Does the design of the NHS Low- Calorie Diet Programme have fidelity to the programme specification? A documentary review of service parameters and behaviour change content in a type 2 diabetes intervention

Background: NHS England commissioned four independent service providers to pilot low- calorie diet programmes to drive weight loss, improve


| INTRODUCTION
Obesity and its associated comorbidities, such as Type 2 Diabetes Mellitus (T2DM), have become a global epidemic. 1,2 Recent evidence has demonstrated that lowcalorie total diet replacement (TDR) approaches can be an effective dietary intervention for achieving sustained T2DM remission (HbA 1c <48 mmol/mol). 3,4 In line with this evidence, National Health Service England (NHSE) launched a low-calorie, TDR pilot intervention for people living with comorbid T2DM and overweight or obesity-The NHS Low-Calorie Diet (NHS-LCD) Programme. 5 Four independent service providers were commissioned by NHSE to deliver the programme across 10 socio-demographically diverse regions across England, to drive weight loss, improve glycaemia and potentially achieve remission of T2DM amongst participants. To be eligible, patients had to be aged 18-65, have received a diagnosis of T2DM within the last 6 years, be non-insulin dependent, and have a BMI of ≥27 kg/ m 2 (or ≥ 25 kg/m 2 for Black, Asian and Minority Ethnic communities).
The 52-week NHS-LCD Programme consists of a TDR phase, such as bars, shakes and soups (estimated 900 kcal/ day) for the first 12 weeks. This is followed by a 6-week structured food reintroduction phase, then a maintenance support phase for the remainder of the 1-year programme. The goals of the last two phases are to encourage healthy eating, increase physical activity and reduce sedentary behaviours following the UK government guidance. 6 This is complemented by ongoing behavioural support, via one of three delivery models: one-to-one, group (both delivered remotely during the COVID-19 pandemic) or digital (web and app-based).
How the programme aims to produce behaviour change, sustained weight loss, and improved T2DM was previously described in a logic model constructed by our group, supplemented by stakeholder input. 7 To achieve the desired outcomes, the intervention includes the use of specific behaviour change techniques (BCTs). BCTs are defined as the observable components of interventions designed to modify the cognitive and psychological processes underlying behaviour, known as 'active ingredients' (e.g., action planning, goal setting). 8 NHSE produced a service specification that stipulated the intervention contents for each of the four providers' NHS-LCD programme designs. 9 Within this, providers delivering the programme were given scope to independently design behaviour change content that is acceptable to their delivery model, available resources, and expertise, whilst aligning with the requirements outlined in the service specification. This flexibility meant the final behaviour change content within each of the providers' programme designs was unknown, including how they may differ.
Assessing intervention fidelity (whether a programme is implemented as intended) 10 is important to distinguish whether unexpected outcomes are due to a lack of intervention effectiveness or a loss of fidelity in its implementation, thus enabling accurate interpretation and scale-up of any intervention outcomes. [11][12][13] The National Institute of Health Behaviour Change Consortium (NIH-BCC) model 13 describes five domains of fidelity: study design (the extent to which the programme design reflects the evidence base); provider training (the extent to which deliverers are trained in a programme's components); treatment delivery (the extent to which the programme is delivered with adherence to the design); treatment receipt (the extent to which programme content is understood by participants); treatment enactment (the extent to which participants apply the programme content in their daily lives). 13 Despite the recommendation for fidelity to be K E Y W O R D S behaviour change, intervention design, intervention fidelity, low-calorie diet, total diet replacement, type 2 diabetes, weight management What's new?
• Assessing intervention fidelity is important, to distinguish whether unexpected outcomes are due to a lack of intervention effectiveness or a loss of fidelity in its implementation. Previous research has illustrated a loss of fidelity during the design phase in other national diabetes programmes. • Some elements of providers' programme designs did not adhere to the service parameters stipulated in the NHS England Low-Calorie Diet service specification. Most but not all (79.5%) of the behaviour change techniques were included in providers' programme plans. • A loss of fidelity during the design phase might have consequences for programme delivery and thus programme outcomes.
considered as multidimensional, numerous reviews report fidelity of delivery to be the most frequently examined domain across behaviour change interventions, whilst fidelity of design remains underexplored. 12,[14][15][16] This is problematic as a loss of fidelity in delivery could be a consequence of a drift in fidelity in the domains that precede it-the design and training phases. For example, if the evidence base is not adequately translated into the design (i.e., programme protocol) and staff training, intervention delivery is likely to be suboptimal. 12 A previous study from the NHS-LCD pilot evaluation examined a recommended component of study design-whether the programmes' 'active ingredients' are reflective of, and mapped onto, behaviour change theory. 11 We found the theoretical underpinnings of providers' programme designs to be unclear, 7 which might lead to a loss in fidelity due to the lack of a clear framework or map of how the programme is expected to work. Furthermore, the design of the NHS Diabetes Prevention Programme (NHS-DPP) (a national programme also commissioned by NHS England and delivered by external providers) was evaluated by extracting information on the service parameters and BCTs from programme protocols. Authors reported high, but not absolute, fidelity of behaviour change content in the providers' programme designs when compared to the NHS-DPP service specification. 17 The aim of this study was therefore to assess the design fidelity of the behaviour change content and service parameters in the NHS-LCD.
Objectives: (1) To describe the service parameters and behaviour change content stipulated in the NHSE pilot service specification, (2) To describe and compare the NHS-LCD programme designs across the four pilot service providers commissioned to deliver across the 10 initial pilot sites, and (3) To assess the fidelity of NHS-LCD programme designs to the NHSE service specification.

| Design
A documentary review comparing the NHS-LCD full programme specification with each of the four providers' programme designs. The following methods were informed by (a) the recommendation to evaluate whether an intervention's key components (i.e., the active ingredients) are fully operationalised, by examining programme protocols and/or manuals, 11 18,19 were included in the analysis as they provided the most comprehensive overview of the behaviour change content to be included in the programme designs. A review of the other guidelines cited in the NHSE service specification found no new BCTs outside of these three documents (Appendix S1).

| Service providers' programme design
The review of the intervention design (describing what providers planned to deliver) comprised: • Provider's programme manuals, describing the programme structure and curriculum, session plans and behaviour change components to be delivered to patients. • STAndardised Reporting of adult behaviouraL weight management InTerventions to aid Evaluation (STAR-LITE) 20 survey responses from the service providers, which provided further detail on the planned programme delivery. • Participant materials, including TDR manuals, takehome workbooks, and in-session worksheets.

| Procedures
The full programme specification and each of the service providers' programme designs were examined using two frameworks. The Template for Intervention Description and Replication (TIDieR) framework 21 was used to extract information on the service parameters, including what, who, how, where, when and how much, and tailoring. This framework aims to standardise and improve intervention reporting. The behaviour change content was coded using The Behaviour Change Technique Taxonomy (BCTTv1), 8 a validated measure and reliable tool for identifying the 'active ingredients' of interventions, 22 consisting of 93 distinct techniques. Only core content was coded for BCTs; optional activities were excluded as these did not describe necessary 'active ingredients'. A TIDieR data extraction sheet was developed by TE. DR and CH conducted TIDieR data extraction independently and in duplicate; TE supported discrepancy resolution through discussion. BCT coding was conducted independently and in duplicate following published guidance by the framework authors. 8 All coding researchers received training in BCT coding 23 (which sets out a number of learning principles to be adhered to when using the BCTTv1) and TE developed standardised data extraction sheets. Following familiarisation with the materials, TE and CD agreed on a set of BCT coding rules to guide the identification of BCT presence and dose (Appendix S2), which modified the coding rules set out in the evaluation of the NHS-DPP. 17 All materials were coded by TE; 2nd coding was completed by PD, CD, or CK. All discrepancies between coders were resolved via discussion with TE to achieve consensus on the final set of BCTs.

| Analysis
TIDieR components extracted from each of the providers' programme designs were tabulated and compared with the corresponding components extracted from the full NHSE service specification. The data extracted for each component was discussed by TE, CH and DR and rated as having fidelity (description meets requirements stipulated in the specification), partial fidelity (description meets some requirements stipulated) or an absence of fidelity (insufficient evidence of meeting requirements).
BCTs were labelled and the dose (frequency) of their delivery per session was reported. The numbers of different BCTs present in the providers' programme designs and the full programme specification documents were calculated and compared. This included analysis of additional BCTs identified in programme designs that were not mandated in the programme specification, as this has frequently occurred in previous studies (e.g., 24,25 ). Frequencies at which the BCTs were intended to be delivered across the programme by each of the service providers were also calculated. Cohen's kappa coefficient and Spearman's Rho analysis were conducted to report the inter-rater reliability of coding the presence of BCTs and their dose, respectively.
To maintain provider anonymity alternative service provider codes are used to present the BCT analysis to those reported for the analysis of service parameters.

| Service parameters
Information extracted on the service parameters stipulated by NHSE 9 and providers' programme designs are described and compared in Table 1 (using the TIDieR framework). 21 A traffic light coding system was used to highlight the degree of fidelity (green-fidelity, amberpartial fidelity, red-absent). Based on the documentation describing providers' programme designs, all four providers secured high fidelity for programme materials and procedures. Instances of lost fidelity included: TDR weight loss targets, training of programme deliverers, physical activity recommendations during TDR, number of sessions during Food Reintroduction, dietary recommendations outside of general healthy eating principles during Weight Maintenance, cultural tailoring, and advice on the intake of foods outside of the TDR products for participants struggling with adherence.
Of the four service providers, three indicated developing their programme based on learnings and user feedback from existing diabetes prevention/remission programmes and other weight management services. One provider also stated that service users provided input on the design and functionality of their accompanying app in its early development.

| Inter-rater reliability
Analysis of inter-rater reliability indicated strong agreement 26 between coders on the presence of a BCT within the full programme specification (k = 0.775, p < 0.001) and providers' A (k = 1.000, p < 0.001) C (k = 0.774, p < 0.001) and D's (k = 0.888, p < 0.001) programme designs. Moderate agreement 26,27 was indicated for BCT presence for provider B (k = 0.404, p < 0.001) and the dose of BCTs per session for Provider D (rs = 0.47, p = <0.001). Agreement on BCT dose was regarded as weak 27 for providers B (rs = 0.22, p = <0.001) and C (rs = 0.23, p = <0.005). As provider A's programme design documents did not include session plans, no information on BCT dose was available and, therefore, could not be analysed; BCT presence was coded based on provider A's logic model, describing planned BCTs and their proposed mechanisms of action.

| Programme specification
BCT coding of the full programme specification identified 22 distinct BCTs and one group of BCTs (those targeting self-belief) ( Table 2). The group targeting self-belief was coded as one BCT as no information was specified on whether one or all four BCTs in this category should be delivered, giving a total of 23 expected BCTs that should be included in providers' programme designs. Definitions of the BCTs identified in the programme specification are reported as a supplement (Appendix S3).

| Intervention design
BCT coding of service providers' programme design documents identified a total of 44, 33, 23, and 30 BCTs in the intervention designs of providers A-D, respectively. 16 BCTs were common across all providers.

| Fidelity of BCT content
Of the 23 BCTs identified in the specification, provider A included 20 (87%) in their programme design, provider B included 19 (83%), and both providers C and D included 17 (74%) in each of their design documents (Table 2). Thus, the overall mean proportion of BCTs in service providers' programme designs to the programme specification was 79.5%. This indicates variation between moderate to high fidelity across providers. 11 Twelve BCTs indicated in the full programme specification documents were included within all provider's programme plans, whilst two BCTs were absent across all providers: 'Behavioural contract'; and 'Social support (emotional)'.

| Additional BCTs
Twenty-nine BCTs were identified across the providers' designs that were not specified within the pilot service specification. The number of additional BCTs identified for each provider ranged from seven to 24. Only three of these BCTs were identified across all providers: 'Information about emotional consequences'; 'Behaviour substitution'; 'reduce negative emotions'. The variation in additional BCTs is described in Table 2.

| Dose of planned BCT content
Analysis of the intended dose of BCTs reported in providers' design documents indicated large variations in the frequency of some BCTs, and consistency in others ( Table 2). The BCT with the largest degree of variability in intended dose was 'Information about health consequences', which varied from 14 to 112 occasions across the 52-week programme between the four providers. On the other hand, the most consistent intended BCT dose was for 'Feedback on outcome(s) of behaviour', which all providers generally intended to deliver by feeding back to participants their body weight at the beginning of each coaching session. A substantial degree of variation was also noted regarding BCT dose per session across the three providers reporting BCT dose. Table 3 reports the number of sessions within which each BCT was included within each provider's programme designs, in addition to the range and mean BCT dose per session across sessions where the BCT was included. The most consistent BCT dosage across providers B-D's session plans were for BCTs 'Action planning' and 'Feedback on outcome(s)'.

| DISCUSSION
The four providers commissioned to deliver the pilot NHS-LCD Programme demonstrated fidelity to most but not all the service parameters stipulated in the NHSE specification. 9 Providers' programme design documents included between 74% and 87% of the 23 BCTs specified by NHSE 9 and NICE, 18,19 whilst an additional 7 to 24 BCTs were included outside of the service specification. Furthermore, there was wide variation between providers in the presence and dosage of BCTs described in their design documents, both within individual sessions and across their programmes as a whole. These findings illustrate a drift in fidelity in the implementation of the NHS-LCD during the design phase, highlighting the complexity of transferring the evidence base into consistent programme plans across providers of large-scale interventions. 28

| Relation to existing research evaluating fidelity of design
Our findings align with the NHS-DPP evaluation, 17 showing comparable (moderate to high) fidelity in planned BCT content to the NHS-DPP. Furthermore, the variation in additional BCTs across providers is similar to that reported by other studies using the BCTTv1 8 to evaluate the design of other health promotion interventions, 17,25 indicating variation in the active ingredients across providers commissioned to deliver the same programme. Our study T A B L E 1 Service parameters outlined in the NHS England programme specification in comparison with each provider's programme design (additional file). This should include TDR products to replace all daily meals, consisting of 'up to' 900 calories a day for up to 12 weeks. Service Users will therefore follow a diet composed solely of nutritionallycomplete TDR products, with total energy intake of up to 900 calories a day, for up to 12 weeks. TDR products can consist of soups, shakes and other suitable products but all TDR products provided on the NHS LCD programme must adhere to all legislation and standards that apply to total diet replacement products. The Provider will be responsible for procuring the TDR products that it supplies to Service Users Note: Degree of provider fidelity to the full programme specification is highlighted through traffic light colour coding (green = complete fidelity, amber = some fidelity, red = low fidelity).Providers 1, 2, 3 and 4 in Table 1 do not correspond to providers A, B,C and D in Table 2 to preserve anonymity for provider organisations.
Abbreviations: NHS, National Health Service; STAR-LITE, STAndardised Reporting of adult behaviouraL weight management InTerventions to aid Evaluation survey; TIDieR, Template for Intervention Description and Replication framework.
is unique in that data was collected and analysed on BCT dose, which like BCT presence, was found to vary across the providers. Importantly, we noted how the degree of fidelity of BCTs to the full programme specification reflected the degree of explicit theory use reported in an analysis of theoretical underpinnings undertaken by the same study team. 7 The provider with the strongest theoretical underpinnings was also found to have the strongest fidelity in their BCT content, whilst the two providers with the weakest theoretical underpinnings were also identified as having the weakest fidelity in their BCT content. This supports the notion that unclear theoretical underpinnings might result in a drift in programme fidelity. 7,28,29 Although many studies evaluating programme designs have focused on coding BCTs, 25,30 we additionally assessed fidelity to service parameters using the TIDieR 21 framework. As we found several important service parameters to not demonstrate fidelity to the service specification, this illustrates other important components of an intervention's protocol 11 outside of BCTs where fidelity might be diluted during the design phase, and not captured using the BCTTv1 8 in isolation. Evaluators of the NHS-DPP also adopted this method, 17 however, they reported stronger fidelity of service parameters than the present study.

| Relation to existing research evaluating the effectiveness of BCTs
Evidence suggests some BCTs included in the NHS-LCD design may be more effective within the context of diet, physical activity and/or T2DM management. A metaanalysis identified four BCTs in diet and physical activity interventions to be associated with clinically significant reductions in HbA 1c . 31 These included 'Action planning' and 'Instruction on how to perform the behaviour' (both of which were included in the full programme specification and all four of the providers' programme designs), in addition to 'Demonstration of the behaviour' and 'Behavioural practice/rehearsal' (neither of which were included in the service specification but were specified in two and three of the providers' programme plans, respectively). This suggests that some of the additional non-prescribed BCTs included in providers' programme designs may be beneficial to achieving programme outcomes. Evidence also supports the effectiveness of self-regulatory BCTs (e.g., goal setting, self-monitoring) for weight loss in participants at risk of or diagnosed with T2DM, 32-34 a variation of which was identified in the full programme specification 9 and each of the four providers' designs. Self-regulatory BCTs were also identified in the NHS-DPP specification documents. 17 Nevertheless, it is important to note that the inclusion of these BCTs within providers' programme plans does not denote the actual delivery of these BCTs. For example, NHS-DPP providers planned to deliver 74% of the 19 BCTs in the NHS programme specification, whilst the research team observed only 7 of those 19 specified BCTs in all eight observation sites, 24 indicating difficulty in translating programme design into programme delivery. Ongoing research will examine programme delivery as part of the evaluation of the NHS-LCD Programme [NIHR132075].

| Strengths and limitations
By building positive stakeholder relationships, all documentation describing the service parameters and behaviour change content of providers' programme designs were obtained. All providers additionally completed the STAR-LITE survey 20 as a supplement to their design documents, to ensure all intervention components were captured. Our methodological approach was informed by that set out by NHS-DPP evaluators, 17 including the use of a validated tool for coding BCTs, 8,22 ensuring clear programme comparisons can be made by researchers and stakeholders.
Another strength of the present study is that by collecting and analysing data on the dose (frequency) of BCTs within NHS-LCD programme designs, the present research reports on and compares the intended BCT dose across providers. Whilst it was interesting to report and compare how dose varied across the different providers, it is important to note: (a) the limitations in terms of a lack of unifying definition for dose, and (b) that the service specification did not specify dose, therefore, the objective was not to examine fidelity to dose. However, by reporting on this we have provided the basis for the variation in dose across providers to be considered when comparing participant outcomes in the future.
High agreement between coders was indicated for BCT presence, but less so for BCT dose. One of the challenges of external evaluation is that BCTs are not always clearly described in the documentation from service providers, making identification difficult when the research team was not involved in the intervention design. For example, 100% agreement was calculated for the coding of SP4, where the intervention description used the labels and definitions outlined in the BCTTv1. 8 This limitation was acknowledged and mitigated through a rigorous approach by double coding all documentation and discussing all discrepancies until consensus was achieved. Furthermore, a limitation of using the BCTTv1 8 is that it is not an exhaustive list of behaviour change strategies. For example, it was noted that many providers drew upon Third-Wave Cognitive Behavioural techniques (e.g., mindfulness), whilst SP1 included many elements of 'Positive Fidelity to the full programme specification is highlighted through traffic light colour coding (green = present, red = absent). Additional BCTs not specified in the specification are not colour coded.
BCTs targeting self-belief were coded as one behaviour change technique in the NHS-LCD service specification as NICE guidelines did not specify which or how many of this BCT group should be included.     Table 1 do not correspond to providers A, B, C and D in this table to preserve anonymity for provider organisations.No information on BCT dose was reported in SPA's programme design.
Numbers in square brackets are corresponding number in BCTTv1. Abbreviations: BCT Behaviour Change Technique; SP Service Provider.

T A B L E 3 (Continued)
Psychology' (e.g., gratitude, resilience) within their TDR session plans that were not captured by the BCT coding. Positive Psychology techniques focus on fostering wellbeing and positive affect as opposed to decreasing negative symptomatology. 35 The absence of this in our coding might help explain why SP1 is reported as including the smallest number of BCTs in their design. 'Increase positive emotions' has been suggested as a BCT for inclusion in future versions of the taxonomy 8 ; based on our findings we would recommend inclusion to ensure BCT coding captures all active ingredients of any intervention.

| Implications for practice
Although providers generally reported good fidelity to the service parameters stipulated by NHSE, the lack of fidelity to some of the service parameters could compromise programme delivery and consequently programme outcomes. For example, one provider's inclusion of physical activity recommendations during the TDR phase might have implications for the safety of participants, as changes to physical activity are not recommended when consuming a low-calorie diet (800-1200 kcal/d) (e.g. 36 ). Furthermore, another provider's lack of reporting on staff training may result in a dilution of training fidelity (i.e., the degree to which delivers are trained in the essential components of the intervention), 13 which may compromise delivery fidelity. Other aspects where fidelity was variable across providers, such as a lack of cultural adaptation, could have consequences for the success of minority group members on the programme. To improve this, the Patient Public Involvement Group for this evaluation have recommended adopting minimum standards for cultural adaptation to strengthen the NHS-LCD Programme specification. As our findings illustrate the difficulty in translating NICE guidelines into BCTs implemented in programme designs, we have recommended to the NHSE that (a) necessitated BCTs be explicitly described in the service specification and (b) that providers be required to have a member of the programme development team with expertise in behaviour change (e.g., a health psychologist). Although three out of four pilot providers included a Clinical Psychologist in their programme development team (Table 1), this profession does not denote expertise in BCTs. If implemented, this could strengthen providers' fidelity to the BCTs specified in the NICE guidance referenced in the NHS-LCD specification.
By extracting and disseminating information on the NHS-LCD Programme design, those interested in implementing a similar programme will have a clear understanding of the service parameters and programme content. The Medical Research Council framework 37 recommends that intervention developers and evaluators clearly articulate an intervention's key components so they can be retained during programme adaptation or scale-up. Readers will have an understanding of the 'active ingredients' necessitated by NHSE 9 and NICE 18,19 and how additional techniques differ between service providers. Moreover, this study is unique in that planned BCT dose-both per session and across programme designswas analysed and reported in addition to BCT presence. This will provide useful insights when comparing participant outcomes across the four providers, as these can be interpreted within the context of their differences in design and planned content. This may also further the evidence base for BCTs in diabetes programmes, as more research is needed on 'how much' of a BCT is necessary to improve T2DM management 31 and whether more BCTs are associated with superior outcomes. Advancements in this area will be supported by a systematic review and meta-analysis of common and effective BCTs in lowcalorie diet interventions. 33 Furthermore, we recommend that service providers developing behavioural weight management programmes use a taxonomy to report their active ingredients in their design documents (e.g., the BCTTv1 8 ) to ensure transparency and minimise the need for interpretation by independent evaluation teams.

| Implications for research
Although fidelity of design is key, equally important is what is subsequentially delivered, and how participants receive and enact the BCTs (i.e., the degree to which BCTs are understood and used by participants in their daily lives). A loss of fidelity at each stage could impact programme outcomes. 33 For example, BCT enactment was associated with improved effectiveness of a diabetes weight management programme (ADDITION-Plus trial); participants using all 16 programme BCTs lost significantly more weight than those using 10 or fewer. 34 However, evidence on the role of BCT delivery and/or enactment is limited, warranting investigation in future evaluations. 33 By reporting each of the providers' key intervention components and active ingredients, the current study provides the basis for assessing these in the later stages of the NIH-BCC model. 13 Importantly, future research should link the data reported by the present study with the clinical outcomes of programme participants to examine whether the degree of fidelity and/or variation in delivery is associated with the degree of weight loss and T2DM remission.

| CONCLUSION
Although the four providers commissioned to deliver the pilot NHS-LCD Programme were identified as having fidelity to most of the service parameters outlined in the NHSE service specification, this study identified some important elements of lost fidelity. This may have consequences for programme delivery and thus programme outcomes, which will be assessed as part of the ongoing evaluation. Furthermore, most but not all (79.5%) of the BCTs specified by NHSE 9 and NICE 18,19 were included in providers' programme plans, in addition to a large number of further, non-prescribed, BCTs. Programmes featured large variations in the use of specific BCTs, as well as variations in their intended dose. We recommend that participant outcomes and experiences are compared across providers to understand how the variation in intervention techniques influences programme engagement and success.