Results
Note Internal NHSE documentation for the results of Phase 1 of the work (March-April 2023) and Phase 2 of the work (April-May 2023) not made public. Outputs from the Toy / Fake scenarios shown below to exemplify outputs.
The "foundation" outputs of the model are the Log of Arrivals (log_arrivals.xlsx) and the Log of Attributes (log_attributes_w.xlsx).
These will be saved if flag_savedata_fulllogs <- TRUE in config.R (a reason to set this to False may be avoiding to save large files unnecessarily).
All relevant graphical visualisations and summary metrics can likely be created from these.
The script post-processing.R already has specified and produces a range of outputs based on those logs, both in table and visual format. These are shown below.
Alternatively, the end-user is free to do their own post-processing with the logs given.
Example outputs from Fake_Data_1A
This pre-created scenario uses a fictitious set of parameters for demand, capacity and job cycle times for a smaller scale toy example. It creates a step disruption in A&E bays available from the 8th day (reducing bays by 8, from 73 to 65).
Example outcomes from parameters/Fake_Data_2/ scenario , with n_AEbays <- 73, T_AE <- 6.5, flag_LoS_file <- TRUE, dn_AEsupplyshock <- -8.
Outputs can be seen in example_outputs/Fake_Scenario2_1.
Demand and Capacity
Figure Red traces represent queues forming. Continuous blue traces represent occupancy, with dashed blue traces showing capacity. The multiple traces relate to the 10 replications run. After 7 days, a disruption is imposed by reducing A&E bays. The effect of this on queues, both for the A&E bays and for the upstream ambulance activity, can be seen.
Response time KPIs
Figure Mean response time KPIs per category and per simulation time-steps. Boxplots showing inter-replication variation. Dashed line is dummy, can be used to benchmark.
Handover KPIs
Figure Handover KPIs per simulation time-steps. Boxplots showing inter-replication variation. Dashed line is dummy, can be used to benchmark.
Events and pathway visualised
NOTE: Contrary to the other outputs shown, for the code used to generate these static and animated pathways, please refer to our Process Mining repository [TBA]. Subfolder [TBA] shows scripts for this AmbModelOpen use case.
Pathway 'mean times' and branching % - week 1 vs week 2
Week 1 (before disruption) - % per branch
Week 1 (before disruption) - mean time
Week 2 (ongoing disruption) - % per branch
Week 2 (ongoing disruption) - mean time
Pathway animated - first week (before disruption)
This animation only relates to days 1-7, so before step disruption. Calendar dates are fictitious. Colours relate to care model.
Embedded below or seen in full screen here.
Pathway animated - scenarios with and without disruption
This animation is similar to the previous one, but animates for the two weeks. Also, the top branch (B-Base scenario) shows the case where no disruption occurs at all. The bottom branch (D-Disrupted scenario) shows the case where step disruption occurs from day 7. Colours relate to breaches (of either category2 response time at 90th %ile, handover of 30+ minutes or both).
Embedded below or seen in full screen here.
Example outputs from Fake_Data_2
This pre-created scenario uses a fictitious yet realistic set of parameters for demand, capacity and job cycle times for a toy ambulance trust.
Example outcomes from parameters/Fake_Data_2/ scenario , with n_AEbays <- 1500, T_AE <- 6.5, flag_LoS_file <- FALSE, dn_AEsupplyshock <- 0.
Outputs can be seen in example_outputs/Fake_Scenario2_1.
Demand
Figure Demand per category and conveyance (colour-coded ; 1 - see and convey; 2 - see and treat; 3 - hear and treat; 99 - direct ED)
Demand and supply
Figure Incident demand per category (colour-coded ; categories 1-4). 3 replications shown. Direct ED demand excluded. Black trace - input demand. Red-trace - input DSV supply
Resource utilisation KPIs
Instantaneous utilisation, capacity and queue traces
Figure Instantaneous resource usage for respectively ambulance and A&E bay. Dashed blue - server capacity ; Solid blue traces - server use ; Solid red traces - queues . Multiple traces due to n=10 replications.
Utilisation KPIs - batch, overall simulation time
Figure Utilisation and queue for A&E bay and ambulance. Average and standard deviation.
Response time KPIs
Figure Mean response time KPIs per category and per simulation time-steps. Boxplots showing inter-replication variation. Dashed line is dummy, can be used to benchmark.
Table Response time KPIs (mean, median, 90th) averaged across simulation time and replications (mean, standard deviation)
Handover KPIs
Figure Handover KPIs per simulation time-steps. Boxplots showing inter-replication variation. Dashed line is dummy, can be used to benchmark.
Table Handover KPIs averaged across simulation time and replications (mean, standard deviation)
Select Job Cycle Time KPIs
Table KPIs 'Time to allocate' , 'Time for pre-handover', 'Total job cycle time', 'Response time', per conveyance. Mean and median.
Call queue KPIs
Figure Average call queue KPI per simulation time-steps (6 hour window). Boxplots showing inter-replication variation.
(Hospital) site queue KPIs
Figure Average site queue KPI per simulation time-steps (24 hour window). Boxplots showing inter-replication variation.
