Page 236 - Data Science Algorithms in a Week

220 Abdulrahman Albar, Ahmad Elshennawy, Mohammed Basingab et al.

and 100, the peak of the surface representing the highest scores for ED workload
becomes smaller, and areas of the surface representing increases in ED workload become
isolated in the plot, as higher values for average patient complexity and occupancy rate
become necessary to achieve high values for ED workload. This represents the impact
that increasing ED staffing to adequate levels has on ED workload, even when average
patient complexity and occupancy rates are high. There are always areas of the surfaces
where ED workload is high, however when ED staffing is increased, ED workload can be
said to decrease even for moderate values of its other two inputs.
Figure 24 consists of surfaces a through k of subsystem IV, showing the impact that
the inputs of boarding and demand have on the output of crowding, when the variable
workload is held at eleven constants, 0, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100. In
surfaces a through c, when workload is low, crowding generally increases with boarding
and demand, however the peak values in surfaces b and c differ from surface a. The peak
of the surface decreases in size and transitions into a plateau in surfaces b and c,
indicating a wider range of input values that lead to the same high level of crowing.
In surfaces d through g when workload is between 30 and 60, the lower values of the
surface become more isolated, and all points on the surfaces appear to rise, representing
an overall increase in crowding for all values of boarding and demand. It can be observed
that increasing the ED workload evenly increases crowding under any condition of
boarding and demand.
As workload approaches values between 70 and 100, surfaces h through k show that
crowding continues to generally increase for all boarding and demand values, and the
surfaces peak at higher values. A plateau emerges in surface h, where crowding remains
constant for boarding values which exceed 0.2, when demand is below 50. Beyond
boarding values of 0.2, crowding will only increase when demand is increased beyond
50. This demonstrates that under high workload, there are consistent levels of crowding
when boarding is high, but demand is low. Only when both boarding and demand are low
does crowding achieve minimum values under high workload.

Figure 23: Sensitivity analysis subsystem III. Figure 24: Sensitivity analysis subsystem IV.