The INCENTIVE study: a mixed-methods evaluation of an innovation in commissioning and delivery of primary dental care compared with traditional dental contracting

Skill mix

One area of potential advantage in the blended incentive-driven model of delivery is more effective use of the dental team. For example, dental hygienists/therapists can carry out courses of treatments recommended by the dentist who has examined the patient. Dental hygienists can carry out treatments such as scaling and polishing, oral health promotion and fissure sealants. A dental therapist can perform additional treatments such as fillings, pulp treatment/stainless steel crowns and extractions on children. Additional skills dental nurses may be trained and competent to give preventative advice and apply preventative fluoride varnishes to teeth. Intuitively, the delegation of treatment to staff who specialise in only a specific range of treatments could reduce costs and increase access to care but this hypothesis needs testing. 36

Skill mix is advocated in several current proposals that continue a trend seen in UK dentistry over the last 20 years. 2,7,37,38 For example, dental therapists may now work in general dental practice7 and their clinical remit has expanded. 39,40 The number of training places has increased and several educational establishments have instituted programmes. The potential contribution of dental therapy is considerable. Evans and colleagues41 found that, within their current remit, therapists could undertake the treatment provided in 35% of dental visits and in 43% of clinical time. Yet dentistry has not harnessed this potential. Dentists may need models to help them employ dental hygienists/therapists profitably and fully use their skills and there have been calls to develop a system that encourages dentists to use dental hygienists/therapists differently. 42,43 The PDS pilots failed to fully involve the wider skill mix available (i.e. dental care professionals or improve their conditions to recruit and retain them). 9

Although there are few hard data to support skill mix in dentistry36 some data are beginning to emerge. A recent practice-based study found the success of fissure sealants to be comparable whether placed by dentists, hygienists or therapists. 44

There is a trend towards greater professional acceptance of therapists,43,45–49 with approximately 60–70% of dentists prepared to consider employing a therapist in the more recent studies. 50,51 Despite this, some dentists remain uncertain of the role of dental therapists. 51,52

There is also some uncertainty about public acceptance of dental therapists. Two recent surveys and a qualitative study suggest that few lay people are aware of dental therapists as a professional group. 53,54 Furthermore, even after the training of dental therapists was explained to them, only 61% of adults were willing to receive simple restorative treatment from a therapist. 53,54 The provision of dental care is influenced by the NHS contract for dentists. The existing UDA contract does not make any allowance for treatment provided by dental therapists, although since 2013 dental hygienists and therapists can carry out their full scope of practice without a prescription or the need for the patient to first see a dentist; this is known as direct access and the impact of this on dental team delivery is yet to be realised. Research is therefore needed to assess if new models of delivery and service design will encourage their use and whether or not they are acceptable to dentists and patients.

Care pathways

The Bradford and Airedale new service delivery model (incentive) was designed to encourage a care pathway approach in which all patients have an OHA on joining the practice and at each subsequent recall. Four sets of information [age group, medical history, social history (self-care, habits/diet) and clinical assessment] are used to inform a traffic light (RAG) risk assessment for patients with high (red), medium (amber) or low (green) risk of oral disease (Table 2).

Within the model, each patient follows a care pathway according to the protocol. The care pathway includes evidence-based preventative treatment and advice, suitable recall interval and restorative care as appropriate (e.g. the red risk category limits patients to stabilisation and lowering their risk status). The care pathway’s evidence base was based on the Department of Health’s Delivering Better Oral Health: An Evidence Based Toolkit for Prevention. 35 Each patient’s status is reviewed on their next OHA. Patients may therefore move between risk categories. Monitoring in practices ensures that evidence-based preventative interventions are delivered in line with identified needs and monitored access to dentistry. OHImp is assessed through the delivery of a performance framework. This framework is based on the transfer of ideas from the general practitioner contract QOF. (It is of note that this contract pre-dated the DQOF. 19)

There is little literature regarding care pathways in primary dental care, though the concept has been around for a number of years. The concepts and benefits of the care pathway approach in dental primary care were described by Hally and Pitts. 55 As a result of recommendations within NHS Dentistry: Options for Change,7 the first widely disseminated care pathway in UK dental primary care was the OHA within the National Institute for Health and Care Excellence (NICE) guidance on dental recall intervals. 56 The OHA care pathway was designed to enable more prevention within personalised care plans taking into account patients’ social and dental histories as well as clinical findings.

The type of risk assessment (the RAG traffic light system) included in the blended/incentive-driven contract in our study has hitherto not been fully evaluated. Examination of different RAG models in other dental settings is ongoing in the north-west of England. 57 Early findings from the NHS dental contract pilots suggest small improvements in risk reduction over the short term. 20

In summary, the incentive dental contracts are aimed at ensuring that evidence-based preventative interventions are delivered in line with identified needs for a defined population; ensuring increased access to dentistry; and ensuring that care is provided by the most appropriate team member to encourage skill mix. Quality indicators linked to contracts and payments have been used widely in other branches of health care, and the results are complex. The indicators can drive organisational change towards best practice, but may also be a disincentive to important but non-rewarded activities. 24 Used alongside demographic data, the indicators can measure practice performance, identify areas for development and assist sharing of best practice. 58 The indicators often increase the quantity of service provision, but not always the quality. 59 Furthermore, the indicators can affect the dynamic of professional relations and the doctor–patient interaction. 60 Although offering great potential, the DQOF with embedded quality indicators has not been comprehensively evaluated in dentistry. A recent systematic review was only able to provide a framework for how such indicators might work. 61 The blended/incentive-driven contract in West Yorkshire provides an opportunity for a comprehensive evaluation to inform the next dental contract reform.

Ethics approval

Ethics approval was obtained from the National Research Ethics Service (NRES) Committee [London – Bromley, Research Ethics Committee (REC) reference 12/LO/0205] on 5 April 2012. Informed consent was obtained from all participants prior to any qualitative enquiry. The study was sponsored by the University of Leeds, Leeds, UK.

Data were collected between August 2012 and February 2014 by two research associates, who were social scientists rather than dentists. Interviews and focus groups followed a topic guide, partly informed by the theoretical framework (see Appendices 1 and 2) but supplemented with themes that emerged from the observations and previous interviews. Interviews with dental team members took place at the dental surgery, while interviews with patients took place in patients’ homes. All were recorded and transcribed verbatim. Interviews lasted between 15 and 70 minutes.

Theoretical framework

The Andersen behavioural model of access was employed as the theoretical framework for the qualitative analysis (Figure 1 shows a simplified depiction). The model sees access as ‘the use of personal health services and everything that facilitates or impedes their use’63 and distinguishes between ‘potential’ and ‘realised’ access. Potential access is measured according to enabling variables such as the availability of care or supply of health-care workers. ‘Realised access’ refers to services actually used or ‘utilisation’. 63

FIGURE 1.

Andersen’s behavioural model of access (adapted from Baker63).

Originally developed 40 years ago as a model of health-care utilisation, the model has evolved in line with advances in understanding that have moved from an individual focus to incorporate the interaction between the individual, the health-care system and the external environment. 63–65 The revisions have not altered its basic foundations, but have added components to it. Later versions introduced health and patient satisfaction as desirable outcomes of health care, which were said to be determined by predisposing and enabling factors, behaviours and need. Findings from studies in many areas of health care, and dentistry in particular, support its use. 66,67

Enabling factors relate to the policies, facilities, staff and the organisation of services that might influence utilisation. 65 It is helpful to consider enabling factors as a spectrum, with different qualities of the same factors either facilitating or challenging service use. Policies, guidelines, rules, practices, contracts, resources and reform define the nature and delivery of care. Financial factors can include the funding and affordability of health care. Organisational factors refer to the structures and processes that influence the availability and distribution of health services, and the personnel and how accommodating they may be to patients’ needs.

From this perspective the blended/incentive-driven contract being evaluated in this study is an enabling factor with health policy, financial and organisational facets. The government has made policy commitments to oral health and dentistry with focus on improving the oral health of the population, particularly children, introducing a new NHS primary dental care contract and increasing access to NHS dentistry. 18,68 Other relevant drivers included the recommendations of the Steele report2 and the evidence-based prevention of Delivering Better Oral Health: An Evidence Based Toolkit for Prevention. 35

The contract replaces some of the financial emphasis on the volume of treatment with incentives for quality and changed outcomes. These incentives were based on the concept of providing care for a population rather than patients via organisational changes such as the use of care pathways, the provision of evidence-based care and the use of skill mix.

Health-care need may be seen as the potential to benefit from care and requires both a health problem and an effective intervention. 69 Health needs may include health education, disease prevention, diagnosis, treatment, rehabilitation and palliative care.

The Andersen behavioural model distinguishes between evaluated (professionally defined or normative) and perceived (personally defined or felt) need70 but recognises that there are social aspects of evaluated need, which can be influenced by technological developments and clinical guidelines.

In some versions of the model predisposing, enabling factors and need are distinguished at the contextual or individual levels. Contextual characteristics operate at a population or combined level, drawing attention to the environment or the circumstances in which health care is delivered. The most recent versions of the model have seen the context and individual levels arranged sequentially with contextual predisposing and enabling factors and need influencing similar factors for individuals.

In turn, these components could influence personal and professional health behaviours. Personal behaviours are activities that shape health status, such as oral hygiene, diet and tobacco use. Health service use is treated as behaviour in itself. Professional behaviours relate to the processes of medical care, such as health education, clinician–patient communication and prescribing behaviour. Both personal and professional behaviours may be amenable to changes via enabling factors and will influence the nature and volume of treatment provided.

The outcomes are (1) perceived and evaluated health status and (2) patient satisfaction. The maintenance and improvement of health should be the primary outcome of access and is a central target of the dental contract reform programme. 68 Perceived health status indicates how much an individual can live a ‘functional, comfortable and pain free existence’. 65 This definition is akin to OHQoL, which may be defined as ‘the impacts of oral disorders on everyday life that are important to people and of sufficient magnitude to affect perception of their life overall’. 71 However, other patient-reported outcomes could be used, such as general health perceptions. Evaluated health status requires a professional assessment of clinical status. Patient satisfaction has a bearing on patient health outcomes, with greater satisfaction being related to health improvement via adherence, involvement with treatment and continued use of services.

An important feature of the model is its recursive nature, with feedback loops so that the outcomes of access have the potential to influence future predisposing and enabling factors, population needs and use of services.

Analysis

Data were analysed in two phases. First, pen portraits of the practices are used to give the data context.

Within the second phase the principal approach used framework analysis to induce the results from the variety of original accounts across stakeholder groups within the structured policy focus of the research. 32,33 The focus of the analysis was to explore the effect of the contract as it interacted as an enabling factor with other stages of the model. Analysis adhered to the following process:

1. familiarisation with the data

2. identifying the thematic framework

3. indexing

4. charting

5. mapping and interpreting. 32

A member of the research team (MH) studied the field notes and transcripts of the post-observation interviews. This process of familiarisation (familiarisation with the data) enabled identification of emerging themes in the data set. 72 Although Andersen’s behavioural model guided the thematic framework in principle, data were not forced into an a priori model. Instead, the framework was refined as required (identifying the thematic framework). Data generated from the post-observation interviews were indexed according to the particular theme to which they corresponded (indexing) and lifted from their original text and placed under subheadings derived from the framework (charting). The themes were flexible and modified as necessary. Data were organised by theme to enable a process of constant comparison across themes and cases. The framework analysis served to either confirm or challenge the model. A form of deviant case analysis was intended to be used to add new categories or revise the model.

The validity of the findings was supported by discussion of interim and final results for triangulation and corrections with participants in focus groups. The results were also compared against existing knowledge, such as the evaluations of the NHS dental contract pilots. 30

Practice 1

Practice 1 is situated in the centre of a large town several miles from Bradford. The practice is based within the one of the 10% most deprived wards in the country, with associated adverse income, living environment, education, health and employment indicators. The practice estimates that approximately 80% of its patients are eligible for benefits and a similar proportion are white British, despite the ethnicity of around 50% of the local population being of Pakistani/Bangladeshi origin. The practice was established in 2008 and is operated by a very large national corporate provider. Its team consists of four GDPs and one dental therapist.

Practice 2

Practice 2 is located at the south-east edge of Bradford, adjacent to a large council estate. Bradford ranks 16th in the most deprived local authority areas in England. The area served by the practice originally had a white British population, although it has become more ethnically diverse in recent years and has a traveller community. The people of this area had experienced difficulties with access to dental care for many years, with limited care available in the area and a local unwillingness to travel for care.

Established in 2009, the practice is owned by an independent contractor with two other practices (not part of the blended/incentive-driven contract). The practice has five surgeries, a separate decontamination room and staff training room. It employs five dentists and two dental therapists. Almost all of its treatment is provided under the NHS (99.5%).

This practice experienced a higher staff turnover than the other study practices, which was attributed anecdotally to the associates being early-career dentists eager to take up a post but not necessarily a commitment to practising in the locality in the long term. Lack of opportunities to carry out complex treatments was also thought to be a factor.

Practice 3

Practice 3 is located in an affluent market town several miles from Bradford, which has lacked NHS dental provision. It was established in 2009 and the provider is part of a social enterprise organisation. It has two surgeries, two dentists and one dental therapist. During the research, one GDP left this practice. The original practice manager was replaced, but later returned.

Three comparison practices (4, 5 and 6) continued to work under the terms of the 2006 NHS dental contract and were matched to the blended/incentive-driven practices according to the number of GDP surgeries, list size, and the deprivation index and ethnicity of the local populace.

Practice 4

Practice 4 is close to practice 1 and part of a group that operates more than 200 practices in the UK. With more than 10,000 NHS patients the practice is not currently taking on any new patients. Almost all (96%) of the work is provided under the NHS. There are seven surgeries, staffed by six GDPs. Two dental hygienists are employed.

Practice 5

Practice 5 is an independent practice on the outskirts of Bradford and is the longest serving practice in the study (over 30 years). It has six dental surgeries with seven GDPs and one dental hygienist, who is available on a private basis. One dentist’s services include non-surgical procedures such as fillers.

Practice 6

Practice 6 is an independent practice run by two dentists. There are no dental hygienists or therapists, but the practice provides a full range of NHS treatments (except orthodontics) to all members of the public. It also provides private treatment including sedation for anxious patients.

Qualitative results

Observations were made during 30 dental appointments. Eighteen lay people, 15 dental team staff and a member of the commissioning team took part in the interviews and focus groups (Tables 4 and 5).

The Andersen behavioural model framework was sustained in the data, with the only revision being the addition of trust as an outcome of access. During the analysis, the implications of the epistemological position of the model became apparent. A wider debate may be required about enabling the use of social resources for health and the position of health services within that. This debate is introduced in the discussion. The results described are supported by anonymised quotes from the data.

The qualitative results are presented in two sections. The first demonstrates the fit between the data and the Andersen model and the second examines the effect of the new contract as interactions between enabling and other factors in the model.

Predisposing factors

Predisposing factors could be characterised as demographic and social characteristics and beliefs. For example, family commitment could facilitate or hamper service utilisation and one participant (Grace) noted the effect of changing attitudes over time.

I can’t remember where I go, I wish I could. My daughter takes me to all my appointments.

Mary, no dentist

I think the danger is when you’re a professional this is what you do for a living, it’s a danger as its making an assumption that everybody else thinks it’s as important as you, and it is down to education and you know we see ourselves very much as a practice set up to educate and to inform. When you’re dealing with the demographic and you know the type of community we are working with, you’ve got to be very careful not to be patronising, you have to be very careful because you know everybody’s need is different and everybody’s circumstances, you know we shouldn’t just assume that because we think it’s important that they will.

Sarah, incentive dentist

I think and it’s something that kind of you’re very aware for your children as well aren’t you? I think when I was younger there was a different kind of attitude towards it and I only went to the dentist if there was a problem and I think attitudes have changed now and I think that its, you know you’re much more proactive in making sure that your children go and they get seen and that type of thing.

Grace, incentive patient

These data suggest that from the patients’ perspectives predisposing factors are things that either hamper or enhance their ability to access care. For the providers it involves fitting the service to what the dentist thinks the patient needs. The last quotation indicates a generational change and changing attitudes to the dentist.

Enabling factors

Compatible with the underlying framework, enabling factors fell into three subthemes of health policy, finance and organisations. The influence of health policy appeared in the data between the extremes of the changes associated with implementation of the blended/incentive-driven contract right through to an apparent lack of policy in some practices.

There’s going to be challenges in terms of its new, so you’ve constantly got perhaps a more demanding type of commissioning process. From a commissioner’s perspective as I mentioned earlier it takes more time.

Service commissioner

We don’t have any guidelines or anything.

Dentist at feedback event, traditional practice

A key part of the contract was the remuneration of dentists and thus finance was a rich theme. On the one hand traditional models of service use were problematic for the treatment of complex cases, whereas the problem for incentive practices became much more focused on the costs of OHAs and building relationships with patients that would enable more prevention.

The current [2006] contract works to an extent but means treating complex cases is difficult. I’m here to make a profit and a living. There’s no point me being here if that’s not going to happen. I’ve trained for a long time, I have laboratory costs to cover, I need to make money. Complex cases are difficult in this contract.

Adam, traditional practice dentist

The cost of doing that [OHAs] is far higher than was first anticipated, so if we were to put that as a separate issue we would say that it’s the costs prediction. The estimation of what it would cost to run this has been very underestimated, because it does take more time. It takes us away from doing, because we’re trying to get patients, encourage them, chase them so much it actually takes an hour for a clinician to build that relationship in the first place and then it’s about compliance and so that was collaboration, its maintaining that, it’s very hard to do.

Business manager, incentive practice

Computing problems featured as organisational factors.

Some of the challenges around things like the software I think have been hard and that’s been symptomatic because we’re using a software system that has to generate, its geared up for payment system as well as a clinical and reception management, patient management system and we’re trying to alter it to continue to provide some of the data that the contract demands around UDA generation but also around capturing some of the data that we want and because we’re such a small customer, it’s hard to influence that.

Service commissioner

The only disadvantage is I think in my opinion is the software, there are so many different varieties of software around and they have not prepared themselves to be really in tune or to deliver the type of things which are needed for this pilot, for this kind of service so that is very generic dental software, dental services software. Well in my opinion, the software should have been tweaked much before implementing this contract.

Amiya, incentive dentist

Need

Evaluated and environmental need and population health indicators were manifest in the data. Two localities in the study are characterised by material deprivation, poor oral health and long-standing undersupply of care. Unsurprisingly, this influenced dental treatment needs in these areas.

It would add up to perhaps eight dental visits just for one course of treatment and that will be just to stabilise the patients. There were an awful lot of extractions that were needed. Some patients do struggle obviously with substance misuse, and smoking is quite a major factor as well . . . We still do get those new patients who come in more or less the same state who require an awful lot of dental treatment.

Jennifer, incentive dentist

Treatment needs are so high and there’s a lot of neglected mouths. Some haven’t seen a dentist in years. Some have lost the motivation to maintain their oral health because of this. So there’s a lot of preparatory work.

Donna, incentive therapist

Health behaviours

Health behaviours involve personal health practices (such as teeth brushing or sugar consumption), the process of care (e.g. the delegation of care or other dental team behaviours) or the use of health resources (such as attending or not attending appointments). Personal practices can influence oral health positively or negatively:

There’s relationship between what you eat and your teeth and your health . . . They told us about what toothpaste to use and gave us some. They told us about drinking fizzy, time of day to brush teeth. I’ve changed it a bit and drink water now.

Jennifer, incentive patient

A novel aspect of the process of care involved the assessing patient’s risk of disease. Dental team members commented on the relative imprecision of the traffic light system and its three categories. However, concerns about imprecise systems were not restricted to the risk assessments.

Patients who are amber may be very red because of their diet and because of other things so a couple of amber should make it red, not amber. The software doesn’t pick out some of the differences.

Suneeta, incentive dentist

Focusing on UDAs, the three bands I think can be an issue, some of the dentists have said they think there should be more bands obviously just lumping all restorative work, apart from that needing lab work, into band 2, it could be anything from a simple filing to a very complex molar re-root canal treatment, it does seem a bit wacky.

Business manager, incentive practice

Attitudes and practices towards prevention varied appreciably among the dental teams:

All the dentists do give advice, you know the preventative advice, to all the patients whether under the contract or not.

Suneeta, incentive dentist

I have been here 12 years and have well-established relationships with my patients. I do find it hard to talk about their health – I’m trained to drill and fill.

Ian, traditional practice dentist

The use of health resources might involve attending or not attending appointments and accessing the wider dental team such as a dental hygienist.

I went through like a stage where I didn’t go for about 3 years but that was, I went back about 5 years ago, I don’t know why I didn’t go, I just stopped going and I think then you get thrown off the register.

Alison, traditional patient

They gave me an appointment because I need a filling and when it came day before I were panicking and worrying, I just cancelled it because at end of day I’m not going through that you know tight chest and I’m not going through all that because at end of day it’s not my fault is it? Sweets I’ve had.

Ann, incentive patient

I’ve started going to this one I’ve actually seen an oral hygienist, whilst I was private dental care I was never even offered to see somebody so this is the first time I’ve actually seen one.

Holly, incentive patient

Outcomes

Evaluated or perceived health and satisfaction with care were present in the data, as was the concept of trust. A dentist noticed his patients’ improved health and patients trusted and were satisfied with their dentist, although one disliked an interval of 2 years before they could have another assessment:

I think according to recalls they have improved their brushing, they are using fluoride toothpaste more, they have started smoking less so there is general improvement in their oral health as well and their attitudes towards oral health.

Manish, incentive dentist

They’re really good. I think they’ve got more of a modern approach there whereas the other ones still a bit, I don’t know seems a bit dated.

Kat, parent of an incentive patient, patient at a traditional practice

I do trust them here – they treated me, gave me root canal treatment and saved my tooth, without them, I’d have been minus a few teeth and my appearance would not have been good.

Nanush, incentive patient

The effects of the blended/incentive-driven contract: interactions in the data

The blended/incentive-driven contract changes the finance and organisation of dental practices to implement health policy. Its effects can therefore be seen as interactions between these enabling factors and other stages in the model.

Enabling and predisposing factors and need

The incentive practices were located in areas of high need, either associated with deprivation and disease or with the poor availability of NHS care.

We’d carried out quite a robust oral health needs assessment prior to commissioning these practices and we’d looked very closely at equity in terms of access to dental care, so we’d looked for places where there was very poor oral health and also looked at areas where there was limited access to services and we wanted a combination of those. The sites of the practices therefore were chosen on oral health needs, current access to NHS care, transport systems and so on.

Service commissioner

For some participants, the incentive practices marked a shift from no dental care, whereas others moved from private to NHS provision. Participants reported the impact of the new services that suited their needs in terms of location, personnel and ease of getting an appointment. Participants referred to the difficulties they had encountered accessing a NHS dentist:

Finding an NHS dentist has been really, really difficult and that’s why I have quite strong views on that to be honest with you. We lived somewhere else and we had an NHS dentist and then when we moved, you know what it’s like and then you lose your place and then you end up on a waiting list waiting for absolutely ages. Then we heard about this one.

Jane, incentive patient

Enabling and behaviour

The effects of the new contract could be detected on the processes of care, on personal health practices and the use of personal health resources. In turn, the process of care appeared to be affected in three ways: by the use of the care pathway, by increasing the amount of prevention and the use of skill mix.

The care pathways form an important feature of the Steele report,2 the reformed NHS contract pilots18 and the blended/incentive-driven model reported here. Initial OHAs for each patient guide treatment and the frequency of recall appointments. Participants reflected on their experiences with the pathways. Benefits included the clear link between the risk assessment and care pathways.

I think some of the people were saying you know if you have a red patient there’s only limited treatment options available for that patient until they start moving from red to amber or green, whichever. And I think that was quite a contentious issue, just to leave them in the state and then wait until they got more progressed through from each stage before you can start doing treatments. But I mean the actual process makes sense in the sense that you know there’s no point carrying out such complex treatment on someone who can’t, who is failing to kind of maintain that level of oral hygiene because it can just make it worse.

Amiya, incentive dentist

The blended/incentive-driven contract formalised Department of Health guidance35 from the perspective of clinicians. As treatment plans incorporated preventative treatment, these approaches became standard procedure.

Red, amber or green and then they do get the fluoride varnish, the smoking cessation and alcohol use is being taken automatically. And then obviously depending on the age groups with the fluoride varnish, depending on the categories, while the schedule of the appointments are set then and the recalls so it’s kind of, it’s part of our contract. We don’t do anything else.

David, incentive dentist

The blended/incentive-driven contract can be contrasted directly with the traditional practices (i.e. those operating under the 2006 contract). In this case the focus of care in the incentive contract had penetrated a traditional practice, causing them to reflect on their processes of care:

We are pushed towards UDAs rather than improving oral health . . . The prevention emphasis is an issue – we are expected to talk about perio disease and smoking and diet and have to squeeze that in. We did that before under the old system and it worked to an extent, but we have to do more and more without getting paid anymore. We have to do more in less time.

Sidney, traditional practice dentist

The feeling about the incentive/blended-driven contract as an enabler of preventative approaches was echoed by clinicians in all three incentive practices. Although such systems are not specific to the incentive model and reflect long-standing guidance,35 the formalisation of such procedures under this model was valued by service providers. Practitioners felt that it gave them time and space to care for patients.

By contrast, communication of the ratings was not always apparent in the observations. Moreover, patients might not be aware of it:

Thinking back to your initial assessment as well, I don’t know if your dentist, do they use the traffic light system?

No, what’s that?

There’s nothing where they kind of rate your oral health or anything like that?

No.

Katie, incentive patient

The incentive practices were not required to use multidisciplinary teams, but the successful bidders employed business models that utilised dental hygienists and dental therapists to deliver prevention-focused care. This wider use of skill mix has been advocated by repeated policy documents but not widely implemented. 7,37,73,74 One novel approach was for dentists to examine the patients and formulate treatment plans, but the practices did not deduct the value of the delegated treatment from dentists’ incomes.

The biggest advantage as well is, is that, it’s about mind set as well, if you are working in a normal practice your dentists, your associate will be charged with using the therapist, the hygienist and so therefore your dentist is, your associate is less likely to use your therapist and hygienist because they don’t want to pay for them. Here they get it for absolutely free.

Anna, incentive practice manager

They readily work with each other. I mean our therapist and the hygienist are generally busy the whole day, which they say they don’t normally get in other practices so, and we pay them on a fixed rate and the therapists are very happy with that because they’ve got full time work, they’re busy, the associates are happy because they’re not having to pay for them . . ., the patients get benefit because they get access to a therapist, . . . it may cost more for us to do it but it’s a more sensible way of running a business because everybody is working together for the same aim.

Claire, incentive practice manager

That’s definitely an advantage obviously because there’s more appointments available now so you can see more patients.

Fiona, Dentist, incentive practice

The new contractual arrangements were seen to influence personal health practices and the use of personal health resources. Lara changed her personal routines and attended the dental hygienist.

I’ve been prescribed Duraphat^® [(Thepenier Pharma Industrie, SAS, Mortange, France)] fluoride toothpaste, I use interdental brushes and see the hygienist. I take advice on board and I don’t want to get told off. So I’ve changed my routine. I didn’t know you could go to a hygienist until I came here.

Lara, incentive patient

Enabling and outcomes

The blended/incentive-driven contract also influenced outcomes of perceived and evaluated health and patient satisfaction.

They managed to sort my mouth so that I can actually smile now and feel confident about my smile, I don’t feel like it’s, or people are looking at my teeth anymore which, I mean it’s a massive improvement, it makes you confidence so much more, it makes you feel better about yourself when you know that someone is helping you sort something that you know is a big problem. You feel self-confident about it.

Shazia, incentive patient

I have been working here for 2 years plus so I can see that the patients from the deprived area, yes, I think according to recalls they have improved their brushing, they are using fluoride toothpaste more, they have started smoking less so there is general improvement in their oral health as well and their attitudes towards oral health.

Jane, incentive dentist

Generally making you feel better at going, I feel much happier going to the dentist now than I did a year ago.

Carol, incentive patient

That’s really good because I don’t know maybe the hygienists have got a bit more time and if it’s something they’re trained to do that’s fine rather than taking up the dentist time which is more specialist work. I mean I know my daughters been to see the hygienist, you know about brushing her teeth and they gave her some Duraphat cream.

Grace, incentive patient

Far-reaching effects

The interactions could ripple throughout the model to have far-reaching effects. For example, the RAG ratings could influence patients’ perceptions of their own needs, leading to personal behaviour changes and satisfaction (an outcome).

I think it’s good because if you know, if someone says to you, you know on this rating you are more at risk, you’re more likely to do something about it aren’t you, as opposed to someone not saying anything to you, you know, I think you are likely to be more active. I would be more active but I think it gives people that, again it’s about having that bit more choice and a bit more involvement in your own kind of care which I think is a good thing.

Grace, incentive patient

Using the ratings to determine recall intervals liberated more time for the process of care and allowed observation of increased health but influenced patient satisfaction both positively and negatively:

The recall intervals will be according to the risk assessments that we are doing and the risk assessment is based on their medical, their social and their clinical domains and patient to understand it and they are really happy, the majority of them 99% are very happy to have the recall intervals as dictated by their risk assessment . . . The recall intervals of 2 years and a year for the green and the amber patients so we are definitely seeing more patients and as the amber ones move towards green or the reds move towards green so we would be having more appointments for the patients to be seen.

Manish, incentive dentist

2 years seemed a long time to wait . . . I sort of think umm would I be prepared to pay a premium price and go back to private care as a result of that and that’s something that I’ll have to consider really.

Martin, incentive patient

These data suggest a need to reconcile contrasting views. From the dentist’s perspective there is the ability to see more patients and hence increase access. At its best, the incentive model seems to enable greater access because it prioritises those who need treatment and rebalances appointments around need. However, the patient was not happy with waiting 2 years for another assessment and may consider seeking alternative or additional care.

In a wholly positive example, a patient satisfied with her own care encouraged her partner to attend so that professional behaviour enhanced satisfaction to change predisposing factors to increase access to care.

. . . fantastic and the dentist themselves are really friendly. They’re really understanding, I mean for me I’m not somewhat scared of the dentist as my partner and it’s made a real improvement for my partner because he’s terrified of them so, and we’ve actually managed to get him there and he’s having work done there which is an improvement, usually trying to get him through the door of the dentist was a real effort, it took a day to get him there, get him sorted and get him home again. So I mean from that point of view they’ve just been absolutely fantastic with both him and me.

Holly, incentive patient

In summary, participants’ observations are compatible with existing knowledge of access to care, but highlight possible effects of the incentive contract (Figure 2). In particular, patients’ needs were seen to influence the siting of practices. The contract had a number of direct effects on practice orientation and costs. However, participants related these effects to better health-related behaviours on the part of patients and changes in dental practice behaviours regarding assessments, prevention, patient communication, the use of skill mix, the number of patients seen and recall intervals. In turn, these changes were related to improvements in perceived and evaluated health, patient satisfaction and trust.

FIGURE 2.

Summary of stakeholders’ perspectives on the incentive contract.

Design and setting

The study uses a non-randomised comparative study design. The three incentive practices were matched with three existing (traditional) practices working under the nGDS contract in West Yorkshire. The matching at practice level considered the practice in terms of deprivation index, age profile, size of practice, ethnicity profile and taking on new patients. Details of the practices are reported in Chapter 3. Patient demographics across the six practices are summarised in Table 6.

From Table 6 we can see that the ethnicity balance in the 3 and 6 pairing differs from the other pairings in as much as the proportion of those who classify themselves ‘white’ is much higher than the other pairings (≈ 90/10 in 3 and 6 compared with ≈ 65/35 in 1 and 4 and in 2 and 5). The loss to follow-up differs across practices; the lowest is 35% in practice 3 and the highest 74% in practice 1. Differences in recruitment between practices are expected and reflect the number of GDPs/size of the practices.

Sample

Sample size was estimated based on powering the primary outcome measure. Based on data from the Clarkson and colleagues’ study,121 we estimated the standard deviation (SD) in percentage sites BoP across a UK cohort to be 27.5%. We assumed a within-patient correlation in baseline to follow percentage sites BoP of 0.5 and a common variance in incentive and traditional practices. We assumed a clinically meaningful mean difference in percentage sites BoP baseline to follow-up in incentive practices of 10%, versus a mean difference in percentage sites BoP baseline to follow-up in traditional practices of 0%. We fixed a type I error rate of 0.05 and a power of 0.8. A design effect was included to account for clustering of patients within the three incentive and three traditional practices, assuming an intracluster correlation of 0.2. A two-sided two independent samples t-test identified a total of 550 patients to be recruited (allowing for a 10% loss to follow-up).

Recruitment and inclusion and exclusion criteria

To achieve a recruitment of 550 new patients in the study project, recruitment was initially planned to take place over a 6-month period, although this was subsequently increased to 8 months. This figure was based on (1) six dental practices included in the study; (2) an average list size of 1000 adult patients per dentist; (3) 10% of patients per year will be new patients to the practice (estimated from the dental public health audit figures of practices in Bradford and Airedale); and (4) of these patients we estimated a minimum of 50% would agree to participate in the study over the recruitment period. Thus, at a practice level, the three incentive practices were matched with three existing traditional practices of similar size, deprivation index, age profile, ethnicity profile and taking on new patients.

At the patient level, the inclusion criteria were that patients must be:

• aged ≥ 16 years (all the patient information leaflets were reviewed by a teenager as suitable to be easily understood for use with a broad age group in the study)

• willing to be followed up for 24 months and give informed consent

• a new patient to the dental practice

• able to complete the patient-completed questionnaires.

If a translator was needed, the availability of provision of a translation service in the spoken language of the participant was via the normal dental practice access routes to such services. Study materials were not translated as the dental practices reported that those accessing the translator services are typically not literate in their mother language. Given that questionnaires were to be completed only at baseline and at 24 months, after taking advice, we concluded that the provision of a translator was sufficient, and indeed the best option to maximise inclusion. All other data collected on the participants were from routinely collected data and were not affected by the language of the participant.

With regard to exclusion criteria, and the specific handling of those who are edentulous, edentulous patients were not excluded from the sample. However, they would be considered supplementary to the core sample of 550 patients and would provide additional specific data. Postcode, age and ethnicity of all patients included within the sample were recorded and profiled during the analysis.

The practices purposefully opened their lists to recruit new patients for our study. All new patients attending the practice for the first time were invited to participate. Prior to their scheduled appointment they were sent a letter of invitation and an information sheet and provided informed consent to be part of the study. Informed consent was taken by members of the dental surgery staff who received training in taking informed consent. Patients were asked to complete the following questionnaires at their first visits and at the follow-up visit 24 months later: the European Quality of Life-5 Dimensions questionnaire, 3-level version (EQ-5D-3L; details of the EQ-5D-3L are presented in Chapter 5, Methods and Results) and the OHIP-14. The dentist undertook the clinical assessment of teeth and gingivae using the ICDAS and BoP at both visits. Family/social history was taken at the first visit only. The OHA, using the RAG system, was completed at baseline and 24 months by the incentive practices.

Patients were contacted by the dental practice around 6–8 weeks prior to their 24-month follow-up date to arrange an appointment. Contact assumed a variety of media, including telephone, short message service (SMS) and letter, in order to optimise follow-up. Patients were contacted a minimum of three times to arrange the appointment.

Ethics approval and archiving

Ethics approval was obtained from the NRES Committee (London – Bromley, REC reference 12/LO/0205) on 5 April 2012. The study was sponsored by the University of Leeds, Leeds, UK. At the end of the study, data will be securely archived in line with the University of Leeds’ procedures for a minimum of 6 years. Data held by the research teams will be locally archived or as instructed by the University of Leeds where it is typically stored in the university’s archive facility and site data and documents. Following authorisation from the sponsor, arrangements for confidential destruction will then be made.

Intervention

Details of the blended/incentive-driven contracts for the incentive practices are described in detail in Chapter 2. In brief, within the incentive practices, the blended/incentive-driven contracts were aimed at ensuring evidence-based preventative interventions are delivered in line with identified needs for a defined population; increased access to dentistry; and that care is provided by the most appropriate team member to encourage skill mix. Within the contracts, 60% of the contract value is apportioned to delivery of a set number of UDAs. The remaining 40% is dependent on the delivery of quality: 20% systems, processes and infrastructure [e.g. cross-infection, policies, Standards for Better Health (latterly becoming Care Quality Commission domains)] and 20% OHImp.

The traditional practices operated under the nGDS contracts which are based on payment for UDAs with the number of UDAs based on historic activity and agreed between PCTs and dental practices. The nGDS contracts meant that the payment mechanism changed from a one-off fee per item of service to a system whereby providers are paid an annual sum in return for delivering an agreed number of ‘courses of treatment’ weighted for complexity.

Main outcome measures

The primary outcome of the clinical effectiveness study was the percentage of points from BoP, assuming 32 teeth per patient and 6 points per tooth, reduced by the number of missing teeth (ICDAS codes ‘97’, ‘98’, ‘99’ or ‘P’) that a patient has.

The secondary outcomes were:

1. Percentage of sound surfaces (defined as free of obvious caries, if any of the following ICDAS codes ‘00’, ‘01’, ‘02’, ‘10’, ‘11’, ‘12’, ‘20’, ‘21’ or ‘22’ are recorded), assuming 32 teeth per patient and five surfaces per tooth (reduced by the number of missing teeth a patient has).

2. Percentage of extracted and filled teeth, assuming 32 teeth per patient (reduced by the number of missing teeth a patient already has). A tooth was defined as filled if the first digit of the ICDAS code was ‘3’, ‘4’, ‘5’, ‘6’, ‘7’ or ‘8’ and the second digit of the ICDAS code was not ‘4’, ‘5’ or ‘6’. A tooth was defined as extracted if the ICDAS code was ‘97’.

3. OHIP-14 total score (integer count between 0 and 56). Each of the 14 individual questions is scored on a 0–4 scale. Any scores outside of this range are regarded as missing. If one or two of the individual questions are missing, then median imputation is used to impute the individual question scores. 122 If three or more individual questions are missing, then the total score is regarded as missing and the patient excluded from analysis.

Exploratory analyses of the RAG system were conducted to study the:

1. distributions across categories (RAG) at baseline and follow-up

2. transitions between categories from baseline to follow-up

3. proportion of risk assessments that were manually over-ridden at baseline and follow-up.

Analysis

Multiple linear regression123 is used to model the primary and secondary outcome measures. In our analysis plan we identified the difference in the measurement from baseline to follow-up as the outcome. However, given the reduced sample size because of loss to follow-up, to improve power, we used an analysis of covariance (ANCOVA) approach with follow-up measurement as the outcome and baseline measurement as a covariate. An important requirement for ANCOVA is that there must be no interaction between the groups and baseline measure. 124 Although it is plausible to assume no interaction within each incentive–traditional practice-matched pair, it is not clear if an interaction will be present across the three matched pairs. For this reason we first analyse the matched pairs separately before combining in a single analysis. Because of the reduced sample size and staff turnover, we have not been able to include practitioner-level variables in our analyses. Imputation was conducted when constructing the OHIP-14 total score as described above.

Sample

Recruitment started on 1 June 2012 and the first patient entered the study on 14 June 2012. Recruitment finished on 31 January 2013. In total 550 participants were recruited to the study at baseline. However, only 291 participants attended both baseline and a follow-up appointment at 24 months. The reasons for non-attendance at the 24-month follow-up were recorded by the dental practices and are grouped as follows.

• Patient non-response: patient did not respond to any letters, text messages (SMS) or voicemails.

• Patient failed to attend: patient failed to attend or cancelled the appointment(s) at the last minute (i.e. ≤ 24 hours’ notice).

• Change in patient status: patient moved away or was unable to attend because of financial difficulties or cancelled the appointment as a result of health reasons (e.g. a recent operation) or changed practice.

• Practice decision: de-registered patients or cases in which the practice cancelled the appointment because of severe weather conditions.

• Practice error: dentist failed to do the follow-up study/assessment.

We should note that for some de-registered patients in the ‘practice decision’ category, no further reason was provided. For others in the same group, the practice de-registered the patient for a variety of reasons that included failure to attend multiple appointments and the patient consecutively cancelling appointments. In addition, this category included de-registration because the patient had moved away. This last reason overlaps with the change in patient status category and, as such, the interpretation of these categories should be treated with caution.

Reasons for no follow-up at 24 months were available for 120 of the patients who did not attend the follow-up. From the answers provided, 27 patients were from the traditional practices and 93 patients were from the incentive practices. As shown in Figure 5, ‘patient non-response’ in both incentive and traditional practices is the most frequent reason for not completing the study (45% and 59%, respectively). In the traditional practices, one patient (4%) did not complete the follow-up because ‘patient failed to attend’, but this number was much higher in the incentive practices – 29% (27 patients). ‘Practice decision’ or ‘centre error’ are not reported in the traditional practices and even for the incentive practices the percentages are very low (4% and 9%, respectively).

FIGURE 5.

Reasons for no follow-up. (a) Incentive practice; and (b) traditional practice.

Of the 550 recruited, a BoP was measure at baseline in 529. A total of 291 participants attended both baseline and follow-up, BoP was measured at both time periods in only 270 (51.04% of those with a BoP measurement at baseline). Following quality assurance, 188 were included in the BoP analysis (incentive, n = 90; traditional practice, n = 98). Full details of the sample number available for analyses for each measure are included in Table 7.

Demographic and baseline outcome measures split by status (completed follow-up/loss to follow-up) are shown in Table 8. Outcomes data were manually reviewed and quality controlled by a clinician prior to the modelling reported in Table 9 and onwards. Patients with implausible data were excluded from modelling. This manual review was conducted only for those with completed follow-up, to enable comparison with those lost to follow-up. Table 8 uses outcomes data prior to this manual review and quality control process.

Those who are lost to follow-up are generally younger and more likely to be male. In terms of outcomes, those lost to follow-up have worse oral health, although this is more variable (a higher mean and SD). This was reflected in the OHIP-14 score, with those lost to follow-up having higher mean scores (indicating poorer OHQoL). In the practices that undertook a RAG rating, those lost to follow-up were more likely to have a red rating.

Bleeding on probing

Figure 6 shows a scatterplot of BoP at baseline and follow-up. The mean (SD) BoP at baseline and follow-up was 41.88 (SD 47.24) and 32.97 (SD 47.66), respectively. Eighty-two patients excluded were following manual review for quality control in addition to those lost to follow-up (see Table 7). This comprised 3, 29, 6, 6, 16 and 22 patients for practices 3, 2, 1, 6, 5 and 4, respectively. In addition, four patients with BoP in excess of 100% had BoP imputed to 100%. This gave a sample for analyses of n = 188 (n = 90, incentive practices; n = 98, traditional practices).

FIGURE 6.

Scatterplot of BoP by practice.

First we analysed the matched pairs separately using a baseline-adjusted ANCOVA. Table 9 relates to matched pair 1 and 4, Table 10 to matched pair 2 and 5 and Table 11 to matched pair 3 and 6.

The model in Table 9 implies that BoP was lower at follow-up for patients in traditional practice 4 (than in incentive practice 1) by an average of 22.55% [95% confidence interval (CI) 4.69% to 40.41%; p = 0.01].

The model in Table 10 implies that BoP was higher at follow-up for patients in traditional practice 5 (than in incentive practice 2) by an average of 14.57% (95% CI 5.55% to 23.59%; p < 0.01).

The model in Table 11 implies that BoP was higher at follow-up for patients in traditional practice 6 (than in incentive practice 3) by an average of 9.98% (95% CI 2.90% to 17.06%; p < 0.01).

Scatterplots of BoP at baseline and follow-up are shown for the matched pairs in Figure 7. Lines of best fit from the models described in Tables 9–11 are superimposed.

FIGURE 7.

Scatterplots of BoP by incentive vs. traditional practice. (a) Matched pair 1 and 4; (b) matched pair 2 and 5; and (c) matched pair 3 and 6.

Combining the matched pairs into a single ANCOVA model produces the output shown in Table 12. This should be interpreted with caution because of the difference between the matched pairs observed in the analyses.

The model in Table 12 implies that BoP was higher at follow-up for patients in traditional practices (than in incentive practices) by an average of 10.24% (95% CI 3.23% to 17.25%; p < 0.01).

A scatterplot of BoP at baseline and follow-up is shown across all the matched pairs in Figure 8. Lines of best fit from the model described in Table 12 are superimposed.

FIGURE 8.

Scatterplot of BoP by incentive vs. traditional practice across all matched pairs.

In summary, for BoP pooled across practices, the 95% CI for the effect size was 3.23 to 17.25, indicating a positive effect for incentive practices but with considerable uncertainty in magnitude. Together with reservations about the validity of pooling (because of the heterogeneity of effects across practice pairs) and differential dropout, the results should be treated with caution.

International Caries Detection and Assessment System change in sound surfaces

Figure 9 shows a scatterplot of sound surfaces at baseline and follow-up. The mean (SD) proportion of sound surfaces at baseline and follow-up was 84.52% (SD 16.45%) and 82.67% (SD 18.09%), respectively. Eighty-three patients excluded were following manual review for quality control in addition to those lost to follow-up (see Table 7). This comprised 3, 28, 5, 8, 16 and 23 patients for practices 3, 2, 1, 6, 5 and 4, respectively.

FIGURE 9.

Scatterplot of sound surfaces by practice.

First we analysed the matched pairs separately using a baseline-adjusted ANCOVA. Table 13 relates to matched pair 1 and 4, Table 14 to matched pair 2 and 5 and Table 15 to matched pair 3 and 6.

The model in Table 13 implies that the proportion of sound surfaces at follow-up was higher for patients in traditional practice 4 (than in incentive practice 1) by an average of 1.03% (95% CI –2.19% to 4.25%). This was not a significant difference between traditional and incentive practices (p = 0.52).

The model in Table 14 implies that the proportion of sound surfaces at follow-up was higher for patients in traditional practice 5 (than in incentive practice 2) by an average of 1.15% (95% CI –2.38% to 4.68%). This was not a significant difference between traditional and incentive practices (p = 0.52).

The model in Table 15 implies that the proportion of sound surfaces at follow-up was higher for patients in traditional practice 6 (than in incentive practice 3) by an average of 13.49% (95% CI 4.38% to 22.60%; p < 0.01).

Scatterplots of the proportion of sound surfaces at baseline and follow-up are shown for the matched pairs in Figure 10. Lines of best fit from the models described in Tables 13–15 are superimposed.

FIGURE 10.

Scatterplots of the proportion of sound surfaces by incentive vs. traditional practice. (a) Matched pair 1 and 4; (b) matched pair 2 and 5; and (c) matched pair 3 and 6.

Combining the matched pairs into a single ANCOVA model produces the output shown in Table 16. This should be interpreted with caution because of the difference between the matched pairs observed in the analyses.

The model in Table 16 implies that the proportion of sound surfaces at follow-up was higher for patients in traditional practices (than in incentive practices) by an average of 4.68% (95% CI 2.18% to 7.18%; p < 0.01).

A scatterplot of the proportion of sound surfaces at baseline and follow-up is shown across all the matched pairs in Figure 11. Lines of best fit from the model described in Table 16 are superimposed.

FIGURE 11.

Scatterplot of the proportion of sound surfaces by incentive vs. traditional practice across all matched pairs.

In summary, the proportion of sound surfaces (defined as caries free and having initial ICDAS codes 1 and 2) at follow-up was higher (4.68%) in traditional practices, although in two of the pairings the difference in the proportion of sound surfaces was non-significant.

International Caries Detection and Assessment System change in extractions and fillings

Figure 12 shows a scatterplot of the proportion of extractions and fillings (E&F) at baseline and follow-up. The mean (SD) of the proportion of E&F at baseline and follow-up was 29.36% (SD 27.02%) and 34.19% (SD 27.89%), respectively. Eighty-one patients were excluded following manual review for quality control in addition to those who were lost to follow-up (see Table 7). This comprised 3, 28, 5, 6, 17 and 22 patients for practices 3, 2, 1, 6, 5 and 4, respectively. In addition, 10 patients with E&F in excess of 100% had E&F imputed to 100%.

FIGURE 12.

Scatterplot of E&F by practice.

First we analysed the matched pairs separately using a baseline-adjusted ANCOVA. Table 17 relates to matched pair 1 and 4, Table 18 relates to matched pair 2 and 5 and Table 19 to matched pair 3 and 6.

The model in Table 17 implies that the proportion of E&F at follow-up was lower for patients in traditional practice 4 (than in incentive practice 1) by an average of 15.11% (95% CI 9.84% to 20.38%; p < 0.01).

The model in Table 18 implies that the proportion of E&F at follow-up was lower for patients in traditional practice 5 (than in incentive practice 2) by an average of 0.95% (95% CI 4.32% to –6.21%). This was not a significant difference between traditional and incentive practices (p = 0.72).

The model in Table 19 implies that the proportion of E&F at follow-up was lower for patients in traditional practice 6 (than in incentive practice 3) by an average of 2.89% (95% CI –6.92% to 1.15%). This was not a significant difference between traditional and incentive practices (p = 0.15).

Scatterplots of E&F at baseline and follow-up are shown for the matched pairs in Figure 13. Lines of best fit from the models described in Tables 17–19 are superimposed.

FIGURE 13.

Scatterplot of E&F by incentive vs. traditional practice. (a) Matched pair 1 and 4; (b) matched pair 2 and 5; and (c) matched pair 3 and 6.

Combining the matched pairs into a single ANCOVA model produces the output shown in Table 20. This should be interpreted with caution because of the difference between the matched pairs observed in the analyses.

The model in Table 20 implies that the proportion of E&F was lower at follow-up for patients in traditional practices (than in incentive practices) by an average of 4.43% (95% CI 1.34% to 7.52%; p = 0.01).

A scatterplot of the proportion of E&F at baseline and follow-up is shown across all the matched pairs in Figure 14. Lines of best fit from the model described in Table 20 are superimposed.

FIGURE 14.

Scatterplot of E&F by incentive vs. traditional practice across all matched pairs.

In summary, the traditional practices had a lower number of E&F at follow-up by 4.43%. However, for two of the three pairings there was no significant difference in E&F.

Oral Health Impact Profile

Figure 15 shows a scatterplot of OHIP-14 at baseline and follow-up. The mean (SD) OHIP-14 score at baseline and follow-up was 9.00 (SD 10.37) points and 6.09 (SD 8.08) points, respectively.

FIGURE 15.

Scatterplot of OHIP-14 scores by practice.

At baseline, there were three patients in whom answers to three or more questions were missing, and thus were excluded, and 12 patients in whom answers to one or two questions were missing (and were imputed as noted previously). At follow-up (excluding those lost to follow-up), there were 10 patients in whom answers to three or more questions were missing, and who were thus excluded, and 10 patients in whom answers to one or two were missing (and who were imputed as noted previously). Overall, 96 patients were excluded following manual review for quality control in addition to those who were lost to follow-up (see Table 7). This comprised 4, 32, 8, 6, 22 and 24 patients for practices 3, 2, 1, 6, 5 and 4, respectively.

First we analysed the matched pairs separately using a baseline-adjusted ANCOVA. Table 21 relates to matched pair 1 and 4, Table 22 relates to matched pair 2 and 5 and Table 23 to matched pair 3 and 6.

The model in Table 21 implies that the OHIP-14 score was higher at follow-up for patients in traditional practice 4 (than in incentive practice 1) by an average of 5.15 points (95% CI 1.17 to 9.14 points; p = 0.01).

The model in Table 22 implies that the OHIP-14 score was higher at follow-up for patients in traditional practice 5 (than in incentive practice 2) by an average of 2.74 points (95% CI 5.68 to –0.18 points). This was not a significant difference between traditional and incentive practices (p = 0.07).

The model in Table 23 implies that the OHIP-14 score was lower at follow-up for patients in traditional practice 6 (than in incentive practice 3) by an average of 1.63 points (95% CI –5.80 to 2.53 points). This was not a significant difference between traditional and incentive practices (p = 0.43).

Scatterplots of the OHIP-14 scores at baseline and follow-up are shown for the matched pairs in Figure 16. Lines of best fit from the models described in Tables 21–23 are superimposed.

FIGURE 16.

Scatterplot of the OHIP-14 score by incentive vs. traditional practice. (a) Matched pair 1 and 4; (b) matched pair 2 and 5; and (c) matched pair 3 and 6.

Combining the matched pairs into a single ANCOVA model produces the output shown in Table 24. This should be interpreted with caution because of the difference between the matched pairs observed in the analyses.

The model in Table 24 implies that the OHIP-14 score was higher at follow-up for patients in traditional practices (than in incentive practices) by an average of 3.05 points (95% CI 1.20 to 4.90 points; p < 0.01).

A scatterplot of OHIP-14 score at baseline and follow-up is shown across all the matched pairs in Figure 17. Lines of best fit from the model described in Table 24 are superimposed.

FIGURE 17.

Scatterplot of OHIP-14 score by incentive vs. traditional practice across all matched pairs.

In summary, overall, OHIP-14 total scores at follow-up score were 3.5 points higher in traditional practices, indicating worse OHQoL. Again, for two of the three pairings there was no significant difference.

Risk assessment traffic light system

Table 25 shows the distribution of risk assessments in the RAG categories at baseline and follow-up. All patients who reported a risk assessment rating are included regardless of whether or not they had BoP assessment at baseline and follow-up, or whether or not they were excluded from modelling work following manual data review for quality control.

Table 25 implies that the most common risk assessment category at both baseline and follow-up was red (73.93% and 44.14%, respectively) and the least common risk assessment category in both time periods was green (4.64% and 17.11%, respectively). The table suggests a possible improvement in risk assessment category from baseline to follow-up, with increasing proportions of patients in amber and green categories as opposed to red. However, the inclusion of patients lost to follow-up in the baseline assessments may also account for these changes.

At the practice level, there were two, one and no incomplete risk assessment traffic lights in practices 1, 2 and 3, respectively. At follow-up (excluding those lost to follow-up) the figures were 7, 14 and 1, respectively.

Table 26 shows the transitions between categories from baseline to follow-up.

Table 26 implies that, in general, patients’ risk assessments improve from baseline to follow-up. Of the 76 patients in the red category at baseline who attended follow-up, only 47 (61.84%) remained red, with 26 (34.21%) moving to amber and three (3.95%) moving to green. Of the 26 patients in the amber category at baseline attending for follow-up, 10 (38.46%) moved to green and only two (7.69%) moved to red. Of the nine patients in the green category at baseline who attended follow-up, six (66.67%) remained green, with the other three (33.33%) moving to amber. Those patients who were red at baseline were less likely to attend follow-up (76 of 207, 36.71%) than those who were amber (26 of 60, 43.33%) or green (9 of 13, 69.23%) at baseline.

Table 27 shows the distribution of risk assessments that were manually over-ridden at baseline and follow-up.

Table 27 implies that very few risk assessments were over-ridden [13 of 280 (4.64%) at baseline and 8 of 111 (7.21%) at follow-up]. Although it would be unwise to overinterpret the data given these small numbers, over-rides to red or amber are more common than over-rides to green (92.31% vs. 7.69% at baseline compared with 87.50% vs. 12.50% at follow-up).

At the practice level, there were 10, 1 and 1 incomplete manual over-ride records at practices 1, 2 and 3, respectively. At follow-up (excluding those lost to follow-up) the figures were 9, 15 and 1, respectively.

Dental practitioners were asked to give reasons for any over-rides made in the risk assessment and thus the changed care pathway. Over-rides to a lower risk tended to be characterised by general comments about improved oral health, the patient’s commitment to improved oral health and reference to caries. The following are typical responses:

• oral health improved

• oral hygiene improved; patient is caries free

• moderate risk for periodontitis and caries

• patient currently amber, but committed to improving oral health

• minor calculus present in two quadrants plus heavy calculus in one gives red, but as everything else is healthy and the patient is highly motivated moved to amber

• minimum calculus present and committed to improving

• more carious teeth last visit.

There were, however, examples of over-rides related to family health conditions or medication that put the patient in a higher risk category than the dentist thought warranted:

• patient on hay fever tablets – computer placed on red pathway, over-ride to amber

• diabetes in the family – placed on high risk, over-ride to medium risk.

The over-rides that placed patients on higher risk than indicated by the RAG system included changes based on the result of investigations within the dental appointment (specifically radiographs) and the patient’s lack of commitment to their OHImp evidenced through attendance at the dentist:

• after further investigation of the bitewing radiographs taken, bone loss found, which makes the patient high risk because of compromised periodontal health

• patient failed to attend for treatments

• irregular attender.

In summary, using the RAG system (incentive practices only), among those who attended both baseline and follow-up there was an improvement with 68% red at baseline and 44% red at follow-up. Those who were red at baseline were less likely to attend follow-up. A total of 36.71% of those who were red at baseline attended follow-up; 69.23% of those who were green at baseline attended follow-up. There were very few decision over-rides on the RAG system.

Outcome measures

Patient health-related quality of life was assessed using the OHIP-14115,116 and the EQ-5D-3L. 125,126 Patients completed both measures at baseline and at the end of the 24-month assessment period.

The OHIP-14 is a non-preference-based measure of the impact of oral disorders on individuals’ well-being. 115 It contains 49 questions (seven questions for each of the seven dimensions of impact of an oral condition) based on a theoretical model of oral health. 130 The OHIP-14 used here is a shorter version based on a subset of two questions for each of the seven dimensions. 116 The OHIP-14 questionnaire asks patients to rate the problems they had with their mouth, teeth or gums in the last 6 months. It consists of 14 items that capture seven dimensions of functional limitation, physical pain, psychological discomfort, physical disability, psychological disability, social disability and handicap. Response options to each of these items ranged between 0 and 4, with 4 = very often, 3 = fairly often, 2 = occasionally, 1 = hardly ever and 0 = never. The total OHIP-14 score was calculated using the additive method (i.e. the sum of the responses to the 14 items). This method was selected because of its operational simplicity and its effectiveness in measuring OHQoL between groups. 117 The total OHIP-14 score ranges between 0 and 56 points, with lower scores indicating better OHQoL.

The QALYs were estimated using the EQ-5D-3L. The EQ-5D-3L is a standardised generic instrument for use as a measure of health outcome. 126 It can be applied to a range of health conditions and treatments, and provides a health-state profile or a utility value index for an individual. In the EQ-5D-3L questionnaire, patients were asked to rate their health on five health-state dimensions. These dimensions were mobility (e.g. walking about); self-care (e.g. washing or dressing); usual activities (e.g. work, study, housework, family or leisure activities); pain/discomfort; and anxiety/depression. Each of the five health states has a three-level response (1 = no problems, 2 = some/moderate problems or 3 = extreme problems/unable). The item responses were converted to health-state utility values using an algorithm derived from a UK tariff. 131 The health-state values could take negative values (i.e. worse than death), but could not exceed 1 (i.e. full health). These utility values were then used to estimate QALYs using the area under the curve approach, in which 1 QALY is the equivalent of 1 year of full health. As any effects were observed over 2 years, discounting the QALY and OHIP-14 score was necessary. The recommended rate of 3.5% was used. 129

For the mapping exercise, we used baseline data from the EQ-5D-3L, OHIP-14 and information on other controls, including date of birth and sex. To simplify the analysis, the responses to each of the OHIP-14 items were recoded into three wide groups: ‘never’, ‘hardly ever’ and ‘occasionally/fairly often/very often’. These groups were also converted into indicator variables with values of zero indicating that there was no response in that group and values of 1 indicating a positive response in that group. Age was calculated in years using the reported year of birth and sex was coded 1 for male and 0 for female.

Resource use and costs

To recap, as set out in Chapter 2, within the incentive practices, 60% of the contract value is apportioned to delivery of a set number of UDAs. The remaining 40% is dependent on the delivery of quality: 20% systems, processes and infrastructure (e.g. cross-infection, policies, Standards for Better Health latterly becoming Care Quality Commission domains) and 20% OHImp. With the traditional practices, contracts are drawn up to deliver a set number of UDAs for an agreed financial value. Thus, all practices were paid the cost of the UDAs claimed, but incentive practices also received a ‘fixed’ cost equivalent to 40% of the total contract value (TCV).

Units of dental activity are based on three treatment bands (band 1: diagnosis, treatment planning and maintenance; band 2: simple treatment, e.g. fillings; and band 3: complex treatment, e.g. bridges or crowns). Each band attracts a set number of UDAs (band 1 = 1 UDA; band 2 = 3 UDAs; band 3 = 12 UDAs) and UDAs awarded for completed treatments. 2

It is important to note that the value of UDAs (i.e. the payment made for each UDA) varies between dental practices and thus UDA values vary from provider to provider. Typically, the greater the need for NHS dentists in an area the higher the UDA value.

For the analysis from the commissioner’s perspective, NHS England provided us with the financial value of UDAs for the traditional practices in the relevant years. The UDA value assigned to the incentive practices in this analysis includes the 40% of the contract value paid to the practices dependent on delivery of quality (systems, processes and infrastructure) and OHImp.

Information on the number, and value, of UDAs claimed per patient within the study was collected from recorded appointment history data from the practices.

The analysis from the service provider’s perspective included information on the number and duration of appointment; the type of treatment; and the dental professional carrying out the appointment/treatment. This information was provided by the dental practices and taken from the appointment history of each patient. Material costs such as the cost of radiographic film or the cost of filling materials are based on the Kent Express Catalogue132 and are shown in Table 28. The salaries of the different staff involved in the treatment were obtained from national sources such as the Pay Circular134 and the NHS Agenda for Change135 and overhead costs were calculated using the same method as in Hulme and colleagues128 (Table 29). Costs and salaries are adjusted for inflation using the CCEMG-EPPI (Campbell and Cochrane Economics Methods Group – Evidence for Policy and Practice Information and Co-ordinating Centre) cost converter. 136 Discounting of costs was also necessary and a discount rate of 3.5% was used. 129

Cost-effectiveness

The outcome of the cost-effectiveness analyses was the incremental cost per unit change in the OHIP-14 score and the cost per QALY. We present ICERs representing the ratios of the incremental cost and incremental benefits (OHIP-14 points/QALYs) between incentive and traditional care practices. The ICER represents the additional cost per one unit of outcome gained, in this case per OHIP-14 point lost/QALY gained for incentive versus traditional practices. As a guideline rule, NICE129 accepts as cost-effective those interventions with an ICER of < £20,000 per QALY. The NICE guidelines129 state that, in general, a treatment that costs > £30,000 per QALY would not be considered cost-effective. In addition, we present cost-effectiveness plane scatterplots showing the uncertainty surrounding the cost-effectiveness estimates. The cost-effectiveness planes were derived using bootstrapping with replacement. This stochastic uncertainty analysis involved running 10,000 bootstrapped estimates of the incremental costs and QALYs/OHIP-14 scores. The bootstrap approach is a non-parametric method that treats the original sample as though it was the population and draws multiple random samples from the original sample. Cost-effectiveness acceptability curves (CEACs) were also generated to illustrate the probability that each treatment would be cost-effective given a range of acceptable threshold values. 137

Sensitivity analyses were further carried out to account for uncertainty in the cost values. For the commissioner’s perspective, we performed one-way sensitivity analyses by assuming either no change or a 3% increase in the TCV per (financial) year or a 0%, 10% increase or 10% decrease in the number of patients treated per year. For the analysis from the service provider’s perspective, costs were altered by ± 20%. Although these values are essentially arbitrary, it was considered likely to represent any uncertainty in the cost values.

Missing data

Following Slade and colleagues,122 we used median imputation if one or two OHIP-14 item scores were missing. Participants in whom more than two components of the OHIP-14 were missing or or for whom EQ-5D-3L at baseline and follow-up was missing are excluded from the analysis. Multiple imputation was also used to attempt to deal with missing EQ-5D-3L/OHIP-14 scores at follow-up; however, the missing values were not missing at random, making this technique not appropriate to use here (see Appendix 3 for more details).

Sample

A total of 550 patients were recruited in this study (see Chapter 4). A total of 210 of those patients were included in the cost-effectiveness analyses. In detail, four patients withdrew prior to their appointment, five patients did not have appointment history data at baseline and/or follow-up (hence costs could not be estimated) and 289 patients did not complete the follow-up assessment. From the remaining 252 patients, 42 patients had more than two components of the OHIP-14 missing or were missing EQ-5D-3L at baseline and follow-up, and were excluded from the analysis (13 patients at baseline and 10 patients at follow-up had two or fewer OHIP-14 item scores missing so we were able to impute the missing data and include these in the analysis). Therefore, for 210 patients (108 from incentive practices and 102 from traditional practices) data were complete [costs, EQ-5D-3L and OHIP-14 (after imputation for those with only one missing item)] and were included in the cost-effectiveness analyses. Baseline characteristics of the patients are presented in Table 30. There are no statistically significant differences in age, sex or ethnicity between those included in these analyses and those lost to follow-up.

Resource use and costs

Table 31 shows the average number of UDAs per person over the 2 years of the study. There are negligible differences in the average number of UDAs per person if the treatment/appointment was carried out by a dentist. It is difficult to draw conclusions for the therapist and hygienist as some of the incentive and/or traditional practices did not have these dental professionals. Table 32 shows the average number of appointments per person over the study. Patients in the incentive practices had more appointments, on average, than those in the traditional practices. The picture is similar when looking at the appointments carried out by a dentist. Patients in both groups had, on average, the same number of appointments with an oral health educator. Again, it is not possible to compare incentive with traditional practices with regards to the number of appointments performed by the rest of the dental professionals.

For the patients who were lost to follow-up (and those for whom we had appointment data), the average number of dental appointments was 7.97 (SD 5.34; n = 152) for the incentive practices and 4.99 (SD 3.53; n = 131) for the traditional practices. Treatment within these appointments included fillings, dentures and crowns. For example, 72% of patients in the incentive practices and 60% in the traditional practices had one or more fillings; 77% and 69%, respectively, had radiographs taken; 8% and 20%, respectively, had dentures made; and 6% and 5%, respectively, had a crown.

For the 210 people included in the health economics study, the mean time spent with the dental professional by group is presented in Table 33. On average, dentists in the incentive practices spent less time with patients than dentists in the traditional practices and therapists in incentive practices spent more time with patients than therapists in traditional practices.

Costs from the perspective of the commissioner

The UDA value paid for each of the practices is shown in Table 34 (the incentive UDA value includes the 40% of the contract dependent on quality and processes).

The mean per-person cost over the study for the incentive practices is £459.77 and for the traditional practices is £281.57 (Table 35).

Costs from the perspective of the dental provider

As highlighted earlier, the costs to the dental provider are made up of the practitioners’ time, materials and laboratory costs. In addition, the dental surgeries receive a payment for UDAs delivered (incentive and traditional practices). The sum of these payments is subtracted from the costs accruing to the dental provider (time/materials/laboratory costs). These payments are the cost to the commissioner detailed in the previous section.

The mean cost of the materials and any laboratory cost per person are presented in Table 36.

The costs are combined with the cost of the time spent by the different dental professionals, and the per-patient cost of UDAs (see Table 35) to find the per-patient cost from the perspective of the service provider. The mean per-person cost from the perspective of the service provider for the incentive practices is –£209.26, whereas for the traditional practices the mean cost is –£116.21 (Table 37).

Outcome (quality-of-life) data

Overall, improvement in the OHIP-14 scores between baseline and follow-up is observed in both the incentive and traditional practice groups (Table 38). However, no statistically significant difference in EQ-5D-3L score was found between groups or over time. There are no statistically significant differences in the OHIP-14 scores between groups, but there are statistically significant differences (at the 5% level) in these scores over time in both groups.

In addition, the changes in the EQ-5D-3L scores by group (or over time) are negligible (Table 39), making it difficult to draw any conclusions about the impact of the new care practice on the patients’ general health-related quality of life.

Cost-effectiveness results

Table 40 shows the costs, from the commissioner’s perspective, and changes in OHIP-14 scores for each of the two groups. The traditional practices had lower costs and higher OHIP-14 scores (i.e. patients in the traditional practices had worse OHQoL). This represents an estimated cost of £199.22 per 1-point decrease (improvement) in OHIP-14 score for the incentive practices.

Similarly, Table 41 shows the costs from the service provider’s perspective. In incentive practices, costs to the dental providers were lower (a greater surplus once costs and income from payments for services delivered had been taken into account) and health outcomes were better. Thus, from the dental provider’s perspective incentive dominates traditional practice.

Similarly, Tables 42 and 43 show the costs and QALYs for each of the two groups. They also provide the incremental costs and benefits (expressed as QALY gains) as well as the ICER. Any interpretation in this case should be tempered given the negligible differences in QALY gains between the two groups.

Figure 19 shows the cost-effectiveness plane for incentive care practice compared with traditional care practice using the QALY outcome measure from the commissioner’s perspective. The sample estimates are spread mainly in the north-west and north-east quadrants, suggesting that incentive care practice is unlikely to be cost-effective from the commissioner’s perspective. Iteration results seem to be equally spread in north-east and north-west quadrants, making it difficult to draw conclusions on which of the two care practices leads to better general health-related quality of life. Similarly, Figure 20 shows the cost-effectiveness plane for incentive care practice compared with traditional care practice using the QALY outcome measure from the perspective of the service provider. Iteration results in this case are spread in the south-west and south-east quadrants, indicating that incentive practices had lower costs than traditional dental practices.

FIGURE 19.

Cost-effectiveness plane for incentive care practice compared with traditional care practice (outcome measure: QALY, commissioner’s perspective).

FIGURE 20.

Cost-effectiveness plane for incentive care practice compared with traditional care practice (outcome measure: QALY, service provider’s perspective).

Figures 21 and 22 show the cost-effectiveness plane for incentive care practice compared with traditional care practice using the OHIP-14, from the commissioner’s and service provider’s perspectives, respectively. From the commissioner’s perspective, the estimated ICERs are mostly located in the north-east quadrant, implying that incentive is more effective with regards to improving OHQoL, but more expensive than traditional care practice. For the service provider, the ICERs are mostly spread in the south-east quadrant, suggesting, again, that incentive is more effective than traditional care practice in terms of improving OHQoL.

FIGURE 21.

Cost-effectiveness plane for incentive care practice compared with traditional care practice (outcome measure: OHIP-14 score, commissioner’s perspective).

FIGURE 22.

Cost-effectiveness plane for incentive care practice compared with traditional care practice (outcome measure: OHIP-14 score, service provider’s perspective).

To account for uncertainty around mean incremental costs and effectiveness, we conducted sensitivity analyses and non-parametric bootstrapping (Tables 44 and 45). From the commissioner’s perspective, the univariate sensitivity analysis varied the TCV per year by 0% or 3%. We also looked at the case that the number of patients treated remained the same per year or increased/decreased by 10%. From the service provider’s perspective, we added and subtracted 20% of the total costs and assessed the subsequent impact on the ICERs. The ICER estimates from the bootstrapping were similar to those of the deterministic base-case scenario from both perspectives.

Figures 23 and 24, the CEACs, show the probability that the incentive care practice will be cost-effective compared with traditional care practice given a range of threshold values between £0 and £60,000 per QALY. The curve is based on 10,000 replications. Using the threshold of £20,000 that is recommended by NICE, from the commissioner’s (service provider’s) perspective the probability that incentive care practice is effective is 0.37 (0.47), compared with a probability of 0.63 (0.53) for the traditional care practice. At lower levels of willingness-to-pay thresholds (< £18,000) there is a lower (higher) probability of the incentive care practice being cost-effective from the commissioner’s (service provider’s) perspective. It is worth reiterating at this point that, although we have presented the sensitivity analysis and bootstrapping results, the results should be interpreted with caution given the negligible difference between the QALYs in incentive and traditional practices.

FIGURE 23.

Cost-effectiveness acceptability curve based on bootstrap replications for incentive compared with traditional practice care (outcome measure: QALY, commissioner’s perspective).

FIGURE 24.

Cost-effectiveness acceptability curve based on bootstrap replications for incentive compared with traditional practice care (outcome measure: QALY, service provider’s perspective).

Characteristics of patients

As described in the earlier Methods section, for this analysis we used only baseline data. We included all participants who had fully completed the EQ-5D-3L and OHIP-14 questionnaires. Details of our sample in this part of the analysis are reported in Table 46. Females accounted for slightly more than half of the sample (50.8%) and the mean age was 40.4 years. The mean EQ-5D-3L was 0.852 with a 95% CI of 0.829 to 0.875. Many patients, more than 30%, reported no symptoms in all OHIP-14 items. Only a few reported that the symptoms occurred ‘occasionally/fairly often/very often’, with responses ranging from 0.77% for ‘trouble in pronouncing words’ to 14.09% for ‘felt self-conscious’. Symptoms were ‘hardly ever’ for a bigger percentage of the sample, ranging from 18.34% for ‘unable to function’ up to 54.74% for ‘uncomfortable eating’. The mean values for the OHIP-14 items ranged from 0.203 for ‘unable to function’ to 0.710 for ‘uncomfortable eating’.

Cost and cost-effectiveness

The cost-effectiveness analyses were undertaken from two perspectives: the commissioner and the service (dental) provider. The analyses from each perspective uses two distinct outcomes: first, use of the OHIP-14 scores and, second, QALYs derived from the EQ-5D-3L. The analyses revealed negligible between-group differences in QALY gains, and no statistically significant difference in the EQ-5D-3L score was found between groups or over time. Although the EQ-5D-3L is the NICE-recommended outcome of choice for economic evaluation,129 the apparent insensitivity of the EQ-5D-3L in oral health led to the use of the OHIP-14. 127 The EQ-5D-3L has been reported to have adequate construct and convergent validity, but may not be as sensitive as specific measures of OHQoL. 142–144

Although we found no statistically significant differences in the OHIP-14 scores between groups, there were statistically significant differences (at the 5% level) in these scores over time in both groups, suggesting an overall improvement in OHQoL for both the incentive and the traditional groups. The magnitude of change in scores did not achieve the level that corresponds to a minimal important difference (MID). The MID is defined as ‘the smallest difference in scores in the domain of interest which patients perceive as beneficial and which would mandate, in the absence of troublesome side effects and excessive cost, a change in the patient’s management’. 145 Locker and colleagues118 suggest a 5-point difference in OHIP-14 score could be considered the MID. In this study, the mean improvement in OHQoL over the 24 months was 3.39 in incentive practices and 1.74 in traditional practices. However, it is important to note that the OHIP-14 score is a secondary outcome; the study was powered on changes in BoP and thus changes in OHIP-14 scores should be treated with caution.

Given the negligible between-group differences, in the EQ-5D-3L and the derived QALY at least, the cost of the two services was driving the cost-effectiveness ratios. The cost of the incentive practices from the commissioner’s perspective is based on the contractual terms, which specify a UDA value and payment for quality indicators (systems, processes, infrastructure) and OHImp. In these analyses the UDA values assigned to the incentive practices included the payment for quality indicators. Traditional practices are based purely on UDAs – no additional payment is made. The value of the UDA is based in part on the practice location, for example whether or not it is in an area of need, and, for the traditional practices, historical data relating to previous work carried out. (In the cases of the incentive practices, these were brand-new practices when the contract terms were agreed and thus there were no historical data to take account of.) The incentive practices were located in areas where there was no, or limited, provision of NHS dentistry with the aim of increasing access to NHS dentistry, and this is likely to have been influential in negotiation of the UDA payment value, which was higher in the incentive practices than in the traditional practices.

In fact, over the study period, for the 210 participants included in the economic analysis, the incentive practices showed, on average, a higher number of UDAs claimed per patient than the traditional practices (11.23 vs. 10.74); this was accompanied by an overall higher number of appointments per patient (8.89 vs. 6.63). The differences in the number of appointments might in part be explained by skill mix within the two different types of practice. The blended/incentive-driven contracts were designed to facilitate provision of dental care by the most appropriate team member to encourage skill mix. It was intended that all the incentive-driven practices would fully utilise skill mix including, for example, dental therapists, hygienists and extended-duty dental nurses. There are indications in our data that more use was made of skill mix within the incentive practices with fewer appointments with the dentists and more appointments with the dental therapist in incentive practices. The dentists within the traditional practices had a marginally higher number of appointments per patient and the mean duration of those appointments was over 40% higher (16.91 minutes in incentive vs. 24.44 minutes in traditional). Within the incentive practices, on average, patients saw the therapist 2.28 times over the 24 months; in the traditional practices, on average, patients saw the hygienist 1.5 times. The difference in use of different practitioners will inevitably be because of staff configuration and thus the availability of, for example, dental therapists and hygienists. For example, one traditional practice did not employ any dental therapists or dental hygienists. However, the incentive practices were actively encouraged to increase skill mix as part of the bidding process for the dental contracts, which might go some way to explain the more prevalent use of dental therapists and the fewer appointments with dentists in the incentive practices.

Mapping from the Oral Health Impact Profile-14 to the European Quality-of-Life-5 Dimensions questionnaire

There are two areas worthy of particular discussion with regards to the mapping exercise. The first is the choice of the model for mapping from the OHIP-14 to the EQ-5D-3L. The second is the strengths and weaknesses of this approach.

We presented results from OLS, Tobit and TPMs in order to account for the prevalence of full health-state values in the sample – the ceiling effect. Estimation results did not seem to differ, at least in terms of significance, but when looking at the plots of the fitted versus the observed health-state values, the TPM seems to be the best way to proceed; the regression line is closer to the observed outcomes especially when the OHIP-14 is continuous (i.e. each subgroup of the 14 questions is a dummy variable). When the OHIP-14 is included as a categorical variable in the estimation, the results between the different techniques are more obscure.

The availability and inclusion of more personal characteristics such as ethnicity or marital status in the estimation analysis can potentially improve the estimation results. For example, it can be argued that black minorities face worse health than their white counterparts146 or that having a partner can improve an individual’s health, especially among the older population, as one partner can act as a carer for the other. In this study, marital status was not available and the response rate to the ethnicity question was low, making it impossible to test these arguments.

The mapping exercise shows that it is possible to perform economic evaluations when a general health-state outcome is not available using mapping but the approach has severe limitations, especially for those patients reporting a low EQ-5D-3L score. This is because, consistent with other mapping studies,138,147 we fail to predict correctly low levels of EQ-5D-3L (i.e. the mean of the predicted EQ-5D-3L is by far greater than the mean of the observed EQ-5D-3L). This, however, may be related to the different time frame set for the two measures; the OHIP-14 asks about problems with oral health in the ‘last 6 months’, whereas the EQ-5D-3L asks about health ‘today’. It may also be related to how well an oral health measure can predict a more general heath outcome such as the EQ-5D-3L. It would be interesting to explore how much information from the OHIP-14 is ‘transferred’ to the EQ-5D-3L, but this remains for future research.

Principal findings

Using a non-randomised study design we compared the effectiveness of treatment in the three newly commissioned blended/incentive-driven (incentive) dental practices with three matched existing (traditional) practices working under the nGDS contract. Gingivitis was selected as the primary outcome because it affects over 90% of the population, is readily measured in a clinical examination, summarises the participant’s personal oral hygiene behaviour over the preceding days, is responsive to interventions (such as oral hygiene advice or using a new toothbrush) within 2 weeks, is a proxy for other self-care behaviours (such as the use of fluoride toothpaste) and is an interim outcome for periodontitis (which is a significant public health problem)107–109 and because changes are readily demonstrated over a period of 2 years. Secondary outcomes included assessment of caries, quality of life and cost-effectiveness.

The results of the quantitative analysis were mixed. We found that the results of the assessment of BoP favoured the blended/incentive-driven model of service delivery, although the results should be treated with caution given the high attrition and issues of data quality.

The health economics analyses showed the blended/incentive-driven contract was more costly for the commissioner. In respect of our secondary measure of quality of life, overall OHQoL, assessed using the OHIP-14, was higher for participants in the blended/incentive-driven practices than it was in practices under the traditional UDA-based contract, but it did not represent a minimal important clinical difference. It is also important to note that the OHIP-14 is a secondary outcome; the study was powered on changes in BoP and thus changes in OHIP-14 scores should be treated with caution. The differences within and between groups for the EQ-5D-3L were negligible. The last result was not entirely unexpected. Although the EQ-5D-3L is the NICE-recommended outcome of choice for economic evaluation,130 the apparent insensitivity of the EQ-5D-3L in oral health led to use of the OHIP-14. 129 The EQ-5D-3L has been reported to have adequate construct and convergent validity, but may not be as sensitive as specific measures of OHQoL. 143,145

Use of the ICDAS was exploratory in the study. It has proved to be illuminating in as much as there are lessons to take forward. Although the majority of charts were completed to an acceptable standard, a number of issues were evident which give cause for caution in interpretation of the results. For example, in some instances teeth with obvious caries or fillings at baseline were charted as sound at follow-up, even allowing for errors in the transposition of adjacent teeth. Lack of confidence in the quality of the ICDAS data means that we have not explored the enamel transitions in the way that we had originally planned to.

As highlighted earlier, in addition to problems with data quality for the ICDAS, data quality for BoP was also an issue, with a substantial number of participants excluded from the analysis following quality assurance. Dentition charting has been highlighted as a problem with the national dental contract pilots. 2 Early on it was found that data on charting were incomplete or absent for many patients. 30 These findings are interesting in as much as in our study charting was completed by the dental practitioners using a paper form. For the ICDAS, in particular, we had intended to use electronic charting, which incorporates built-in automatic error checks. However, in the pilots there was a requirement for practices to keep patient records in electronic form and they reported that ‘the evidence to date, however, suggests there have been inconsistencies in the quality of electronic data recording from practice to practice and within practices, from clinician to clinician’. 20 It would seem that charting and dentist recording of data is a challenge whether paper based or electronic and despite the training given in the INCENTIVE study, many of the issues the pilot sites experienced appear to be replicated in our study, including misinterpretation and lack of understanding of what is required.

In relation to access to dental services, within our qualitative study we identified perceptions that the blended/incentive-driven contract increased access to dental care, with the contract determining dentists’ and patients’ perceptions of need, their behaviours, evaluated and subjective health outcomes and patient satisfaction. These outcomes were then seen to feed back to shape people’s predispositions to visit the dentist. However, an important, and unexpected, finding of the study was the high ‘fail to return’ rate. A large proportion of people in the study who had access to a dentist did not follow up on oral care. These individuals were more likely to be younger males and have poorer oral health. Within this patient group, although access to dental services increased, this did not appear to facilitate continued use of services. Although the reasons for non-attendance recorded by the dental practices was incomplete, of the reasons recorded the most frequently cited was no response to contact from the dental practices and, in the blended/incentive-driven practices, a substantial number also failed to attend pre-booked appointments. Interestingly, those lost to follow-up had multiple appointments and treatments, including fillings, crowns and bridges. One possible explanation for this may be that the problem they had visited the dental surgery for had been rectified and, in the absence of any current dental problems, they did not wish to visit the dentist for a ‘check-up’. Indeed, one patient talks about attitudes in the past which meant she only went to the dentist if there was a problem, although she goes on to say that she thinks attitudes have changed now. A dental therapist talks about high levels of need and there being ‘a lot of neglected mouths. Some haven’t seen a dentist in years. Some have lost their motivation because of this’. Another patient recalls not going to the dentist for about 3 years, saying ‘I don’t know why I didn’t go, I just stopped going and I think then you get thrown off the register’. This aligns with previous figures that suggest around one-third of the population in the UK go to the dentist only when they have a problem. 2

The failure of some individuals to attend pre-booked appointments may, in part, be attributed to dental anxiety, a recognised barrier to dental attendance. 11 One patient spoke about cancelling an appointment for a filling because she was ‘panicking and worrying’. Another talked about her partner being ‘scared to death of the dentist’ and how ‘trying to get him through the door of the dentist was a real effort’.

Another possible explanation for the unexpected high loss to follow-up within our study, and indeed the difference between estimated and actual loss to follow-up, is that all the patients in the study were new patients. This suggests that we had inadvertently selected a sample that were not regular dental attenders, although this may still be representative as one-third of the population falls into this classification. 2 The study sample size (including allowance for attrition) was powered on a previous study. 121 In the study by Clarkson and colleagues,121 the authors outline their inclusion criteria; eligible patients were dentate adults who had already made an appointment for a routine check-up and had in whom probing of gingiva was not contraindicated at the time of the appointment. This indicates that their sample included primarily regular dental practice attenders rather than new patients who may not have attended a dental appointment for some time. It also indicates, given that contrandication to probing of gingiva was an exclusion criterion, that their sample had better oral health than those included in our study.

Furthermore, groups of new patients might be likely to include a larger proportion of patients who are migrants. Most research on the dental attendance of migrants has been conducted in the USA or Australia. This work consistently describes the barriers to access among migrants. 149,150 Brennan and Spencer149 noted that service utilisation patterns were related to cultural factors within populations and it may be that groups who have little history of dental attendance do not adopt this behaviour immediately.

The low rate of reattendance for regular dental care (thereby implying a pattern of emergency care) provides support for the care pathway approach recommended in the Steele report,2 which legitimises irregular dental attendance for those who choose it. High proportions of patients adopting this pattern will shape the practices in these areas. Such practices will be characterised by the provision of more emergency care and more of the time-consuming initial assessments. This burden might be considered in the commissioning of services in these areas.

Skill mix was of particular interest in our study given one of the aims of the blended/incentive-driven contract is to improve health outcomes through wider skill mix utilisation. In England there has been a steady shift in the make-up of the dental practice team with an expansion of numbers of, for example, dental therapists and dental hygienists. 151 Our analyses indicate that skill mix was utilised in the blended/incentive-driven practices to a greater extent than the practices under the traditional nGDS contract. While on average incentive patients saw dental practitioners more often over the study period, in the incentive practices the number and duration of appointments with dentists was lower, as use was made of dental therapists. The difference in use of different practitioners will inevitably be influenced by current staff configuration and thus the availability of, for example, dental therapists and hygienists. However, the incentive practices were actively encouraged to increase skill mix as part of the bidding process for the dental contracts, which may go some way to explain the more prevalent use of dental therapists and the lower number of appointments with dentists in the incentive practices. Collectively, the patients in incentive practices had, on average, more appointments than patients in the traditional practices. In this regard the commissioning strategy was seen to be effective.

There are obstacles to overcome to realise any benefits of the greater deployment of skill mix. Dentists may need support in these areas and to recognise the differences between caring for individual patients and caring for segments of the population, such as that formed by the patient base of a practice. In addition, one incentive practice experienced a higher turnover of its dentists than the other practices, which was attributed in part to lack of opportunities to carry out complex treatments. This echoes evidence from elsewhere of tension between ‘acknowledging that less qualified practitioners can contribute directly to dental treatment and the unwelcome consequences of a modularised approach’. 152

Intuitively, the delegation of treatment to staff specialised in only a specific range of treatments could reduce costs and increase access to care. 132 The acceptability of the blended/incentive-driven contract to dental providers is clearly important, not least because of the financial implications. Primary care dentist practices, unlike many other medical providers in the UK, are operated as businesses,84 with NHS provision of dental services governed using quasi-market principles. 90 It has been suggested that under the current nGDS contract remuneration structure there are financial barriers that ‘prevent the profitability and effective use of skill mix’. 84 Our analyses included assessment of costs from the perspective of the dental provider. This included the time spent in appointments with patients and the materials and laboratory costs. The analysis showed that, despite the increased number of appointments per patient, overall, the new contract was estimated to provide greater financial returns than those observed in the traditional practices. However, for the incentive practices in this study, although the analysis included overheads (included in the practitioner time element of the analysis), the costs associated with setting up the new dental practices were not included and are likely to be substantial.

The unit of analysis for the costs in this study was the practice; the financial implications for the individual GDPs in the practices are unclear given that one of the most important barriers is referral by the dentist to, for example, a dental therapist, preventing further earning potential for additional band 1 treatments. 84 However, in the qualitative study dental therapists were perceived to increase the availability of care and patient satisfaction. Practices increased the utilisation of dental therapists by not reducing the payments to individual dentists who referred their patients to them. Although seemingly paying two staff members for the same treatments, this approach incentivised referral, thus liberating dentists’ time for patient OHAs and more remunerative complex treatments. This might be further explored and evaluated in the evaluations of the forthcoming contract prototypes. 75

The data hint at appreciable challenges related to a general refocusing of care and especially to perceptions about preventative dentistry and use of the risk assessments and care pathways. The quantitative data from the RAG assessment showed > 30% of the participants in our sample to be in the highest risk category (red) and < 7% in the lowest category (green) at baseline. There was an improvement from baseline to follow-up which suggests a degree of responsivity. In addition, there were very few over-rides of the RAG system (4.64% at baseline and 7.21% at follow-up). Although it would be inappropriate to make direct comparisons with findings from the dental contract pilots with regard to changes over time, the proportion of patients in each category at the initial OHA (6% in green and 26% in red for the national pilots and 7% and 30% in the INCENTIVE study) indicate a similar distribution of patients. 30 One of the concerns a priori in our study was that for individuals with particular irreversible medical conditions it was not possible to move out of the amber category. Unfortunately, because of the small number of pathways that had been over-ridden, we are unable to ascertain whether or not this a priori concern was borne out. Explanations for over-rides, especially moves away from red, suggest other reasons. Only one patient score was over-ridden because of a family medical condition (diabetes). It is of note that those who were red at baseline were less likely to attend follow-up.

The OHA RAG system is key to the current national dental contract pilots and integral to the incentive contract. Within our study, practitioners were concerned that it was not sufficiently precise and, as illustrated above, rarely overrode the assessments, even though they could. In practice applications, dentists’ assessments may be prone to allocation bias as treatments are apportioned according to the ratings, and to measurement bias changes in status which are used for contract monitoring. Further research is required to validate the RAG assessment, but in its primary guise for risk assessment, and also in this application as a tool for evaluation.

Identification and/or prioritisation of the research questions and development of the research design

We worked closely with two PPI contributors, each bringing a unique perspective as current users of dental services but covering different age and societal groups – middle-aged participants and parents (Susan) and working young adults/students (Alex) – that we considered may have differing expectations and use of dental services that would be of relevance to INCENTIVE. We also sought specific advice from a third PPI advisor (Rosie, 84 years of age) on an ad hoc basis; she did not attend the advisory group meetings.

The qualitative enquiry was informed from discussions with the PPI contributors and the sampling matrix was codesigned to include criteria linked to the objectives of the programme including demographic factors (age, sex, ethnicity, socioeconomic status). Access to dental services was seen as a priority by the PPI contributors and they were central in ensuring that we also collected data from non-patients to understand barriers to access to NHS dentistry. Rosie helped shape the inclusion of elderly patient groups, lobbying for a broad entry criteria and to ensure that patients > 65 years were included in the cohort. Susan and Alex welcomed the idea that we were addressing the utilisation of the wider skill mix in our analysis of the contracts.

Our PPI contributors regularly attended the project advisory group meetings and had additional input on a needs basis; for example, they made a significant contribution to the preparation and internal development and review of the ethics application to ensure it was understandable yet not patronising, to ensure informed consent was readily achievable and the study not too onerous for participants by carefully considering the volume of paperwork participants would need to complete.

Dissemination

The PPI contributors codeveloped the participant information leaflets and are currently assisting in the preparation of materials for the dissemination event and ensuring that the study material will improve participant understanding of the importance of INCENTIVE in redesigning the dental contracts as we move forward. Our PPI contributors have ensured our dissemination strategies are inclusive and accessible to research participants, patients and the public.

Patient and public involvement identification and training

Two INCENTIVE PPI advisors sit on the study advisory group, and all PPI advisors were willing to be contacted by the researchers to contribute to individual work packages as required.

Patient and public involvement contributor training and mentorship

The INCENTIVE PPI contributors were offered training and mentorship to increase their confidence in working in partnership. Similarly, the INCENTIVE research team were also trained in how to work with PPI contributors as equal partners. We provided mentorship and personal support to PPI contributors before and after meetings, addressing queries, language and documentation. The PPI contributors were also provided with a ‘jargon buster’. The researchers received training to facilitate inclusive project meetings with PPI representatives by knowledge transfer from the ‘getting involved in shaping research and building partnerships’ workshop.

Throughout we adhered to INVOLVE guidelines and reimbursed PPI members according to INVOLVE recommendations. 154

Access and inequalities

Access can be assessed using list size as a proxy (as in the national dental contract pilot evaluation). 20 The incentive practices were new practices and, therefore, still establishing their patient base, thus we were unable to assess increased access in this way. However, the siting of the incentive practices in areas of high need associated either with deprivation and disease or with the poor availability of NHS care meant that access to dental services increased within these communities. But many people who had access to a dentist in these practices, and many of the new patients included in the study in traditional activity-based practices, did not continue with their oral care. These individuals were more likely to be young men and have poorer oral health.

The qualitative data suggested that this was in part a consequence of taking on many new patients who might not wish to receive continued care. Some patients may also have come from minority ethnic groups not used to attending the dentist regularly. Thus, although the new practices increased access, further work is required to understand how best to promote and encourage appropriate dental service attendance especially among those with high level of need to avoid increasing health inequalities.

Oral health assessment and skill mix

As described earlier in the chapter (see Principal findings), use of the RAG risk assessment in our study showed improvement from baseline to follow-up which suggests a degree of responsivity. Although the RAG ratings were used in risk assessment, participants were concerned that it was not sufficiently precise and rarely over-rode the assessments, even though they could. In practice, dentists’ assessments may be prone to allocation bias as treatments are apportioned according to the ratings, and to measurement bias where changes in status are used for contract monitoring. The risk assessments were less frequently used as an aid to patient communication. Within the national pilots sites, communication of the care pathway from practitioner to patient was also mixed. 20 Further research is required to validate the RAG assessment, in the following guises: risk assessment, communication aid, contract-monitoring tool and also as a tool for evaluation.

The evaluation of the national pilots also reports that the pathway model takes longer to operate per patient than the previous model. Within the INCENTIVE study, although patients had a greater number of appointments in incentive practices, the duration of the appointment was on average shorter than the practices operating under the UDA-based contract. The incentive practices were set up in areas of high need and poor access to dental services and therefore it is not surprising that they had more appointments and a higher level of treatment. Although it is not possible to draw any firm conclusions from the figures, the qualitative data hint at appreciable challenges related to a general refocusing of care and especially to perceptions about preventative dentistry and use of the risk assessments and care pathways. There are also obstacles to overcome to realise any benefits of the greater deployment of skill mix. One potential solution would be for contractual drivers to change to be supplemented with educational efforts to support the refocusing of care, perceptions about preventative dentistry, internal practice business models on the use of skill mix, the role of evidence-based dentistry, working with care pathways, communication skills and the need for a greater understanding of the difference between caring for individual patients and a population.

Clinical effectiveness and clinical outcomes

The findings with respect to the clinical effectiveness of the blended/incentivised contracts are mixed. Although the results of the primary outcome of gingivitis (BoP) favour the blended/incentivised model, and there is some reassurance that the effect size is similar to that included in the original power calculation, the results should be treated with caution given the reduced sample size and data quality issues (discussed in more detail in Data quality), and reservations about the validity of pooling data.

Few studies have assessed the clinical effectiveness of contracting in dentistry; for example, the national pilots’ evaluation looks at clinical response using the RAG status between assessment and review. 20 Should services consider an oral health-related outcome measure of clinical effectiveness, despite the challenges, this study would support the use of BoP. BoP is readily measured and responsive to change. It can be changed by dental teams’ behaviours including preventative advice and treatment but is also subject to changes in patients’ behaviours and to measurement error. Random changes in patient behaviours and measurement error would attenuate the apparent relationships between the contract types and BoP identified in this study. The estimates of the treatment effect may therefore be underestimated. It is possible that measurement bias may have favoured one or other contract type, but this seems unlikely. Few other studies have assessed clinical outcomes in relation to dental contracts. Nevertheless, the utility of BoP as a surrogate for other oral health outcomes warrants further investigation.

Data quality

The data quality issues with BoP were also seen within the ICDAS. In respect of caries assessment, which favoured the traditional practices, although the majority of ICDAS charts were completed to an acceptable standard, a number of issues were evident that give cause for caution in interpretation of the results.

Use of the ICDAS was exploratory and changes in the ICDAS scores may be the result of multiple factors, such as attrition of study participants, diagnostic inaccuracies, poor completion, complexity of the ICDAS, difficulty in diagnosing enamel caries in general dental practice, errors in syntax or the relatively short time frame of the study. Further work is indicated on both the utility and the validity of recording dental caries and treatment experience with an indicator such as the ICDAS as a contract outcome. The precision of this system (the number of categories and levels) mean that it is time-consuming to use and dental caries and treatment experience may not change sufficiently quickly for contract monitoring. Standardisation of the dental practice software could facilitate the use of this and other clinical outcomes.

Data quality and dentist data recording have recently been flagged as challenges within the dental contract pilots, with the recommendation that there should be a strong driver in the contract for data to be collected accurately and for appropriate training and support provided for practices. 20 Evidence from the INCENTIVE study supports this.

Cost-effectiveness

The economic analysis showed the INCENTIVE arm of the study to attract a higher cost for the dental commissioner but to be financially attractive for the dental provider at the practice level. Caveats also apply to these findings given the high attrition rate and further research that is required to assess the financial impact of the contract, in particular the impact of an increase of skill mix on the individual practitioner, in order to support the model.

Finally, within dentistry, the OHIP-14 measure has been used extensively in cost-effectiveness analyses in oral health but cannot be used to produce QALYs. Within this study the OHIP-14 scores were mapped using regression techniques to the baseline EQ-5D-3L scores to explore the potential for estimating utility scores that can be used to produce QALYs based on responses to the OHIP-14. Unfortunately, consistent with other mapping studies, we fail to predict correctly low levels of EQ-5D-3L. This suggests that there is a need for the development of a condition-specific utility measure for use in oral health for economic evaluation.

Recommendations for future research

Further research is required to:

• understand how best to promote and encourage appropriate dental service attendance, especially among those with a high level of need to avoid increasing health inequalities

• assess the financial impact of the contract and particularly the increase of skill mix on the individual practitioner in order to support the model

• validate the RAG assessment as a risk assessment, communication aid and contract-monitoring tool and as a tool for evaluation

• explore further the utility of BoP as a surrogate for other oral health outcomes

• explore both the utility and the validity of recording dental caries and treatment experience with an indicator such as the ICDAS as a contract outcome

• develop a condition-specific utility measure for use in oral health for economic evaluation.

Implications for practice

• The blended/incentive-driven contracts were perceived to increase access to dental care, with the contract determining dentists’ and patients’ perceptions of need, their behaviours, evaluated and subjective health outcomes and patient satisfaction.

• A large proportion of people in the study who had access to a dentist did not follow up on oral care. This supports the care pathway approach recommended in the Steele report,2 which legitimises irregular dental attendance for those who choose it.

• For dental practitioners, there are challenges within the blended/incentive-driven contracts related to a general refocusing of care around preventative dentistry, risk assessment and a care pathway approach rather than the focus on treatment inherent in the UDA-based contract.

• There are obstacles to overcome to realise the benefits of the greater deployment of skill mix. Intuitively, the delegation of treatment to staff specialising in only a specific range of treatments could reduce costs and increase access to care but, as previous research indicates, there may be financial barriers that prevent the profitability and effective use of skill mix.

• Data quality and dentist data recording, particularly dentition charting, was challenging. This supports the view that there should be a strong driver in the contract for it to be collected accurately and appropriate training and support provided for practices.