NeoNet: Providing a national demand/capacity model for neonatal care in England

South West PeninsulaReproductive Health and Childbirth
Start Date: 1 Sep 2015 End Date: 30 Jun 2017


Funded by the NIHR Health Services and Delivery Research (HS&DR) Follow-on Studies programme from the 1st September 2015 to 28th February 2017, this project aims to understand the current and potential future national patterns of neonatal care in England. This will be achieved by studying location of demand and care, acuity of care, and cost of care to service providers and parents. Modelling techniques will be used to investigate the trade-offs between different service configurations with varying degrees of centralisation. Extensive analysis will be conducted to assess the economic aspects of different scenarios, with patient consultation forming a key part of the project. The planned research is a collaboration between PenCHORD and the Health Economics Group at the University of Exeter Medical School, the Patient-Public Involvement group based at the University of the West of England, the Neonatal Data Analysis Unit (NDAU) based at Imperial College, London, with the Neonatal Critical Care Clinical Reference Group acting as steering body.

A previously funded NIHR HS&DR project developed a discrete event simulation model to mimic the behaviour of a regional neonatal network. From this model, we were able to predict:

  • Average and peak workloads in units

  • Distances from parents home

  • Time spent above nominal BAPM guideline work levels

  • Resource utilisation

  • Number and distances of transfers

  • Nursing costs

  • Demand derived from ‘Badger’ neonatal data or nationally available GP profile data

  • Performance with different network configurations (e.g. closure or re-designation of units)

This model has been shown to have very good predictive ability.  However, whilst answering key questions at a regional level, it raised further questions at the national level, which this current project aims to answer.


Flow of patients through care levels. Patients are cared for in the closest hospital to home that can meet the clinical needs of the patient and that has spare capacity to accept new admissions.


Research questions

1. What is the nature of the trade-off between unit throughput and parent travel distances?

Centralisation of services has potential benefits; i) increased throughput of infants, leading to increased specialism and expertise; and ii) reduction in ‘spare capacity’ needed to deal with peaks and troughs in workload. However, centralisation will increase the average distance parents must travel to the point of care. The effects of centralisation in neonatal care is complicated by the transition of the infant through different levels of care (each of which may have a different degree of centralisation) and by the organisation of units into networks of units.

To explore these complexities, this project will addressing the following:

  • What effect does network size have on travel distances and times?

  • How would service reconfiguration affect the trade-off between throughput and parent travel distances? How would the number and distance of transfers be affected?

  • What is the average and maximum planned distance and travel time from parents’ home to point of care? How does this vary across the country? How does this compare with actual distances and travel times?

  • What effect do network boundaries have on travel distances and times? What happens to travel distances if network boundaries are removed?

  • How may conflicting objectives be balanced against each other?

  • What is the relationship between the number of units and the expected travel times and throughputs? Given any fixed number of units, which locations would minimise travel times?

  • What is the expected impact regionally and nationally of applying population projections?

2. How will costs per infant and outcome change with reconfiguration of services?

During our regional project we found the use of Healthcare Resource Group (HRG) reference costs had limited value, as they assume a fixed infant cost regardless of unit size. In the regional model we allowed for variation in nursing costs, depending on configuration. However, it is necessary to understand neonatal costs in significantly more detail to better predict the likely relationship between service configuration and NHS costs. Using individual responses to a recent BLISS survey on costs to parents, we will seek to better model the relationship between network configuration and costs to parents.

We will address the following issues:

  • What components, and in what proportion, contribute to the costs of the different types of neonatal unit?

  • How would changes in the degree of centralisation of services affect the ‘spare capacity’ needed to deal with peaks and troughs in workload? How would total costs be affected?

  • How would changes in the degree of centralisation of services affect parent travel distances and costs?

  • How does the degree of centralisation affect the requirement for local accommodation for parents?

Project activity

The model will be based on discrete event simulation. In this type of simulation each computer-generated infant exists as an independent object in the model and has associated details (gestational age at birth, entry level of care, home hospital, etc). The simulation runs through time and takes into account the variability experienced in the system (e.g. random occurrence of births with varying needs, fluctuating availability of staff). For each hospital the number of cots, the number of nurses and the highest level of care each cot is capable of supporting will be specified.

The distribution of gestational age, entry level and duration of care initially required will be as observed in the NDAU historic data set. Geographic location of demand will be based on the home location of the mother. Infants will be divided into categories based on gestational age. The probabilities of requiring particular levels of care and the duration of care will be assigned according to gestational age category. Lengths of stay will be sampled from distributions so that infant-to-infant variability (even at the same gestational age) is incorporated into the model.

The model will then seeks to find a suitable cot in the following order:

  • Cot in hospital closest to mother’s home location.

  • Closest available cot in mother’s neonatal network.

  • Closest cot outside of mother’s neonatal network.

In order for a hospital in the model to accept an infant there must be an available cot capable of the level of care required and sufficient nursing staff. Infants stay at a defined level of care for a given time (sampled from a distribution based on their gestational age), before transitioning to another level of care or exiting the network. There are opportunities to move infants closer to their home hospital when space becomes available.

The model will be run under different configurations (e.g. what if the number of intensive care units were reduced, or what if the number of total neonatal units were reduced). The effect of possible changes on both units and parents will be forecast.

Anticipated outputs

The key outputs of the model will be:

  • Daily variation in workload at each hospital (for each care level). This includes number of infants and the nurse workload.

  • The proportion of time units work above British Association of Perinatal Medicine (BAPM) workload guidelines and the proportion of time units are closed to new admissions.

  • The proportion of time infants are cared for away from the most appropriate hospital (the closest hospital to home that is able and where resources are available, to care for the infant).

  • The distances from mother’s home location to the location of care (including the number of infants cared for more than 50km from home; assumed to be a maximum reasonable daily commuting distance).

  • The number of infants that cannot be cared for within the mother’s home neonatal network.

  • The number and distances of transfers.

We are also developing costs of care into the model. At a minimum this is the nursing cost for any planned configuration of care, but we are seeking to include as many additional costs as possible.

Professor Martin Pitt