Add "why you should not compare 22/23 outputs from a 19/20 based model to actual 22/23 activity volumes"

[yiwen-h]

The NHP model is a baseline adjustment model. Such models take a baseline dataset and apply a series of changes to that baseline based on expectations about a range of factors that might be expected to affect that baseline. In our model primarily these are the impact of demographics and the impact of strategies or interventions aimed at avoiding certain types of admissions or improving the efficiency (LOS) of certain types of activity.

Baseline adjustment models are predicated on the use of a baseline that is considered “normal”. Such models are intended to provide medium to long term estimates of future activity and cannot and are not intended to take account of short term or one off unexpected or unusual factors that result in material shifts in the size or shape of activity (the Covid pandemic would be a good example of this). Therefore we would expect that any results for 22/23 resulting from our model using a 19/20 baseline will likely be very different from the actual size and shape of activity seen in 22/23 as our model does not incorporate any adjustment factors designed to replicate the impact of the pandemic.

Importantly this does not mean that the longer term model estimates are wrong as long as we assume that 19/20 remains a reasonably “normal” year.

Arguably the pandemic may have had an enduring impact on the size and shape of activity such that 19/20 may no longer be a reasonably normal year as a basis for future planning. However even if the view is taken that 19/20 is no longer a reasonably "normal " year for the purposes of future planning this still does not mean that it cannot be used as a baseline as long as any understanding of recent changes are considered and taken into account as part of setting mitigator assumptions. For example average LOS compared to 19/20 baseline has increased since the pandemic. If the view is taken that this is an enduring change then a Trust may wish to moderate its assumptions about how much it might be able to reduce LOS from the 19/20 baseline. Similarly a large increase in the use of virtual outpatient appointments has been since the pandemic so any assumptions about future levels of virtual appointments should clearly need to take account of the changes that have already occured since 2019/20.

There may be changes to how data is recorded that again baseline adjustment models cannot account for . One example of this is changes in recording of SDEC activity that in 2019 was typically recorded as an inpatient spell but is now recorded within the ECDS dataset. Local knowledge of how and where activity was recorded within the baseline year should allow the model outputs to be used to identify and accomodate that activity that would be considered SDEC.