Modificación del artículo 9

Our study had several strengths, including the recruitment of the required number of participants and clusters, testing a single primary outcome, achievement of a low attrition rate at an individual level and an a priori analysis plan, all of which reduce the risk of bias. Furthermore, the ICC for the primary outcome is smaller than originally assumed for the sample size calculation and the sensitivity analyses support the main findings of the study, suggesting that the findings are robust.

Our design aimed to maximise the external validity and generalisability of the study. The trial was conducted within the NHS in 23 community ID services in different parts of England, covering urban, semi-rural and rural areas, from which we recruited 246 service users. Despite the geographical spread of the services, we found that services were not very different. Although there were some minor differences in baseline participant characteristics, adjusting for these variables in the main analysis had no bearing on the study outcomes. Our inclusion and exclusion criteria closely resembled the cases that were likely to be referred for assessment and treatment to those services. The overall median level of challenging behaviour

of the participants was higher than that reported in a trial of antipsychotic medication by Tyrer et al.61

In order to guard against the tendency for the impact of training to dissipate over time, we set up

mentoring and peer support during the study period.1,28,93_{Our process evaluation indicated that training in}

PBS was well received by the therapists and the delivery of PBS was welcome and highly valued by the service managers of participating services, carers and participants with ID.

The study also has limitations, including low fidelity and a less than optimal engagement of the therapists with the intervention. We found that 30% of participants received all elements of PBS as specified in the training and 43.5% received only partial input, mainly observations. In addition, there was a low uptake of the mentoring support that was offered by the PBS trainers. A reason for poor implementation was the difficulty experienced by some of the participating NHS organisations to assist their staff in undertaking the additional responsibilities of the study procedures by failing to reduce therapist caseloads. Therefore, the therapists either undertook the delivery of PBS in addition to their normal workload or were unable to implement it. Furthermore, therapists were embedded in the ordinary community ID team and, therefore, clinical procedures including managing emergencies and interprofessional referrals may have also hindered the delivery of the intervention as intended. This may be a reflection of the realities of implementing PBS in community ID services without the provision of additional resources, such as specific posts for accredited behavioural therapists.

Regarding the lack of impact on the subgroup with broadly defined ASD, it may be that interventions that are not specific to ASD are less likely to help improve behaviour. However, the findings in this respect are preliminary but worth considering in the context of investigating treatment approaches for this population. There was a considerable turnover of paid carers in the study. This may also have affected some of the delivery of PBS, as the therapists were meant to work closely with family and paid carers, and any change in paid carers would have required the therapists to continue this work with a new carer, possibly having an impact on the continuity of their work. Also, because of the considerable turnover in paid carers and the small number of family carers in the intervention arm, we were unable to further explore the secondary outcomes in paid and family carers.

As a result of the nature of the intervention, it was not possible to blind carers and participants to

treatment arm allocations. There were six cases in which researchers were unblinded to a cluster’s

treatment arm allocation, which indicates a small possible risk of bias. In such cases, another researcher

subsequently collected data from those sites. Moreover, our analysis showed that researchers’ predictions

of treatment arm allocation were considered to be no better than chance.

Costing for accommodation was based on very broad assumptions because of poor data on accommodation

costs. There was no evidence of changes in participants’ accommodation throughout the study. We were

unable to provide any estimates of the impact on employment or state benefits as a result of the intervention due to poor reporting and a large number of missing or inconsistent data. This points to the need for better methods of collecting data on accommodation, employment and state benefits for people with ID. The self-reported questionnaire for unpaid carer time was limited to four questions and it is potentially unreliable

as it has not been extensively validated.94_{It was reported that 14 participants did not have challenging}

behaviour at the time of first contact with the therapist. This may have been attributable to a number of

acutely present at the time; (2) other interventions may have been delivered by other team members with positive outcomes; and (3) the behaviour being mild enough not to cause any harm or severely reduce quality of life and so treatment was not seen as priority by the carers. Paid carer turnover was significant and limited the undertaking of some of the PBS-related tasks such as goodness of fit, which requires discussion

of the plans and their implementation in the patient’s home or activity setting. Finally, because of the design

and logistics of the study, we were unable to train new staff to undertake PBS in place of the therapists who left. Therefore, the therapist cohorts were fairly static and any changes meant that participants allocated to the intervention arm did not receive it as we were unable to replace therapists. The study began at a time of increasing awareness of PBS as a preferred psychosocial approach to manage challenging behaviour and

ended at a time when training in PBS at scale was implemented.95_{These changes in health and social care}

service approaches may have closed the gap between the intervention and TAU. There is currently a debate on whether PBS is delivered as it should be or if cursory attention is paid to its delivery across multiple services. Although we had considered this issue at the inception of the trial, we believe that we had offered significant incentives and support to motivate and entice services to participate. However, we were

unprepared for the wide variation in delivery that we encountered and the opposition at times from services allocated to the intervention to make adjustments to accommodate research activity. Although our results may be disappointing to many in the PBS community, they nevertheless present a fair account of how PBS

as an intervention may be corrupted96_{and the extent to which it requires further modification in both}

content and delivery in order to provide the benefits that its proponents advocate. At 12 months, the intervention appeared to increase QALYs and, in this context, society may be willing to pay a higher price given the intractable problem of challenging behaviour. However, it is also important to take into account the fact that the measures used for cost-effectiveness (i.e. EQ-5D-Y) may have uncertain validity in this population with long-term disability.