The latest changes to the Mental Health Act mean there is an increasing need for National Health Service (NHS) services to carefully review their provision of health-based places of safety (HBPOS). As little as 15 years ago, when police used their emergency powers under section 136 (s136), more than one-third of people were held in police cells while awaiting assessment. ¹ After years of effort to reduce this practice, the latest amendments remove police cells as options for detention. It is more important than ever that NHS capacity is available to meet demand for these emergency assessments.

There is increasing national support for effective demand and capacity modelling by services, but common approaches can require detailed historical data, significant ongoing project support, or access to advanced technical skills to implement methods such as discrete event simulations. ^{Reference Paton and Tiffin2} Thankfully, the situation for s136 assessments is sufficiently simple that useful capacity estimates can be made more simply, assuming that s136 detentions occur independently of one another, and that the average rate of detentions is constant over time. Modelling before implementation can guide innovative approaches to improve provision without unnecessary expenditure.

A free online calculator (https://psychoxmodels-poscapacity.share.connect.posit.cloud) allows estimation of how many assessments will exceed available capacity based on an average monthly level of demand and the number of HBPOS available. This information may well be readily available from NHS providers or other stakeholders such as Approved Mental Health Professional services or police. The Poisson model code is also available for review or adjustment (https://github.com/psychoxmodels/POScapacity).

A notional locality expecting on average one assessment daily might seem well resourced with two HBPOS available. However, using the calculator, we find that despite this locality having the capacity to see twice the average daily demand, inherent randomness means that local services should actually expect to be unable to accept 37 assessments annually, having to divert referrals elsewhere roughly once a fortnight. Various approaches might reduce the unmet demand, and their probable effects can be modelled in advance. Local solutions to reduce unmet demand could be considered: commissioning one further HBPOS could improve matters substantially, with only nine assessment requests needing diversion per annum, although this would have substantial resource implications. Alternatively, to achieve similar benefits, initiatives to reduce demand for assessments would need to generate a reduction of 40% in s136 detentions. Such a substantial change in demand could prove challenging to implement – such benefits have been reported for the introduction of street triage systems, ^{Reference Rodgers, Thomas, Dalton, Harden and Eastwood3} but such schemes are already widely in place, so their benefits may already have been realised. ^{Reference Kirubarajan, Puntis, Perfect, Tarbit, Buckman and Molodynski4}

A possible system solution would be to operate in concert with a neighbouring locality by developing a shared facility ⁵ or – as is more likely – making operational changes with a shared single point of access for referrals and allocation to available HBPOS across either area. Strikingly, if two such similar areas worked in concert, this change alone should mean that each area would expect to reduce missed assessments by 27% before provision of any additional capacity or reduction in external demand. Three localities combining forces could expect a reduction of more than 80% in the need for diversion elsewhere.

The assumptions underlying the simple model may not always apply; for example, differing availability of alternative crisis care pathways by day of the week may cause predictable fluctuations in rate, with periods of higher excess demand than expected from the overall monthly levels. Estimates can be updated readily as monthly rates change. In practice, a combination of approaches might well be preferred depending on local circumstances. The time is right for services to review their HBPOS provision to meet this ongoing demand, and the tools are now available to model the probable effects of changes before implementation to support rational service design.