistently.

Health economic analysis

Delivery of SIT incurred significant additional costs. When intervention costs were not included in total NHS/PSS costs, NHS/PSS costs of care for usual care-only participants were significantly higher than for SIT participants. The mean difference in ABC-I scores between arms did not achieve statistical significance at 6 or 12 months, suggesting no additional clinical benefit of SIT. Economic evidence suggests that SIT has a low probability of being a cost-effective option compared with usual care.

Summary

In summary, the SenITA trial was a robustly conducted evaluation of the clinical effectiveness and cost-effectiveness of SIT for behavioural, functional and quality-of-life outcomes. Targets for recruitment, retention and other key parameters used to estimate sample size were met, maintaining 90% power at the primary analysis time point. Usual care was significantly different from the intervention offered as part of the trial and was delivered with good fidelity and adherence. In addition, there is no evidence of significant contamination in the control arm or of those allocated to SIT receiving meaningful intervention via other routes. No adverse effects were reported. The population recruited are likely to be representative of young children with a diagnosis of autism and presenting to services, although this is distinct from being truly representative of the target population, given the well-documented underdiagnosis of autism in girls and minority ethnic children.

Trial procedures and outcome measures were acceptable to carers, and therapists generally found involvement in the trial to be a positive experience. However, many therapists reported issues with management of trial documentation and technology and many felt that the time allowed for assessment and recording clinical notes was insufficient. Carers in the intervention arm reported high levels of satisfaction and benefit of SIT, although no statistically significant quantitative change in carer stress was observed. Both carers and therapists tended to report improvements in well-being and daily functioning, although there is no evidence to suggest that these effects were maintained following the end of therapy. Therapists’ experience of delivering SIT was also generally positive, and most felt it to be an approach that could offer benefit to some children, depending on their needs.

There were, however, no statistically significant effects of SIT on any behavioural, functional, quality-of-life or well-being outcomes at either 6 or 12 months (with the exception of the VABS communication subdomain, as discussed above). Sensitivity analyses do suggest that scores on the ABC-I at 6 months were generally better in participants who received SIT and that intervention may work better for boys and participants with a comorbid ADHD diagnosis, although these were exploratory analyses and should not be overinterpreted. Carer-rated goal performance and satisfaction did increase significantly with SIT exposure. However, there is no evidence to suggest that improvements were maintained, and data were available for only a subset of intervention arm participants. Health economic evaluation also indicates that SIT is unlikely to represent a cost-effective treatment option when compared with usual care.

Strengths

The key strength of the SenITA trial is that it was adequately powered to answer the key question in the original commissioning bid of ‘What is the effectiveness and cost-effectiveness of SIT for children with autism?’. Important outcomes, as detailed in the commissioning brief, were challenging behaviours, socialisation, engagement with activities and sensory sensitivities. Other potential outcomes listed included adherence to the therapy manual, process indicators, health-related quality of life, carer burden, adverse effects and cost-effectiveness. The trial was sufficiently powered to determine effectiveness of manualised SIT on the primary outcome, that is, a measure of challenging behaviour. The intervention was also delivered with good fidelity and adherence and was acceptable to both families and therapists; no adverse effects were reported. Inclusion of longer-term (12-month) outcome assessment is also a strength, as previous studies33,34 of SIT have assessed outcomes post intervention only (or at an equivalent time point in the comparator arm).

This trial has also made a significant contribution to knowledge around what is generally delivered as usual care to children with autism and SPDs. In addition, the trial has highlighted the gap between services generally available to families (e.g. advice, strategies to improve functional/daily living skills) and services that carers would most value (e.g. intervention to support behavioural problems).

Limitations

Although SIT was generally delivered with good fidelity and all therapists met the required minimum threshold for training and experience, most therapists in Wales were employed by the NHS and had little recent experience of delivering full-fidelity SIT within their NHS roles. In England, trial therapists were working privately, as we were unable to recruit sufficiently trained and experienced therapists in England for a variety of reasons, including a reluctance of therapists to offer therapy via participation in the trial that was not otherwise available and, therefore, setting a level of expectation among families that could not be sustained. Subsequently, therapists at English sites were more used to delivering SIT with this population and did not have to juggle the competing demands of local service targets. Therapists also reported that the time allowed for assessment and clinical reasoning was insufficient, particularly at initiation of therapy, meaning that the total costs of intervention would probably be higher than indicated in practice.

For a variety of pragmatic reasons, including therapist availability and workload, some participants allocated to receive SIT did not start intervention within the intended time frame and had, therefore, received a potentially lower dose of SIT by the primary outcome assessment time point. Sensitivity analyses, however, indicate that this is highly unlikely to have significantly affected any key findings. A further limitation is the artificial ceiling effect observed for domains of the VABS. Interpretation of the positive effect on carer-reported goal performance is also limited by the fact that not all intervention therapists reported this consistently and that it was not possible to measure this in the control arm, as participants were not generally in contact with services.