It is critical that PCNs are not set up to fail by taking on too many tasks too quickly
Primary care networks (PCNs) are a key part of the NHS Long Term Plan. Based on the plan, all practices are required to be in a network of around 30-50,000 registered patients by June 2019, and clinical commissioning groups (CCGs) are required to commit recurrent funding to develop and maintain them. 
There are also numerous expectations of general practice working at scale. Networks will employ multi-disciplinary and integrated community-based teams, alleviating some GP and nurse shortage pressures, and provide proactive care for the patients with the most complex needs. These changes, along with better use of digital consultations and social prescribing, are expected to improve care and outcomes.
We have reviewed the research on GP collaborations, and there are three areas which we believe policymakers should consider before more detailed guidance is developed on PCNs. [2-5]
What can networks achieve, and when?
Surveys of existing GP collaborations have revealed that they have a wide array of aims – many stating multiple goals including improving access, sustainability, and shifting services into the community. [6,7]
Evaluations of the various archetypes of GP collaboration have found that the length of time to form spans years and exceeds expectations, and that many collaborations find it difficult to define and achieve their aims. [2,5,8]
This suggests that, rather than overwhelming PCNs with becoming employers, budget holders, and new service providers, there may be value in networks initially focusing on one or two clearly defined projects (and having discretion over governance structures and local performance levers). As new projects are added, layer by layer over many years, most PCNs should appear more like the networks described in the plan.
Should they be geographically mandated?
While the formation of GP networks elsewhere in the UK has been mandated by local health boards, England has allowed voluntary formation, resulting in a mixed market of collaborations of various geographies and sizes. [6,7]
There are both pros and cons in mandating membership of PCNs. Mandating membership can provide clear accountability chains and ensure coverage across the country. But it can also lead to GP disengagement and could disturb the ambition and goodwill of existing collaborations that are based on like-mindedness and trusting relationships.
Evidence from the development of integrated care systems shows that key factors for successful collaborations include developing trusting relationships, a shared vision, common values, and good leadership. 
If PCN membership is geographically mandated, it would be wise to allow networks to democratically define their own aims and activities, and for practices to opt-in to shared network activities.
What’s the right size?
The evaluation literature suggests that there is no clear consistent relationship between the size of healthcare collaboration and performance or quality. [4,5]
The Long Term Plan suggests that 30,000-50,000 registered patients is the right size, borrowing from the primary care home model. 
We traced the evidence for this number back to studies from the experience of GP fundholding and total purchasing pilots, which suggested this was the risk pool needed for the purchase of secondary care services. 
It was not related to provision of primary care itself. Some also point to Hill and Dunbar’s maximum number of around 150 members in a social network to retain familiarity with one another, presumably leading to better working.  However, the optimal size of a GP network is more likely to depend on its intended functions and on how performance is measured, so flexibility may be needed.
It is critical that PCNs are not set up to fail by taking on too many tasks too quickly. Local leaders must be supported to deal with issues such as risk, indemnity, governance, data-sharing and organisational development. The risk otherwise is that disillusionment following over-inflated expectations, as well as time-limited funding and support, leaves general practice in a worse position.
Stephanie Kumpunen, senior fellow in health policy at the Nuffield Trust.
Beccy Baird, senior fellow in policy at the King’s Fund.
Competing interests: None declared.
2. Kumpunen, S. Rosen, R. Kossarova, L. Sherlaw-Johnson, C. (2017). “Primary Care Home: Evaluating a new model of primary care” Research report. Nuffield Trust.
3. Pettigrew, L.M., Kumpunen, S., Mays, N., Rosen, R. and Posaner, R. (2018). The impact of new forms of large-scale general practice provider collaborations on England’s NHS: a systematic review. Br J Gen Pract, p.bjgp18X694997.
4. Pettigrew, L.M., Kumpunen, S., Rosen, R., Posaner, R. and Mays, N. (2019). Lessons for ‘large-scale’ general practice provider organisations in England from other inter-organisational healthcare collaborations. Health Policy. 123(1), pp. 51-61.
5. Rosen, R., Kumpunen, S., Curry, N., Davies, A., Pettigrew, L., and Kossarova, L. (2016). Is bigger better? Lessons for large-scale general practice. Research report. Nuffield Trust.
6. Kumpunen, S., Curry, N., Farnworth, M., Rosen, R. (2017). Collaboration in general practice: Surveys of GP practice and clinical commissioning groups. Slide pack. Nuffield Trust, Royal College of General Practitioners.
7. Kumpunen, S., Curry, N., Ballard, T., Price, H., Holmes, M., Edwards, N. (2015). Collaboration in general practice: surveys of GPs and CCGs. Slide pack. Nuffield Trust, Royal College of General Practitioners.
8. Goodwin, N., Mays, N., McLeod, H., Malbon, G., Raftery, J. (1998). Evaluation of total purchasing pilots in England and Scotland and implications for primary care groups in England: personal interviews and analysis of routine data. BMJ, 317(7153), pp.256-259.
9.Charles, A., Wenzel, L., Kershaw, M., Ham, C., Walsh, N. (2018). A year of integrated care systems. Reviewing the journey so far. Research report. King’s Fund.
11. Ham, C. (2010). GP budget holding: Lessons from across the pond and from the NHS. Research report. University of Birmingham.
12. Hill, R.A. and Dunbar, R.I. (2003). Social network size in humans. Human nature, 14(1), pp.53-72.