Scale, scope and impact of skill mix change in primary care in England: a mixed-methods study

Research question 1 (work package 1): what is the scale and distribution of skill mix changes in primary care and how is skill mix change associated with outcomes and costs?

• How has the workforce changed and where has any change occurred?

• How are compositional changes to the workforce associated with later changes in a range of outcomes, including patient and practitioner satisfaction?

• How are workforce changes associated with later changes in costs and practice efficiency?

Research question 2 (work packages 2 and 3): what motivations drive skill mix deployment at the practice level and what is delivered by the deployment of different practitioner types?

• What motivates practices to choose/not choose increased skill mix deployment? (WP2.)

• Which aspects of health care are undertaken by different practitioner types? (WP3.)

Research question 3 (work package 3): how do skill mix changes affect the experiences of employers, practitioners and patients?

• How are new ways of working being negotiated in general practices where skill mix changes have occurred?

• How is the implementation of change in skill mix associated with the achievement of organisational objectives at practice level?

• How does increased skill mix affect patients’ experiences when accessing primary care services? (WP1 and WP3.)

Work package 4 drew together data from all WPs to develop a comprehensive understanding of the implementation of skill mix changes in general practices and the consequences of these changes.

NHS Digital data (work package 1)

We used data about the range of practitioners employed in practices across England, which were available from NHS Digital. Using panel data regression techniques, we estimated practice-level associations between changes in patterns of skill mix and changes in costs and outcomes measured using national data sets.

Growth in the number and range of different workforce professionals employed by practices in recent years has facilitated new ways to deliver services for patients and offered the potential to lower costs and improve outcomes. 89 As practices are primarily small, employee-owned enterprises, they are likely able to respond to these changes more rapidly than larger organisations, such as hospitals, to contain costs and prioritise specific outcomes.

Details of data sources and a full account of our analysis are described in Chapters 3 and 6. In broad terms, we created and analysed a longitudinal practice-level workforce data set using the practice-level wMDS available from NHS Digital for 2015–19. The resulting workforce composition data were analysed against selected outcome data sets to look for associations between workforce composition and outcomes, including overall costs.

In addition to examining workforce changes by geographical region and over time, we looked for indications of potential substitution between the staff groups. Modelling was used to generate scenarios about potential variation in outcomes produced by changes in the number of practitioners employed.

Practice manager survey (work package 2)

We conducted an online survey with practice managers in general practices in England between August and December 2019. We targeted practice managers because it is typically their responsibility to report workforce data to NHS England, and they are generally involved in implementing decisions about staff employment.

Our initial plan was to conduct a postal survey to be sent to all general practices in England using publicly available practice addresses, but negotiated an agreement to conduct an online survey for the following reasons.

First, by using an online platform, we were able to use data that practices had previously supplied to NHS Digital to prepopulate the survey with practice-specific workforce data for managers to confirm or correct according to their current workforce. This allowed us to adapt survey questions so that they were appropriate for each general practice and to avoid duplication or unnecessary questions. For example, we asked participants to indicate the reasons for employing the types of practitioners in their current workforce.

Second, it became evident that local clinical research networks (LCRNs) were interested in supporting us in distributing the survey. The online version was seen as the most effective and appropriate way to reach practice managers.

Third, the use of an online survey was a more cost-effective way of gathering and extracting data for analysis.

We developed and pilot-tested an online questionnaire to be completed by practice managers. In addition to facilitating a check for accuracy of the NHS Digital wMDS, managers were asked about motivations for hiring staff from selected practitioner groups, their future workforce plans and their ideal workforce.

Survey administration

We worked with 15 LCRNs to e-mail the survey link to practice managers in England. Practices that had not opted out of receiving research invitations from their LCRNs were invited to participate in the study, receiving three e-mail invitations/reminders to participate in the study. We requested assistance from the Practice Management Network and used social media to publicise the survey to reach additional practices.

Practice managers were asked to enter a practice identification (ID) (Figure 2), which prevented duplication and allowed the survey software to automatically assign numerical IDs based on the order in which the respondent had started to complete the questionnaire. We did not store the response data with the practice IDs and the key was stored separately. All data were stored on our secured servers. The practice’s identity was needed to ensure that we presented the practice with appropriate questions related to their current staff and provided information for LCRN records about responding practices.

FIGURE 2.

Practice manager survey front page.

The online questionnaire was created using Lighthouse Studio (version 9.7.2; Sawtooth Software Inc., Provo, UT, USA) and was hosted on University of Manchester (Manchester, UK) servers. Data analysis was conducted using Stata^® 15.1 (StataCorp LP, College Station, TX, USA).

Survey content

The study focused on six practitioners (ANPs, specialist nurses, HCAs, PAs, paramedics and CPs) because they represented a mix of:

• staff whose employment in practices largely predates the new GP contract (e.g. HCAs and advanced nurses)

• staff whose employment has often been linked to financial incentives (e.g. CPs)

• staff whose employment can be used to offer additional services (e.g. specialist nurses and CPs)

• additional roles that can also be funded through PCNs (e.g. PAs and paramedics).

The questionnaire consisted of three sets of questions. The first set of questions focused on verifying publicly available practice workforce data. Practice managers were shown the workforce data [both headcount and full-time equivalent (FTE)] that NHS Digital released for their practice, as of 31 March 2019. This question is shown in Figure 3.

FIGURE 3.

Practice manager questionnaire: is the data correct?

Practice managers were then asked to enter the correct current headcount or FTE figures for any incorrect practitioner groups (Figure 4).

FIGURE 4.

Practice manager questionnaire: entering correct data.

The second set of questions asked those who employed staff in one of the six roles of interest what factors influenced their decision to employ staff in that role. Respondents were presented with a list of 12 predefined factors, the selection of which was informed by our knowledge of the literature and previous workforce research. Respondents were also permitted to add factors not captured in the list using free text. They were asked to select all factors that applied to their decision for each type of worker. They were then asked if any additional funding was explicitly supplied to support the employment of these staff, from which funding organisation and whether or not they were still receiving funding.

Regardless of their current workforce composition, the final set of questions asked all participants whether or not they would wish to employ additional staff from a list of roles in the future and, if so, whether they would prefer them to be directly employed by the practice or through a PCN. Finally, participants were asked to indicate their ideal workforce composition by selecting the percentage of their total clinical workforce made up of each of the listed roles. GPs were included in the list of roles for this question. Respondents were required to use slider bars to indicate the percentages for each worker group (Figure 5). These bars were programmed so that total percentages were automatically adjusted to 100%.

FIGURE 5.

Practice manager questionnaire: ideal workforce slider bars.

Case study (work package 3)

Triangulation (work package 4)

Our methodological approaches have been combined using the principle of ‘triangulation’. 95,96 In this context, the term triangulation describes a process that employs mixed methods to gain a complete picture of a problem. 97 Qualitative and quantitative data are not only used to cross-corroborate findings (i.e. to confirm descriptions of how particular roles are implemented); qualitative findings are also used to complement the quantitative analysis. For example, qualitative elements of the research are used to inform the measurements or analyses undertaken in the quantitative research, providing context to assist interpretation while at the same time inductively generating insights into the way in which phenomena are related in practice. 96 In this way, qualitative and qualitative research work together to facilitate a richer understanding of the phenomenon, generating a multifaceted portrayal of a complex social situation with breadth and depth. 98

Although the WPs (i.e. WP1, WP2 and WP3) were analysed separately, ongoing engagement between researchers on each WP was maintained during ad hoc conversations and monthly team meetings throughout the project to ensure that findings were integrated as specified in WP4. For example, insights from study sites and clinical members of the research team helped shape analysis and interpretation of how the employment of CPs may influence prescribing patterns.

As more structured support for this engagement, one researcher (IM) produced thematic codes based on a review of literature and the research questions. Themes were used to help structure an iterative dialogue between members of the research team, with regular meetings held to share findings as they emerged in each of the WPs. The thematic codes were grouped into meta-themes that cut across findings from different methods and integrated using a ‘convergence coding matrix’. 99 Meta-themes were considered for agreement and disagreement based on convergence (i.e. where findings directly agree), complementarity (i.e. where findings offer complementary information on the same issue), dissonance (i.e. where findings appear to contradict one another) and silence (i.e. where themes arise from one component study but not others). 100 Figure 6 depicts the triangulation process.

FIGURE 6.

Triangulation process.

Workforce data

As a result of changes made by NHS Digital to their collection, analysis and publication of the wMDS, our analysis was delayed while NHS Digital revised statistics from the earlier period of analysis.

GP Patient Survey

Changes in access requirements prevented access to the most detailed level of the most recent GP Patient Survey. We were, however, able to analyse data suitable for addressing our original research questions.

Hospital Episode Statistics

Delays in the processes necessary to obtain access to Hospital Episode Statistics after 2016–17 have not been resolvable within the study period. Nonetheless, our analysis of these data makes a useful contribution to our proposed overall analysis of outcomes.

Adjustments in WP2 to change from a postal to online survey, as described above, gained momentum following initial contact with the lead Clinical Research Network. Imelda McDermott set up contacts with LCRNs. Jon Gibson managed the running of the Practice Manager Survey and responded to queries from the LCRNs.

For WP3, the team involved in this WP (i.e. IM, SS, MG, KC, AM and DH) met fortnightly during the data analysis period and shared nuanced findings across the remaining members of the research team.

Meetings of the Study Steering Committee were arranged at the intended 6-monthly intervals. Full attendance was achieved on each occasion using virtual platforms, and minutes were produced and agreed.

Expert Advisory Group meetings were similarly arranged at 6-monthly intervals, but it proved challenging to maintain attendance by individuals from different practitioner groups. However, those who attended these meetings made valuable contributions and our patient and public involvement representatives remained engaged with the project throughout.

Composition of the practice workforce

Since September 2015, for the sample practices with usable data, there has been an increase in total workforce FTE from around 47,600 to 51,600 (Table 1). Within that increase in workforce FTE, there has been a decline in the FTE of partner GPs and an increase in the FTE of advanced nurse roles and newer DPC roles, such as CPs and PAs. Table 1 also presents figures for workforce FTE PTP. These figures highlight workforce change in relation to the number of registered patients to whom the workforce are delivering health care, with implications for the capacity of the workforce to deal with changing levels of workload. For example, where an unchanging workforce has to serve an increasing number of patients, then this potentially reduces the practitioner time that can be focused on the needs of each patient. Table 1 shows that, although the number of FTEs in these workforce roles has increased by 8.44%, FTEs PTP have increased by 2.87% after accounting for the growing number of patients. These figures show the same pattern as the raw workforce figures. The number of GP partner FTEs PTP are declining, whereas salaried GPs and locum FTEs are increasing. The number of practice nurse FTEs PTP remains roughly constant, whereas the number of ANP FTEs PTP is increasing. This confirms a gradual shift in the overall workforce composition during this period. Graphical representations of this analysis can be found in Appendix 1, Figures 28–30.

There has been an expansion of skill mix in a greater proportion of practices in some regions than in others. Table 2 shows the proportion of practices in each region with at least some FTE of the workforce roles. This demonstrates that, for instance, paramedics have been hired predominantly by practices in the south of England, whereas CPs are more evenly geographically distributed but have the highest uptake by practices in the east of England, North East and Yorkshire and the Midlands. ANPs are employed by 57% of practices in the east of England, but are employed by only 23% of practices in London. Paramedics are employed by 12% of practices in the south-east of England, but are employed by only 1% of practices in the north of England. These results highlight the variation in workforce composition across England.

Figure 7 presents a map of England, highlighting practices that employ the different workforce roles. In addition to the pattern of paramedic and CP employment described above, Figure 7 shows that the practices employing PAs are located in the main population centres (i.e. London, the Midlands and the north-west of England) and, also, that there are a very small number of practices currently employing physiotherapists.

FIGURE 7.

Maps of England highlighting practices that employ different workforce roles. Map data ©2021 Google.

Modelling workforce

In addition to plotting the geographical distribution of the workforce, we estimated regression models to further examine the association between workforce and practice characteristics.

Our statistical analyses were based on a sample of 6287 general practices. Despite not being able to analyse data for all practices in England, our core sample was highly representative and allowed us to confidently extrapolate our results to the situation of English primary care more broadly. The characteristics of the typical practice in our sample are described in Table 3, which indicates that the mean number of GP FTE PTP is 0.57, whereas the mean numbers of nurse and other DPC (i.e. pharmacists, paramedics, PAs and physiotherapists) FTE PTPs are 0.27 and 0.20 respectively. The mean distance to the nearest hospital for a practice was 8.57 km and 17% of the practices in our sample were in rural areas. Moreover, 16% of practices in the sample were dispensing practices and 72% were on a GMS contract. The mean percentage of patients aged ≥ 65 years in the sample was 17.73%.

To study the association between staffing and practice-level characteristics, we estimated a seemingly unrelated regressions model. We opted for a seemingly unrelated regressions model instead of a set of independent standard linear regressions because the employment decisions for the three workforce groups are likely to be connected. For instance, employing extra GPs is likely to reduce the workforce budget for employing staff from the other groups. In this case, the three separate models will likely have a correlated error term. The seemingly unrelated regressions model accounts for this contemporaneous correlation. 103 The Breusch–Pagan test is a test for independence of residual vectors between the model equations. As shown in Table 3, the null hypothesis of independence of residuals is rejected and, therefore, our model choice is justified.

Patient list size enters these models in several ways. First, the outcome variables are workforce FTE PTP to ensure comparability across practices with different catchment populations. Second, patient list size is included as covariates as a cubic function to capture non-linearities in the effect of patient population on staffing (e.g. if the contribution to the practice drops or jumps above a certain patient number for certain roles). The parameter estimates, shown in Table 4, suggest that the number of patients is not significantly associated with workforce PTP. This relationship is presented differently in Figure 8, which shows that the level of staff PTP is stable across different patient population sizes.

FIGURE 8.

Workforce predictive margins of patient list size.

Table 4 also shows that distance to the nearest hospital and medical school (in kilometres) are both positively associated with nurse and DPC FTE PTP, whereas distance to nearest medical school (in kilometres) is negatively associated with GP workforce. Distance to nearest medical school is potentially a supply effect where newly qualified GPs are more likely to locate near to where they have trained, whereas the distance to nearest hospital is a measure of the availability of secondary care. Practices that are located further away from a hospital have more nursing and DPC staff employed.

The proportion of patients aged ≥ 65 years, the amount of payments received through the minimum practice income guarantee (MPIG), rurality, patient income deprivation, the payments for extended hours access and being a dispensing practice are positively associated with workforce groups. These factors indicate that patient demand factors and income payments are positively associated with staff employment.

The market forces factor, which measures the unavoidable cost differences between geographical locations, is negatively associated with nurse and DPC workforce, but is not significantly different from zero for GPs. This is potentially a supply factor where staff on lower salaries (e.g. DPC/nurses) are unable to locate to more expensive areas. Alternatively, there could be a greater demand for GP services in more expensive locations and less demand for appointments with DPC and nursing staff.

General practitioners

During site visits, we observed that most GPs were predominantly engaged in providing health care directly to patients. GPs were also supplying supervision and support to other practitioners who were conducting separate consultations with patients, as well as providing advice to reception staff and attending to a wide variety of clinical and non-clinical responsibilities. The increase in skill mix was perceived as changing the role of some GPs in that non-GP practitioners tended to deal with less complex cases, whereas GPs dealt with the more complex cases. This led GPs to comment that they were increasingly operating in a similar manner to hospital consultants (i.e. seeing more complex cases as if they were ‘GP consultants’ whereas non-GPs selectively took on less complex tasks that would formerly have been part of a more varied GP caseload):

But because we’ve got so many nurse practitioners it means that us GPs see all the much more complicated things and [. . . ] we are becoming more like a GP consultant.

ID44: site D, GP partner (bold text highlights the relevance of the quote)

However, not every GP was sufficiently experienced to take on the consultant role. Experienced GPs recognised that less experienced colleagues needed proper support and time for development to progress their careers within general practice:

So our new GP can’t jump into that complex multimorbidity role; and that’s a real challenge for us, so we’re having to create tiers now of GPs, which is where the GP consultant role comes in, actually I can support . . . the junior GP has the training to deal with the complexity but perhaps not the experience; now how can we work with them to support them.

ID28: site C, GP partner (bold text highlights the relevance of the quote)

In addition, although GPs were traditionally always available to provide advice for non-GPs, with the addition of newer roles, some practices in our study decided to allocate an on-call/triage GP who did not see patients face to face but who was fully available to support non-GPs (e.g. providing advice or signing prescriptions on behalf of practitioners, such as PAs, who were not able to prescribe).

Owing to multiple different responsibilities and a lack of designated time for each task, GPs indicated that they found it difficult or impossible to complete all their work and responsibilities within office hours.

Nurses (including advanced nurse practitioners)

As noted in Chapter 1, the employment of practice nurses represented an early change in general practice skill mix some decades ago and, for many practice nurses, their roles have not changed from dealing with chronic disease management and care. Subsequently, different nurse roles have developed, such as specialist nurses and ANPs.

In our sites, ANPs were often responsible for the initial assessment of patients requesting health care for a wide range of problems. ANPs often dealt with same-day care, sometimes as part of dedicated acute illness teams, with a remit for minor illnesses and ‘low-end’ mental health issues, with more complex health issues excluded (particularly at site E). Many ANPs would request tests and assess blood results as part of managing same-day caseloads or in relation to regular reviews of patients with chronic diseases. Some ANPs were responsible for reviewing hospital letters and thereby reduced pressure on GPs who might otherwise review them. In an evolving hierarchy of diverse practitioners, less qualified staff could escalate problems they were unable to deal with to ANPs, whereas ANPs themselves sometimes escalated problems to GPs. Some ANPs, as senior nurses, had an educational role in training or supervising other nurses and HCAs.

The ability to prescribe was seen as an essential component of ANPs’ scope of practice, enabling them to deal with minor illnesses independently and reduce the need to interrupt GP/patient consultations to get prescriptions signed. Of particular interest in this study was that this enabled some ANPs to see patients with undifferentiated symptoms in the same way as GPs, rather than being delegated specific tasks. Therefore, some ANPs viewed themselves as ‘generalists’ and engaged in a patient-facing work very similar to that of GPs. In recognition of this, the ANP in site C was deployed on a GP rota and paid at the same rate as GPs. In another site, ANPs were regarded as ‘80% of a doctor’ (ID43: site D, managing partner), with the distinction that GPs also did managerial work and GP partners had legal responsibilities.

As noted in the next section, some practices preferred to recruit advanced practitioners from a range of different occupations, noting that they were not appointing to a particular occupational role, but appointing the ‘right person’ who could be upskilled and moulded into how the practice wanted them to work. As illustrated below, this meant that, although an advanced nurse might be appointed as an advanced practitioner, the practice preferred to name the role as an ACP. Consequently, the non-occupational specific role title of ‘ACP’ introduced some opaqueness about which regulatory body the role-holder was working under (such as nursing, paramedic or allied health professional) and which was relevant for their individual requirements of advanced practice.

Advanced clinical/care practitioners

Three of our sites (i.e. sites A, B and C) employed ACPs. Site A had two ACPs (one was a trained nurse and the other a trained paramedic). Both ACPs were employed by a community health service as part of a local pilot to put advanced practitioners into various surgeries. Commitment to complete ACP training was attached to the job and, as such, both ANPs were in the process of completing the master’s-level course to attain ACP qualification. The practice received funding to mentor these two practitioners in exchange for their training in primary care. The nurse ACP dealt with frail and elderly patients and patients in care homes, particularly those at risk of hospital admission, whereas the paramedic ACP dealt with same-day appointments at the practice and urgent home visits.

The paramedic ACP at site A had a remit to look after patients with frailty and those aged ≥ 65 years. The practice served four nursing homes, at which the ACP attended for half a day each week in rotation. From care homes, she received advance notification of patients who needed to be seen, but encouraged patients who were mobile enough to come into the surgery for minor treatments, such as ear syringing (which the nurses would do). The paramedic ACP deferred women’s health issues to colleagues because she lacked experience in this area. Similarly, the paramedic ACP did not deal with asthma or diabetic reviews or any other chronic illness and, as a consequence, experienced a lack of variation in her role. Owing to the demand for same-day appointments, the paramedic ACP was involved in seeing mainly acute cases and claimed that her paramedic training meant that she had experiences of dealing with whatever she was presented with, although, as noted above, this did not include the treatment of certain conditions.

Site B had two ACPs from nursing backgrounds employed as part of a local project to reduce hospital admissions. The remit of the ACPs was to care for frail elderly patients, including those in care homes, which involved collaboration and coordination with GPs and the district nurses co-located at the practice site. One ACP nurse dealt with patients aged ≥ 55 years and the other ACP focused on frailty for older patients (i.e. mostly patients aged ≥ 65 years, but with an increasing number of patients aged > 85 or 90 years). The nurse ACP who dealt with frailty for older patients previously worked as a district nurse and then a community matron before training as an ACP and joining the practice. Her work was supported by telephone triage that she carried out early each morning to assess all requests for home visits by older adults. She dealt with home visits for housebound older people (‘domiciliary care’) and ward rounds at care homes, which included medication reviews, symptom management, care plan updates, blood pressure checks and monitoring for delirium.

Site C had one trainee ACP who was doing a master’s in advanced clinical practice and another staff member about to begin training. Both staff had background experience in urgent care and worked in an urgent care setting to deal with same-day urgent appointments.

Clinical pharmacists

The new title of CP was introduced in 2015 as part of an NHS pilot to address the scarcity of GPs and nurses when there was an oversupply of qualified pharmacists. 105

Clinical pharmacists in our case study sites were allocated medication-related work. We found variation across CPs and sites regarding the broad range of activities that CPs were performing (listed in Box 1).

Prior qualifications and experience allowed CPs to form links between general practices and community pharmacies. For example, CPs could readily identify and redirect appropriate patients to community pharmacists for help with symptom control (e.g. for coughs and colds) and signpost patients to services, such as smoking cessation services. Experienced CPs also managed a group of medicine management pharmacists working within nearby practices and acted in a supervisory role by managing and/or overseeing less experienced pharmacists and training and supporting pharmacy technicians. CPs consulted with the GPs about more complex areas of medicine, such as mental health, and did not generally perform physical examinations of patients.

Clinical pharmacists’ roles generally evolved to fit the practice’s needs and this could involve exploring different configurations of work in terms of how CPs balance time in various clinical and non-clinical aspects of the role (as seen, for example, in the appointment of the CP at site C as education lead for the practice).

The role of each CP reflected who employed them and to what extent their role and time were split between different organisations. The CP at site E was employed directly by the practice, whereas at sites A and B the CPs were employed by the local CCGs but based at the practice. Therefore, at sites A and B, the CPs’ roles were split between doing CCG work and practice work. CPs who were employed by CCGs were able to exploit their access to wider knowledge of best practice and the latest evidence gained from this boundary-spanning network and apply this to clinical reviews with patients. However, one of the challenges of being employed by the CCG while working and based at a practice was balancing competing time and work requirements.

The CP at site D was employed by an on-site pharmacy owned by the GP partners. Therefore, arrangements were made for them to carry out work, such as NHS health checks, on behalf of the practice, in addition to running the pharmacy. By accessing patients’ records, the CPs could ensure that patients had the appropriate clinical checks and blood tests. CPs also delivered same-day care for minor ailments, such as urinary tract infections and simple skin infections. Site D also envisaged that their CP would in future (i.e. after training) handle prescribing, prescription reauthorisations and care home medications management.

Paramedics: urgent care practitioners

Site A employed one paramedic and site C employed two urgent care practitioners. These practitioners were referred to using the non-occupational specific role title of ‘advanced clinical/care practitioners’ in all sites (see Advanced clinical/care practitioners).

Physiotherapists

To date, there are limited numbers of physiotherapists delivering ‘first-contact physiotherapy’ in general practice with a potential role in the assessment and independent management of patients with musculoskeletal conditions. Although site C had a physiotherapist, access was through GP referral rather than direct access and, therefore, it was not within the remit of this study.

Physician associates

Physician associates are a non-regulated profession. This means that there is no centrally agreed scope of practice and no regulatory body overseeing their training. Therefore, it is up to each practice to negotiate the work carried out by the PAs that they employ. The patient-facing work carried out by PAs in our study appeared similar to much of the patient-facing work carried out at the first point of contact by GPs and ANPs, with a focus on being the first point of contact for acute problems. PAs did not generally deal with the long-term management of health conditions, palliative care and ‘behind the scenes’ work (such as administrative work, home visits and ‘on-call’ work) that GPs do in addition to their acute work.

An exception was that site E employed two PAs whose remit included medication reviews and chronic disease management, in addition to seeing patients with mental health problems. The more experienced PA also reviewed test results and hospital letters. However, much of the PA’s work involved dealing with acute conditions and minor illnesses. These PAs were not as involved in long-term management; however, for some patients, they attempted to provide continuity of care through follow-up appointments and, in one case, specifically employing skills gained through prior experience of working in a hospital psychiatric unit.

The PAs in site E claimed that they could spend more time with their patients because they felt under less time pressure and had fewer responsibilities than GPs. The PAs perceived that this allowed their care to be more effective, thorough and more ‘holistic’ than care provided by GPs. However, as PAs were unable to prescribe, their ability to progress through consultations was hampered by the need to ask GPs to sign prescriptions, reducing their consulting rate.

Others often viewed PAs as non-independent practitioners who had been fast-tracked in their training for their role, but who had limited experience in health care. To build knowledge and confidence, PAs in site E regularly asked for help and spoke of picking up ‘tips and tricks’ from more experienced practitioners. As their clinical experience developed, PAs had gradually been able to see more ‘complex’ patients. As such, the scope of their practice was defined in terms of GPs’ level of confidence in their perceived abilities.

At site D, the role of the GPs in determining the contribution of PAs resulted in a less favourable assessment of this role. Site D’s experience of employing PAs in the past was that they needed a lot of support, particularly during training. GPs who reported poor opinions of PAs perceived them as highly protocol driven and lacking in clinical experience. In addition, PAs were believed to offer limited value to the practice because of their inability to prescribe or request imaging investigations and, therefore, they were unable to independently manage a significant proportion of their cases. As a result of these difficulties, PAs were no longer employed at this practice.

Health-care assistants

In general practice, HCAs/health-care support workers are employed to carry out a wide range of duties. The work of HCAs often involves background tasks (e.g. sterilising equipment, restocking clinical rooms), but can also involve work with patients, including conducting health checks, phlebotomy and health promotional activities. 106 The work of HCAs supports the delivery of the Quality and Outcomes Framework (QOF) and health promotion. 106 However, as HCAs have limited training or qualifications, their work is supervised and supported by qualified nurses or other professional staff.

Social prescribers and care co-ordinators

The social prescriber role was non-regulated, and work was delegated by referrals from GPs. Site C was the only site to employ a social prescriber and their remit was to work with patients to identify support for social problems that were barriers to their health and well-being. In particular, the social prescriber dealt with social isolation, loneliness and mental health, which tended to overlap with other co-existing medical issues. This overlap meant that part of the social prescriber’s role was to clarify ways of approaching difficulties that crossed the boundaries of health and social care:

And we take a holistic approach to social prescribing, in that we gather background information and then work in a client-centred way to identify goals in collaboration with the client. And identifying barriers to their well-being in terms of isolation or loneliness, or identifying positive activities, meaningful activity that might help their well-being. Or identifying sources of, potential sources, my advice or support that might help their well-being. So it’s quite open ended in terms of what we provide.

ID31: site C, social prescriber

At sites A and B, the roles of ‘care coordinators’ had much in common with elements of the social prescriber role, and their work was practice-based, although they were employed by a local community health service. The care coordinators worked to coordinate with and/or signpost patients to social care and community services and resources, and sought to address the social dimensions of health directly and indirectly through preventing or ameliorating associated clinical conditions. Improving coordination between primary care and social care organisations and addressing social care needs were addressed, reducing the need for GPs to be personally involved in these issues.

Although these roles are increasing the skill mix in the practice, they are not taking tasks away from the GP and are an illustration of a service improvement.

Summary

Discussion of the above roles indicates the extent to which general practices have become more multidisciplinary, with practitioners from a more diverse range of disciplinary backgrounds. Our research confirmed the continuing development and enhancement of nursing roles that sees many nurses now working at an advanced level. In addition, some newer practitioner types, such as CPs and paramedics, work independently within their regulated scope of practice (i.e. innovation). The introduction of newer non-regulated roles, such as the PA and social prescriber, further broadens the diversity of workers within general practice.

Although these skill mix changes increase opportunities to distribute work differently within general practices, decisions must now be made about which tasks to transfer from GPs to non-GPs. Training, supervision and regulatory arrangements are all different for each occupational group, and often for each individual; therefore, there is no automatic association between one occupation group and whether the transfer of tasks is through substitution or delegation. In addition, as indicated above, a number of these roles appear to be providing service improvements and role supplementation rather than the role substitution of GPs.

The increased diversity of practitioners makes it more challenging to effectively match practitioners with conditions and problems appropriate for their role and experience. The next section indicates the processes of categorisation that practices used to meet this challenge and the flexibility and adaptability required to achieve effective matching and resolve instances of unsuccessful matching.

Categorisation

Practitioner categorisation refers to how practices categorised their practitioners to effectively match a practitioner with a patient. The detailed study of these categorisation processes within five general practices indicates the complexity of determining an effective categorisation at the first point of contact between the patient and the practice.

The first challenge relates to the need to consider the difference between competencies gained through training qualifications and those gained through acquired previous experience and developing expertise in the role. Practitioner categorisation rests on the notion that different practitioners are trained/qualified to deal with a subset of work, which we would call ‘role competencies’. However, our studies also indicated that practitioners can become experts in an area of work because of upskilling, specialisation and/or past experiences, which we would refer to as ‘role-holder competencies’ (i.e. competencies that go beyond role competencies derived from the original training course). For example, site E employed two PAs who had received training around specific role competencies. Both PAs, therefore, had the same qualifications and were largely trained in the same way. However, although both PAs dealt with same-day appointments, one had an interest in mental health and had gained expertise when working in psychiatry (i.e. role-holder competencies). Therefore, patients with mental health problems would preferentially be offered appointments with this PA.

Likewise, some practitioners have expertise that is separate from their current role. For example, site A employed two ACPs (one ACP with a background in nursing and one ACP with a background as a paramedic). The ACP qualification (i.e. role competencies) confirmed a set of standards for both practitioners to work at an advanced level. However, as one ACP was a nurse and the other a paramedic, the ACPs had different specialisations and past experiences (i.e. role-holder competencies) and dealt with different types of patients.

Our study indicated that the broader the spectrum of practitioner competencies, the greater the need for a more sophisticated process categorise practitioners and to address the differences between role competencies and role-holder competencies. This spectrum requires practices to utilise knowledge of role competencies and individual strengths and interests when evaluating role-holder competencies.

A second challenge arose if broad practitioner categorisations led to duplication of work. This GP describes the difficulty with avoiding duplication and managing overlaps between practitioners:

I think there’s a fair amount of duplication going on [ . . . ] I think there’s duplication sometimes with pharmacists and nurses, and we’re trying to get that rationalised. We keep working on it, to make sure that people aren’t being seen too many times, and talk to too many people. But I think you do often end up talking through something on the phone, that you realise halfway through, that somebody else is doing it, or you know, there’s really nothing to do, until something else happens, and it’s that kind of thing. And I don’t quite know how to reduce that [ . . . ] so that’s a difficult balance.

ID11: site A, GP partner (bold text highlights the relevance of the quote)

Although it may be possible to have an agreed overlap for minor tasks, there is a risk that practitioners feel that they have infringed on someone else’s role, as described by our participants:

So, I know certainly there’ve been times where I felt like I’ve stepped on one of the practice nurse’s toes because I’m getting . . . you know, I’ve got a diabetic patient coming in to see me and I’m doing their med[ication] review but then they’re still booking in, you know, to have their feet checked, and those bits are the sort of overlap of how we do it.

ID17: site B, CP (bold text highlights the relevance of the quote)

A separate process within the matching exercise, but equally important to effective matching, was the patient and problem categorisation. Again, our study indicated the changes that practices had made to ensure that this was as accurate as possible.

To facilitate problem categorisation, the role of receptionists had significantly changed and expanded, as they no longer simply booked appointments, but also guided patients to appointments with different types of practitioners. This required receptionists to explain to patients why they needed to ask about their symptoms or health issues and reassure them that this was to get them to the right appointment. Receptionists needed to acquire new knowledge and skills and use more complex processes than had previously been required, developing a nuanced understanding of practitioners’ changing skills and competencies. Provision of clinically informed advice from doctors or other practitioners to assist reception staff when these decisions were unclear was part of the previously mentioned supporting mechanisms.

Matching

For many patients, receptionists quickly understood what was required and confidently guided them to a suitable problem. However, matching did not always lead to an appropriate appointment (even in cases that appeared straightforward):

There’s always going to be the odd error in the system but that’s where you look and you think, well I can’t prescribe something for that infected toe, it needs to be seen by someone else.

ID13: site A, practice nurse

Accurate categorisation of problems also depended on how patients articulated the problem. In the context of primary care, the innumerable array of undifferentiated symptoms and unfiltered health and other concerns that were routinely encountered could make categorisation difficult. Although experienced receptionists could develop and apply tacit knowledge, as well as formal training, incidents would occur where patients had appointments with practitioners unable to deal with their problem:

. . . with the pressure of appointments and the demand of the patient, just sometimes the receptionist will book an inappropriate appointment. And we learn, you know. So where that came from was the nurses’ meeting last week . . . where the reception manager was in there and the nurses were saying, well, you know, this appointment was made, and it wasn’t right, you know, so we learn.

ID3: site A, practice manager

In some cases, an apparently simple problem becomes recognised as a more serious problem or complex issue. Whatever the reason, prompt resolution of the mismatching of problem and practitioner is particularly important for the patient. Such situations highlight the need for short-term flexibility and longer-term adaptability within the organisation and in how practitioners work.

Flexibility and adaptability

The use of processes of categorisation and matching can have an impact on patients seeking health care. Instead of simply booking an appointment with the practitioner of their choice, they will be asked to explain the problems behind their request. Based on this information, a receptionist will propose a course of action that may be different from their expectations, such as the type of practitioner or appointment. A more detailed exploration of these elements is explored in Research question 3(iii): how does increased skill mix affect patients’ experiences when accessing primary care services?.

In addition, general practices, as organisations, needed to facilitate longer-term adaptability by setting up and revising processes and procedures that accommodate and keep pace with changes in categorisation and matching. Adaptations vary from reviews of changes in care that are deliverable by individual practitioners to ‘work arounds’ that are necessary to accommodate the requirements of medical indemnity insurance cover for advanced practitioners delivering specific services:

. . . so for example we’ve got all these different skill mixes and some of them going out to see vulnerable people at home; and then there were daft things like the insurance didn’t cover certain health professionals doing a flu jab, they could do it in practice but not at home; and it’s little bits like that that suddenly then you’ve now got to send out a second health professional.

ID28: site C, GP partner

We observed other examples of organisational adaptability in the different systems general practices put in place to monitor whether or not categorisation ‘works’ (i.e. whether or not patients were or would be booked in with the ‘right’ practitioners). Site C designated a GP as workforce lead with responsibility for such monitoring. In other sites, recognising this issue was dependent on individual staff raising it at staff meetings rather than through a more systematic process at an organisational level. Further details of the longer-term organisational adaptability are provided in Supporting mechanisms.

Last, practitioners needed to demonstrate short-term flexibility by, for example, finding solutions for patients when a mismatch of problems and practitioners occurred. This required organisations to consider supporting mechanisms that facilitated practitioners seeking advice or needing to transfer patients to other practitioners. Similarly, the need for staff flexibility and adaptability extended to reception staff. In some sites, reception staff carried out some administrative duties. In others sites, administrative duties were the responsibility of back-office administrative staff. In site D, rotation of front-desk and back-office staff ensured greater operational flexibility. Allocation of duties at other sites demonstrated various processes and working practices, as non-clinical tasks were variously allocated to receptionists, medical secretaries and others.

To reduce pressure on GPs and other clinicians, site C adopted a ‘lean journey’ approach whereby an administrative team followed a protocol to deal with prescription management and blood pressure checks, with provision for alerting GPs about abnormal results from tests (i.e. an exception handling approach that takes much of the work of processing results away from GPs). The administrative team were also responsible for the arrangement of pre-appointment blood pressure checks and selecting those patients who were suitable for face-to-face consultations.

Supporting mechanisms

Organisations had made some longer-term adjustments to their processes to support categorisation processes, matching and addressing any mid-matching. These supporting mechanisms are discussed under Managerial and administrative processes, Training, Supervision and Team meetings and information.

Summary

Against a background of gradual change in the composition of the primary care workforce, the recent introduction of financial support for workforce expansion through PCNs is set to accelerate the employment of a more diverse workforce.

Our in-depth exploration of observed clinical consultations (both GP and non-GP consultations) and interviews with staff working at general practices found variation in the degree to which roles are established and the familiarity of patients, GPs and other staff with newer roles. We also found variation in practitioners’ scope of practice within particular roles due to the inconsistency of individuals’ competencies and familiarity with general practice.

The generalist nature of general practice means that categorisation is needed to allow non-clinical reception staff to distribute work effectively. In our study, practice staff attempted to match the ‘right’ patients to the ‘right’ clinicians. However, in practice, the accuracy of categorisation is dependent of patients ‘correctly’ reporting their problem and the receptionist ‘correctly’ understanding and matching the problem with the ‘right’ practitioner. Improved outcomes from categorisation and matching can be achieved where organisational structures, staff training, practitioner flexibility and patient acceptance work effectively together to resolve deficiencies. Underpinning the implementation of skill mix, we identified adaptations in how practitioners are supported by management and administration, training, supervision, and formal and informal interactions.

Implementation of skill mix is an evolving process rather than an end in itself. Our study suggests that, rather than attempting to achieve an ‘optimal’ skill mix, general practices need to engage in ongoing management of the skills of their workforce using the processes outlined above. The next section describes several unintended consequences of new ways of working, including the risk of duplication of work and the impact on GP workloads of ongoing supervision for newer roles, which can adversely affect their work–life balance.

Motivations driving skill mix change

Participants at our study sites reported two dominant motivations for implementing skill mix change: (1) increasing capacity and (2) achieving cost-efficiencies. This GP partner articulates the manner in which some participants linked these two motivations:

The motivation was partly around trying to find a more efficient way of doing some of the work. [ . . . ] So part of it’s about dealing with the workload efficiently. [ . . . ] recruiting doctors can sometimes be difficult. We’ve never failed to do it when we needed to, but sometimes we’ve had to work quite hard to get someone of the quality we want. And obviously, lots of other practices have struggled. Part of it’s about dealing with the workload efficiently. [ . . . ] part of it is workforce, part of it is efficiency, both in terms of cost and clinically. And part of it is that the NHS system is encouraging things like this, and so there’s the ability to access funding and support for it.

ID21: site B, GP partner

Practice managers tended to focus on the need for cost-efficiencies. However, the examples provided related to potential cost-efficiencies for task transfer in particular cases, rather than for the practice overall. For example, the practice manager at site A thought that for some tasks the GPs in their practice were working below their licence, such as doing care home visits that could be carried out by ‘lesser-qualified staff’, such as an ANP. Therefore, the transfer of home care visits was an explicit substitution (for that task) of an ANP for a GP. Another explicit substitution was the transfer of patients from GPs to physiotherapists and CPs who could use their specialist knowledge to deal with ‘complex’ cases, thereby releasing GPs to do other work. Therefore, increased skill mix was conceptualised as supporting cost-efficient patient interactions while also enabling the provision of additional services and high quality of care for patients.

Some of those participants expressing these motivations also referred to the consultant GP model as the future model of general practice workforce, whereby GPs develop specialist skills (as in hospitals) and retain the value of long-term relationships with patients and the provision of ‘holistic’ care.

In addition to direct employment of newer practitioners, our participants argued that employing practitioners through PCNs could support practices by providing PCN-employed practitioners to cover sickness or periods of leave, thereby enabling a more standardised approach to doing things. However, these respondents could also envisage a risk that practitioners working in and across different practices would need to juggle different, and possibly conflicting, priorities. Moreover, our participants highlighted a potential tension between PCNs and general practices in relation to establishing the ‘most efficient way of providing care’ (ID21: site B, GP partner).

Notwithstanding the above, it must be noted that these were not universally accepted views. Many GP partners expressed that they would still like to have more GPs and, therefore, viewed recruitment of non-GPs as a pragmatic response and short-term solution to GP recruitment difficulties.

The next section discusses the extent to which practices indicated that they had achieved their objectives by increasing skill mix.

Meeting objectives of increasing capacity and achieving cost-efficiencies (while improving the quality of the service)

All case study sites claimed that skill mix changes had improved access to appointments [see also Research question 3(iii): how does increased skill mix affect patients’ experiences when accessing primary care services?]. Participants said that the introduction of newer practitioners had enabled new ways of organising care and increasing capacity. In addition, the introduction of newer practitioners had allowed practices to move away from firefighting, or ‘just getting by’ in the context of heavy demand, to operating at a more sustainable pace.

Our participants claimed that skill mix could enable a ‘better’ standard of care for particular groups, such as frail elderly people, people who need same-day appointments or people who need medication reviews. ‘Better’ was defined as making use of a practitioner who has more specialist skills in a particular area (e.g. medication reviews or specialist care for chronic illness) than GPs and who is able to perform the task thoroughly because they have more time to devote to the exercise. Medication safety was identified by our participants as an area of clear improvement in the quality of care when practices had a CP.

Cost-efficiencies were described in site C enabling them to deliver services differently, including making sure that GPs were only dealing with ‘complex’ cases and not just about taking out chunks of GP work and replacing GPs with a ‘specialist’. Site C had undertaken internal modelling and looked at CPs in terms of the administrative prescription changes CPs made and how much work was being taken away from GPs. The practice noted that significant extra hours were added to the GPs’ working time when their CPs were absent. From this, the practice was confident that its model was working, although the practice found it difficult to describe in economic terms:

. . . we know it’s working, but if you are speaking to a health economist, I would struggle to give the absolute in-depth figures.

ID28: site C, GP partner

The practice manager in site D claimed that, although it was cheaper to employ non-GPs such as ANPs, the practice would need to reinvest the money elsewhere. Therefore, skill mix was not about saving money, but about using existing resources and potentially employing another ANP. As the practice manager put it:

. . . it’s purely just been redeployment of skills.

ID36: site D, practice manager

The efficiencies noted above appear to depend on finding the right ‘place’ for staff (i.e. the fit between the skills of staff and their work). For example, upskilling staff enabled staff to do more, but practice managers were aware that this could result in demands for higher pay and/or staff leaving to work elsewhere. Practices found it challenging to determine the cost-efficiencies of skill mix because of unanticipated extra work and uncertainty about how sustainable certain roles would be. For example, a GP partner in site E felt that there were many unknown costs in terms of throughput of patients. The GP partner felt that she was doing extra work due to unnecessary additional work generated by PAs bringing people back too often and using up appointments. Likewise, several GPs noted that, although some tasks were taken away from GPs, additional workload arose from the requirements that GPs must authorise prescriptions generated by non-prescribing colleagues, make referrals and make decisions based on letters from specialists or investigation results (see Supervision). Such comments demonstrate that multiple additional factors and consequences should also be considered alongside the calculation of cost-efficiencies.

Impact on practice staff

Practice staff needed to manage uncertainties and anxieties surrounding their own and their patients’ lack of understanding of what newer practitioners could do. Limited understanding of roles and the performance of similar tasks by different practitioners made it difficult for patients to distinguish between practitioner types and patients would find themselves being corrected when they mistakenly referred to newer practitioners as ‘doctors’:

But there obviously will be times during the consultation, especially with the elderly or anyone with mental health issues, that they might say, doctor, thank you. And I’ll be like, no, you can only correct people so many times without it being a bit, no, I’m not. And sometimes they call me nurse and sometimes I try and say, oh, actually I’m not a nurse, it all depends whether I’m in a bad mood.

ID26: site C, trainee ACP

Skill mix had also increased the need for staff to legitimise their roles in relation to others. This particularly affected newer roles, such as urgent care practitioners and PAs. A trainee ACP with a paramedic background described it as being caught between the doctors’ and nurses’ ‘camps’ (ID26: site C, trainee ACP).

Other practitioners viewed PAs as seeing themselves as similar to a GP. This was attributed to the need to defend their newly acquired status in primary care and being ‘ambassadors’ for their role (ID34: site C, PA). As a consequence, PAs were often viewed less favourably by other practitioners:

I think it’s possibly more problematic for physician associates because they’re non-clinicians [ . . . ]. But the problem is they have no clinical experience which is very tricky for them I think, the differences in the way that they’re regulated and under the purview of a doctor so to speak. [ . . . ] they really do believe they’re almost equal, which I find, you know, completely disagreeable [ . . . ] PAs have trained in the medical schools, and I think some of the golden mantle effect rubs off on them, we’re almost a doctor, we’re almost that confident, we’re almost that good, but they’re only nurses, so there’s a perception of status amongst the PAs going, well, we’re equal. In reality, you know, I think there’s always a situation of they don’t know what they don’t know, but I think that profession will mature over time [ . . . ] You know, the difficulty is you read a physician associate’s notes and fantastic assessment, but then the plan is then discussed with the doctor or more likely is dictated by the doctor.

ID32: site C, ANP (bold text highlights the relevance of the quotation)

Unintended implications for general practitioners

In addition to the impact on staff, several unintended implications of skill mix changes were observed specifically for GPs in the case study sites.

First, the need for ongoing GP supervision of other practitioners created a new and, for many, unexpected workload that was found onerous by some, particularly where there were fewer GPs performing this role and where there was a lack of peer support. GPs found that supervision could be particularly demanding when several practitioners needed support simultaneously. This caused ‘debrief fatigue’ (ID1: site A, GP partner):

There can be times when I have about three or four people queued up just waiting for my time really, and sometimes I just get them all in and say, look, let’s just take this as a learning opportunity, come in and listen to what’s going on with the debrief.

ID33: site C, GP partner

Our GP participants noted an increased need to provide safety nets when working with more non-GPs. Patient safety was protected when each practitioner took a cautious approach to care, knew when to ask for help, and used debriefing/supervision. Patient safety was also assured through processes that provided checks and balances, including a policy for GPs to review patient cases.

The supervisory role meant that GPs needed to juggle their clinical work with the need to review that of other practitioners and with responding to instant messages requesting help, which, consequently, affected GPs’ work–life balance:

So I’m constantly supporting even throughout the lunch break, so it’s very rare that I even get to look at any of my own clinical workload until the end of the day [ . . . ] And then, unfortunately, you get your . . . [newly qualified] GPs who are staying late, because they can’t keep on top of just the clinical workload from the day, who then come in with a list of questions and you’re trying to support them and you’re thinking, I’m trying to get them home, but I might be here another hour later. [ . . . ] it doesn’t affect work, but home life I’ve had to set up a laptop that I take home with me, so that I can see the kids and then I can log on and I can do all my clinical work. But, occasionally, I’ll come across something and it’ll make me cross, because I thought, I should have seen that earlier today and if the patient is going to need a phone call and it’s 8 o’clock at night now, it’s just . . . that’s not right.

ID33: site C, GP partner

Second, a GP partner in site C argued that GPs were feeling pressure in a ‘squeezed middle’:

I feel very much in the squeezed middle group, so I’m an experienced clinician but I’m not close to retirement, it’s not an option, so I need to support everybody, so I’ll need to try and prevent my partners from retiring too early, so you’re protecting the retiring cohort, you’re trying to support the newly trained. I wasn’t trained to do the job I’m now doing, so it’s evolved.

ID28: site C, GP partner

Last, with practitioners upskilling and specialising in some areas, our participants highlighted the risk of GPs being de-skilled because of the reduced range of issues they would deal with. Moreover, our case study sites had supported experienced practitioners to undertake such training only to have them request higher salaries or take their skills elsewhere, exacerbating local workforce retention. Upskilling could also increase the duplication of work. This was an area where some practices struggled to get a clear idea of how to avoid and manage overlaps between practitioners dealing with the same patient.

Summary

General practitioners most often stated that their motivation for skill mix change was a shortage of GPs. In contrast, practice managers tended to focus on the need for cost-efficiencies and saw skill mix as enabling new ways of organising care, which benefited the quality of care.

Skill mix was also said to improve access to appointments; however, as noted below, this requires that patients are happy with the choice of practitioners they are offered. Reference was made to the tension between enabling continuity of care (i.e. seeing the same practitioners) and prompt access to appointments with practitioners who might not be their ‘usual’ practitioner or their practitioner of choice.

Changing the skill mix in the case study sites has enabled the transfer of tasks from GPs to others (e.g. through either enhancement or innovation). However, whether this is through delegation or role substitution is sometimes unclear when specialists or advanced practitioners are taking on these tasks wholly unsupervised.

In terms of achieving cost-efficiencies from skill mix, this was difficult for practice managers to evaluate in purely economic terms because of unanticipated extra work, such as GP supervision for non-GPs. In terms of safety and quality of care, skill mix also created a need to provide more safety nets through ongoing GP supervision. The extra work for GPs associated with skill mix change was frequently discussed.

These case studies also raise questions about whether practices are transferring tasks as a means of role substitution or supplementation. Where a specialist is involved, or a discrete cohort of patients (e.g. frail and elderly patients) are concerned, it would appear that there is an explicit full substitution of tasks. For other interventions, however, such as the first contact appointment, it appears that skill mix increases the number of occupational groups who could perform that task, with GPs continuing to do so too (unless a practice was operating a consultant GP model). If the GPs continue to perform these tasks, the practices arguably operate an explicit role supplementation approach to improve timely access. Again, this appears to be achieved through explicit supplementation, where there is no intention that an alternative practitioner will take up all of the GPs’ tasks. Therefore, practices enable task substitution rather than role substitution, with associated implications for extra work for GPs.

Usefulness of patient consultation

Patients were asked to complete a survey based on their consultation on that same day. Survey questions included asking how useful the appointment was, whether or not the patient thought they saw the right practitioner to deal with their problem and whether or not it was the type of practitioner they wanted. Half of the respondents (63/125) had consulted with a non-GP. As noted in Table 15, a large majority of patients indicated that everything had been dealt with (82%), that the health practitioner was definitely or probably the right practitioner (94%) and that they had seen the type of practitioner they wanted (77%), with 23% of respondents indicating no preference. In addition, 5% of respondents reported that their practitioner had ‘dealt with something they had not intended to discuss’.

The focus groups provided further information about the accessibility of appointments at these practices. As noted above, 77% of patient respondents indicated that they had seen the practitioner they wanted. Patients in the focus groups reported their perception that appointments were more readily available because of skill mix. However, these findings need to be supplemented with reference to the ‘work’ required by patients in booking appointments in practices with a diverse workforce, which is identified below as (1) information-gathering, (2) navigating information technology (IT) and telephone appointment systems, (3) making compromises and (4) developing new relationships.

Patient experience of categorisation and matching

Summary

A large majority (77%) of patients surveyed believed that their appointment had been useful and that they had seen the right practitioner to deal with their problem(s). However, the focus groups did raise concerns about the patient ‘work’ required to understand, access and develop relationships with new practitioners. The introduction of newer roles had brought new systems of access/triage systems to practices, but changes were not communicated well, and concerns were raised about how less-informed patients, particularly older adults, infrequent attendees and vulnerable groups, could navigate the new system. Although patients had limited understanding of newer practitioner types and their capabilities, many patients were reassured after they had consulted with a new non-GP. A contributing factor in this was that a longer appointment slot provided time for more ‘holistic’ consultations than was typically available with GPs, which does have implications for costs. In addition, the visibility and availability of GP ‘supervision’, when needed, improved trust in, and acceptance of, the newer roles among patients.

Structure

In March 2020, at the beginning of the pandemic, support work was needed to keep up to date with rapidly changing guidelines and operating procedures to allow general practices to continue working safely. Practices reported that this included procuring laptops and setting up IT systems to allow staff to work from home. Deficient communication through official channels meant that GP partners and practice managers would watch evening news bulletins for government announcements about what must be carried out the next day with regard to contacting shielding patients (i.e. patients who had been identified as clinically extremely vulnerable by NHS England, which may not be the same list of patients as identified by local general practices). The support provided by local organisations varied and often related to the meeting of IT needs.

At the practice, reception staff had to manage the throughput of patients differently. Entrances to practices were colour coded to manage patient flow. In site E, the reception team kept a record every 30 minutes of how many people were in the waiting room to ensure that social distancing was maintained. Although site E had concern over space during our fieldwork, during COVID-19 it managed to designate rooms for (potential) COVID-19 cases. However, the site also set up an in-hours and out-of-hours COVID-19 hub with neighbouring general practices.

All sites encouraged their staff to work from home. However, some staff were less comfortable with making full use of digital technology than others. For example, in site E, older GPs were judged to be ‘stuck in their ways’ and worried about how they would manage home visits if only a limited number of GPs were in the building. In site B, nurses were doing relatively fewer remote consultations than GPs. In site C, non-GPs were generally less comfortable when dealing with risks associated with telephone triage and consultations.

Processes

During the pandemic, there was a move towards total triage and remote consultations to prevent the spread of the virus by people coming to general practices. A GP partner in site C described this as ‘COVID medicine’:

I think best practice medicine outwith a pandemic would’ve been to go and see the patient but you’re all the time we’re weighing up the risk that we pose by exposing them to more clinicians and having more footfall either through the surgery or going out to the nursing home.

Site C

For triage, all practices used COVID-19 screening questions as part of a process to determine which patients needed a consultation via telephone/video and which patients could be seen face to face. Site C prepared a complete list of specialist skills within the team and booked patients based on skills rather than practitioners’ roles (e.g. in relation to women’s health or mental health). The system was described as similar to a hospital referral letter, which provides information about the nature of the patient’s problem. In site D, the receptionist did a basic triage in terms of whether the patient’s problem was ongoing or new. Patients were then allocated to the on-call list of practitioners. When the practitioner did an initial telephone consultation, they decided whether or not patients needed to be seen, and those requiring a face-to-face consultation were generally seen on the same day. In sites A, B and D, doctors dealt with telephone triage. However, in site B, triage for frail and elderly patients, home visits and care home visits were carried out by the ACPs, as was the case before COVID-19 started. In site E, the site had a ‘doctor first’ triage model where all patients were triaged by a GP. The GP would speak to the patient on the telephone and would decide whether or not the patient needed to be seen face to face.

For remote consultations, all general practices used software that enabled two-way text messaging and video consultations. Patients were able to attach photographs in a text message or upload several pictures to e-mails and images were then attached to the patient’s record.

In terms of supervision, staff could not meet with a wider group of colleagues. Informal meetings were staggered or arranged in shifts, allowing only four to six practitioners in a room together. However, GPs were more available for debriefing or advice, as they were mostly using the telephone rather than seeing patients face to face.

Outcomes

Practices reported variation in patients’ acceptance of remote consultations. Site D found that the elderly population in its area was not averse to using technology and site B found less resistance from elderly patients than expected. In site C, remote consultations worked well for working-age adults, as they did not have to take time off work to see their GP. Site A had not carried out many video consultations, but found that photographs and telephone consultations worked well.

Remote consultations worked for different groups of patients in different sites. Site C found that remote consultations worked well for patients with mental health conditions. Most mental health examinations were based on verbal communication, rather than requiring examinations, and were, therefore, suitable for a video or telephone consultation. Site B found that remote consultations worked well for children when parents only needed reassurance. Site A developed a way of conducting ward rounds at care home:

. . . it’s very useful for care homes, so they’re finding they’re doing their ward rounds at care homes by a video. So, basically somebody at the care home is walking round with a tablet and showing the doctor at that patient’s bedside the wound or whatever the issue is, and then walking onto the next patient. So it works very well with care homes and that’s what the video consultations is an improvement for, and that’s great – care homes.

Site A

However, there were challenges with the use of remote consultations. Video consultations were described as ‘a little bit of an uphill battle’ and ‘cumbersome’ because of difficulty with coordinating the doctor, patient and functioning video cameras. There were also difficulties with connectivity. Despite having Wi-Fi connections in site C, video-consulting was streamed through practitioners’ telephones. Some practice rooms had a worse connection than others, and unreliable connectivity adversely affected video quality. Similar problems were encountered by practitioners working from home. Site C found it difficult for practitioners to communicate complex issues digitally or on the telephone, especially for elderly patients with multiple comorbidities.

When asked about arrangements made during COVID-19 that will stay in the future, all practices envisaged continuing with total triage whereby patients were assessed in terms of whether they need to be seen face to face or remotely (using telephone or video) and whether or not they need to be seen by a GP.

Reflecting generally on their COVID-19 experiences, site A spoke about introducing a shift change conversation where all staff had a 5-minute uninterrupted catch-up, which is similar to a handover in hospitals, to help practice staff feel like they are part of a multidisciplinary team. Site B had learned more about managing patient demand and shifting patient expectations and had thought about which non-GPs they will need in the future. Initially, site B had wanted to recruit a paramedic to deal with home visits; however, as the number of home visits had declined following firmer demand management, the site decided that it wanted to invest more in practitioners who can deal with mental health, specifically a community psychiatric nurse, because the site had seen an increase in low-grade mental health issues related to COVID-19, rather than chronic mental health issues per se. Site C was moving towards a total triage model using artificial intelligence to ensure that it can meet what are assessed as the clinical needs of patients, rather than responding to what could be perceived as patient demands. The proposed change was based on shifting patients’ expectations from access because they wanted to be seen to access because they were telling practices why they needed to be seen. The practice will then decide the ‘right’ next step, which may be a telephone call, advice-giving, signposting elsewhere or face-to-face appointments. Although previously the practice’s reception team did not do any clinical or health navigation, with the new tool, receptionists will be gathering clinical information, including an assessment of urgency, which will then be forwarded to a clinician to assess the appropriateness of the appointment. Site D found that it had carried out more shared decision-making during COVID-19 regarding if and how they should see patients, which brought staff closer together. Site E envisaged that there would be more flexible working from home than in the past, with doctors doing telephone triage and issuing prescriptions from home while practitioners and attached staff (e.g. midwives) who do hands-on treatments would return to working in the practice. Staff at site E had also started communicating more with their neighbouring practices, could access each other’s systems (for practices within the same PCN) and had learned about what activities can be amalgamated and coordinated with neighbouring practices and what activities can be done in-house.

During the COVID-19 pandemic, all sites had daily online and/or socially distanced meetings in the reception area. Site A had moved all meetings online, whereas site C had a daily online meeting to convey information from an announcement made by the government the previous evening. Every morning staff at site C met to discuss issues, such as the workforce problems caused by many of the non-clinical managers working from home, daily statistics on COVID-19 rates, nursing home issues and any changes in government announcements. Their regular monthly face-to-face meetings had been replaced with an hour-long weekly ‘power catch-up’ between clinicians and the management team.

Our study participants also highlighted specific concerns following the move to ‘COVID medicine’. Practitioners felt that they could not build up the face-to-face rapport with patients via remote consultations, which they felt was a cornerstone of GP work. A GP partner in site C reported feeling anonymous to patients, a situation that was described as similar to patients telephoning into NHS 111. This GP further argued that it was hard to measure the unintended consequences of remote consultations.

Practitioners also reported concerns about patients delaying seeking help for conditions such as cancer, heart attack, stroke and chronic obstructive pulmonary disease. In addition, practitioners were unable to do preventative health checks or face-to-face components of health checks for patients with long-term conditions and were dealing with a 6-month backlog of these routine appointments.

Workforce data

As in Chapter 3, we used the workforce data obtained from NHS Digital on the range of practitioners employed in practices across England. We created a longitudinal, practice-level workforce data set using the wMDS. 107 As of October 2020, the data sets are available monthly; however, for the duration of the analysis we present here, data were available annually or quarterly. We use data from 2015 to 2019.

As with 2019 data, these data contain both headcounts and numbers of FTEs for 35 categories of staff (i.e. 11 categories of GPs, 7 categories of nurses and 17 categories of DPC staff). We did not consider variations in other types of staff, as our focus was to look at the outcomes for patients as a consequence of clinical decisions made by and direct interactions with different types of practitioners.

The practice-level workforce data sets are an evolving data collection, with different staff members being added at different times and a dedicated payment now offered to recognise the costs involved in reporting the data. Table 16 presents a list of workforce roles included in the data sets in 2015 and in 2019. The number of roles recorded in the DPC categories has expanded from nine categories to 17 categories, plus eight categories for trainees and apprentices to better reflect the expanding skill mix in general practice. Detailed descriptions of the categories are provided in the National Workforce Reporting System data entry guidance document. 108 We worked closely with NHS Digital to understand the properties of the data. We identified some errors that were fixed and some observations that contained imputed values. Out of the 6867 practices for whom workforce data were made available in the September 2019 data set, there were usable data for 6296 practices.

Outcomes data

We obtained outcomes from the following range of sources:

• the GP Patient Survey (GPPS)

• QOF

• prescribing data

• Hospital Episode Statistics

• NHS payments to general practices

• the GP Worklife Survey.

With the exception of the GP Worklife Survey, all of these sources have universal coverage of practices in England and we obtained them at practice level. The GP Worklife Survey includes data from a sample of individual GP respondents. The outcomes are listed in Table 17. 109

Covariates

From the payments to general practices data set, we obtained figures on the weighted population for each practice. By dividing by the total population from the same source, we created a measure of per-capita workload. We also obtained the amount paid to the practice under the MPIG scheme from this same source. The MPIG payment is an additional payment to the practice over what would be expected under the capitation formula because the practice historically received this level of payment. We divided the payment by the registered population.

We classified practices by contract type as PMS practices tend to have higher payment levels than practices holding the General Medical Services (GMS) contract and may have a wider scope of services than GMS practices. We also identified practices that could dispense for at least some of their patients because of a lack of pharmacies in their area of residence.

We obtained the Market Forces Factor index for each practice from the NHS England CCG allocations 2019/20 to 2023/24 technical documents. 120 This is a measure of relative wages in the local area and is used in the capitation formula to compensate practices for variations in expected input prices due to cost of living.

Scenario analysis

In addition to the margin plots, we also present the results as scenario analysis. We calculated how each outcome would change if the average practice (i.e. with average population size and an average staff level mix) was to increase the number of FTE of each staff type by one unit. We undertook the same exercise for expenditure, using the estimated function for practice payments plus prescribing costs and the unit costs applied to the estimated effects on the volumes of hospital care.

Longitudinal analysis

We estimated long differences using differences between the first and last periods available in the data. This approach has two main advantages. First, it examines the effect of more substantial differences in staffing at the level of practices. Second, this differencing eliminates time-invariant practice characteristics, which may affect efficiency and quality. The coefficients estimated are, therefore, more likely to isolate the independent association between outcomes and staffing levels, reducing the risk of bias due to unobserved heterogeneity.

The longitudinal analysis is presented in Appendix 2 (see Table 23 and Figures 31–42). Owing to limited availability of data, in the longitudinal analysis, we assessed changes in outcomes and expenditure PTP as functions of polynomials of staffing indicators (i.e. changes in levels of FTE staffing PTP, plus interaction terms between GPs and nurses and between GPs and DPC, as well as total number of FTE). For the analysis of changes in the logarithm of total costs, we used the following explanatory variables: polynomials of changes in staffing with interaction terms (GP with nurses and DPC FTE), total FTE and total number of patients.

Estimation

We used Stata for the analyses. We weighted all analyses by the denominator of the dependent variable (most often, the size of the registered population). We used robust standard errors to allow for heteroscedasticity.

Research question 1(ii): how are compositional changes to the workforce associated with later changes in a range of outcomes, including patient and practitioner satisfaction?

Descriptive statistics for each of the outcomes used in the cross-sectional analysis are provided in Table 18. These are shown only for the 6287 practices on which usable workforce data were available. The median practice was awarded 98.49% of the total QOF points available. One-quarter of practices were awarded < 96% of available QOF points. Descriptive statistics for the staffing variables and other covariates are shown in Table 19.

Table 20 presents the fit statistics for the outcome regression models. In addition, tests of joint significance (F-tests) for the workforce parameters are also presented. The p-values for these tests show that for the majority of outcomes the staff variables are significant at p < 0.05 and mostly at p < 0.01. The interaction terms between the GP and nurse/DPC categories are generally not significant.

In the following sections, we discuss the results of the regression models. Each figure (see Figures 13–27) has three or four subgraphs that plot the effect of workforce on the outcome measure. For instance, an upwards sloping curve would signify a positive relationship between the amount of workforce PTP in that staff category and the outcome measure. Figures 13–27 present 95% confidence intervals and the relationship between workforce and the outcome across the range of workforce levels. Given the different scales of outcomes, we focus here on the direction of the relationship before focusing on the magnitude of the relationship in the next section (see Scenario analysis).

How is time since last practice visit with a specific practitioner type associated with workforce composition?

Our analysis of GPPS data on the time since patients last saw a GP (Figure 13) and a nurse (Figure 14) show that practices with more GPs have shorter time periods since respondents last saw a GP, whereas practices with more nurses have lengthier time periods since respondents last saw a GP. Practices with more nurses and DPC staff have shorter times since respondents last saw a nurse.

FIGURE 13.

Estimated effects of staff levels on average time (months) since last saw a GP. (a) GPs; (b) nurses; and (c) DPC staff.

FIGURE 14.

Estimated effects of staff levels on average time (months) since last saw a nurse. (a) GPs; (b) nurses; and (c) DPC staff.

How is patient satisfaction associated with workforce composition?

Practices with more GP FTE PTP have a higher proportion of respondents reporting having had a very good or fairly good experience of making an appointment. Figures 15 and 16 show that the other two broad workforce groups (i.e. nurse and DPC staff) show an initial decline in the proportion of patients selecting the two good responses, followed by a levelling off. Respondents were more likely to report very good or fairly good overall experience of their general practice in practices with more GP FTE PTP. Practices with more DPC staff had lower proportions of respondents indicating a good overall experience.

FIGURE 15.

Question 18 of the GPPS: overall experience of making an appointment. (a) GPs; (b) nurses; and (c) DPC staff.

FIGURE 16.

Question 28 of the GPPS: overall experience. (a) GPs; (b) nurses; and (c) DPC staff.

How is achievement across the clinical sections of Quality and Outcomes Framework associated with workforce composition?

Our analysis looked at the association of the QOF points achieved by practices and their workforce composition PTP. Figure 17 shows increasing and then levelling off trends for all three worker groups. This indicates that practices with more staff FTE PTP achieve, on average, a higher percentage of QOF points.

FIGURE 17.

Quality and Outcomes Framework: total achievement (%). (a) GPs; (b) nurses; and (c) DPC staff.

How are selected prescribing measures associated with workforce composition?

A larger number of CPs PTP is associated with a lower ratio of broad- to narrow-spectrum antibiotics (Figure 18), fewer total items prescribed per ASTRO-PU (Figure 19) and lower average costs per item (Figure 20). There is no significant relationship between the level of GP staffing PTP and the three prescribing outcome measures.

FIGURE 18.

Percentage of broad- to narrow-spectrum antibiotics. (a) GPs; (b) nurses; (c) DPC staff (without pharmacists); and (d) pharmacists.

FIGURE 19.

Volume of prescribing: total items prescribed per ASTRO-PUs. (a) GPs; (b) nurses; (c) DPC staff (without pharmacists); and (d) pharmacists.

FIGURE 20.

Prescribing cost per item ( = cost per ASTRO-PU/items per ASTRO-PU). (a) GPs; (b) nurses; (c) DPC staff (without pharmacists); and (d) pharmacists.

How is practice population utilisation of hospital services associated with workforce composition?

Analysis of secondary care activity (using Hospital Episode Statistics data) shows that the relationship between GPs and nurses FTE and A&E attendances is relatively flat, whereas increased numbers of DPC staff is associated with higher levels of A&E attendance (Figure 21). There is a positive relationship for GP and nursing staff and hospital admissions (i.e. elective and emergency), whereas there is a negative relationship with DPC staff (Figure 22). There is a positive relationship with GP staffing for outpatient appointments, but a negative relationship with both nursing and DPC staff (Figure 23).

FIGURE 21.

Accident and emergency attendances PTP. (a) GPs; (b) nurses; and (c) DPC staff.

FIGURE 22.

Admissions PTP. (a) GPs; (b) nurses; and (c) DPC staff.

FIGURE 23.

Outpatient attendances PTP. (a) GPs; (b) nurses; and (c) DPC staff.

How are general practitioners’ working lives associated with workforce composition?

The analysis of GP Worklife Survey (looking at GP job satisfaction, hours of work and percentage of work that could be delegated) shows that GP job satisfaction is positively associated with the number of GPs PTP at the respondent’s practice. The relationship between GP job satisfaction and the other two staff categories is slightly negative, but not statistically significant (Figure 24), indicating that GP job satisfaction is unaffected by an increase in the number of these staff. There is a negative relationship between the number of GPs PTP at the respondent’s practice and the number of hours that the respondent GP reports working, which suggests that GPs work longer hours in practices where there are fewer GPs PTP (Figure 25). The association between GP working hours and the two other staff categories is flat or slightly negative, suggesting that the employment of the two other staff has a limited impact on the workload undertaken by GPs (Figure 26).

FIGURE 24.

General practitioner job satisfaction. (a) GPs; (b) nurses; and (c) DPC staff.

FIGURE 25.

General practitioner hours of work per week. (a) GPs; (b) nurses; and (c) DPC staff.

FIGURE 26.

Percentage of work that could be delegated to new roles. (a) GPs; (b) nurses; and (c) DPC staff.

Research question 1(iii): how are workforce changes associated with later changes in costs and practice efficiency?

Our analysis of the association between workforce changes and changes in costs and efficiency looked at NHS expenditure in the form of payments to practices and prescribing costs. In Figure 27, we report the analysis for both total NHS expenditure PTP and for log-transformed values of total NHS expenditure. On the one hand, rescaling the NHS expenditure by the number of patients allows the comparison of practices of different size. On the other hand, the analysis on the log-transformed variable help adjusting the skewness in distribution of (raw) total NHS expenditure, and to allow an intuitive interpretation of distances on the scale of the vertical axis as percentage changes in total NHS expenditures. There is a positive association between the log of total NHS expenditure on activity delivered or generated by general practices and levels of staffing in all three workforce groups (i.e. GPs, nurses and DPC staff). The result is not entirely surprising, as it suggests that additional staff generally increase the total NHS expenditure for the practice. Therefore, our cross-sectional analysis does not show a clear pattern of cost savings related to differences in skill mix.

FIGURE 27.

Expenditure (payments to practice + prescribing costs) PTP. (a) GPs; (b) nurses; and (c) DPC staff.

Summary of findings

Our approach represents a cross-sectional assessment of associations between different levels of employment of different types of professionals and a variety of outcome indicators, followed by a scenario analysis modelling potential changes that may be associated with marginal changes in employment mix.

We have found associations in a wide range of indicators associated with different levels of skill mix employment in primary care. Figures 13–27 demonstrate levels of outcomes associated with differing levels of employment of particular types of practitioners, controlling for a range of contextual factors (as listed in Methods and in Table 19). We found that, unsurprisingly, a larger number of FTE of GPs or nurses PTP is associated with a lower reported time since seeing a GP or a nurse. Perhaps more interestingly, we found that a higher FTE of nurses PTP is associated with a longer reported time since seeing a GP. This suggests that there is at least some degree of substitutability between these two groups. Similarly, a higher FTE of GPs PTP is associated with a slightly longer time since last seeing a nurse.

In terms of patient satisfaction with making an appointment, we have shown that a higher FTE of GPs PTP is associated with relatively higher patient satisfaction with their experience of making an appointment, whereas a higher FTE of both nurses and other DPC staff is associated with lower satisfaction. We found that a higher FTE of GPs PTP is associated with a higher level of overall satisfaction with a general practice, whereas a higher FTE of other DPC staff is associated with a relatively lower overall satisfaction. A higher FTE of nurses PTP is not associated with any difference in overall satisfaction. These findings may reflect the issues, which we have described in Chapter 5, relating to the complexity of the systems required to support the categorisation of patients to see appropriate professionals. It would appear that, notwithstanding their satisfaction with their individual clinical encounters, patients may find these processes burdensome. Furthermore, patients’ limited awareness of what newer types of practitioners could provide could contribute to lower satisfaction with a more diverse workforce and their desire to achieve continuity of care from the same practitioner would be likely to be less achievable.

In terms of clinical quality, we found that a higher FTE PTP of GPs, nurses and other DPC staff are all associated with higher levels of total QOF performance, with the highest level associated with a higher FTE of GPs PTP. This is interesting, as it is often nurses who undertake the majority of QOF-related work. It might be expected that a higher FTE of nurses might be the biggest driver of increase in QOF performance. However, our study suggests that having more GPs PTP supports better QOF performance, perhaps due to GPs in such practices having more time to spend overseeing the QOF process.

Because of their role in monitoring and influencing prescribing, we separated CPs from other DPC staff when examining prescribing. We found that a higher FTE of CPs PTP was associated with relatively higher prescribing quality (as measured by the percentage of broad- to narrow-spectrum antibiotics prescribed), with no change associated with higher FTE of GPs, nurses or other DPC staff. In terms of prescribing volume, we found that a higher FTE of GPs PTP is not associated with any difference in numbers of items prescribed per weighted population, whereas a higher FTE PTP of both nurses and other DPC staff (excluding CPs) is associated with more items prescribed. A higher FTE of CPs PTP is associated with fewer items prescribed per weighted population.

Finally, in terms of prescribing costs, we found that a higher FTE of GPs PTP is associated with a higher cost per item prescribed, whereas a higher FTE of nurses and other DPC staff is associated with a lower cost per item. A higher FTE of CPs PTP is not associated with any difference in cost per item. Therefore, it would seem that employing more CPs is associated with better prescribing quality and lower prescribing volume, with no association with cost per item. A higher employment level of nurses and other DPC staff has no association with prescribing quality, but is associated with higher prescribing volume, albeit of lower-cost items. This may relate to the discovery of unmet needs or an improvement in chronic disease management. However, if so, one might expect that a higher FTE of GPs PTP would also be associated with higher prescribing volume. Alternatively, this finding may reflect the complex processes, which we have described in Research question 3(i): how are new ways of working being negotiated in general practices where skill mix changes have occurred?, by which non-prescribing professionals seek support from GPs or other prescribers. If a nurse, PA or HCA without prescribing rights feels that a patient needs a prescription, then they have to request that a GP or other prescriber authorise the script. It is possible that such prescribing activity is subject to less scrutiny or inclination to consider a non-pharmaceutical alternative (e.g. counselling, physiotherapy) than is the case when a GP personally initiates a prescription, leading to a relative increase in prescribing overall.

Considering hospital utilisation, we found that a higher FTE of GPs or nurses PTP was associated with a relatively lower rate of A&E attendances, but a higher FTE of other DPC staff PTP was associated with a relatively higher rate of A&E attendances. This may be related to the increased complexity of making an appointment or could reflect a feeling among patients that consulting a non-GP did not bring them sufficient assurance or resolution of their problem. In addition, owing to perceived or formal constraints on which practitioners are permitted to refer patients to hospital departments, some non-GPs may recommend that patients attend A&E as a self-referral for the same (and possibly appropriate) reasons as a GP would process a formal referral to an A&E or non-A&E department. Associations between changes in workforce composition and hospital admissions and outpatient attendances are difficult to interpret. Higher FTE of GPs PTP would seem to be associated with more admissions and outpatient attendances, perhaps reflecting the uncovering of unmet need, whereas higher FTE of nurses PTP is associated with more admissions but fewer outpatient attendances. Higher FTE of other DPC staff PTP is, similarly, associated with a lower outpatient activity. It may be that increasing the available nursing workforce and DPC workforce allows practices to more effectively manage patients with chronic diseases, such as diabetes or asthma, and thereby reduce acute events requiring hospital care. Similarly, effective delivery of health promotion may reduce patients’ requirements from hospital services.

We explored associations between the FTE of different professionals with GP satisfaction using GP Worklife Survey data. We found that higher FTE of GPs PTP is associated with higher overall GP job satisfaction, but higher FTE of nurses and other DPC staff is associated with slightly lower overall GP satisfaction. In terms of GP hours of work, higher FTE of GPs PTP is associated with a slightly smaller number of hours of work, whereas higher FTE of both nurses and other DPC staff is associated with a negative and then positive relationship with GP hours of work (i.e. a slight U-shaped curve). This suggests that there may be an optimum ratio of other professionals to GPs and that there is a threshold of employment of other types of practitioners beyond which higher levels of employment are associated with a larger number of hours of work for GPs, perhaps reflecting higher supervision requirements. GPs were asked what proportion of their workload could be safely delegated to other professionals. GPs working in practices with a higher FTE of both nurses and other DPC staff suggested a higher percentage of work that they felt could be delegated, perhaps indicating that experience of working alongside other types of professionals increases confidence in delegation.

Finally, we explored costs for GP-based prescribing and payable to practices (including QOF, salaries and business) costs associated with higher FTE PTP of all types of professionals and found that higher FTE of GPs, nurses and other DPC staff are all associated with higher costs.

Overall, the patterns identified with the longitudinal analysis are broadly consistent with the cross-sectional results. This is reassuring, as it attenuates concerns regarding potential unobserved practice characteristics driving the cross-sectional results. The most interesting exceptions are represented by A&E attendances, overall admissions and outpatient admissions, which have been found to grow stronger in facilities showing higher increases in FTE across all staff categories (i.e. GP, nurse and DPC). Total cost also reveals an interesting longitudinal pattern, growing less in practices where GP FTE PTP increased the most.

Strengths and limitations

Our analysis is based on national data. We used the most up-to-date data available, although we did not use data for 2020 because the COVID-19 pandemic has had profound impacts on health-care activity levels and patient outcomes. It would be of future interest to consider whether or not practices with some forms of skill mix were more resilient to the pandemic and could adapt and respond more flexibly. It would also be interesting to discover if workforce composition helped practices to maintain high performance on quality indicators and if skill mix contributed to a more rapid recovery of performance.

No national data are available at the practice level on the volume of consultations delivered. The appointments in general practice series currently provides data at only the CCG level. We created a measure of how recently a practice’s patients had seen a GP and how recently they had seen a nurse based on data from the 2017 GPPS. This is based on a sample of patients’ responses about each practice and is only a crude proxy for the utilisation of primary care.

There is potential for skill mix to be endogenous in the outcome equations if there are unmeasured factors associated with both skill mix and outcomes. This would occur, for example, if a broader skill mix was associated with innovation and if innovation led to better patient care. We included a range of measured characteristics of practices and their populations to minimise the risk of omitted variable bias. We also examined changes over time, which controls for unmeasured but time-invariant characteristics of practices. Nonetheless, the risk of confounding from time-varying factors remains. We experimented with potentially exogenous variables (i.e. whether or not the CCG has offered incentives to employ new staff types, local labour and housing market conditions) in an instrumental variables framework, but these variables were not sufficiently predictive to provide accurate results. The results should be interpreted with this limitation in mind, and future work should focus on identifying better exogenous predictors of variation in skill mix.

Implications

We will consider the wider implications of these findings in our discussion. However, our quantitative findings would appear to align with the qualitative findings from our case studies. In particular, we find that patient satisfaction with their overall interaction with their practices would seem to be sensitive to changes in skill mix, which may reflect the increasing complexity of appointment systems that we have described. This may also be one of the underlying mechanisms driving the association between higher FTE of other DPC staff and higher A&E attendances. CPs are associated with better prescribing quality and lower volumes, suggesting a particular role in medication management, whereas higher levels of employment of both nurses and other professionals are associated with higher prescribing volume. In our case studies, we found that non-GPs may respond to patients’ needs by generating a prescription that is then presented to a colleague who is a prescriber for authorisation/signing. It is possible that this split responsibility (between indicating to the patient that a prescription is indicated, generating a request for a prescription and signing off the prescription) drives the higher prescribing volume that we found. Alternatively, it may represent a more direct effect of nurses and other professionals undertaking chronic disease management and identifying unmet needs. More worrying for current policy in England, we found that higher FTE of the non-GP workforce is associated with lower GP job satisfaction, which suggests that, although the current policy is driven by difficulties in recruiting and retaining GPs, it is not a forgone conclusion that increasing skill mix provides an answer to the problem. Overall costs are higher when the number of practitioners PTP is larger, and our findings suggest that these differences are only marginally smaller for professionals other than GPs.

How has the workforce changed and where has any change occurred?

Skill mix in general practice has been expanding to include roles such as CPs and paramedics. However, these roles still represent a very small proportion of the workforce recorded during the study period (i.e. 2015–19). The roll-out of these roles into general practice has not been uniform across the country, with paramedics mainly employed by practices in the south of England and PAs being employed by practices in larger cities. Analysis of practice-level workforce data shows that a range of demand and supply factors are associated with the practice workforce. These factors include geographical factors (e.g. distance to the nearest hospital that may provide initial training for health professionals), supply factors (e.g. the market forces factor, which may affect the affordability of living near the practice for less highly paid workers) and demand factors (e.g. the proportion of patients aged ≥ 65 years whose health-care needs are likely to increase).

Practice managers’ reasons for implementing skill mix changes and vision for the future

Our national practice manager survey, which received 1205 responses (approximately 17% of all general practices in England) between August and December 2019, showed that the motivation for employing new roles varied between types of practitioners. In addition, the range of motivations included to increase overall appointment availability, release GP time and provide a better match between what patients need and what the general practice team can deliver. Funding was not reported to be a key motivating factor for practices’ current employment choices. The survey also showed that the ideal practice workforce would comprise > 70% GPs and nurses, whereas newer practitioners would make up < 20% of the ideal practice workforce. Our case studies showed that practice managers held different views from many interviewed practitioners, which may explain why ‘unable to recruit a GP’ was not frequently selected in the online survey (WP2) by practice managers as a motivation for skill mix. Our case study findings added more nuance around these motivations, suggesting that the recruitment of new roles was a pragmatic response and a short-term solution to GP recruitment difficulties. Moreover, although accessing new roles via PCNs was broadly welcomed, practices expressed concern over the potential for conflicting priorities between PCNs and practices. During our case studies, details of how practitioners would be employed via PCNs continued to emerge and develop. Many of the arrangements were complex and variable, and it was unclear how this would work in practice. We return to this issue below.

New ways of working

The case studies explored in depth the way in which skill mix change has been accommodated in five practices. Our findings further complemented the survey by showing that, although newer roles can increase overall appointment availability and release GP time, some of the newer roles required longer appointment times, reducing the number of patients who can be seen. Moreover, we found that most roles required some ongoing GP supervision, creating an unexpected new workload for GPs and raising questions about the cost-effectiveness of skill mix. The case studies further described the processes that general practices had put in place to organise new ways of working and gathered the experiences of practice staff and patients in navigating the new system. There was variability in the breadth of practitioners’ scope of practice, in the degree to which roles were established and in the familiarity of patients, GPs and other staff with new roles. There was also variability of staff holding these roles in terms of the consistency of competencies they held and their familiarity with general practice. Role competencies (i.e. those competencies derived from training programmes) often differed from role-holder competencies (i.e. those competencies derived from individuals’ personal experiences and acquired skills). This means that practices cannot assume that practitioners holding the same title, qualifications and training will necessarily undertake the same tasks. In our case study sites, general practices engaged in categorisation (of practitioner and problem) to inform appointment matching. Practices categorised their practitioners according to their qualifications, training, upskilling, specialisation and/or past experiences. Categorisation took the form of a competency framework, skill mix matrix or internal directory of services. Receptionists used these categories to match patients’ problems with what practitioners could deal with. However, the problem(s) patients presented to receptionists during triage could be different from those which patients subsequently presented during consultations. This could lead to imperfect matching and required practices to put in place an adaptable system to allow practitioners to respond with flexibility when patients’ needs demanded a prompt resolution of mismatching. Practices also needed to monitor, review and adjust processes where necessary. Patient education/communication and availability of GP supervision for new roles were important in ensuring patient acceptance of skill mix change. Patients reflected that relational continuity (i.e. seeing the same practitioner more than once) was helpful in developing trust and confidence.

The arrival of the COVID-19 pandemic changed the way in which practices delivered care. All practices adopted a total triage approach to determine if a telephone/video consultation was adequate or if patients needed to be seen face to face. In various ways, general practices also changed the way in which they categorised their practitioners and problems. The inevitable and accelerating use of remote consultations brought challenges in terms of accessibility of equipment and IT connectivity and also created difficulties for patients and staff in adjusting to consultations using technology. It was envisaged that, in future, a greater proportion of work would be carried out from home and that increased use of remote consultations would continue. The format of meetings also changed, with more regular, shorter meetings with the wider team and more shared decision-making and coordination of work with neighbouring practices.

Outcomes and expenditure

Our quantitative analysis utilised a production function approach. This approach explores associations between an increase in FTE of particular practitioner groups and a range of outcomes, controlling for factors such as deprivation, practice size and patient demographics. We also modelled change in expenditure associated with increases in FTE of different practitioners and changes in outcomes associated with a unit change in FTE of different practitioner groups.

We found that higher FTE PTP of nurses and other DPC staff are associated with relatively lower patient satisfaction with their experience of making an appointment, perhaps reflecting the complexities associated with categorisation and matching, which we have elucidated. Higher levels of employment of other DPC staff are also associated with a relatively lower overall satisfaction with the practice. We found that a higher FTE PTP of GPs was associated with higher GP satisfaction, whereas a higher FTE PTP of other DPC staff and nurses is associated with a slightly lower overall GP satisfaction. GP hours of work describe a U-shaped curve, with an apparent threshold effect where a higher FTE of other DPC staff PTP above a certain level is associated with larger number of GP hours of work.

Clinical pharmacists would seem to have a particular role in improving prescribing quality and reducing prescribing volume, reflecting their particular skills and the ways in which we found their time being used in practices. We found evidence of a higher prescribing volume associated with higher FTE of both nurses and other DPC staff. This may reflect increasing chronic disease management, but it may also reflect a loosening of control of prescribing associated with the separation that we observed between prescription initiation and signing when practitioners do not hold prescribing qualifications. In terms of health services use, higher FTE of other DPC staff is associated with a slightly higher A&E attendance. Higher levels of employment of all practitioner types PTP is associated with higher costs, with the largest increase associated with higher numbers of GPs. However, the magnitude of the differences between types of practitioners observed is not large.

In summary, we have shown an increase in skill mix diversity in general practices, with an increasing proportion of practices employing a wide range of newer types of practitioners. Patients are generally accepting of these changes, but their implementation is not straightforward and does not necessarily result in overall increases in efficiency because of the complicated matching processes required to make sure patients see the right practitioners, the inefficiencies associated with mismatching and the need for ongoing supervision of some of the newer practitioners. There is some evidence of lower patient satisfaction with the practical steps of making an appointment that are a necessary part of matching their problem with a suitable practitioner. However, our qualitative work also reveals the dynamic nature of skill mix change, with practices, practitioners and patients all adapting over time to the changes that are occurring. Therefore, any evaluation of their impact will represent only a snapshot of an evolving situation.

The complexity of the task of skill mix implementation, and the need for ongoing negotiation and adaptation over time, raises important questions for the implementation of skill mix change via PCNs. In particular, our findings suggest that practitioners who work across multiple practices may find themselves working in different ways in different practices and that the supportive mechanisms required to facilitate the necessary adaptations over time may be more difficult to implement with practitioners who move between practices. Therefore, it is possible that the complex ongoing adaptive work required to make these changes work in practice may be difficult to achieve with practitioners who spend short periods of time in more than one practice.

Conceptualisation

There are various definitions attributed to the term ‘skill mix’. Moreover, the plethora of role titles and inconsistency in the definition of roles have created confusion, for both patients and practice staff, in terms of what health-care workers can deliver in practice. 67,122–125 To counter this, it has been suggested that skill mix should be considered within the broader organisational and health-care system contexts. 17 Our study offers a complementary view by suggesting that skill mix should be seen as an evolving process rather than an end in itself. Rather than attempting to achieve an ‘optimal’ skill mix, general practices need to engage in ongoing categorisation work (of practitioner and problem) to match each problem with a suitable practitioner and to maintain flexibility and adaptability of organisational processes. Our study also suggests that the ultimate expression of skill mix in general practices will be one of diversity, in which different practitioners work in different ways in different contexts. Although this brings with it some complications, it is in fact consonant both with the historical experience of practices with the employment of practice nurses in general practice (in that practice nurses have always undertaken varied and developing roles) and specialist nurses in hospital settings. 126,127 What is crucial in all settings is that appropriate organisational processes exist for the induction and supervision of staff and that processes allow flexibility over time so that staff acquiring new skills can adapt their scope of practice.

Task reallocation

In their early work on this topic, Sibbald et al. 16 delineated skill mix change in terms of enhancement, substitution, delegation and/or innovation. Enhancement refers to individuals increasing the scope of their role, whereas substitution references the ideal that practitioners will be interchangeable, with a newer type of worker substituting for a GP. It is this formulation that, to a large extent, underpins the assumptions of the policy context that we have elucidated (i.e. newer types of practitioners will compensate for the shortage of GPs, both by direct substitution and also by delegation, with GPs passing on some tasks, thereby freeing them up to undertake more complex work). However, recent studies have suggested that enhancement and innovation most appropriately refer to whether changes are through existing or new roles, and delegation and substitution refer to how tasks are transferred. 16,20,21 Our study shows that non-GPs, specifically CPs and PAs, viewed themselves as providing work that is complementary to GPs’ work rather than as acting GP substitutes.

The potential for skill mix change to complement or supplement existing roles (to improve services) is not new and has been referenced over the last couple of decades. 24,128 This study, however, has enabled us to see how these outcomes of supplementation (rather than substitution) emerge from the complexity of multidisciplinary working where patients presenting with (often) undifferentiated symptoms need to be seen in a time-critical environment and multiple practitioners are enabled to do a range of tasks for them. This apparent (and relative) openness to performing the same tasks across professions is arguably different from other studies in which resistance to changing skill mix and transferring tasks has been seen as emerging from the protection of more specialist professional boundaries, and maybe this too will evolve over time in general practice. 64

The study has confirmed the benefits (identified in Chapter 1) of developing Sibbald et al. ’s16 framework to better delineate whether skills are changed through enhancement of someone’s existing role or innovation of a new role (with associated training), the basis on which tasks are transferred (substitution or delegation), and how the GP and alternative practitioner are subsequently deployed (substitution or supplementary).

The need for supervision, which we observed in the case studies, suggests that strict substitution will rarely be possible. Furthermore, the continued evolution of service delivery, as staff develop new skills and practices adapt to accommodate a wider range of professionals, suggests that it is unlikely ever to be possible to straightforwardly distinguish between these categories.

Our quantitative work indicates that larger numbers of other DPC staff may be associated with larger numbers of GP hours of work. This finding requires further elucidation; however, taken together, these findings suggest that the assumptions underlying the current skill mix change policy may be overly simplistic.

Other studies reviewing the implementation of skill mix have commented on the lack of clear role definitions and several studies have called for clearer role definitions for practitioners. 22,23,49,88,89,129 Our study shows that attempts to delineate roles between practitioners need to be seen as part of an ongoing process that will be locally specific and contingent. Moreover, we would suggest that roles will continue to adapt over time as newer practitioners settle into their roles, undertake additional training and learn from their colleagues. Importantly, we have highlighted that role competencies and role-holder competencies may differ, requiring practices to adapt to the evolving capabilities of different practitioners over time.

It has also been proposed that much of the time GPs spend on administrative tasks could be freed up by transferring such tasks to administrative staff. 9 Administrative staff were found to be able to deal with referrals and results-chasing if supported by simplified processes. 37 Studies have suggested that non-clinical staff should be responsible for e-prescribing and dealing with technical difficulties. 130 However, it is difficult to clearly distinguish between ‘administrative’ and ‘clinical’ tasks, with many so-called administrative tasks requiring a degree of clinical knowledge. 131 General practices in our study had identified elements of work that had formerly fallen to GPs, but could be carried out by other practitioners or non-clinical staff (e.g. administrative or reception staff), such as signing prescriptions, reviewing blood test results and reviewing hospital letters. However, other studies have highlighted the realities of how complex these tasks can be, with practices developing quite convoluted routines that are also context specific. 132 This suggests that it may not be straightforward to identify a general category of work that can be transferred in every situation. Moreover, our finding that higher levels of employment of other DPC staff is associated with a higher number of GP hours of work suggests that changing skill mix may not straightforwardly reduce the workloads of GPs.

Factors such as legal limits in the scope of practice (e.g. around prescribing) and lack of trust in non-GPs’ abilities have been found to affect task reallocation. 21 Our study shows that the lack of prescribing rights for some new roles, specifically PAs, could disrupt the consultations’ flow and may be associated with increased prescribing volume. On the other hand, the availability of GPs to provide supervision for non-GPs increased patients’ trust and confidence in consultations with non-GPs, albeit at the expense of hoped-for increased efficiency.

Multidisciplinary working in general practice

Issues of interprofessional competition and professionals’ attempts to protect occupational jurisdiction in their work are well documented. 62–65 Studies have revealed tensions between staff in relation to authority, legitimacy, expertise and efforts made to gain professional recognition in multidisciplinary working. 43,66–68 Earlier studies found that the following factors facilitated multidisciplinary working in primary care: co-location, a stable organisational structure, having clearly defined roles and workflow, good communication through ‘huddles’, team meetings, informal ‘handoffs’ of patients, shared goals, respect and trust. 49–54 Our study adds to the existing literature by providing a detailed description of different ways in which meetings can be conducted and the purpose of these meetings. In addition, our study has highlighted a significant degree of variation in approaches to professional boundaries in practice. Although some practitioners defended boundaries in traditional ways by, for example, explaining how and why their role was different from others, at the same time we found a willingness to be flexible and an acknowledgement that professional work may change over time.

Incentives

Earlier studies found that financial incentives influence the willingness to adopt skill mix. 41,42,89,129 Although additional funding to encourage new ways of working can incentivise skill mix change, such changes could be unsustainable without longer-term funding. 42,133 Our survey shows that funding and incentives were not the main reasons for employing non-GPs. With the exception of CPs and, to a much lesser extent, PAs, most general practices employed advanced nurses, specialist nurses, paramedics or HCAs without incentives. However, our case studies show that the source of funding determined which practitioners the practice could ‘experiment’ with without taking on additional financial risk. Therefore, the availability of funding could affect how practices categorised their practitioners.

Care navigation

The availability of wider practitioners in general practice required reception staff to have training about their various capabilities, and some were more highly trained than others in care navigation. Care navigation was seen as an extension to a receptionist’s job whereby they signposted individual patients to external non-clinical services. This could include generating personalised action plans linked with resources in the community to support patients using local groups and support services to address their non-medical needs. 134,135 HEE developed three levels for care navigation competencies, with the receptionist at the ‘essential’ level, the social prescriber at the ‘enhanced’ level and the navigator team leader at the ‘expert’ level. 136 In our study, care navigation was carried out by either receptionists or social prescribers. Although receptionists signposted patients internally to different practitioners or externally to other primary care services, social prescribers coordinated and/or signposted patients to social care and community services and resources. In addition, social prescribers sought to address the social dimensions of health directly and also indirectly through preventing or ameliorating associated clinical conditions.

Supervision

Studies have identified GP supervision as an unintended consequence of skill mix. 137 Our study provides a detailed description of what supervision involves and the importance of informal coffee break meetings for knowledge exchange and staff development. The purpose of these informal meetings was to provide supervision, develop less siloed working and facilitate multidisciplinary working. For this to work, it requires practices to have a relatively stable staff make up with several longstanding staff. Our study also found that supervision created a new additional workload for GPs that was causing ‘debrief fatigue’. Our quantitative work suggests reduced work-related satisfaction for GPs in practices with higher employment of other professionals. This finding requires further study.

Cost-effectiveness

There were mixed views about whether non-GPs provided good value for money. The perceived cost-effectiveness of employing ACPs was found to be pertinent to its sustainability. 45 However, non-GPs may take longer to complete tasks than GPs, as was the case for practice-employed CPs. 138 There was also evidence of increased transaction costs and a decrease in the continuity of care as staff spent increasing amounts of time conferring with each other and, therefore, had less time available for DPC. 139 Although cost-effectiveness did not prompt change, without careful attention from managers skill mix changes might lead to service improvements at a greater overall cost. 128 If this is the case, a proportion of these additional costs will require continuous funding through realignment of budgets for populations rather than within practices. Moreover, our interview participants claimed that cost-effectiveness of skill mix was difficult to measure in economic terms because of unanticipated extra work (e.g. GP supervision) involved when employing new roles, which has created an additional workload for GPs. This is consistent with our finding that higher FTE of all types of practitioners is associated with higher expenditures. The perceived cost-effectiveness of employing ACPs was found to be pertinent to its sustainability. 45

Patient experiences

Patient experiences of skill mix were referred to in the literature in terms of patterns of care, service utilisation, patient attitudes or consultations with single practitioner types. 140–143 Although patients reported having limited knowledge of what new roles can deliver, patient acceptability of non-GPs was influenced by patients’ expectation of what new roles can deliver, practitioners’ accessibility and promotion of the role. 43,144 Skill mix has been associated with increased access to appointments, longer consultation times and improvement in patient satisfaction, quality of life, follow-up appointments and continuity of practitioners (with non-GPs, as they were more available than GPs). 55,57,58,72,73,75,144,145 Similar to these studies, we found that patient-reported satisfaction with skill mix changes was related to improved waiting times for access to appointments. However, our quantitative work suggests that overall satisfaction with the process of obtaining an appointment is lower when there are more different types of practitioners employed. Patients in our study also had a limited understanding of newer practitioner types and their capabilities. However, some patients reported being happy with their non-GP consultations and contributing factors included longer appointment slots, which were seen as providing patients with more ‘holistic’ consultations than with GPs, and, consequently, improved trust in and acceptance of the new roles among patients. Patients in our study appreciated practitioner accessibility, indicating a trade-off for some between access to care and seeing their preferred practitioner. The introduction of new roles resulted in new systems of access/triage, which could raise issues for less informed patients, particularly older adults, vulnerable groups and infrequent attendees. Therefore, patient communication was highlighted in our study as an important factor in gaining patients’ acceptance and understanding of new roles. In addition, respondents in our focus groups told us that their acceptance of skill mix change was enhanced by a feeling that newer practitioners were well supported, and that GPs were available for expert advice if required. This suggests that, although surveys of patient satisfaction provide important evidence about patient experiences, the picture may be somewhat more nuanced than it initially appears. Qualitative interviews provide an important avenue for exploring in more depth the factors that underpin satisfaction with different types of appointments and, most importantly, the work involved for patients in skill mix changes in general practice.

Impact on outcomes

Although our own research (see Chapter 5) and previous research141,146 have shown that patients are generally satisfied with their encounters with nurses and other professionals, our study suggests that a higher FTE of professionals other than GPs is associated with overall lower satisfaction with patients’ interaction with their practices. 141,146 This may reflect the increasingly complex organisational processes required to ensure that patients are categorised correctly and matched with the correct professional. As we have described in Research question 3(iii): how does increased skill mix affect patients’ experiences when accessing primary care services?, such processes may also be burdensome for patients.

A study examining the correlation between nurse staffing and a range of indicators from the QOF, using data from the second year of the scheme, found that a higher level of nursing staff was associated with better performance on several clinical domains of the QOF. 147 We examined more recent data from the QOF and included a wider range of staff. Although we did not examine QOF achievement in particular disease areas, we considered a wider range of outcomes.

We found that a higher FTE of CPs is associated with higher prescribing quality and lower prescribing volume. There is a considerable volume of previous research (see Rationale for employing newer roles) exploring the role of CPs in primary care practices. Our findings are in accord with a Cochrane review148 from 2010, which explored the evidence relating to the impact of pharmacist-led interventions in outpatient and community settings. The review148 found that pharmacist-led services reduced the total number of medications prescribed and therapeutic duplication. Ethnographic research into influences on prescribing in general practice found that practice-level ‘macro’ policies (e.g. practice-wide policies about what to prescribe in certain circumstances) effectively improved prescribing quality, and it may be that the presence of CPs supports the development of such policies. 149 An evaluation of a programme in one CCG in the West Midlands found that CPs employed in practices were able to undertake activities such as medication reviews, audits and planned medication changes to improve cost-effectiveness. 150 These interventions were deemed to have both saved money and saved GPs time. However, data collection was subjective (relying on practice reports). Estimations of cost were based solely on cost reductions arising out of medication changes, rather than taking into account potential later costs associated with those changes (e.g. hospital admissions or need for new prescribing). A Cochrane review151 of interventions to reduce polypharmacy in older adults found no clear evidence of a beneficial effect of pharmaceutical care delivered by physicians, pharmacists or geriatricians. A realist synthesis of the conduct of post-discharge medication reviews by pharmacists in primary care concluded that such programmes work best when they are designed to promote collaboration and trust between professionals and when pharmacists take responsibility for outcomes. 152 This suggests (in keeping with our qualitative evidence) that the value of employing CPs in general practice is context dependent and requires careful service design.

Previous research153 has shown that increasing access to primary care may be associated with a decline in A&E attendance. Our study suggests that, although an increase in access to GPs and nurses is associated with a reduction in A&E attendances, the same is not true for other DPC staff.

Although we are not aware of any previous research that has looked at the association of GP satisfaction with the employment of a wider range of professionals, previous research154,155 has considered associations between GP burnout and other aspects of their practices. Workplace factors, such as workload and perceived control, were found to be drivers of primary care physician burnout. 154 Feeling unsupported and having a higher burden of administrative tasks were associated with an increased risk of burnout. 155 Our findings are congruent with this, in that it is likely that GPs working in an environment with a larger number of FTE GPs PTP will feel more supported, driving increased satisfaction. Moreover, it is possible that GPs working in an environment with a wider variety of DPC staff and fewer GPs have a relatively higher burden of administrative work (e.g. dealing with clinical decisions arising from patient letters and blood test results), which cannot be eased by delegation.

In terms of costs, previous research156 has suggested that employing ANPs may reduce costs and health-care utilisation; however, we are not aware of any other studies that have looked at overall primary care costs.

Study overall

Our study took place at a time when, as we have demonstrated, skill mix in general practice is changing rapidly. Therefore, it may be that both our case studies and our outcomes analysis are capturing the effects of early turbulence as practices and individuals get used to new ways of working. However, some of our case study practices had been working with a more diverse workforce for several years, and we found that the need for adaptation and flexibility were ongoing. Delays in obtaining outcomes data limited our ability to undertake more in-depth analyses of change over time. Our study also took place when general practice was on the cusp of a change from having newer types of practitioner predominantly employed by practices to a policy-driven shift towards financially supported, centrally employed staff moving between practices. Although we are unable to offer direct evidence about the impact of this shift, we can make some inferences, which are discussed below. In addition, this means that our examination of outcomes associated with different levels of skill mix may be capturing outcomes associated with newly employed practitioners. We have shown that there is a development over time as practitioners settle into their roles and develop new skills, and it will be important in future to explore whether or not this has an impact on outcomes.

Workforce data sets and outcomes data

Our analysis is based on national data, although we did not use data for 2020 because the COVID-19 pandemic had profound impacts on health-care activity levels and patient outcomes. No national data are currently available at practice level on the volume of consultations delivered. We created a measure of how recently a practice’s patients had seen a GP and how recently they had seen a nurse, on average, based on data from the 2017 GPPS. This is based on a sample of patients’ responses about each practice and is only a crude proxy for the utilisation of primary care.

There is potential for skill mix to be endogenous in the outcome equations if there are unmeasured factors associated with both skill mix and outcomes. We included a range of measured characteristics of practices and their populations to minimise the risk of omitted variable bias. We also examined changes over time, which controls for unmeasured but time-invariant characteristics of practices. Nonetheless, the risk of confounding from time-varying factors remains. We experimented with potentially exogenous variables (e.g. whether or not the CCG has offered incentives to employ new staff types, local labour and housing market conditions) in an instrumental variables framework, but these were not sufficiently predictive to provide accurate results. The results should be interpreted with this limitation in mind, and future work should focus on identifying better exogenous predictors of variation in skill mix.

Practice managers’ survey

The strengths of this study are in the collection of large-scale data on general practices’ motivation for employing skill mix and a detailed description of how skill mix policy is enacted in practice.

A limitation of the national practice manager survey was the response rate (i.e. 17%). However, we have implemented weighting to minimise any bias.

Another limitation is in the suspected inaccuracy of wMDS. However, our analysis has demonstrated that systematic bias is not present in our sample. 157

Case studies

As with all case study approaches, our aim was not for generalisation based on representativeness. Rather, our aim was to use theory to support generalisation from the particulars of our case study sites to the wider context of the NHS. Using this approach, we have shown how the particularities of each case required the development of locally specific routines, and we have drawn out from this a number of broader issues that practices need to consider in implementing skill mix, including categorisation and maintaining adaptability. The strength of this work lies in the multifaceted and in-depth data collected from observation in both clinical and administrative areas, which allowed us to capture a fully rounded picture of the implementation task.

A practical limitation of the case study work was the difficulty in getting patients involved with the study. We did not manage to collect as many patient surveys as planned, despite seeking practices’ support in distributing the survey. There was variable capacity and/or willingness within practices for getting reception staff to do this. Moreover, many patients needed the questions read out to them or more explanation of the purpose of the survey than reception staff were able to give. In addition, one of the sites (i.e. site C) did not have an active PPG nor were we able to recruit patients for a focus group at that site. We contacted all patients who indicated in the survey that they were willing to participate in the focus groups. However, only patients in site D agreed to participate in the focus group.

For research

• Further research is required to analyse changes in outcomes (and related overall costs) over time as larger numbers of newer practitioners settle into practice and develop their skills. It will be particularly important to track impacts on other measures of overall health service use.

• It will be important to monitor closely the implementation of skill mix via PCNs, with a close examination of the ways in which practices are adapting to staff who are only present for short periods of time, and identify factors that support this process and monitor the impact on outcomes.

• There is a need to track GP job satisfaction over time and to consider the factors affecting this. We have highlighted that newly qualified GPs may not feel ready to take on the required supervisory roles and other responsibilities.

• There is some evidence that the experience of working in a more diverse workforce affects attitudes, and it seems likely that GPs and managers exposed to newer practitioners may adjust their expectations about what their workforce should look like. These changes need to be explored over time, as they have implications for training and the supply of workers.

• The current ARRS offers incentives to employ a more diverse workforce at group level. Future research should consider the impact of ARRS in comparison with more direct investment in practices.

• Wider professional issues that have been identified in previous research (see Multidisciplinary working) and surfaced in our research remain partially understood:

  • Further research should explore the mechanisms that help mixed groups of professionals retain what is valued in their own professional identities and maintain engagement and motivation while working across professional boundaries.

  • It has been proposed that some GPs prefer to continue in roles closer to the comprehensive, generalist roles that attracted them to general practice, rather than working in roles more typical of consultants with an emphasis on dealing with complex cases, supervision, etc. Further research is needed to explore how different caseload preferences may be accommodated in GP settings.

  • The processes by which newly qualified but relatively inexperienced GPs progress to become more experienced and eventually capable of senior clinical roles are unclear. Further information about this is needed to inform the development of appropriate career pathways.

  • Further research is needed to determine how the categorisation of patients’ problems can be improved. It has been suggested that artificial intelligence may offer some solutions for this issue. However, our study suggests that patients may express their problems in ways that do not facilitate categorisation and, therefore, it is possible that artificial intelligence-based solutions will make this worse. The impact of such innovations in practice needs to be carefully studied.

For policy

Current policy is driving the implementation of skill mix across groups of practices within PCNs via a direct subsidy of the employment of a variety of newer types of practitioners. Our study demonstrates that there are complex challenges associated with the implementation of skill mix when it occurs at the practice level. It is not at all clear how these challenges will be managed when practitioners are employed centrally to work across more than one practice. Specific issues which will need to be considered:

• Policy should not assume that the deployment of newer practitioners across PCNs will provide a straightforward or complete solution to the current GP workforce and workload crisis. GPs in practices will need to supervise peripatetic workers carefully and the benefits for practices in terms of workload reduction are not currently clear.

• General practices need to manage the distribution of work through, for example, categorisation and matching of their practitioners and patients. Greater adaptability of various practitioners can be supported by having both formal and informal meetings (e.g. regular coffee breaks) between practitioners. However, this may be more challenging for peripatetic practitioners, such as PCN-employed staff who are only present in each practice on 1 or 2 days per week. PCNs could seek to limit the number of practices with which each practitioner interacts by, for example, deploying practitioners over a small number of closely linked practices rather than across the whole group. Moreover, PCNs should prioritise retaining staff so that there is time for practices and patients to adapt. It takes time for new roles to bed in, and this is likely to take longer when practitioners are only present in the practice for short periods of time.

• In facilitating the management of staff, there will need to be close interactions between practices and PCNs. Practices will need to agree on how individual practitioners will be expected to work, including their responsibilities and scope of practice, with close liaison if the scope of practice is to be changed. Line management and the respective responsibilities of practices and the PCN for oversight and performance management of staff will need to be clearly worked out and managed. There will need to be a clear understanding of who is liable with regard to practitioner performance. Regulation of new roles needs to be clarified. Moreover, dedicated funding or development support for PCN management should be prioritised.

• Change in outcomes associated with the ARRS will be very difficult to determine. The wMDS needs to clearly capture these new roles and demarcate the length of time that each practitioner employed across the PCN spends in each practice.

• Patients value continuity of practitioner and satisfaction may be affected by complex categorisation and matching processes. There may, therefore, be a trade-off between patient and practice satisfaction with new arrangements. Moreover, our study found increased attendance levels at A&E for practices with higher levels of non-GPs. It is important that policy allows time for adjustments to be made and accommodations reached before making changes in response to this.

• There is stronger evidence of benefit from the deployment of CPs and ACPs, suggesting that these roles should be prioritised.

For practice

• General practices need to understand how they can choose between practitioner types (if there is sufficient supply in their locality) and how they can gain the greatest benefit from skill mix.

• Practice managers need to understand how to manage a diverse skill mix workforce and develop functional processes to optimise categorisation and matching. Practice managers may require additional training for this. PCNs could play a role in training practice managers and working with practices to support them in integrating new roles into their practice, but PCNs will require dedicated management resources to achieve this.

• The training and development of newer types of practitioners needs to be carefully managed and optimised over time. Those responsible for training different practitioners need to work together, understanding the difference between role competencies and role-holder competencies. Preparation for working in an isolated setting and for working collaboratively within a diverse team need to be prioritised.

• Regulatory arrangements need to be clarified and made more transparent, particularly in relation to practitioners employed across more than one practice.

• Practices need dedicated time to allow training and team development. It is unclear how this will work with regard to peripatetic members of staff who may need to attend meetings in multiple practices. The creation of small locality groups of two to three practices within PCNs may help this.

For practitioners

• Practitioners need to communicate about, and feel comfortable with, their scope of practice, and practices should seek to generate an ethos in which it feels ‘safe’ for each individual to be candid about knowledge, skills and aptitudes.

• In the context of dealing with undifferentiated caseloads, practitioners in all roles need to recognise their strengths and limitations. Practitioners may find it difficult to identify unconscious gaps, but participation in sensitive and supportive case reviews can be used to reveal deficits and improve future performance.

• Both acceptability of seeking assistance from colleagues and the practicalities of getting assistance need to be fully supported by all practitioners to make this a normal and everyday practice that underpins patient safety.

• Experiential learning is a long-established facet of clinical work. The value and service enhancement of such learning should be recognised and appropriately rewarded for all practitioners.

For patients

• Increasing diversification of the practitioner workforce will, in time, become more familiar to patients. Discussions with our patient and public involvement representatives about how this might look have recognised that the deeply contextual nature of how individual practitioners work in specific settings makes it difficult to design a universally applicable solution. Individual practices will need to work closely with their PPG to generate ways of effectively communicating how appointment booking systems work or what different practitioners can do.

• Categorisation depends on effective communication by patients of the nature of their problem, which is likely to present greater difficulties for those not familiar with medical conditions and a more significant problem for patients who, for any reason, are more vulnerable to increased health inequalities. Systems to identify such patients should be considered for inclusion in the initial assessment. Once such difficulties have been confirmed, they can be allocated to a receptionist with enhanced training for an individually tailored assessment.

• Building patients’ satisfaction, confidence and trust in new systems requires time and the willingness of both practitioners and patients to adapt to the new arrangements. Factors that can increase patient confidence and trust include longer appointment times and ensuring that non-GPs have access to advice from more experienced colleagues when uncertain of how to proceed.

Participants

We would like to thank all the patients and staff who took time out of their extremely busy workdays to participate in our study.

Academic colleagues

We would like to acknowledge the collaborative engagement and support of academic colleagues, particularly whose expertise and specialist knowledge have assisted our quantitative analysis.

Expert Advisory Group

We would like to thank members of the Expert Advisory Group who offered support and feedback on the content and progress of the study: Kirsty Roberts (Workforce Lead for Primary Care, Greater Manchester Health and Social Care Partnership), Stephen Heaney (Practice Manager, Oldham CCG Board Member), Myra Upton (President of the Association of Medical Secretaries, Practice Managers, Administrators and Receptionists), Rachel Souter (Head of the General Practice Workforce Programme, NHS England), Dan Smith (Paramedic Practitioner, Manchester), Kellie Johnson (Practice Nursing and Care Navigation Training and Deployment, Stoke CCG), Zoe Ahearne (ANP), Sunil Aggarwal (deputy for Drew Tarmey, PA Training Course, Manchester), Ceinwen Mannall (National Lead for the General Practice Pharmacist Training Pathway), Judith Hogg (PRIMER member) and Ailsa Donnelly (PRIMER member).

Patient and public involvement

We would like to thank PRIMER and the PPG for providing a patient and public perspective. We would especially like to thank two PRIMER representatives, Ailsa Donnelly and Judith Hogg, for providing feedback throughout the project.

Administrative support

Kevin Sanderson Shortt (project manager) provided project management support (September 2018–September 2019).

Danielle Harris (project support administrator) provided administrative support (December 2019–December 2020).